: (section contents) * * 1 byte number from 0 to 254 identifying the section * * 1 byte SNF_REQUIRED: this section is required for correct * spell checking * * 4 bytes length of section contents, MSB first * * 1 byte SN_END * * * sectionID == SN_INFO: * N bytes free format text with spell file info (version, * website, etc) * * sectionID == SN_REGION: ... * 2 bytes Up to 8 region names: ca, au, etc. Lower case. * First is region 1. * * sectionID == SN_CHARFLAGS: * * 1 byte Number of bytes in (should be 128). * N bytes List of flags (first one is for character 128): * 0x01 word character CF_WORD * 0x02 upper-case character CF_UPPER * 2 bytes Number of bytes in . * N bytes Folded characters, first one is for character 128. * * sectionID == SN_MIDWORD: * N bytes Characters that are word characters only when used * in the middle of a word. * * sectionID == SN_PREFCOND: ... * 2 bytes Number of items following. * : * 1 byte Length of . * N bytes Condition for the prefix. * * sectionID == SN_REP: ... * 2 bytes number of items, MSB first. * : * 1 byte length of * N bytes "from" part of replacement * 1 byte length of * N bytes "to" part of replacement * * sectionID == SN_REPSAL: ... * just like SN_REP but for soundfolded words * * sectionID == SN_SAL: ... * 1 byte flags for soundsalike conversion: * SAL_F0LLOWUP * SAL_COLLAPSE * SAL_REM_ACCENTS * 2 bytes number of items following * : * 1 byte length of * N bytes "from" part of soundsalike * 1 byte length of * N bytes "to" part of soundsalike * * sectionID == SN_SOFO: * 2 bytes length of * N bytes "from" part of soundfold * 2 bytes length of * N bytes "to" part of soundfold * * sectionID == SN_SUGFILE: * 8 bytes time in seconds that must match with .sug file * * sectionID == SN_NOSPLITSUGS: nothing * * sectionID == SN_WORDS: ... * N bytes NUL terminated common word * * sectionID == SN_MAP: * N bytes String with sequences of similar characters, * separated by slashes. * * sectionID == SN_COMPOUND: * ... * 1 byte Maximum nr of words in compound word. * 1 byte Minimal word length for compounding. * 1 byte Maximum nr of syllables in compound word. * 2 bytes COMP_ flags. * 2 bytes number of following * N bytes Flags from COMPOUNDRULE items, separated by * slashes. * * : * 1 byte length of * N bytes end or begin chars from CHECKCOMPOUNDPATTERN * * sectionID == SN_NOBREAK: (empty, its presence is what matters) * * sectionID == SN_SYLLABLE: * N bytes String from SYLLABLE item. * * : * * : * * : * * * : ... * * 4 bytes Number of nodes following. MSB first. * * : ... * * 1 byte Number of siblings in this node. The siblings * follow in sorted order. * * : [ * | [] [] [] * | [] ] * * 1 byte Byte value of the sibling. Special cases: * BY_NOFLAGS: End of word without flags and for all * regions. * For PREFIXTREE and * follow. * BY_FLAGS: End of word, follow. * For PREFIXTREE , * and follow. * BY_FLAGS2: End of word, and * follow. Not used in PREFIXTREE. * BY_INDEX: Child of sibling is shared, * and follow. * * 3 bytes Index of child for this sibling, MSB first. * * 1 byte byte value of the sibling. * * 1 byte bitmask of: * WF_ALLCAP word must have only capitals * WF_ONECAP first char of word must be capital * WF_KEEPCAP keep-case word * WF_FIXCAP keep-case word, all caps not allowed * WF_RARE rare word * WF_BANNED bad word * WF_REGION follows * WF_AFX follows * * 1 byte Bitmask of: * WF_HAS_AFF >> 8 word includes affix * WF_NEEDCOMP >> 8 word only valid in compound * WF_NOSUGGEST >> 8 word not used for suggestions * WF_COMPROOT >> 8 word already a compound * WF_NOCOMPBEF >> 8 no compounding before this word * WF_NOCOMPAFT >> 8 no compounding after this word * * 1 byte bitmask of: * WFP_RARE rare prefix * WFP_NC non-combining prefix * WFP_UP letter after prefix made upper case * * 1 byte Bitmask for regions in which word is valid. When * omitted it's valid in all regions. * Lowest bit is for region 1. * * 1 byte ID of affix that can be used with this word. In * PREFIXTREE used for the required prefix ID. * * 2 bytes Prefix condition number, index in list * from HEADER. * * All text characters are in 'encoding', but stored as single bytes. */ /* * Vim .sug file format: * * * * : * * 6 bytes "VIMsug" * 1 byte VIMSUGVERSION * 8 bytes timestamp that must match with .spl file * * * : (see above, no flags or region used) * * * : ... * * 4 bytes number of following * * : ... NUL * * : X bytes word number that results in this soundfolded word, * stored as an offset to the previous number in as * few bytes as possible, see offset2bytes()) */ #include "vim.h" #if defined(FEAT_SPELL) || defined(PROTO) #ifndef UNIX /* it's in os_unix.h for Unix */ # include /* for time_t */ #endif #define MAXWLEN 250 /* Assume max. word len is this many bytes. Some places assume a word length fits in a byte, thus it can't be above 255. */ /* Type used for indexes in the word tree need to be at least 4 bytes. If int * is 8 bytes we could use something smaller, but what? */ #if SIZEOF_INT > 3 typedef int idx_T; #else typedef long idx_T; #endif #ifdef VMS # define SPL_FNAME_TMPL "%s_%s.spl" # define SPL_FNAME_ADD "_add." # define SPL_FNAME_ASCII "_ascii." #else # define SPL_FNAME_TMPL "%s.%s.spl" # define SPL_FNAME_ADD ".add." # define SPL_FNAME_ASCII ".ascii." #endif /* Flags used for a word. Only the lowest byte can be used, the region byte * comes above it. */ #define WF_REGION 0x01 /* region byte follows */ #define WF_ONECAP 0x02 /* word with one capital (or all capitals) */ #define WF_ALLCAP 0x04 /* word must be all capitals */ #define WF_RARE 0x08 /* rare word */ #define WF_BANNED 0x10 /* bad word */ #define WF_AFX 0x20 /* affix ID follows */ #define WF_FIXCAP 0x40 /* keep-case word, allcap not allowed */ #define WF_KEEPCAP 0x80 /* keep-case word */ /* for , shifted up one byte to be used in wn_flags */ #define WF_HAS_AFF 0x0100 /* word includes affix */ #define WF_NEEDCOMP 0x0200 /* word only valid in compound */ #define WF_NOSUGGEST 0x0400 /* word not to be suggested */ #define WF_COMPROOT 0x0800 /* already compounded word, COMPOUNDROOT */ #define WF_NOCOMPBEF 0x1000 /* no compounding before this word */ #define WF_NOCOMPAFT 0x2000 /* no compounding after this word */ /* only used for su_badflags */ #define WF_MIXCAP 0x20 /* mix of upper and lower case: macaRONI */ #define WF_CAPMASK (WF_ONECAP | WF_ALLCAP | WF_KEEPCAP | WF_FIXCAP) /* flags for */ #define WFP_RARE 0x01 /* rare prefix */ #define WFP_NC 0x02 /* prefix is not combining */ #define WFP_UP 0x04 /* to-upper prefix */ #define WFP_COMPPERMIT 0x08 /* prefix with COMPOUNDPERMITFLAG */ #define WFP_COMPFORBID 0x10 /* prefix with COMPOUNDFORBIDFLAG */ /* Flags for postponed prefixes in "sl_pidxs". Must be above affixID (one * byte) and prefcondnr (two bytes). */ #define WF_RAREPFX (WFP_RARE << 24) /* rare postponed prefix */ #define WF_PFX_NC (WFP_NC << 24) /* non-combining postponed prefix */ #define WF_PFX_UP (WFP_UP << 24) /* to-upper postponed prefix */ #define WF_PFX_COMPPERMIT (WFP_COMPPERMIT << 24) /* postponed prefix with * COMPOUNDPERMITFLAG */ #define WF_PFX_COMPFORBID (WFP_COMPFORBID << 24) /* postponed prefix with * COMPOUNDFORBIDFLAG */ /* flags for */ #define COMP_CHECKDUP 1 /* CHECKCOMPOUNDDUP */ #define COMP_CHECKREP 2 /* CHECKCOMPOUNDREP */ #define COMP_CHECKCASE 4 /* CHECKCOMPOUNDCASE */ #define COMP_CHECKTRIPLE 8 /* CHECKCOMPOUNDTRIPLE */ /* Special byte values for . Some are only used in the tree for * postponed prefixes, some only in the other trees. This is a bit messy... */ #define BY_NOFLAGS 0 /* end of word without flags or region; for * postponed prefix: no */ #define BY_INDEX 1 /* child is shared, index follows */ #define BY_FLAGS 2 /* end of word, byte follows; for * postponed prefix: follows */ #define BY_FLAGS2 3 /* end of word, and bytes * follow; never used in prefix tree */ #define BY_SPECIAL BY_FLAGS2 /* highest special byte value */ /* Info from "REP", "REPSAL" and "SAL" entries in ".aff" file used in si_rep, * si_repsal, sl_rep, and si_sal. Not for sl_sal! * One replacement: from "ft_from" to "ft_to". */ typedef struct fromto_S { char_u *ft_from; char_u *ft_to; } fromto_T; /* Info from "SAL" entries in ".aff" file used in sl_sal. * The info is split for quick processing by spell_soundfold(). * Note that "sm_oneof" and "sm_rules" point into sm_lead. */ typedef struct salitem_S { char_u *sm_lead; /* leading letters */ int sm_leadlen; /* length of "sm_lead" */ char_u *sm_oneof; /* letters from () or NULL */ char_u *sm_rules; /* rules like ^, $, priority */ char_u *sm_to; /* replacement. */ #ifdef FEAT_MBYTE int *sm_lead_w; /* wide character copy of "sm_lead" */ int *sm_oneof_w; /* wide character copy of "sm_oneof" */ int *sm_to_w; /* wide character copy of "sm_to" */ #endif } salitem_T; #ifdef FEAT_MBYTE typedef int salfirst_T; #else typedef short salfirst_T; #endif /* Values for SP_*ERROR are negative, positive values are used by * read_cnt_string(). */ #define SP_TRUNCERROR -1 /* spell file truncated error */ #define SP_FORMERROR -2 /* format error in spell file */ #define SP_OTHERERROR -3 /* other error while reading spell file */ /* * Structure used to store words and other info for one language, loaded from * a .spl file. * The main access is through the tree in "sl_fbyts/sl_fidxs", storing the * case-folded words. "sl_kbyts/sl_kidxs" is for keep-case words. * * The "byts" array stores the possible bytes in each tree node, preceded by * the number of possible bytes, sorted on byte value: * ... * The "idxs" array stores the index of the child node corresponding to the * byte in "byts". * Exception: when the byte is zero, the word may end here and "idxs" holds * the flags, region mask and affixID for the word. There may be several * zeros in sequence for alternative flag/region/affixID combinations. */ typedef struct slang_S slang_T; struct slang_S { slang_T *sl_next; /* next language */ char_u *sl_name; /* language name "en", "en.rare", "nl", etc. */ char_u *sl_fname; /* name of .spl file */ int sl_add; /* TRUE if it's a .add file. */ char_u *sl_fbyts; /* case-folded word bytes */ idx_T *sl_fidxs; /* case-folded word indexes */ char_u *sl_kbyts; /* keep-case word bytes */ idx_T *sl_kidxs; /* keep-case word indexes */ char_u *sl_pbyts; /* prefix tree word bytes */ idx_T *sl_pidxs; /* prefix tree word indexes */ char_u *sl_info; /* infotext string or NULL */ char_u sl_regions[17]; /* table with up to 8 region names plus NUL */ char_u *sl_midword; /* MIDWORD string or NULL */ hashtab_T sl_wordcount; /* hashtable with word count, wordcount_T */ int sl_compmax; /* COMPOUNDWORDMAX (default: MAXWLEN) */ int sl_compminlen; /* COMPOUNDMIN (default: 0) */ int sl_compsylmax; /* COMPOUNDSYLMAX (default: MAXWLEN) */ int sl_compoptions; /* COMP_* flags */ garray_T sl_comppat; /* CHECKCOMPOUNDPATTERN items */ regprog_T *sl_compprog; /* COMPOUNDRULE turned into a regexp progrm * (NULL when no compounding) */ char_u *sl_comprules; /* all COMPOUNDRULE concatenated (or NULL) */ char_u *sl_compstartflags; /* flags for first compound word */ char_u *sl_compallflags; /* all flags for compound words */ char_u sl_nobreak; /* When TRUE: no spaces between words */ char_u *sl_syllable; /* SYLLABLE repeatable chars or NULL */ garray_T sl_syl_items; /* syllable items */ int sl_prefixcnt; /* number of items in "sl_prefprog" */ regprog_T **sl_prefprog; /* table with regprogs for prefixes */ garray_T sl_rep; /* list of fromto_T entries from REP lines */ short sl_rep_first[256]; /* indexes where byte first appears, -1 if there is none */ garray_T sl_sal; /* list of salitem_T entries from SAL lines */ salfirst_T sl_sal_first[256]; /* indexes where byte first appears, -1 if there is none */ int sl_followup; /* SAL followup */ int sl_collapse; /* SAL collapse_result */ int sl_rem_accents; /* SAL remove_accents */ int sl_sofo; /* SOFOFROM and SOFOTO instead of SAL items: * "sl_sal_first" maps chars, when has_mbyte * "sl_sal" is a list of wide char lists. */ garray_T sl_repsal; /* list of fromto_T entries from REPSAL lines */ short sl_repsal_first[256]; /* sl_rep_first for REPSAL lines */ int sl_nosplitsugs; /* don't suggest splitting a word */ /* Info from the .sug file. Loaded on demand. */ time_t sl_sugtime; /* timestamp for .sug file */ char_u *sl_sbyts; /* soundfolded word bytes */ idx_T *sl_sidxs; /* soundfolded word indexes */ buf_T *sl_sugbuf; /* buffer with word number table */ int sl_sugloaded; /* TRUE when .sug file was loaded or failed to load */ int sl_has_map; /* TRUE if there is a MAP line */ #ifdef FEAT_MBYTE hashtab_T sl_map_hash; /* MAP for multi-byte chars */ int sl_map_array[256]; /* MAP for first 256 chars */ #else char_u sl_map_array[256]; /* MAP for first 256 chars */ #endif hashtab_T sl_sounddone; /* table with soundfolded words that have handled, see add_sound_suggest() */ }; /* First language that is loaded, start of the linked list of loaded * languages. */ static slang_T *first_lang = NULL; /* Flags used in .spl file for soundsalike flags. */ #define SAL_F0LLOWUP 1 #define SAL_COLLAPSE 2 #define SAL_REM_ACCENTS 4 /* * Structure used in "b_langp", filled from 'spelllang'. */ typedef struct langp_S { slang_T *lp_slang; /* info for this language */ slang_T *lp_sallang; /* language used for sound folding or NULL */ slang_T *lp_replang; /* language used for REP items or NULL */ int lp_region; /* bitmask for region or REGION_ALL */ } langp_T; #define LANGP_ENTRY(ga, i) (((langp_T *)(ga).ga_data) + (i)) #define REGION_ALL 0xff /* word valid in all regions */ #define VIMSPELLMAGIC "VIMspell" /* string at start of Vim spell file */ #define VIMSPELLMAGICL 8 #define VIMSPELLVERSION 50 #define VIMSUGMAGIC "VIMsug" /* string at start of Vim .sug file */ #define VIMSUGMAGICL 6 #define VIMSUGVERSION 1 /* Section IDs. Only renumber them when VIMSPELLVERSION changes! */ #define SN_REGION 0 /* section */ #define SN_CHARFLAGS 1 /* charflags section */ #define SN_MIDWORD 2 /* section */ #define SN_PREFCOND 3 /* section */ #define SN_REP 4 /* REP items section */ #define SN_SAL 5 /* SAL items section */ #define SN_SOFO 6 /* soundfolding section */ #define SN_MAP 7 /* MAP items section */ #define SN_COMPOUND 8 /* compound words section */ #define SN_SYLLABLE 9 /* syllable section */ #define SN_NOBREAK 10 /* NOBREAK section */ #define SN_SUGFILE 11 /* timestamp for .sug file */ #define SN_REPSAL 12 /* REPSAL items section */ #define SN_WORDS 13 /* common words */ #define SN_NOSPLITSUGS 14 /* don't split word for suggestions */ #define SN_INFO 15 /* info section */ #define SN_END 255 /* end of sections */ #define SNF_REQUIRED 1 /* : required section */ /* Result values. Lower number is accepted over higher one. */ #define SP_BANNED -1 #define SP_OK 0 #define SP_RARE 1 #define SP_LOCAL 2 #define SP_BAD 3 /* file used for "zG" and "zW" */ static char_u *int_wordlist = NULL; typedef struct wordcount_S { short_u wc_count; /* nr of times word was seen */ char_u wc_word[1]; /* word, actually longer */ } wordcount_T; static wordcount_T dumwc; #define WC_KEY_OFF (unsigned)(dumwc.wc_word - (char_u *)&dumwc) #define HI2WC(hi) ((wordcount_T *)((hi)->hi_key - WC_KEY_OFF)) #define MAXWORDCOUNT 0xffff /* * Information used when looking for suggestions. */ typedef struct suginfo_S { garray_T su_ga; /* suggestions, contains "suggest_T" */ int su_maxcount; /* max. number of suggestions displayed */ int su_maxscore; /* maximum score for adding to su_ga */ int su_sfmaxscore; /* idem, for when doing soundfold words */ garray_T su_sga; /* like su_ga, sound-folded scoring */ char_u *su_badptr; /* start of bad word in line */ int su_badlen; /* length of detected bad word in line */ int su_badflags; /* caps flags for bad word */ char_u su_badword[MAXWLEN]; /* bad word truncated at su_badlen */ char_u su_fbadword[MAXWLEN]; /* su_badword case-folded */ char_u su_sal_badword[MAXWLEN]; /* su_badword soundfolded */ hashtab_T su_banned; /* table with banned words */ slang_T *su_sallang; /* default language for sound folding */ } suginfo_T; /* One word suggestion. Used in "si_ga". */ typedef struct suggest_S { char_u *st_word; /* suggested word, allocated string */ int st_wordlen; /* STRLEN(st_word) */ int st_orglen; /* length of replaced text */ int st_score; /* lower is better */ int st_altscore; /* used when st_score compares equal */ int st_salscore; /* st_score is for soundalike */ int st_had_bonus; /* bonus already included in score */ slang_T *st_slang; /* language used for sound folding */ } suggest_T; #define SUG(ga, i) (((suggest_T *)(ga).ga_data)[i]) /* TRUE if a word appears in the list of banned words. */ #define WAS_BANNED(su, word) (!HASHITEM_EMPTY(hash_find(&su->su_banned, word))) /* Number of suggestions kept when cleaning up. We need to keep more than * what is displayed, because when rescore_suggestions() is called the score * may change and wrong suggestions may be removed later. */ #define SUG_CLEAN_COUNT(su) ((su)->su_maxcount < 130 ? 150 : (su)->su_maxcount + 20) /* Threshold for sorting and cleaning up suggestions. Don't want to keep lots * of suggestions that are not going to be displayed. */ #define SUG_MAX_COUNT(su) (SUG_CLEAN_COUNT(su) + 50) /* score for various changes */ #define SCORE_SPLIT 149 /* split bad word */ #define SCORE_SPLIT_NO 249 /* split bad word with NOSPLITSUGS */ #define SCORE_ICASE 52 /* slightly different case */ #define SCORE_REGION 200 /* word is for different region */ #define SCORE_RARE 180 /* rare word */ #define SCORE_SWAP 75 /* swap two characters */ #define SCORE_SWAP3 110 /* swap two characters in three */ #define SCORE_REP 65 /* REP replacement */ #define SCORE_SUBST 93 /* substitute a character */ #define SCORE_SIMILAR 33 /* substitute a similar character */ #define SCORE_SUBCOMP 33 /* substitute a composing character */ #define SCORE_DEL 94 /* delete a character */ #define SCORE_DELDUP 66 /* delete a duplicated character */ #define SCORE_DELCOMP 28 /* delete a composing character */ #define SCORE_INS 96 /* insert a character */ #define SCORE_INSDUP 67 /* insert a duplicate character */ #define SCORE_INSCOMP 30 /* insert a composing character */ #define SCORE_NONWORD 103 /* change non-word to word char */ #define SCORE_FILE 30 /* suggestion from a file */ #define SCORE_MAXINIT 350 /* Initial maximum score: higher == slower. * 350 allows for about three changes. */ #define SCORE_COMMON1 30 /* subtracted for words seen before */ #define SCORE_COMMON2 40 /* subtracted for words often seen */ #define SCORE_COMMON3 50 /* subtracted for words very often seen */ #define SCORE_THRES2 10 /* word count threshold for COMMON2 */ #define SCORE_THRES3 100 /* word count threshold for COMMON3 */ /* When trying changed soundfold words it becomes slow when trying more than * two changes. With less then two changes it's slightly faster but we miss a * few good suggestions. In rare cases we need to try three of four changes. */ #define SCORE_SFMAX1 200 /* maximum score for first try */ #define SCORE_SFMAX2 300 /* maximum score for second try */ #define SCORE_SFMAX3 400 /* maximum score for third try */ #define SCORE_BIG SCORE_INS * 3 /* big difference */ #define SCORE_MAXMAX 999999 /* accept any score */ #define SCORE_LIMITMAX 350 /* for spell_edit_score_limit() */ /* for spell_edit_score_limit() we need to know the minimum value of * SCORE_ICASE, SCORE_SWAP, SCORE_DEL, SCORE_SIMILAR and SCORE_INS */ #define SCORE_EDIT_MIN SCORE_SIMILAR /* * Structure to store info for word matching. */ typedef struct matchinf_S { langp_T *mi_lp; /* info for language and region */ /* pointers to original text to be checked */ char_u *mi_word; /* start of word being checked */ char_u *mi_end; /* end of matching word so far */ char_u *mi_fend; /* next char to be added to mi_fword */ char_u *mi_cend; /* char after what was used for mi_capflags */ /* case-folded text */ char_u mi_fword[MAXWLEN + 1]; /* mi_word case-folded */ int mi_fwordlen; /* nr of valid bytes in mi_fword */ /* for when checking word after a prefix */ int mi_prefarridx; /* index in sl_pidxs with list of affixID/condition */ int mi_prefcnt; /* number of entries at mi_prefarridx */ int mi_prefixlen; /* byte length of prefix */ #ifdef FEAT_MBYTE int mi_cprefixlen; /* byte length of prefix in original case */ #else # define mi_cprefixlen mi_prefixlen /* it's the same value */ #endif /* for when checking a compound word */ int mi_compoff; /* start of following word offset */ char_u mi_compflags[MAXWLEN]; /* flags for compound words used */ int mi_complen; /* nr of compound words used */ int mi_compextra; /* nr of COMPOUNDROOT words */ /* others */ int mi_result; /* result so far: SP_BAD, SP_OK, etc. */ int mi_capflags; /* WF_ONECAP WF_ALLCAP WF_KEEPCAP */ win_T *mi_win; /* buffer being checked */ /* for NOBREAK */ int mi_result2; /* "mi_resul" without following word */ char_u *mi_end2; /* "mi_end" without following word */ } matchinf_T; /* * The tables used for recognizing word characters according to spelling. * These are only used for the first 256 characters of 'encoding'. */ typedef struct spelltab_S { char_u st_isw[256]; /* flags: is word char */ char_u st_isu[256]; /* flags: is uppercase char */ char_u st_fold[256]; /* chars: folded case */ char_u st_upper[256]; /* chars: upper case */ } spelltab_T; static spelltab_T spelltab; static int did_set_spelltab; #define CF_WORD 0x01 #define CF_UPPER 0x02 static void clear_spell_chartab __ARGS((spelltab_T *sp)); static int set_spell_finish __ARGS((spelltab_T *new_st)); static int spell_iswordp __ARGS((char_u *p, win_T *wp)); static int spell_iswordp_nmw __ARGS((char_u *p)); #ifdef FEAT_MBYTE static int spell_mb_isword_class __ARGS((int cl)); static int spell_iswordp_w __ARGS((int *p, win_T *wp)); #endif static int write_spell_prefcond __ARGS((FILE *fd, garray_T *gap)); /* * For finding suggestions: At each node in the tree these states are tried: */ typedef enum { STATE_START = 0, /* At start of node check for NUL bytes (goodword * ends); if badword ends there is a match, otherwise * try splitting word. */ STATE_NOPREFIX, /* try without prefix */ STATE_SPLITUNDO, /* Undo splitting. */ STATE_ENDNUL, /* Past NUL bytes at start of the node. */ STATE_PLAIN, /* Use each byte of the node. */ STATE_DEL, /* Delete a byte from the bad word. */ STATE_INS_PREP, /* Prepare for inserting bytes. */ STATE_INS, /* Insert a byte in the bad word. */ STATE_SWAP, /* Swap two bytes. */ STATE_UNSWAP, /* Undo swap two characters. */ STATE_SWAP3, /* Swap two characters over three. */ STATE_UNSWAP3, /* Undo Swap two characters over three. */ STATE_UNROT3L, /* Undo rotate three characters left */ STATE_UNROT3R, /* Undo rotate three characters right */ STATE_REP_INI, /* Prepare for using REP items. */ STATE_REP, /* Use matching REP items from the .aff file. */ STATE_REP_UNDO, /* Undo a REP item replacement. */ STATE_FINAL /* End of this node. */ } state_T; /* * Struct to keep the state at each level in suggest_try_change(). */ typedef struct trystate_S { state_T ts_state; /* state at this level, STATE_ */ int ts_score; /* score */ idx_T ts_arridx; /* index in tree array, start of node */ short ts_curi; /* index in list of child nodes */ char_u ts_fidx; /* index in fword[], case-folded bad word */ char_u ts_fidxtry; /* ts_fidx at which bytes may be changed */ char_u ts_twordlen; /* valid length of tword[] */ char_u ts_prefixdepth; /* stack depth for end of prefix or * PFD_PREFIXTREE or PFD_NOPREFIX */ char_u ts_flags; /* TSF_ flags */ #ifdef FEAT_MBYTE char_u ts_tcharlen; /* number of bytes in tword character */ char_u ts_tcharidx; /* current byte index in tword character */ char_u ts_isdiff; /* DIFF_ values */ char_u ts_fcharstart; /* index in fword where badword char started */ #endif char_u ts_prewordlen; /* length of word in "preword[]" */ char_u ts_splitoff; /* index in "tword" after last split */ char_u ts_splitfidx; /* "ts_fidx" at word split */ char_u ts_complen; /* nr of compound words used */ char_u ts_compsplit; /* index for "compflags" where word was spit */ char_u ts_save_badflags; /* su_badflags saved here */ char_u ts_delidx; /* index in fword for char that was deleted, valid when "ts_flags" has TSF_DIDDEL */ } trystate_T; /* values for ts_isdiff */ #define DIFF_NONE 0 /* no different byte (yet) */ #define DIFF_YES 1 /* different byte found */ #define DIFF_INSERT 2 /* inserting character */ /* values for ts_flags */ #define TSF_PREFIXOK 1 /* already checked that prefix is OK */ #define TSF_DIDSPLIT 2 /* tried split at this point */ #define TSF_DIDDEL 4 /* did a delete, "ts_delidx" has index */ /* special values ts_prefixdepth */ #define PFD_NOPREFIX 0xff /* not using prefixes */ #define PFD_PREFIXTREE 0xfe /* walking through the prefix tree */ #define PFD_NOTSPECIAL 0xfd /* highest value that's not special */ /* mode values for find_word */ #define FIND_FOLDWORD 0 /* find word case-folded */ #define FIND_KEEPWORD 1 /* find keep-case word */ #define FIND_PREFIX 2 /* find word after prefix */ #define FIND_COMPOUND 3 /* find case-folded compound word */ #define FIND_KEEPCOMPOUND 4 /* find keep-case compound word */ static slang_T *slang_alloc __ARGS((char_u *lang)); static void slang_free __ARGS((slang_T *lp)); static void slang_clear __ARGS((slang_T *lp)); static void slang_clear_sug __ARGS((slang_T *lp)); static void find_word __ARGS((matchinf_T *mip, int mode)); static int match_checkcompoundpattern __ARGS((char_u *ptr, int wlen, garray_T *gap)); static int can_compound __ARGS((slang_T *slang, char_u *word, char_u *flags)); static int can_be_compound __ARGS((trystate_T *sp, slang_T *slang, char_u *compflags, int flag)); static int match_compoundrule __ARGS((slang_T *slang, char_u *compflags)); static int valid_word_prefix __ARGS((int totprefcnt, int arridx, int flags, char_u *word, slang_T *slang, int cond_req)); static void find_prefix __ARGS((matchinf_T *mip, int mode)); static int fold_more __ARGS((matchinf_T *mip)); static int spell_valid_case __ARGS((int wordflags, int treeflags)); static int no_spell_checking __ARGS((win_T *wp)); static void spell_load_lang __ARGS((char_u *lang)); static char_u *spell_enc __ARGS((void)); static void int_wordlist_spl __ARGS((char_u *fname)); static void spell_load_cb __ARGS((char_u *fname, void *cookie)); static slang_T *spell_load_file __ARGS((char_u *fname, char_u *lang, slang_T *old_lp, int silent)); static char_u *read_cnt_string __ARGS((FILE *fd, int cnt_bytes, int *lenp)); static int read_region_section __ARGS((FILE *fd, slang_T *slang, int len)); static int read_charflags_section __ARGS((FILE *fd)); static int read_prefcond_section __ARGS((FILE *fd, slang_T *lp)); static int read_rep_section __ARGS((FILE *fd, garray_T *gap, short *first)); static int read_sal_section __ARGS((FILE *fd, slang_T *slang)); static int read_words_section __ARGS((FILE *fd, slang_T *lp, int len)); static void count_common_word __ARGS((slang_T *lp, char_u *word, int len, int count)); static int score_wordcount_adj __ARGS((slang_T *slang, int score, char_u *word, int split)); static int read_sofo_section __ARGS((FILE *fd, slang_T *slang)); static int read_compound __ARGS((FILE *fd, slang_T *slang, int len)); static int byte_in_str __ARGS((char_u *str, int byte)); static int init_syl_tab __ARGS((slang_T *slang)); static int count_syllables __ARGS((slang_T *slang, char_u *word)); static int set_sofo __ARGS((slang_T *lp, char_u *from, char_u *to)); static void set_sal_first __ARGS((slang_T *lp)); #ifdef FEAT_MBYTE static int *mb_str2wide __ARGS((char_u *s)); #endif static int spell_read_tree __ARGS((FILE *fd, char_u **bytsp, idx_T **idxsp, int prefixtree, int prefixcnt)); static idx_T read_tree_node __ARGS((FILE *fd, char_u *byts, idx_T *idxs, int maxidx, idx_T startidx, int prefixtree, int maxprefcondnr)); static void clear_midword __ARGS((win_T *buf)); static void use_midword __ARGS((slang_T *lp, win_T *buf)); static int find_region __ARGS((char_u *rp, char_u *region)); static int captype __ARGS((char_u *word, char_u *end)); static int badword_captype __ARGS((char_u *word, char_u *end)); static void spell_reload_one __ARGS((char_u *fname, int added_word)); static void set_spell_charflags __ARGS((char_u *flags, int cnt, char_u *upp)); static int set_spell_chartab __ARGS((char_u *fol, char_u *low, char_u *upp)); static int spell_casefold __ARGS((char_u *p, int len, char_u *buf, int buflen)); static int check_need_cap __ARGS((linenr_T lnum, colnr_T col)); static void spell_find_suggest __ARGS((char_u *badptr, int badlen, suginfo_T *su, int maxcount, int banbadword, int need_cap, int interactive)); #ifdef FEAT_EVAL static void spell_suggest_expr __ARGS((suginfo_T *su, char_u *expr)); #endif static void spell_suggest_file __ARGS((suginfo_T *su, char_u *fname)); static void spell_suggest_intern __ARGS((suginfo_T *su, int interactive)); static void suggest_load_files __ARGS((void)); static void tree_count_words __ARGS((char_u *byts, idx_T *idxs)); static void spell_find_cleanup __ARGS((suginfo_T *su)); static void onecap_copy __ARGS((char_u *word, char_u *wcopy, int upper)); static void allcap_copy __ARGS((char_u *word, char_u *wcopy)); static void suggest_try_special __ARGS((suginfo_T *su)); static void suggest_try_change __ARGS((suginfo_T *su)); static void suggest_trie_walk __ARGS((suginfo_T *su, langp_T *lp, char_u *fword, int soundfold)); static void go_deeper __ARGS((trystate_T *stack, int depth, int score_add)); #ifdef FEAT_MBYTE static int nofold_len __ARGS((char_u *fword, int flen, char_u *word)); #endif static void find_keepcap_word __ARGS((slang_T *slang, char_u *fword, char_u *kword)); static void score_comp_sal __ARGS((suginfo_T *su)); static void score_combine __ARGS((suginfo_T *su)); static int stp_sal_score __ARGS((suggest_T *stp, suginfo_T *su, slang_T *slang, char_u *badsound)); static void suggest_try_soundalike_prep __ARGS((void)); static void suggest_try_soundalike __ARGS((suginfo_T *su)); static void suggest_try_soundalike_finish __ARGS((void)); static void add_sound_suggest __ARGS((suginfo_T *su, char_u *goodword, int score, langp_T *lp)); static int soundfold_find __ARGS((slang_T *slang, char_u *word)); static void make_case_word __ARGS((char_u *fword, char_u *cword, int flags)); static void set_map_str __ARGS((slang_T *lp, char_u *map)); static int similar_chars __ARGS((slang_T *slang, int c1, int c2)); static void add_suggestion __ARGS((suginfo_T *su, garray_T *gap, char_u *goodword, int badlen, int score, int altscore, int had_bonus, slang_T *slang, int maxsf)); static void check_suggestions __ARGS((suginfo_T *su, garray_T *gap)); static void add_banned __ARGS((suginfo_T *su, char_u *word)); static void rescore_suggestions __ARGS((suginfo_T *su)); static void rescore_one __ARGS((suginfo_T *su, suggest_T *stp)); static int cleanup_suggestions __ARGS((garray_T *gap, int maxscore, int keep)); static void spell_soundfold __ARGS((slang_T *slang, char_u *inword, int folded, char_u *res)); static void spell_soundfold_sofo __ARGS((slang_T *slang, char_u *inword, char_u *res)); static void spell_soundfold_sal __ARGS((slang_T *slang, char_u *inword, char_u *res)); #ifdef FEAT_MBYTE static void spell_soundfold_wsal __ARGS((slang_T *slang, char_u *inword, char_u *res)); #endif static int soundalike_score __ARGS((char_u *goodsound, char_u *badsound)); static int spell_edit_score __ARGS((slang_T *slang, char_u *badword, char_u *goodword)); static int spell_edit_score_limit __ARGS((slang_T *slang, char_u *badword, char_u *goodword, int limit)); #ifdef FEAT_MBYTE static int spell_edit_score_limit_w __ARGS((slang_T *slang, char_u *badword, char_u *goodword, int limit)); #endif static void dump_word __ARGS((slang_T *slang, char_u *word, char_u *pat, int *dir, int round, int flags, linenr_T lnum)); static linenr_T dump_prefixes __ARGS((slang_T *slang, char_u *word, char_u *pat, int *dir, int round, int flags, linenr_T startlnum)); static buf_T *open_spellbuf __ARGS((void)); static void close_spellbuf __ARGS((buf_T *buf)); /* * Use our own character-case definitions, because the current locale may * differ from what the .spl file uses. * These must not be called with negative number! */ #ifndef FEAT_MBYTE /* Non-multi-byte implementation. */ # define SPELL_TOFOLD(c) ((c) < 256 ? (int)spelltab.st_fold[c] : (c)) # define SPELL_TOUPPER(c) ((c) < 256 ? (int)spelltab.st_upper[c] : (c)) # define SPELL_ISUPPER(c) ((c) < 256 ? spelltab.st_isu[c] : FALSE) #else # if defined(HAVE_WCHAR_H) # include /* for towupper() and towlower() */ # endif /* Multi-byte implementation. For Unicode we can call utf_*(), but don't do * that for ASCII, because we don't want to use 'casemap' here. Otherwise use * the "w" library function for characters above 255 if available. */ # ifdef HAVE_TOWLOWER # define SPELL_TOFOLD(c) (enc_utf8 && (c) >= 128 ? utf_fold(c) \ : (c) < 256 ? (int)spelltab.st_fold[c] : (int)towlower(c)) # else # define SPELL_TOFOLD(c) (enc_utf8 && (c) >= 128 ? utf_fold(c) \ : (c) < 256 ? (int)spelltab.st_fold[c] : (c)) # endif # ifdef HAVE_TOWUPPER # define SPELL_TOUPPER(c) (enc_utf8 && (c) >= 128 ? utf_toupper(c) \ : (c) < 256 ? (int)spelltab.st_upper[c] : (int)towupper(c)) # else # define SPELL_TOUPPER(c) (enc_utf8 && (c) >= 128 ? utf_toupper(c) \ : (c) < 256 ? (int)spelltab.st_upper[c] : (c)) # endif # ifdef HAVE_ISWUPPER # define SPELL_ISUPPER(c) (enc_utf8 && (c) >= 128 ? utf_isupper(c) \ : (c) < 256 ? spelltab.st_isu[c] : iswupper(c)) # else # define SPELL_ISUPPER(c) (enc_utf8 && (c) >= 128 ? utf_isupper(c) \ : (c) < 256 ? spelltab.st_isu[c] : (FALSE)) # endif #endif static char *e_format = N_("E759: Format error in spell file"); static char *e_spell_trunc = N_("E758: Truncated spell file"); static char *e_afftrailing = N_("Trailing text in %s line %d: %s"); static char *e_affname = N_("Affix name too long in %s line %d: %s"); static char *e_affform = N_("E761: Format error in affix file FOL, LOW or UPP"); static char *e_affrange = N_("E762: Character in FOL, LOW or UPP is out of range"); static char *msg_compressing = N_("Compressing word tree..."); /* Remember what "z?" replaced. */ static char_u *repl_from = NULL; static char_u *repl_to = NULL; /* * Main spell-checking function. * "ptr" points to a character that could be the start of a word. * "*attrp" is set to the highlight index for a badly spelled word. For a * non-word or when it's OK it remains unchanged. * This must only be called when 'spelllang' is not empty. * * "capcol" is used to check for a Capitalised word after the end of a * sentence. If it's zero then perform the check. Return the column where to * check next, or -1 when no sentence end was found. If it's NULL then don't * worry. * * Returns the length of the word in bytes, also when it's OK, so that the * caller can skip over the word. */ int spell_check(wp, ptr, attrp, capcol, docount) win_T *wp; /* current window */ char_u *ptr; hlf_T *attrp; int *capcol; /* column to check for Capital */ int docount; /* count good words */ { matchinf_T mi; /* Most things are put in "mi" so that it can be passed to functions quickly. */ int nrlen = 0; /* found a number first */ int c; int wrongcaplen = 0; int lpi; int count_word = docount; /* A word never starts at a space or a control character. Return quickly * then, skipping over the character. */ if (*ptr <= ' ') return 1; /* Return here when loading language files failed. */ if (wp->w_s->b_langp.ga_len == 0) return 1; vim_memset(&mi, 0, sizeof(matchinf_T)); /* A number is always OK. Also skip hexadecimal numbers 0xFF99 and * 0X99FF. But always do check spelling to find "3GPP" and "11 * julifeest". */ if (*ptr >= '0' && *ptr <= '9') { if (*ptr == '0' && (ptr[1] == 'x' || ptr[1] == 'X')) mi.mi_end = skiphex(ptr + 2); else mi.mi_end = skipdigits(ptr); nrlen = (int)(mi.mi_end - ptr); } /* Find the normal end of the word (until the next non-word character). */ mi.mi_word = ptr; mi.mi_fend = ptr; if (spell_iswordp(mi.mi_fend, wp)) { do { mb_ptr_adv(mi.mi_fend); } while (*mi.mi_fend != NUL && spell_iswordp(mi.mi_fend, wp)); if (capcol != NULL && *capcol == 0 && wp->w_s->b_cap_prog != NULL) { /* Check word starting with capital letter. */ c = PTR2CHAR(ptr); if (!SPELL_ISUPPER(c)) wrongcaplen = (int)(mi.mi_fend - ptr); } } if (capcol != NULL) *capcol = -1; /* We always use the characters up to the next non-word character, * also for bad words. */ mi.mi_end = mi.mi_fend; /* Check caps type later. */ mi.mi_capflags = 0; mi.mi_cend = NULL; mi.mi_win = wp; /* case-fold the word with one non-word character, so that we can check * for the word end. */ if (*mi.mi_fend != NUL) mb_ptr_adv(mi.mi_fend); (void)spell_casefold(ptr, (int)(mi.mi_fend - ptr), mi.mi_fword, MAXWLEN + 1); mi.mi_fwordlen = (int)STRLEN(mi.mi_fword); /* The word is bad unless we recognize it. */ mi.mi_result = SP_BAD; mi.mi_result2 = SP_BAD; /* * Loop over the languages specified in 'spelllang'. * We check them all, because a word may be matched longer in another * language. */ for (lpi = 0; lpi < wp->w_s->b_langp.ga_len; ++lpi) { mi.mi_lp = LANGP_ENTRY(wp->w_s->b_langp, lpi); /* If reloading fails the language is still in the list but everything * has been cleared. */ if (mi.mi_lp->lp_slang->sl_fidxs == NULL) continue; /* Check for a matching word in case-folded words. */ find_word(&mi, FIND_FOLDWORD); /* Check for a matching word in keep-case words. */ find_word(&mi, FIND_KEEPWORD); /* Check for matching prefixes. */ find_prefix(&mi, FIND_FOLDWORD); /* For a NOBREAK language, may want to use a word without a following * word as a backup. */ if (mi.mi_lp->lp_slang->sl_nobreak && mi.mi_result == SP_BAD && mi.mi_result2 != SP_BAD) { mi.mi_result = mi.mi_result2; mi.mi_end = mi.mi_end2; } /* Count the word in the first language where it's found to be OK. */ if (count_word && mi.mi_result == SP_OK) { count_common_word(mi.mi_lp->lp_slang, ptr, (int)(mi.mi_end - ptr), 1); count_word = FALSE; } } if (mi.mi_result != SP_OK) { /* If we found a number skip over it. Allows for "42nd". Do flag * rare and local words, e.g., "3GPP". */ if (nrlen > 0) { if (mi.mi_result == SP_BAD || mi.mi_result == SP_BANNED) return nrlen; } /* When we are at a non-word character there is no error, just * skip over the character (try looking for a word after it). */ else if (!spell_iswordp_nmw(ptr)) { if (capcol != NULL && wp->w_s->b_cap_prog != NULL) { regmatch_T regmatch; /* Check for end of sentence. */ regmatch.regprog = wp->w_s->b_cap_prog; regmatch.rm_ic = FALSE; if (vim_regexec(®match, ptr, 0)) *capcol = (int)(regmatch.endp[0] - ptr); } #ifdef FEAT_MBYTE if (has_mbyte) return (*mb_ptr2len)(ptr); #endif return 1; } else if (mi.mi_end == ptr) /* Always include at least one character. Required for when there * is a mixup in "midword". */ mb_ptr_adv(mi.mi_end); else if (mi.mi_result == SP_BAD && LANGP_ENTRY(wp->w_s->b_langp, 0)->lp_slang->sl_nobreak) { char_u *p, *fp; int save_result = mi.mi_result; /* First language in 'spelllang' is NOBREAK. Find first position * at which any word would be valid. */ mi.mi_lp = LANGP_ENTRY(wp->w_s->b_langp, 0); if (mi.mi_lp->lp_slang->sl_fidxs != NULL) { p = mi.mi_word; fp = mi.mi_fword; for (;;) { mb_ptr_adv(p); mb_ptr_adv(fp); if (p >= mi.mi_end) break; mi.mi_compoff = (int)(fp - mi.mi_fword); find_word(&mi, FIND_COMPOUND); if (mi.mi_result != SP_BAD) { mi.mi_end = p; break; } } mi.mi_result = save_result; } } if (mi.mi_result == SP_BAD || mi.mi_result == SP_BANNED) *attrp = HLF_SPB; else if (mi.mi_result == SP_RARE) *attrp = HLF_SPR; else *attrp = HLF_SPL; } if (wrongcaplen > 0 && (mi.mi_result == SP_OK || mi.mi_result == SP_RARE)) { /* Report SpellCap only when the word isn't badly spelled. */ *attrp = HLF_SPC; return wrongcaplen; } return (int)(mi.mi_end - ptr); } /* * Check if the word at "mip->mi_word" is in the tree. * When "mode" is FIND_FOLDWORD check in fold-case word tree. * When "mode" is FIND_KEEPWORD check in keep-case word tree. * When "mode" is FIND_PREFIX check for word after prefix in fold-case word * tree. * * For a match mip->mi_result is updated. */ static void find_word(mip, mode) matchinf_T *mip; int mode; { idx_T arridx = 0; int endlen[MAXWLEN]; /* length at possible word endings */ idx_T endidx[MAXWLEN]; /* possible word endings */ int endidxcnt = 0; int len; int wlen = 0; int flen; int c; char_u *ptr; idx_T lo, hi, m; #ifdef FEAT_MBYTE char_u *s; #endif char_u *p; int res = SP_BAD; slang_T *slang = mip->mi_lp->lp_slang; unsigned flags; char_u *byts; idx_T *idxs; int word_ends; int prefix_found; int nobreak_result; if (mode == FIND_KEEPWORD || mode == FIND_KEEPCOMPOUND) { /* Check for word with matching case in keep-case tree. */ ptr = mip->mi_word; flen = 9999; /* no case folding, always enough bytes */ byts = slang->sl_kbyts; idxs = slang->sl_kidxs; if (mode == FIND_KEEPCOMPOUND) /* Skip over the previously found word(s). */ wlen += mip->mi_compoff; } else { /* Check for case-folded in case-folded tree. */ ptr = mip->mi_fword; flen = mip->mi_fwordlen; /* available case-folded bytes */ byts = slang->sl_fbyts; idxs = slang->sl_fidxs; if (mode == FIND_PREFIX) { /* Skip over the prefix. */ wlen = mip->mi_prefixlen; flen -= mip->mi_prefixlen; } else if (mode == FIND_COMPOUND) { /* Skip over the previously found word(s). */ wlen = mip->mi_compoff; flen -= mip->mi_compoff; } } if (byts == NULL) return; /* array is empty */ /* * Repeat advancing in the tree until: * - there is a byte that doesn't match, * - we reach the end of the tree, * - or we reach the end of the line. */ for (;;) { if (flen <= 0 && *mip->mi_fend != NUL) flen = fold_more(mip); len = byts[arridx++]; /* If the first possible byte is a zero the word could end here. * Remember this index, we first check for the longest word. */ if (byts[arridx] == 0) { if (endidxcnt == MAXWLEN) { /* Must be a corrupted spell file. */ EMSG(_(e_format)); return; } endlen[endidxcnt] = wlen; endidx[endidxcnt++] = arridx++; --len; /* Skip over the zeros, there can be several flag/region * combinations. */ while (len > 0 && byts[arridx] == 0) { ++arridx; --len; } if (len == 0) break; /* no children, word must end here */ } /* Stop looking at end of the line. */ if (ptr[wlen] == NUL) break; /* Perform a binary search in the list of accepted bytes. */ c = ptr[wlen]; if (c == TAB) /* is handled like */ c = ' '; lo = arridx; hi = arridx + len - 1; while (lo < hi) { m = (lo + hi) / 2; if (byts[m] > c) hi = m - 1; else if (byts[m] < c) lo = m + 1; else { lo = hi = m; break; } } /* Stop if there is no matching byte. */ if (hi < lo || byts[lo] != c) break; /* Continue at the child (if there is one). */ arridx = idxs[lo]; ++wlen; --flen; /* One space in the good word may stand for several spaces in the * checked word. */ if (c == ' ') { for (;;) { if (flen <= 0 && *mip->mi_fend != NUL) flen = fold_more(mip); if (ptr[wlen] != ' ' && ptr[wlen] != TAB) break; ++wlen; --flen; } } } /* * Verify that one of the possible endings is valid. Try the longest * first. */ while (endidxcnt > 0) { --endidxcnt; arridx = endidx[endidxcnt]; wlen = endlen[endidxcnt]; #ifdef FEAT_MBYTE if ((*mb_head_off)(ptr, ptr + wlen) > 0) continue; /* not at first byte of character */ #endif if (spell_iswordp(ptr + wlen, mip->mi_win)) { if (slang->sl_compprog == NULL && !slang->sl_nobreak) continue; /* next char is a word character */ word_ends = FALSE; } else word_ends = TRUE; /* The prefix flag is before compound flags. Once a valid prefix flag * has been found we try compound flags. */ prefix_found = FALSE; #ifdef FEAT_MBYTE if (mode != FIND_KEEPWORD && has_mbyte) { /* Compute byte length in original word, length may change * when folding case. This can be slow, take a shortcut when the * case-folded word is equal to the keep-case word. */ p = mip->mi_word; if (STRNCMP(ptr, p, wlen) != 0) { for (s = ptr; s < ptr + wlen; mb_ptr_adv(s)) mb_ptr_adv(p); wlen = (int)(p - mip->mi_word); } } #endif /* Check flags and region. For FIND_PREFIX check the condition and * prefix ID. * Repeat this if there are more flags/region alternatives until there * is a match. */ res = SP_BAD; for (len = byts[arridx - 1]; len > 0 && byts[arridx] == 0; --len, ++arridx) { flags = idxs[arridx]; /* For the fold-case tree check that the case of the checked word * matches with what the word in the tree requires. * For keep-case tree the case is always right. For prefixes we * don't bother to check. */ if (mode == FIND_FOLDWORD) { if (mip->mi_cend != mip->mi_word + wlen) { /* mi_capflags was set for a different word length, need * to do it again. */ mip->mi_cend = mip->mi_word + wlen; mip->mi_capflags = captype(mip->mi_word, mip->mi_cend); } if (mip->mi_capflags == WF_KEEPCAP || !spell_valid_case(mip->mi_capflags, flags)) continue; } /* When mode is FIND_PREFIX the word must support the prefix: * check the prefix ID and the condition. Do that for the list at * mip->mi_prefarridx that find_prefix() filled. */ else if (mode == FIND_PREFIX && !prefix_found) { c = valid_word_prefix(mip->mi_prefcnt, mip->mi_prefarridx, flags, mip->mi_word + mip->mi_cprefixlen, slang, FALSE); if (c == 0) continue; /* Use the WF_RARE flag for a rare prefix. */ if (c & WF_RAREPFX) flags |= WF_RARE; prefix_found = TRUE; } if (slang->sl_nobreak) { if ((mode == FIND_COMPOUND || mode == FIND_KEEPCOMPOUND) && (flags & WF_BANNED) == 0) { /* NOBREAK: found a valid following word. That's all we * need to know, so return. */ mip->mi_result = SP_OK; break; } } else if ((mode == FIND_COMPOUND || mode == FIND_KEEPCOMPOUND || !word_ends)) { /* If there is no compound flag or the word is shorter than * COMPOUNDMIN reject it quickly. * Makes you wonder why someone puts a compound flag on a word * that's too short... Myspell compatibility requires this * anyway. */ if (((unsigned)flags >> 24) == 0 || wlen - mip->mi_compoff < slang->sl_compminlen) continue; #ifdef FEAT_MBYTE /* For multi-byte chars check character length against * COMPOUNDMIN. */ if (has_mbyte && slang->sl_compminlen > 0 && mb_charlen_len(mip->mi_word + mip->mi_compoff, wlen - mip->mi_compoff) < slang->sl_compminlen) continue; #endif /* Limit the number of compound words to COMPOUNDWORDMAX if no * maximum for syllables is specified. */ if (!word_ends && mip->mi_complen + mip->mi_compextra + 2 > slang->sl_compmax && slang->sl_compsylmax == MAXWLEN) continue; /* Don't allow compounding on a side where an affix was added, * unless COMPOUNDPERMITFLAG was used. */ if (mip->mi_complen > 0 && (flags & WF_NOCOMPBEF)) continue; if (!word_ends && (flags & WF_NOCOMPAFT)) continue; /* Quickly check if compounding is possible with this flag. */ if (!byte_in_str(mip->mi_complen == 0 ? slang->sl_compstartflags : slang->sl_compallflags, ((unsigned)flags >> 24))) continue; /* If there is a match with a CHECKCOMPOUNDPATTERN rule * discard the compound word. */ if (match_checkcompoundpattern(ptr, wlen, &slang->sl_comppat)) continue; if (mode == FIND_COMPOUND) { int capflags; /* Need to check the caps type of the appended compound * word. */ #ifdef FEAT_MBYTE if (has_mbyte && STRNCMP(ptr, mip->mi_word, mip->mi_compoff) != 0) { /* case folding may have changed the length */ p = mip->mi_word; for (s = ptr; s < ptr + mip->mi_compoff; mb_ptr_adv(s)) mb_ptr_adv(p); } else #endif p = mip->mi_word + mip->mi_compoff; capflags = captype(p, mip->mi_word + wlen); if (capflags == WF_KEEPCAP || (capflags == WF_ALLCAP && (flags & WF_FIXCAP) != 0)) continue; if (capflags != WF_ALLCAP) { /* When the character before the word is a word * character we do not accept a Onecap word. We do * accept a no-caps word, even when the dictionary * word specifies ONECAP. */ mb_ptr_back(mip->mi_word, p); if (spell_iswordp_nmw(p) ? capflags == WF_ONECAP : (flags & WF_ONECAP) != 0 && capflags != WF_ONECAP) continue; } } /* If the word ends the sequence of compound flags of the * words must match with one of the COMPOUNDRULE items and * the number of syllables must not be too large. */ mip->mi_compflags[mip->mi_complen] = ((unsigned)flags >> 24); mip->mi_compflags[mip->mi_complen + 1] = NUL; if (word_ends) { char_u fword[MAXWLEN]; if (slang->sl_compsylmax < MAXWLEN) { /* "fword" is only needed for checking syllables. */ if (ptr == mip->mi_word) (void)spell_casefold(ptr, wlen, fword, MAXWLEN); else vim_strncpy(fword, ptr, endlen[endidxcnt]); } if (!can_compound(slang, fword, mip->mi_compflags)) continue; } else if (slang->sl_comprules != NULL && !match_compoundrule(slang, mip->mi_compflags)) /* The compound flags collected so far do not match any * COMPOUNDRULE, discard the compounded word. */ continue; } /* Check NEEDCOMPOUND: can't use word without compounding. */ else if (flags & WF_NEEDCOMP) continue; nobreak_result = SP_OK; if (!word_ends) { int save_result = mip->mi_result; char_u *save_end = mip->mi_end; langp_T *save_lp = mip->mi_lp; int lpi; /* Check that a valid word follows. If there is one and we * are compounding, it will set "mi_result", thus we are * always finished here. For NOBREAK we only check that a * valid word follows. * Recursive! */ if (slang->sl_nobreak) mip->mi_result = SP_BAD; /* Find following word in case-folded tree. */ mip->mi_compoff = endlen[endidxcnt]; #ifdef FEAT_MBYTE if (has_mbyte && mode == FIND_KEEPWORD) { /* Compute byte length in case-folded word from "wlen": * byte length in keep-case word. Length may change when * folding case. This can be slow, take a shortcut when * the case-folded word is equal to the keep-case word. */ p = mip->mi_fword; if (STRNCMP(ptr, p, wlen) != 0) { for (s = ptr; s < ptr + wlen; mb_ptr_adv(s)) mb_ptr_adv(p); mip->mi_compoff = (int)(p - mip->mi_fword); } } #endif c = mip->mi_compoff; ++mip->mi_complen; if (flags & WF_COMPROOT) ++mip->mi_compextra; /* For NOBREAK we need to try all NOBREAK languages, at least * to find the ".add" file(s). */ for (lpi = 0; lpi < mip->mi_win->w_s->b_langp.ga_len; ++lpi) { if (slang->sl_nobreak) { mip->mi_lp = LANGP_ENTRY(mip->mi_win->w_s->b_langp, lpi); if (mip->mi_lp->lp_slang->sl_fidxs == NULL || !mip->mi_lp->lp_slang->sl_nobreak) continue; } find_word(mip, FIND_COMPOUND); /* When NOBREAK any word that matches is OK. Otherwise we * need to find the longest match, thus try with keep-case * and prefix too. */ if (!slang->sl_nobreak || mip->mi_result == SP_BAD) { /* Find following word in keep-case tree. */ mip->mi_compoff = wlen; find_word(mip, FIND_KEEPCOMPOUND); #if 0 /* Disabled, a prefix must not appear halfway a compound word, unless the COMPOUNDPERMITFLAG is used and then it can't be a postponed prefix. */ if (!slang->sl_nobreak || mip->mi_result == SP_BAD) { /* Check for following word with prefix. */ mip->mi_compoff = c; find_prefix(mip, FIND_COMPOUND); } #endif } if (!slang->sl_nobreak) break; } --mip->mi_complen; if (flags & WF_COMPROOT) --mip->mi_compextra; mip->mi_lp = save_lp; if (slang->sl_nobreak) { nobreak_result = mip->mi_result; mip->mi_result = save_result; mip->mi_end = save_end; } else { if (mip->mi_result == SP_OK) break; continue; } } if (flags & WF_BANNED) res = SP_BANNED; else if (flags & WF_REGION) { /* Check region. */ if ((mip->mi_lp->lp_region & (flags >> 16)) != 0) res = SP_OK; else res = SP_LOCAL; } else if (flags & WF_RARE) res = SP_RARE; else res = SP_OK; /* Always use the longest match and the best result. For NOBREAK * we separately keep the longest match without a following good * word as a fall-back. */ if (nobreak_result == SP_BAD) { if (mip->mi_result2 > res) { mip->mi_result2 = res; mip->mi_end2 = mip->mi_word + wlen; } else if (mip->mi_result2 == res && mip->mi_end2 < mip->mi_word + wlen) mip->mi_end2 = mip->mi_word + wlen; } else if (mip->mi_result > res) { mip->mi_result = res; mip->mi_end = mip->mi_word + wlen; } else if (mip->mi_result == res && mip->mi_end < mip->mi_word + wlen) mip->mi_end = mip->mi_word + wlen; if (mip->mi_result == SP_OK) break; } if (mip->mi_result == SP_OK) break; } } /* * Return TRUE if there is a match between the word ptr[wlen] and * CHECKCOMPOUNDPATTERN rules, assuming that we will concatenate with another * word. * A match means that the first part of CHECKCOMPOUNDPATTERN matches at the * end of ptr[wlen] and the second part matches after it. */ static int match_checkcompoundpattern(ptr, wlen, gap) char_u *ptr; int wlen; garray_T *gap; /* &sl_comppat */ { int i; char_u *p; int len; for (i = 0; i + 1 < gap->ga_len; i += 2) { p = ((char_u **)gap->ga_data)[i + 1]; if (STRNCMP(ptr + wlen, p, STRLEN(p)) == 0) { /* Second part matches at start of following compound word, now * check if first part matches at end of previous word. */ p = ((char_u **)gap->ga_data)[i]; len = (int)STRLEN(p); if (len <= wlen && STRNCMP(ptr + wlen - len, p, len) == 0) return TRUE; } } return FALSE; } /* * Return TRUE if "flags" is a valid sequence of compound flags and "word" * does not have too many syllables. */ static int can_compound(slang, word, flags) slang_T *slang; char_u *word; char_u *flags; { regmatch_T regmatch; #ifdef FEAT_MBYTE char_u uflags[MAXWLEN * 2]; int i; #endif char_u *p; if (slang->sl_compprog == NULL) return FALSE; #ifdef FEAT_MBYTE if (enc_utf8) { /* Need to convert the single byte flags to utf8 characters. */ p = uflags; for (i = 0; flags[i] != NUL; ++i) p += mb_char2bytes(flags[i], p); *p = NUL; p = uflags; } else #endif p = flags; regmatch.regprog = slang->sl_compprog; regmatch.rm_ic = FALSE; if (!vim_regexec(®match, p, 0)) return FALSE; /* Count the number of syllables. This may be slow, do it last. If there * are too many syllables AND the number of compound words is above * COMPOUNDWORDMAX then compounding is not allowed. */ if (slang->sl_compsylmax < MAXWLEN && count_syllables(slang, word) > slang->sl_compsylmax) return (int)STRLEN(flags) < slang->sl_compmax; return TRUE; } /* * Return TRUE when the sequence of flags in "compflags" plus "flag" can * possibly form a valid compounded word. This also checks the COMPOUNDRULE * lines if they don't contain wildcards. */ static int can_be_compound(sp, slang, compflags, flag) trystate_T *sp; slang_T *slang; char_u *compflags; int flag; { /* If the flag doesn't appear in sl_compstartflags or sl_compallflags * then it can't possibly compound. */ if (!byte_in_str(sp->ts_complen == sp->ts_compsplit ? slang->sl_compstartflags : slang->sl_compallflags, flag)) return FALSE; /* If there are no wildcards, we can check if the flags collected so far * possibly can form a match with COMPOUNDRULE patterns. This only * makes sense when we have two or more words. */ if (slang->sl_comprules != NULL && sp->ts_complen > sp->ts_compsplit) { int v; compflags[sp->ts_complen] = flag; compflags[sp->ts_complen + 1] = NUL; v = match_compoundrule(slang, compflags + sp->ts_compsplit); compflags[sp->ts_complen] = NUL; return v; } return TRUE; } /* * Return TRUE if the compound flags in compflags[] match the start of any * compound rule. This is used to stop trying a compound if the flags * collected so far can't possibly match any compound rule. * Caller must check that slang->sl_comprules is not NULL. */ static int match_compoundrule(slang, compflags) slang_T *slang; char_u *compflags; { char_u *p; int i; int c; /* loop over all the COMPOUNDRULE entries */ for (p = slang->sl_comprules; *p != NUL; ++p) { /* loop over the flags in the compound word we have made, match * them against the current rule entry */ for (i = 0; ; ++i) { c = compflags[i]; if (c == NUL) /* found a rule that matches for the flags we have so far */ return TRUE; if (*p == '/' || *p == NUL) break; /* end of rule, it's too short */ if (*p == '[') { int match = FALSE; /* compare against all the flags in [] */ ++p; while (*p != ']' && *p != NUL) if (*p++ == c) match = TRUE; if (!match) break; /* none matches */ } else if (*p != c) break; /* flag of word doesn't match flag in pattern */ ++p; } /* Skip to the next "/", where the next pattern starts. */ p = vim_strchr(p, '/'); if (p == NULL) break; } /* Checked all the rules and none of them match the flags, so there * can't possibly be a compound starting with these flags. */ return FALSE; } /* * Return non-zero if the prefix indicated by "arridx" matches with the prefix * ID in "flags" for the word "word". * The WF_RAREPFX flag is included in the return value for a rare prefix. */ static int valid_word_prefix(totprefcnt, arridx, flags, word, slang, cond_req) int totprefcnt; /* nr of prefix IDs */ int arridx; /* idx in sl_pidxs[] */ int flags; char_u *word; slang_T *slang; int cond_req; /* only use prefixes with a condition */ { int prefcnt; int pidx; regprog_T *rp; regmatch_T regmatch; int prefid; prefid = (unsigned)flags >> 24; for (prefcnt = totprefcnt - 1; prefcnt >= 0; --prefcnt) { pidx = slang->sl_pidxs[arridx + prefcnt]; /* Check the prefix ID. */ if (prefid != (pidx & 0xff)) continue; /* Check if the prefix doesn't combine and the word already has a * suffix. */ if ((flags & WF_HAS_AFF) && (pidx & WF_PFX_NC)) continue; /* Check the condition, if there is one. The condition index is * stored in the two bytes above the prefix ID byte. */ rp = slang->sl_prefprog[((unsigned)pidx >> 8) & 0xffff]; if (rp != NULL) { regmatch.regprog = rp; regmatch.rm_ic = FALSE; if (!vim_regexec(®match, word, 0)) continue; } else if (cond_req) continue; /* It's a match! Return the WF_ flags. */ return pidx; } return 0; } /* * Check if the word at "mip->mi_word" has a matching prefix. * If it does, then check the following word. * * If "mode" is "FIND_COMPOUND" then do the same after another word, find a * prefix in a compound word. * * For a match mip->mi_result is updated. */ static void find_prefix(mip, mode) matchinf_T *mip; int mode; { idx_T arridx = 0; int len; int wlen = 0; int flen; int c; char_u *ptr; idx_T lo, hi, m; slang_T *slang = mip->mi_lp->lp_slang; char_u *byts; idx_T *idxs; byts = slang->sl_pbyts; if (byts == NULL) return; /* array is empty */ /* We use the case-folded word here, since prefixes are always * case-folded. */ ptr = mip->mi_fword; flen = mip->mi_fwordlen; /* available case-folded bytes */ if (mode == FIND_COMPOUND) { /* Skip over the previously found word(s). */ ptr += mip->mi_compoff; flen -= mip->mi_compoff; } idxs = slang->sl_pidxs; /* * Repeat advancing in the tree until: * - there is a byte that doesn't match, * - we reach the end of the tree, * - or we reach the end of the line. */ for (;;) { if (flen == 0 && *mip->mi_fend != NUL) flen = fold_more(mip); len = byts[arridx++]; /* If the first possible byte is a zero the prefix could end here. * Check if the following word matches and supports the prefix. */ if (byts[arridx] == 0) { /* There can be several prefixes with different conditions. We * try them all, since we don't know which one will give the * longest match. The word is the same each time, pass the list * of possible prefixes to find_word(). */ mip->mi_prefarridx = arridx; mip->mi_prefcnt = len; while (len > 0 && byts[arridx] == 0) { ++arridx; --len; } mip->mi_prefcnt -= len; /* Find the word that comes after the prefix. */ mip->mi_prefixlen = wlen; if (mode == FIND_COMPOUND) /* Skip over the previously found word(s). */ mip->mi_prefixlen += mip->mi_compoff; #ifdef FEAT_MBYTE if (has_mbyte) { /* Case-folded length may differ from original length. */ mip->mi_cprefixlen = nofold_len(mip->mi_fword, mip->mi_prefixlen, mip->mi_word); } else mip->mi_cprefixlen = mip->mi_prefixlen; #endif find_word(mip, FIND_PREFIX); if (len == 0) break; /* no children, word must end here */ } /* Stop looking at end of the line. */ if (ptr[wlen] == NUL) break; /* Perform a binary search in the list of accepted bytes. */ c = ptr[wlen]; lo = arridx; hi = arridx + len - 1; while (lo < hi) { m = (lo + hi) / 2; if (byts[m] > c) hi = m - 1; else if (byts[m] < c) lo = m + 1; else { lo = hi = m; break; } } /* Stop if there is no matching byte. */ if (hi < lo || byts[lo] != c) break; /* Continue at the child (if there is one). */ arridx = idxs[lo]; ++wlen; --flen; } } /* * Need to fold at least one more character. Do until next non-word character * for efficiency. Include the non-word character too. * Return the length of the folded chars in bytes. */ static int fold_more(mip) matchinf_T *mip; { int flen; char_u *p; p = mip->mi_fend; do { mb_ptr_adv(mip->mi_fend); } while (*mip->mi_fend != NUL && spell_iswordp(mip->mi_fend, mip->mi_win)); /* Include the non-word character so that we can check for the word end. */ if (*mip->mi_fend != NUL) mb_ptr_adv(mip->mi_fend); (void)spell_casefold(p, (int)(mip->mi_fend - p), mip->mi_fword + mip->mi_fwordlen, MAXWLEN - mip->mi_fwordlen); flen = (int)STRLEN(mip->mi_fword + mip->mi_fwordlen); mip->mi_fwordlen += flen; return flen; } /* * Check case flags for a word. Return TRUE if the word has the requested * case. */ static int spell_valid_case(wordflags, treeflags) int wordflags; /* flags for the checked word. */ int treeflags; /* flags for the word in the spell tree */ { return ((wordflags == WF_ALLCAP && (treeflags & WF_FIXCAP) == 0) || ((treeflags & (WF_ALLCAP | WF_KEEPCAP)) == 0 && ((treeflags & WF_ONECAP) == 0 || (wordflags & WF_ONECAP) != 0))); } /* * Return TRUE if spell checking is not enabled. */ static int no_spell_checking(wp) win_T *wp; { if (!wp->w_p_spell || *wp->w_s->b_p_spl == NUL || wp->w_s->b_langp.ga_len == 0) { EMSG(_("E756: Spell checking is not enabled")); return TRUE; } return FALSE; } /* * Move to next spell error. * "curline" is FALSE for "[s", "]s", "[S" and "]S". * "curline" is TRUE to find word under/after cursor in the same line. * For Insert mode completion "dir" is BACKWARD and "curline" is TRUE: move * to after badly spelled word before the cursor. * Return 0 if not found, length of the badly spelled word otherwise. */ int spell_move_to(wp, dir, allwords, curline, attrp) win_T *wp; int dir; /* FORWARD or BACKWARD */ int allwords; /* TRUE for "[s"/"]s", FALSE for "[S"/"]S" */ int curline; hlf_T *attrp; /* return: attributes of bad word or NULL (only when "dir" is FORWARD) */ { linenr_T lnum; pos_T found_pos; int found_len = 0; char_u *line; char_u *p; char_u *endp; hlf_T attr; int len; # ifdef FEAT_SYN_HL int has_syntax = syntax_present(wp); # endif int col; int can_spell; char_u *buf = NULL; int buflen = 0; int skip = 0; int capcol = -1; int found_one = FALSE; int wrapped = FALSE; if (no_spell_checking(wp)) return 0; /* * Start looking for bad word at the start of the line, because we can't * start halfway a word, we don't know where it starts or ends. * * When searching backwards, we continue in the line to find the last * bad word (in the cursor line: before the cursor). * * We concatenate the start of the next line, so that wrapped words work * (e.g. "etcetera"). Doesn't work when searching backwards * though... */ lnum = wp->w_cursor.lnum; clearpos(&found_pos); while (!got_int) { line = ml_get_buf(wp->w_buffer, lnum, FALSE); len = (int)STRLEN(line); if (buflen < len + MAXWLEN + 2) { vim_free(buf); buflen = len + MAXWLEN + 2; buf = alloc(buflen); if (buf == NULL) break; } /* In first line check first word for Capital. */ if (lnum == 1) capcol = 0; /* For checking first word with a capital skip white space. */ if (capcol == 0) capcol = (int)(skipwhite(line) - line); else if (curline && wp == curwin) { /* For spellbadword(): check if first word needs a capital. */ col = (int)(skipwhite(line) - line); if (check_need_cap(lnum, col)) capcol = col; /* Need to get the line again, may have looked at the previous * one. */ line = ml_get_buf(wp->w_buffer, lnum, FALSE); } /* Copy the line into "buf" and append the start of the next line if * possible. */ STRCPY(buf, line); if (lnum < wp->w_buffer->b_ml.ml_line_count) spell_cat_line(buf + STRLEN(buf), ml_get_buf(wp->w_buffer, lnum + 1, FALSE), MAXWLEN); p = buf + skip; endp = buf + len; while (p < endp) { /* When searching backward don't search after the cursor. Unless * we wrapped around the end of the buffer. */ if (dir == BACKWARD && lnum == wp->w_cursor.lnum && !wrapped && (colnr_T)(p - buf) >= wp->w_cursor.col) break; /* start of word */ attr = HLF_COUNT; len = spell_check(wp, p, &attr, &capcol, FALSE); if (attr != HLF_COUNT) { /* We found a bad word. Check the attribute. */ if (allwords || attr == HLF_SPB) { /* When searching forward only accept a bad word after * the cursor. */ if (dir == BACKWARD || lnum != wp->w_cursor.lnum || (lnum == wp->w_cursor.lnum && (wrapped || (colnr_T)(curline ? p - buf + len : p - buf) > wp->w_cursor.col))) { # ifdef FEAT_SYN_HL if (has_syntax) { col = (int)(p - buf); (void)syn_get_id(wp, lnum, (colnr_T)col, FALSE, &can_spell, FALSE); if (!can_spell) attr = HLF_COUNT; } else #endif can_spell = TRUE; if (can_spell) { found_one = TRUE; found_pos.lnum = lnum; found_pos.col = (int)(p - buf); #ifdef FEAT_VIRTUALEDIT found_pos.coladd = 0; #endif if (dir == FORWARD) { /* No need to search further. */ wp->w_cursor = found_pos; vim_free(buf); if (attrp != NULL) *attrp = attr; return len; } else if (curline) /* Insert mode completion: put cursor after * the bad word. */ found_pos.col += len; found_len = len; } } else found_one = TRUE; } } /* advance to character after the word */ p += len; capcol -= len; } if (dir == BACKWARD && found_pos.lnum != 0) { /* Use the last match in the line (before the cursor). */ wp->w_cursor = found_pos; vim_free(buf); return found_len; } if (curline) break; /* only check cursor line */ /* Advance to next line. */ if (dir == BACKWARD) { /* If we are back at the starting line and searched it again there * is no match, give up. */ if (lnum == wp->w_cursor.lnum && wrapped) break; if (lnum > 1) --lnum; else if (!p_ws) break; /* at first line and 'nowrapscan' */ else { /* Wrap around to the end of the buffer. May search the * starting line again and accept the last match. */ lnum = wp->w_buffer->b_ml.ml_line_count; wrapped = TRUE; if (!shortmess(SHM_SEARCH)) give_warning((char_u *)_(top_bot_msg), TRUE); } capcol = -1; } else { if (lnum < wp->w_buffer->b_ml.ml_line_count) ++lnum; else if (!p_ws) break; /* at first line and 'nowrapscan' */ else { /* Wrap around to the start of the buffer. May search the * starting line again and accept the first match. */ lnum = 1; wrapped = TRUE; if (!shortmess(SHM_SEARCH)) give_warning((char_u *)_(bot_top_msg), TRUE); } /* If we are back at the starting line and there is no match then * give up. */ if (lnum == wp->w_cursor.lnum && (!found_one || wrapped)) break; /* Skip the characters at the start of the next line that were * included in a match crossing line boundaries. */ if (attr == HLF_COUNT) skip = (int)(p - endp); else skip = 0; /* Capcol skips over the inserted space. */ --capcol; /* But after empty line check first word in next line */ if (*skipwhite(line) == NUL) capcol = 0; } line_breakcheck(); } vim_free(buf); return 0; } /* * For spell checking: concatenate the start of the following line "line" into * "buf", blanking-out special characters. Copy less then "maxlen" bytes. * Keep the blanks at the start of the next line, this is used in win_line() * to skip those bytes if the word was OK. */ void spell_cat_line(buf, line, maxlen) char_u *buf; char_u *line; int maxlen; { char_u *p; int n; p = skipwhite(line); while (vim_strchr((char_u *)"*#/\"\t", *p) != NULL) p = skipwhite(p + 1); if (*p != NUL) { /* Only worth concatenating if there is something else than spaces to * concatenate. */ n = (int)(p - line) + 1; if (n < maxlen - 1) { vim_memset(buf, ' ', n); vim_strncpy(buf + n, p, maxlen - 1 - n); } } } /* * Structure used for the cookie argument of do_in_runtimepath(). */ typedef struct spelload_S { char_u sl_lang[MAXWLEN + 1]; /* language name */ slang_T *sl_slang; /* resulting slang_T struct */ int sl_nobreak; /* NOBREAK language found */ } spelload_T; /* * Load word list(s) for "lang" from Vim spell file(s). * "lang" must be the language without the region: e.g., "en". */ static void spell_load_lang(lang) char_u *lang; { char_u fname_enc[85]; int r; spelload_T sl; #ifdef FEAT_AUTOCMD int round; #endif /* Copy the language name to pass it to spell_load_cb() as a cookie. * It's truncated when an error is detected. */ STRCPY(sl.sl_lang, lang); sl.sl_slang = NULL; sl.sl_nobreak = FALSE; #ifdef FEAT_AUTOCMD /* We may retry when no spell file is found for the language, an * autocommand may load it then. */ for (round = 1; round <= 2; ++round) #endif { /* * Find the first spell file for "lang" in 'runtimepath' and load it. */ vim_snprintf((char *)fname_enc, sizeof(fname_enc) - 5, #ifdef VMS "spell/%s_%s.spl", #else "spell/%s.%s.spl", #endif lang, spell_enc()); r = do_in_runtimepath(fname_enc, FALSE, spell_load_cb, &sl); if (r == FAIL && *sl.sl_lang != NUL) { /* Try loading the ASCII version. */ vim_snprintf((char *)fname_enc, sizeof(fname_enc) - 5, #ifdef VMS "spell/%s_ascii.spl", #else "spell/%s.ascii.spl", #endif lang); r = do_in_runtimepath(fname_enc, FALSE, spell_load_cb, &sl); #ifdef FEAT_AUTOCMD if (r == FAIL && *sl.sl_lang != NUL && round == 1 && apply_autocmds(EVENT_SPELLFILEMISSING, lang, curbuf->b_fname, FALSE, curbuf)) continue; break; #endif } #ifdef FEAT_AUTOCMD break; #endif } if (r == FAIL) { smsg((char_u *) #ifdef VMS _("Warning: Cannot find word list \"%s_%s.spl\" or \"%s_ascii.spl\""), #else _("Warning: Cannot find word list \"%s.%s.spl\" or \"%s.ascii.spl\""), #endif lang, spell_enc(), lang); } else if (sl.sl_slang != NULL) { /* At least one file was loaded, now load ALL the additions. */ STRCPY(fname_enc + STRLEN(fname_enc) - 3, "add.spl"); do_in_runtimepath(fname_enc, TRUE, spell_load_cb, &sl); } } /* * Return the encoding used for spell checking: Use 'encoding', except that we * use "latin1" for "latin9". And limit to 60 characters (just in case). */ static char_u * spell_enc() { #ifdef FEAT_MBYTE if (STRLEN(p_enc) < 60 && STRCMP(p_enc, "iso-8859-15") != 0) return p_enc; #endif return (char_u *)"latin1"; } /* * Get the name of the .spl file for the internal wordlist into * "fname[MAXPATHL]". */ static void int_wordlist_spl(fname) char_u *fname; { vim_snprintf((char *)fname, MAXPATHL, SPL_FNAME_TMPL, int_wordlist, spell_enc()); } /* * Allocate a new slang_T for language "lang". "lang" can be NULL. * Caller must fill "sl_next". */ static slang_T * slang_alloc(lang) char_u *lang; { slang_T *lp; lp = (slang_T *)alloc_clear(sizeof(slang_T)); if (lp != NULL) { if (lang != NULL) lp->sl_name = vim_strsave(lang); ga_init2(&lp->sl_rep, sizeof(fromto_T), 10); ga_init2(&lp->sl_repsal, sizeof(fromto_T), 10); lp->sl_compmax = MAXWLEN; lp->sl_compsylmax = MAXWLEN; hash_init(&lp->sl_wordcount); } return lp; } /* * Free the contents of an slang_T and the structure itself. */ static void slang_free(lp) slang_T *lp; { vim_free(lp->sl_name); vim_free(lp->sl_fname); slang_clear(lp); vim_free(lp); } /* * Clear an slang_T so that the file can be reloaded. */ static void slang_clear(lp) slang_T *lp; { garray_T *gap; fromto_T *ftp; salitem_T *smp; int i; int round; vim_free(lp->sl_fbyts); lp->sl_fbyts = NULL; vim_free(lp->sl_kbyts); lp->sl_kbyts = NULL; vim_free(lp->sl_pbyts); lp->sl_pbyts = NULL; vim_free(lp->sl_fidxs); lp->sl_fidxs = NULL; vim_free(lp->sl_kidxs); lp->sl_kidxs = NULL; vim_free(lp->sl_pidxs); lp->sl_pidxs = NULL; for (round = 1; round <= 2; ++round) { gap = round == 1 ? &lp->sl_rep : &lp->sl_repsal; while (gap->ga_len > 0) { ftp = &((fromto_T *)gap->ga_data)[--gap->ga_len]; vim_free(ftp->ft_from); vim_free(ftp->ft_to); } ga_clear(gap); } gap = &lp->sl_sal; if (lp->sl_sofo) { /* "ga_len" is set to 1 without adding an item for latin1 */ if (gap->ga_data != NULL) /* SOFOFROM and SOFOTO items: free lists of wide characters. */ for (i = 0; i < gap->ga_len; ++i) vim_free(((int **)gap->ga_data)[i]); } else /* SAL items: free salitem_T items */ while (gap->ga_len > 0) { smp = &((salitem_T *)gap->ga_data)[--gap->ga_len]; vim_free(smp->sm_lead); /* Don't free sm_oneof and sm_rules, they point into sm_lead. */ vim_free(smp->sm_to); #ifdef FEAT_MBYTE vim_free(smp->sm_lead_w); vim_free(smp->sm_oneof_w); vim_free(smp->sm_to_w); #endif } ga_clear(gap); for (i = 0; i < lp->sl_prefixcnt; ++i) vim_free(lp->sl_prefprog[i]); lp->sl_prefixcnt = 0; vim_free(lp->sl_prefprog); lp->sl_prefprog = NULL; vim_free(lp->sl_info); lp->sl_info = NULL; vim_free(lp->sl_midword); lp->sl_midword = NULL; vim_free(lp->sl_compprog); vim_free(lp->sl_comprules); vim_free(lp->sl_compstartflags); vim_free(lp->sl_compallflags); lp->sl_compprog = NULL; lp->sl_comprules = NULL; lp->sl_compstartflags = NULL; lp->sl_compallflags = NULL; vim_free(lp->sl_syllable); lp->sl_syllable = NULL; ga_clear(&lp->sl_syl_items); ga_clear_strings(&lp->sl_comppat); hash_clear_all(&lp->sl_wordcount, WC_KEY_OFF); hash_init(&lp->sl_wordcount); #ifdef FEAT_MBYTE hash_clear_all(&lp->sl_map_hash, 0); #endif /* Clear info from .sug file. */ slang_clear_sug(lp); lp->sl_compmax = MAXWLEN; lp->sl_compminlen = 0; lp->sl_compsylmax = MAXWLEN; lp->sl_regions[0] = NUL; } /* * Clear the info from the .sug file in "lp". */ static void slang_clear_sug(lp) slang_T *lp; { vim_free(lp->sl_sbyts); lp->sl_sbyts = NULL; vim_free(lp->sl_sidxs); lp->sl_sidxs = NULL; close_spellbuf(lp->sl_sugbuf); lp->sl_sugbuf = NULL; lp->sl_sugloaded = FALSE; lp->sl_sugtime = 0; } /* * Load one spell file and store the info into a slang_T. * Invoked through do_in_runtimepath(). */ static void spell_load_cb(fname, cookie) char_u *fname; void *cookie; { spelload_T *slp = (spelload_T *)cookie; slang_T *slang; slang = spell_load_file(fname, slp->sl_lang, NULL, FALSE); if (slang != NULL) { /* When a previously loaded file has NOBREAK also use it for the * ".add" files. */ if (slp->sl_nobreak && slang->sl_add) slang->sl_nobreak = TRUE; else if (slang->sl_nobreak) slp->sl_nobreak = TRUE; slp->sl_slang = slang; } } /* * Load one spell file and store the info into a slang_T. * * This is invoked in three ways: * - From spell_load_cb() to load a spell file for the first time. "lang" is * the language name, "old_lp" is NULL. Will allocate an slang_T. * - To reload a spell file that was changed. "lang" is NULL and "old_lp" * points to the existing slang_T. * - Just after writing a .spl file; it's read back to produce the .sug file. * "old_lp" is NULL and "lang" is NULL. Will allocate an slang_T. * * Returns the slang_T the spell file was loaded into. NULL for error. */ static slang_T * spell_load_file(fname, lang, old_lp, silent) char_u *fname; char_u *lang; slang_T *old_lp; int silent; /* no error if file doesn't exist */ { FILE *fd; char_u buf[VIMSPELLMAGICL]; char_u *p; int i; int n; int len; char_u *save_sourcing_name = sourcing_name; linenr_T save_sourcing_lnum = sourcing_lnum; slang_T *lp = NULL; int c = 0; int res; fd = mch_fopen((char *)fname, "r"); if (fd == NULL) { if (!silent) EMSG2(_(e_notopen), fname); else if (p_verbose > 2) { verbose_enter(); smsg((char_u *)e_notopen, fname); verbose_leave(); } goto endFAIL; } if (p_verbose > 2) { verbose_enter(); smsg((char_u *)_("Reading spell file \"%s\""), fname); verbose_leave(); } if (old_lp == NULL) { lp = slang_alloc(lang); if (lp == NULL) goto endFAIL; /* Remember the file name, used to reload the file when it's updated. */ lp->sl_fname = vim_strsave(fname); if (lp->sl_fname == NULL) goto endFAIL; /* Check for .add.spl (_add.spl for VMS). */ lp->sl_add = strstr((char *)gettail(fname), SPL_FNAME_ADD) != NULL; } else lp = old_lp; /* Set sourcing_name, so that error messages mention the file name. */ sourcing_name = fname; sourcing_lnum = 0; /* *

: */ for (i = 0; i < VIMSPELLMAGICL; ++i) buf[i] = getc(fd); /* */ if (STRNCMP(buf, VIMSPELLMAGIC, VIMSPELLMAGICL) != 0) { EMSG(_("E757: This does not look like a spell file")); goto endFAIL; } c = getc(fd); /* */ if (c < VIMSPELLVERSION) { EMSG(_("E771: Old spell file, needs to be updated")); goto endFAIL; } else if (c > VIMSPELLVERSION) { EMSG(_("E772: Spell file is for newer version of Vim")); goto endFAIL; } /* * :

... *

: (section contents) */ for (;;) { n = getc(fd); /* or */ if (n == SN_END) break; c = getc(fd); /* */ len = get4c(fd); /* */ if (len < 0) goto truncerr; res = 0; switch (n) { case SN_INFO: lp->sl_info = read_string(fd, len); /* */ if (lp->sl_info == NULL) goto endFAIL; break; case SN_REGION: res = read_region_section(fd, lp, len); break; case SN_CHARFLAGS: res = read_charflags_section(fd); break; case SN_MIDWORD: lp->sl_midword = read_string(fd, len); /* */ if (lp->sl_midword == NULL) goto endFAIL; break; case SN_PREFCOND: res = read_prefcond_section(fd, lp); break; case SN_REP: res = read_rep_section(fd, &lp->sl_rep, lp->sl_rep_first); break; case SN_REPSAL: res = read_rep_section(fd, &lp->sl_repsal, lp->sl_repsal_first); break; case SN_SAL: res = read_sal_section(fd, lp); break; case SN_SOFO: res = read_sofo_section(fd, lp); break; case SN_MAP: p = read_string(fd, len); /* */ if (p == NULL) goto endFAIL; set_map_str(lp, p); vim_free(p); break; case SN_WORDS: res = read_words_section(fd, lp, len); break; case SN_SUGFILE: lp->sl_sugtime = get8ctime(fd); /* */ break; case SN_NOSPLITSUGS: lp->sl_nosplitsugs = TRUE; /* */ break; case SN_COMPOUND: res = read_compound(fd, lp, len); break; case SN_NOBREAK: lp->sl_nobreak = TRUE; break; case SN_SYLLABLE: lp->sl_syllable = read_string(fd, len); /* */ if (lp->sl_syllable == NULL) goto endFAIL; if (init_syl_tab(lp) == FAIL) goto endFAIL; break; default: /* Unsupported section. When it's required give an error * message. When it's not required skip the contents. */ if (c & SNF_REQUIRED) { EMSG(_("E770: Unsupported section in spell file")); goto endFAIL; } while (--len >= 0) if (getc(fd) < 0) goto truncerr; break; } someerror: if (res == SP_FORMERROR) { EMSG(_(e_format)); goto endFAIL; } if (res == SP_TRUNCERROR) { truncerr: EMSG(_(e_spell_trunc)); goto endFAIL; } if (res == SP_OTHERERROR) goto endFAIL; } /* */ res = spell_read_tree(fd, &lp->sl_fbyts, &lp->sl_fidxs, FALSE, 0); if (res != 0) goto someerror; /* */ res = spell_read_tree(fd, &lp->sl_kbyts, &lp->sl_kidxs, FALSE, 0); if (res != 0) goto someerror; /* */ res = spell_read_tree(fd, &lp->sl_pbyts, &lp->sl_pidxs, TRUE, lp->sl_prefixcnt); if (res != 0) goto someerror; /* For a new file link it in the list of spell files. */ if (old_lp == NULL && lang != NULL) { lp->sl_next = first_lang; first_lang = lp; } goto endOK; endFAIL: if (lang != NULL) /* truncating the name signals the error to spell_load_lang() */ *lang = NUL; if (lp != NULL && old_lp == NULL) slang_free(lp); lp = NULL; endOK: if (fd != NULL) fclose(fd); sourcing_name = save_sourcing_name; sourcing_lnum = save_sourcing_lnum; return lp; } /* * Read a length field from "fd" in "cnt_bytes" bytes. * Allocate memory, read the string into it and add a NUL at the end. * Returns NULL when the count is zero. * Sets "*cntp" to SP_*ERROR when there is an error, length of the result * otherwise. */ static char_u * read_cnt_string(fd, cnt_bytes, cntp) FILE *fd; int cnt_bytes; int *cntp; { int cnt = 0; int i; char_u *str; /* read the length bytes, MSB first */ for (i = 0; i < cnt_bytes; ++i) cnt = (cnt << 8) + getc(fd); if (cnt < 0) { *cntp = SP_TRUNCERROR; return NULL; } *cntp = cnt; if (cnt == 0) return NULL; /* nothing to read, return NULL */ str = read_string(fd, cnt); if (str == NULL) *cntp = SP_OTHERERROR; return str; } /* * Read SN_REGION: ... * Return SP_*ERROR flags. */ static int read_region_section(fd, lp, len) FILE *fd; slang_T *lp; int len; { int i; if (len > 16) return SP_FORMERROR; for (i = 0; i < len; ++i) lp->sl_regions[i] = getc(fd); /* */ lp->sl_regions[len] = NUL; return 0; } /* * Read SN_CHARFLAGS section: * * Return SP_*ERROR flags. */ static int read_charflags_section(fd) FILE *fd; { char_u *flags; char_u *fol; int flagslen, follen; /* */ flags = read_cnt_string(fd, 1, &flagslen); if (flagslen < 0) return flagslen; /* */ fol = read_cnt_string(fd, 2, &follen); if (follen < 0) { vim_free(flags); return follen; } /* Set the word-char flags and fill SPELL_ISUPPER() table. */ if (flags != NULL && fol != NULL) set_spell_charflags(flags, flagslen, fol); vim_free(flags); vim_free(fol); /* When is zero then must also be zero. */ if ((flags == NULL) != (fol == NULL)) return SP_FORMERROR; return 0; } /* * Read SN_PREFCOND section. * Return SP_*ERROR flags. */ static int read_prefcond_section(fd, lp) FILE *fd; slang_T *lp; { int cnt; int i; int n; char_u *p; char_u buf[MAXWLEN + 1]; /* ... */ cnt = get2c(fd); /* */ if (cnt <= 0) return SP_FORMERROR; lp->sl_prefprog = (regprog_T **)alloc_clear( (unsigned)sizeof(regprog_T *) * cnt); if (lp->sl_prefprog == NULL) return SP_OTHERERROR; lp->sl_prefixcnt = cnt; for (i = 0; i < cnt; ++i) { /* : */ n = getc(fd); /* */ if (n < 0 || n >= MAXWLEN) return SP_FORMERROR; /* When is zero we have an empty condition. Otherwise * compile the regexp program used to check for the condition. */ if (n > 0) { buf[0] = '^'; /* always match at one position only */ p = buf + 1; while (n-- > 0) *p++ = getc(fd); /* */ *p = NUL; lp->sl_prefprog[i] = vim_regcomp(buf, RE_MAGIC + RE_STRING); } } return 0; } /* * Read REP or REPSAL items section from "fd": ... * Return SP_*ERROR flags. */ static int read_rep_section(fd, gap, first) FILE *fd; garray_T *gap; short *first; { int cnt; fromto_T *ftp; int i; cnt = get2c(fd); /* */ if (cnt < 0) return SP_TRUNCERROR; if (ga_grow(gap, cnt) == FAIL) return SP_OTHERERROR; /* : */ for (; gap->ga_len < cnt; ++gap->ga_len) { ftp = &((fromto_T *)gap->ga_data)[gap->ga_len]; ftp->ft_from = read_cnt_string(fd, 1, &i); if (i < 0) return i; if (i == 0) return SP_FORMERROR; ftp->ft_to = read_cnt_string(fd, 1, &i); if (i <= 0) { vim_free(ftp->ft_from); if (i < 0) return i; return SP_FORMERROR; } } /* Fill the first-index table. */ for (i = 0; i < 256; ++i) first[i] = -1; for (i = 0; i < gap->ga_len; ++i) { ftp = &((fromto_T *)gap->ga_data)[i]; if (first[*ftp->ft_from] == -1) first[*ftp->ft_from] = i; } return 0; } /* * Read SN_SAL section: ... * Return SP_*ERROR flags. */ static int read_sal_section(fd, slang) FILE *fd; slang_T *slang; { int i; int cnt; garray_T *gap; salitem_T *smp; int ccnt; char_u *p; int c = NUL; slang->sl_sofo = FALSE; i = getc(fd); /* */ if (i & SAL_F0LLOWUP) slang->sl_followup = TRUE; if (i & SAL_COLLAPSE) slang->sl_collapse = TRUE; if (i & SAL_REM_ACCENTS) slang->sl_rem_accents = TRUE; cnt = get2c(fd); /* */ if (cnt < 0) return SP_TRUNCERROR; gap = &slang->sl_sal; ga_init2(gap, sizeof(salitem_T), 10); if (ga_grow(gap, cnt + 1) == FAIL) return SP_OTHERERROR; /* : */ for (; gap->ga_len < cnt; ++gap->ga_len) { smp = &((salitem_T *)gap->ga_data)[gap->ga_len]; ccnt = getc(fd); /* */ if (ccnt < 0) return SP_TRUNCERROR; if ((p = alloc(ccnt + 2)) == NULL) return SP_OTHERERROR; smp->sm_lead = p; /* Read up to the first special char into sm_lead. */ for (i = 0; i < ccnt; ++i) { c = getc(fd); /* */ if (vim_strchr((char_u *)"0123456789(-<^$", c) != NULL) break; *p++ = c; } smp->sm_leadlen = (int)(p - smp->sm_lead); *p++ = NUL; /* Put (abc) chars in sm_oneof, if any. */ if (c == '(') { smp->sm_oneof = p; for (++i; i < ccnt; ++i) { c = getc(fd); /* */ if (c == ')') break; *p++ = c; } *p++ = NUL; if (++i < ccnt) c = getc(fd); } else smp->sm_oneof = NULL; /* Any following chars go in sm_rules. */ smp->sm_rules = p; if (i < ccnt) /* store the char we got while checking for end of sm_lead */ *p++ = c; for (++i; i < ccnt; ++i) *p++ = getc(fd); /* */ *p++ = NUL; /* */ smp->sm_to = read_cnt_string(fd, 1, &ccnt); if (ccnt < 0) { vim_free(smp->sm_lead); return ccnt; } #ifdef FEAT_MBYTE if (has_mbyte) { /* convert the multi-byte strings to wide char strings */ smp->sm_lead_w = mb_str2wide(smp->sm_lead); smp->sm_leadlen = mb_charlen(smp->sm_lead); if (smp->sm_oneof == NULL) smp->sm_oneof_w = NULL; else smp->sm_oneof_w = mb_str2wide(smp->sm_oneof); if (smp->sm_to == NULL) smp->sm_to_w = NULL; else smp->sm_to_w = mb_str2wide(smp->sm_to); if (smp->sm_lead_w == NULL || (smp->sm_oneof_w == NULL && smp->sm_oneof != NULL) || (smp->sm_to_w == NULL && smp->sm_to != NULL)) { vim_free(smp->sm_lead); vim_free(smp->sm_to); vim_free(smp->sm_lead_w); vim_free(smp->sm_oneof_w); vim_free(smp->sm_to_w); return SP_OTHERERROR; } } #endif } if (gap->ga_len > 0) { /* Add one extra entry to mark the end with an empty sm_lead. Avoids * that we need to check the index every time. */ smp = &((salitem_T *)gap->ga_data)[gap->ga_len]; if ((p = alloc(1)) == NULL) return SP_OTHERERROR; p[0] = NUL; smp->sm_lead = p; smp->sm_leadlen = 0; smp->sm_oneof = NULL; smp->sm_rules = p; smp->sm_to = NULL; #ifdef FEAT_MBYTE if (has_mbyte) { smp->sm_lead_w = mb_str2wide(smp->sm_lead); smp->sm_leadlen = 0; smp->sm_oneof_w = NULL; smp->sm_to_w = NULL; } #endif ++gap->ga_len; } /* Fill the first-index table. */ set_sal_first(slang); return 0; } /* * Read SN_WORDS: ... * Return SP_*ERROR flags. */ static int read_words_section(fd, lp, len) FILE *fd; slang_T *lp; int len; { int done = 0; int i; int c; char_u word[MAXWLEN]; while (done < len) { /* Read one word at a time. */ for (i = 0; ; ++i) { c = getc(fd); if (c == EOF) return SP_TRUNCERROR; word[i] = c; if (word[i] == NUL) break; if (i == MAXWLEN - 1) return SP_FORMERROR; } /* Init the count to 10. */ count_common_word(lp, word, -1, 10); done += i + 1; } return 0; } /* * Add a word to the hashtable of common words. * If it's already there then the counter is increased. */ static void count_common_word(lp, word, len, count) slang_T *lp; char_u *word; int len; /* word length, -1 for upto NUL */ int count; /* 1 to count once, 10 to init */ { hash_T hash; hashitem_T *hi; wordcount_T *wc; char_u buf[MAXWLEN]; char_u *p; if (len == -1) p = word; else { vim_strncpy(buf, word, len); p = buf; } hash = hash_hash(p); hi = hash_lookup(&lp->sl_wordcount, p, hash); if (HASHITEM_EMPTY(hi)) { wc = (wordcount_T *)alloc((unsigned)(sizeof(wordcount_T) + STRLEN(p))); if (wc == NULL) return; STRCPY(wc->wc_word, p); wc->wc_count = count; hash_add_item(&lp->sl_wordcount, hi, wc->wc_word, hash); } else { wc = HI2WC(hi); if ((wc->wc_count += count) < (unsigned)count) /* check for overflow */ wc->wc_count = MAXWORDCOUNT; } } /* * Adjust the score of common words. */ static int score_wordcount_adj(slang, score, word, split) slang_T *slang; int score; char_u *word; int split; /* word was split, less bonus */ { hashitem_T *hi; wordcount_T *wc; int bonus; int newscore; hi = hash_find(&slang->sl_wordcount, word); if (!HASHITEM_EMPTY(hi)) { wc = HI2WC(hi); if (wc->wc_count < SCORE_THRES2) bonus = SCORE_COMMON1; else if (wc->wc_count < SCORE_THRES3) bonus = SCORE_COMMON2; else bonus = SCORE_COMMON3; if (split) newscore = score - bonus / 2; else newscore = score - bonus; if (newscore < 0) return 0; return newscore; } return score; } /* * SN_SOFO: * Return SP_*ERROR flags. */ static int read_sofo_section(fd, slang) FILE *fd; slang_T *slang; { int cnt; char_u *from, *to; int res; slang->sl_sofo = TRUE; /* */ from = read_cnt_string(fd, 2, &cnt); if (cnt < 0) return cnt; /* */ to = read_cnt_string(fd, 2, &cnt); if (cnt < 0) { vim_free(from); return cnt; } /* Store the info in slang->sl_sal and/or slang->sl_sal_first. */ if (from != NULL && to != NULL) res = set_sofo(slang, from, to); else if (from != NULL || to != NULL) res = SP_FORMERROR; /* only one of two strings is an error */ else res = 0; vim_free(from); vim_free(to); return res; } /* * Read the compound section from the .spl file: * * Returns SP_*ERROR flags. */ static int read_compound(fd, slang, len) FILE *fd; slang_T *slang; int len; { int todo = len; int c; int atstart; char_u *pat; char_u *pp; char_u *cp; char_u *ap; char_u *crp; int cnt; garray_T *gap; if (todo < 2) return SP_FORMERROR; /* need at least two bytes */ --todo; c = getc(fd); /* */ if (c < 2) c = MAXWLEN; slang->sl_compmax = c; --todo; c = getc(fd); /* */ if (c < 1) c = 0; slang->sl_compminlen = c; --todo; c = getc(fd); /* */ if (c < 1) c = MAXWLEN; slang->sl_compsylmax = c; c = getc(fd); /* */ if (c != 0) ungetc(c, fd); /* be backwards compatible with Vim 7.0b */ else { --todo; c = getc(fd); /* only use the lower byte for now */ --todo; slang->sl_compoptions = c; gap = &slang->sl_comppat; c = get2c(fd); /* */ todo -= 2; ga_init2(gap, sizeof(char_u *), c); if (ga_grow(gap, c) == OK) while (--c >= 0) { ((char_u **)(gap->ga_data))[gap->ga_len++] = read_cnt_string(fd, 1, &cnt); /* */ if (cnt < 0) return cnt; todo -= cnt + 1; } } if (todo < 0) return SP_FORMERROR; /* Turn the COMPOUNDRULE items into a regexp pattern: * "a[bc]/a*b+" -> "^$a[bc]\|a*b\+$$". * Inserting backslashes may double the length, "^$" is 7 bytes. * Conversion to utf-8 may double the size. */ c = todo * 2 + 7; #ifdef FEAT_MBYTE if (enc_utf8) c += todo * 2; #endif pat = alloc((unsigned)c); if (pat == NULL) return SP_OTHERERROR; /* We also need a list of all flags that can appear at the start and one * for all flags. */ cp = alloc(todo + 1); if (cp == NULL) { vim_free(pat); return SP_OTHERERROR; } slang->sl_compstartflags = cp; *cp = NUL; ap = alloc(todo + 1); if (ap == NULL) { vim_free(pat); return SP_OTHERERROR; } slang->sl_compallflags = ap; *ap = NUL; /* And a list of all patterns in their original form, for checking whether * compounding may work in match_compoundrule(). This is freed when we * encounter a wildcard, the check doesn't work then. */ crp = alloc(todo + 1); slang->sl_comprules = crp; pp = pat; *pp++ = '^'; *pp++ = '\\'; *pp++ = '('; atstart = 1; while (todo-- > 0) { c = getc(fd); /* */ if (c == EOF) { vim_free(pat); return SP_TRUNCERROR; } /* Add all flags to "sl_compallflags". */ if (vim_strchr((char_u *)"?*+[]/", c) == NULL && !byte_in_str(slang->sl_compallflags, c)) { *ap++ = c; *ap = NUL; } if (atstart != 0) { /* At start of item: copy flags to "sl_compstartflags". For a * [abc] item set "atstart" to 2 and copy up to the ']'. */ if (c == '[') atstart = 2; else if (c == ']') atstart = 0; else { if (!byte_in_str(slang->sl_compstartflags, c)) { *cp++ = c; *cp = NUL; } if (atstart == 1) atstart = 0; } } /* Copy flag to "sl_comprules", unless we run into a wildcard. */ if (crp != NULL) { if (c == '?' || c == '+' || c == '*') { vim_free(slang->sl_comprules); slang->sl_comprules = NULL; crp = NULL; } else *crp++ = c; } if (c == '/') /* slash separates two items */ { *pp++ = '\\'; *pp++ = '|'; atstart = 1; } else /* normal char, "[abc]" and '*' are copied as-is */ { if (c == '?' || c == '+' || c == '~') *pp++ = '\\'; /* "a?" becomes "a\?", "a+" becomes "a\+" */ #ifdef FEAT_MBYTE if (enc_utf8) pp += mb_char2bytes(c, pp); else #endif *pp++ = c; } } *pp++ = '\\'; *pp++ = ')'; *pp++ = '$'; *pp = NUL; if (crp != NULL) *crp = NUL; slang->sl_compprog = vim_regcomp(pat, RE_MAGIC + RE_STRING + RE_STRICT); vim_free(pat); if (slang->sl_compprog == NULL) return SP_FORMERROR; return 0; } /* * Return TRUE if byte "n" appears in "str". * Like strchr() but independent of locale. */ static int byte_in_str(str, n) char_u *str; int n; { char_u *p; for (p = str; *p != NUL; ++p) if (*p == n) return TRUE; return FALSE; } #define SY_MAXLEN 30 typedef struct syl_item_S { char_u sy_chars[SY_MAXLEN]; /* the sequence of chars */ int sy_len; } syl_item_T; /* * Truncate "slang->sl_syllable" at the first slash and put the following items * in "slang->sl_syl_items". */ static int init_syl_tab(slang) slang_T *slang; { char_u *p; char_u *s; int l; syl_item_T *syl; ga_init2(&slang->sl_syl_items, sizeof(syl_item_T), 4); p = vim_strchr(slang->sl_syllable, '/'); while (p != NULL) { *p++ = NUL; if (*p == NUL) /* trailing slash */ break; s = p; p = vim_strchr(p, '/'); if (p == NULL) l = (int)STRLEN(s); else l = (int)(p - s); if (l >= SY_MAXLEN) return SP_FORMERROR; if (ga_grow(&slang->sl_syl_items, 1) == FAIL) return SP_OTHERERROR; syl = ((syl_item_T *)slang->sl_syl_items.ga_data) + slang->sl_syl_items.ga_len++; vim_strncpy(syl->sy_chars, s, l); syl->sy_len = l; } return OK; } /* * Count the number of syllables in "word". * When "word" contains spaces the syllables after the last space are counted. * Returns zero if syllables are not defines. */ static int count_syllables(slang, word) slang_T *slang; char_u *word; { int cnt = 0; int skip = FALSE; char_u *p; int len; int i; syl_item_T *syl; int c; if (slang->sl_syllable == NULL) return 0; for (p = word; *p != NUL; p += len) { /* When running into a space reset counter. */ if (*p == ' ') { len = 1; cnt = 0; continue; } /* Find longest match of syllable items. */ len = 0; for (i = 0; i < slang->sl_syl_items.ga_len; ++i) { syl = ((syl_item_T *)slang->sl_syl_items.ga_data) + i; if (syl->sy_len > len && STRNCMP(p, syl->sy_chars, syl->sy_len) == 0) len = syl->sy_len; } if (len != 0) /* found a match, count syllable */ { ++cnt; skip = FALSE; } else { /* No recognized syllable item, at least a syllable char then? */ #ifdef FEAT_MBYTE c = mb_ptr2char(p); len = (*mb_ptr2len)(p); #else c = *p; len = 1; #endif if (vim_strchr(slang->sl_syllable, c) == NULL) skip = FALSE; /* No, search for next syllable */ else if (!skip) { ++cnt; /* Yes, count it */ skip = TRUE; /* don't count following syllable chars */ } } } return cnt; } /* * Set the SOFOFROM and SOFOTO items in language "lp". * Returns SP_*ERROR flags when there is something wrong. */ static int set_sofo(lp, from, to) slang_T *lp; char_u *from; char_u *to; { int i; #ifdef FEAT_MBYTE garray_T *gap; char_u *s; char_u *p; int c; int *inp; if (has_mbyte) { /* Use "sl_sal" as an array with 256 pointers to a list of wide * characters. The index is the low byte of the character. * The list contains from-to pairs with a terminating NUL. * sl_sal_first[] is used for latin1 "from" characters. */ gap = &lp->sl_sal; ga_init2(gap, sizeof(int *), 1); if (ga_grow(gap, 256) == FAIL) return SP_OTHERERROR; vim_memset(gap->ga_data, 0, sizeof(int *) * 256); gap->ga_len = 256; /* First count the number of items for each list. Temporarily use * sl_sal_first[] for this. */ for (p = from, s = to; *p != NUL && *s != NUL; ) { c = mb_cptr2char_adv(&p); mb_cptr_adv(s); if (c >= 256) ++lp->sl_sal_first[c & 0xff]; } if (*p != NUL || *s != NUL) /* lengths differ */ return SP_FORMERROR; /* Allocate the lists. */ for (i = 0; i < 256; ++i) if (lp->sl_sal_first[i] > 0) { p = alloc(sizeof(int) * (lp->sl_sal_first[i] * 2 + 1)); if (p == NULL) return SP_OTHERERROR; ((int **)gap->ga_data)[i] = (int *)p; *(int *)p = 0; } /* Put the characters up to 255 in sl_sal_first[] the rest in a sl_sal * list. */ vim_memset(lp->sl_sal_first, 0, sizeof(salfirst_T) * 256); for (p = from, s = to; *p != NUL && *s != NUL; ) { c = mb_cptr2char_adv(&p); i = mb_cptr2char_adv(&s); if (c >= 256) { /* Append the from-to chars at the end of the list with * the low byte. */ inp = ((int **)gap->ga_data)[c & 0xff]; while (*inp != 0) ++inp; *inp++ = c; /* from char */ *inp++ = i; /* to char */ *inp++ = NUL; /* NUL at the end */ } else /* mapping byte to char is done in sl_sal_first[] */ lp->sl_sal_first[c] = i; } } else #endif { /* mapping bytes to bytes is done in sl_sal_first[] */ if (STRLEN(from) != STRLEN(to)) return SP_FORMERROR; for (i = 0; to[i] != NUL; ++i) lp->sl_sal_first[from[i]] = to[i]; lp->sl_sal.ga_len = 1; /* indicates we have soundfolding */ } return 0; } /* * Fill the first-index table for "lp". */ static void set_sal_first(lp) slang_T *lp; { salfirst_T *sfirst; int i; salitem_T *smp; int c; garray_T *gap = &lp->sl_sal; sfirst = lp->sl_sal_first; for (i = 0; i < 256; ++i) sfirst[i] = -1; smp = (salitem_T *)gap->ga_data; for (i = 0; i < gap->ga_len; ++i) { #ifdef FEAT_MBYTE if (has_mbyte) /* Use the lowest byte of the first character. For latin1 it's * the character, for other encodings it should differ for most * characters. */ c = *smp[i].sm_lead_w & 0xff; else #endif c = *smp[i].sm_lead; if (sfirst[c] == -1) { sfirst[c] = i; #ifdef FEAT_MBYTE if (has_mbyte) { int n; /* Make sure all entries with this byte are following each * other. Move the ones that are in the wrong position. Do * keep the same ordering! */ while (i + 1 < gap->ga_len && (*smp[i + 1].sm_lead_w & 0xff) == c) /* Skip over entry with same index byte. */ ++i; for (n = 1; i + n < gap->ga_len; ++n) if ((*smp[i + n].sm_lead_w & 0xff) == c) { salitem_T tsal; /* Move entry with same index byte after the entries * we already found. */ ++i; --n; tsal = smp[i + n]; mch_memmove(smp + i + 1, smp + i, sizeof(salitem_T) * n); smp[i] = tsal; } } #endif } } } #ifdef FEAT_MBYTE /* * Turn a multi-byte string into a wide character string. * Return it in allocated memory (NULL for out-of-memory) */ static int * mb_str2wide(s) char_u *s; { int *res; char_u *p; int i = 0; res = (int *)alloc(sizeof(int) * (mb_charlen(s) + 1)); if (res != NULL) { for (p = s; *p != NUL; ) res[i++] = mb_ptr2char_adv(&p); res[i] = NUL; } return res; } #endif /* * Read a tree from the .spl or .sug file. * Allocates the memory and stores pointers in "bytsp" and "idxsp". * This is skipped when the tree has zero length. * Returns zero when OK, SP_ value for an error. */ static int spell_read_tree(fd, bytsp, idxsp, prefixtree, prefixcnt) FILE *fd; char_u **bytsp; idx_T **idxsp; int prefixtree; /* TRUE for the prefix tree */ int prefixcnt; /* when "prefixtree" is TRUE: prefix count */ { int len; int idx; char_u *bp; idx_T *ip; /* The tree size was computed when writing the file, so that we can * allocate it as one long block. */ len = get4c(fd); if (len < 0) return SP_TRUNCERROR; if (len > 0) { /* Allocate the byte array. */ bp = lalloc((long_u)len, TRUE); if (bp == NULL) return SP_OTHERERROR; *bytsp = bp; /* Allocate the index array. */ ip = (idx_T *)lalloc_clear((long_u)(len * sizeof(int)), TRUE); if (ip == NULL) return SP_OTHERERROR; *idxsp = ip; /* Recursively read the tree and store it in the array. */ idx = read_tree_node(fd, bp, ip, len, 0, prefixtree, prefixcnt); if (idx < 0) return idx; } return 0; } /* * Read one row of siblings from the spell file and store it in the byte array * "byts" and index array "idxs". Recursively read the children. * * NOTE: The code here must match put_node()! * * Returns the index (>= 0) following the siblings. * Returns SP_TRUNCERROR if the file is shorter than expected. * Returns SP_FORMERROR if there is a format error. */ static idx_T read_tree_node(fd, byts, idxs, maxidx, startidx, prefixtree, maxprefcondnr) FILE *fd; char_u *byts; idx_T *idxs; int maxidx; /* size of arrays */ idx_T startidx; /* current index in "byts" and "idxs" */ int prefixtree; /* TRUE for reading PREFIXTREE */ int maxprefcondnr; /* maximum for */ { int len; int i; int n; idx_T idx = startidx; int c; int c2; #define SHARED_MASK 0x8000000 len = getc(fd); /* */ if (len <= 0) return SP_TRUNCERROR; if (startidx + len >= maxidx) return SP_FORMERROR; byts[idx++] = len; /* Read the byte values, flag/region bytes and shared indexes. */ for (i = 1; i <= len; ++i) { c = getc(fd); /* */ if (c < 0) return SP_TRUNCERROR; if (c <= BY_SPECIAL) { if (c == BY_NOFLAGS && !prefixtree) { /* No flags, all regions. */ idxs[idx] = 0; c = 0; } else if (c != BY_INDEX) { if (prefixtree) { /* Read the optional pflags byte, the prefix ID and the * condition nr. In idxs[] store the prefix ID in the low * byte, the condition index shifted up 8 bits, the flags * shifted up 24 bits. */ if (c == BY_FLAGS) c = getc(fd) << 24; /* */ else c = 0; c |= getc(fd); /* */ n = get2c(fd); /* */ if (n >= maxprefcondnr) return SP_FORMERROR; c |= (n << 8); } else /* c must be BY_FLAGS or BY_FLAGS2 */ { /* Read flags and optional region and prefix ID. In * idxs[] the flags go in the low two bytes, region above * that and prefix ID above the region. */ c2 = c; c = getc(fd); /* */ if (c2 == BY_FLAGS2) c = (getc(fd) << 8) + c; /* */ if (c & WF_REGION) c = (getc(fd) << 16) + c; /* */ if (c & WF_AFX) c = (getc(fd) << 24) + c; /* */ } idxs[idx] = c; c = 0; } else /* c == BY_INDEX */ { /* */ n = get3c(fd); if (n < 0 || n >= maxidx) return SP_FORMERROR; idxs[idx] = n + SHARED_MASK; c = getc(fd); /* */ } } byts[idx++] = c; } /* Recursively read the children for non-shared siblings. * Skip the end-of-word ones (zero byte value) and the shared ones (and * remove SHARED_MASK) */ for (i = 1; i <= len; ++i) if (byts[startidx + i] != 0) { if (idxs[startidx + i] & SHARED_MASK) idxs[startidx + i] &= ~SHARED_MASK; else { idxs[startidx + i] = idx; idx = read_tree_node(fd, byts, idxs, maxidx, idx, prefixtree, maxprefcondnr); if (idx < 0) break; } } return idx; } /* * Parse 'spelllang' and set w_s->b_langp accordingly. * Returns NULL if it's OK, an error message otherwise. */ char_u * did_set_spelllang(wp) win_T *wp; { garray_T ga; char_u *splp; char_u *region; char_u region_cp[3]; int filename; int region_mask; slang_T *slang; int c; char_u lang[MAXWLEN + 1]; char_u spf_name[MAXPATHL]; int len; char_u *p; int round; char_u *spf; char_u *use_region = NULL; int dont_use_region = FALSE; int nobreak = FALSE; int i, j; langp_T *lp, *lp2; static int recursive = FALSE; char_u *ret_msg = NULL; char_u *spl_copy; /* We don't want to do this recursively. May happen when a language is * not available and the SpellFileMissing autocommand opens a new buffer * in which 'spell' is set. */ if (recursive) return NULL; recursive = TRUE; ga_init2(&ga, sizeof(langp_T), 2); clear_midword(wp); /* Make a copy of 'spellang', the SpellFileMissing autocommands may change * it under our fingers. */ spl_copy = vim_strsave(wp->w_s->b_p_spl); if (spl_copy == NULL) goto theend; /* loop over comma separated language names. */ for (splp = spl_copy; *splp != NUL; ) { /* Get one language name. */ copy_option_part(&splp, lang, MAXWLEN, ","); region = NULL; len = (int)STRLEN(lang); /* If the name ends in ".spl" use it as the name of the spell file. * If there is a region name let "region" point to it and remove it * from the name. */ if (len > 4 && fnamecmp(lang + len - 4, ".spl") == 0) { filename = TRUE; /* Locate a region and remove it from the file name. */ p = vim_strchr(gettail(lang), '_'); if (p != NULL && ASCII_ISALPHA(p[1]) && ASCII_ISALPHA(p[2]) && !ASCII_ISALPHA(p[3])) { vim_strncpy(region_cp, p + 1, 2); mch_memmove(p, p + 3, len - (p - lang) - 2); len -= 3; region = region_cp; } else dont_use_region = TRUE; /* Check if we loaded this language before. */ for (slang = first_lang; slang != NULL; slang = slang->sl_next) if (fullpathcmp(lang, slang->sl_fname, FALSE) == FPC_SAME) break; } else { filename = FALSE; if (len > 3 && lang[len - 3] == '_') { region = lang + len - 2; len -= 3; lang[len] = NUL; } else dont_use_region = TRUE; /* Check if we loaded this language before. */ for (slang = first_lang; slang != NULL; slang = slang->sl_next) if (STRICMP(lang, slang->sl_name) == 0) break; } if (region != NULL) { /* If the region differs from what was used before then don't * use it for 'spellfile'. */ if (use_region != NULL && STRCMP(region, use_region) != 0) dont_use_region = TRUE; use_region = region; } /* If not found try loading the language now. */ if (slang == NULL) { if (filename) (void)spell_load_file(lang, lang, NULL, FALSE); else { spell_load_lang(lang); #ifdef FEAT_AUTOCMD /* SpellFileMissing autocommands may do anything, including * destroying the buffer we are using... */ if (!buf_valid(wp->w_buffer)) { ret_msg = (char_u *)"E797: SpellFileMissing autocommand deleted buffer"; goto theend; } #endif } } /* * Loop over the languages, there can be several files for "lang". */ for (slang = first_lang; slang != NULL; slang = slang->sl_next) if (filename ? fullpathcmp(lang, slang->sl_fname, FALSE) == FPC_SAME : STRICMP(lang, slang->sl_name) == 0) { region_mask = REGION_ALL; if (!filename && region != NULL) { /* find region in sl_regions */ c = find_region(slang->sl_regions, region); if (c == REGION_ALL) { if (slang->sl_add) { if (*slang->sl_regions != NUL) /* This addition file is for other regions. */ region_mask = 0; } else /* This is probably an error. Give a warning and * accept the words anyway. */ smsg((char_u *) _("Warning: region %s not supported"), region); } else region_mask = 1 << c; } if (region_mask != 0) { if (ga_grow(&ga, 1) == FAIL) { ga_clear(&ga); ret_msg = e_outofmem; goto theend; } LANGP_ENTRY(ga, ga.ga_len)->lp_slang = slang; LANGP_ENTRY(ga, ga.ga_len)->lp_region = region_mask; ++ga.ga_len; use_midword(slang, wp); if (slang->sl_nobreak) nobreak = TRUE; } } } /* round 0: load int_wordlist, if possible. * round 1: load first name in 'spellfile'. * round 2: load second name in 'spellfile. * etc. */ spf = curwin->w_s->b_p_spf; for (round = 0; round == 0 || *spf != NUL; ++round) { if (round == 0) { /* Internal wordlist, if there is one. */ if (int_wordlist == NULL) continue; int_wordlist_spl(spf_name); } else { /* One entry in 'spellfile'. */ copy_option_part(&spf, spf_name, MAXPATHL - 5, ","); STRCAT(spf_name, ".spl"); /* If it was already found above then skip it. */ for (c = 0; c < ga.ga_len; ++c) { p = LANGP_ENTRY(ga, c)->lp_slang->sl_fname; if (p != NULL && fullpathcmp(spf_name, p, FALSE) == FPC_SAME) break; } if (c < ga.ga_len) continue; } /* Check if it was loaded already. */ for (slang = first_lang; slang != NULL; slang = slang->sl_next) if (fullpathcmp(spf_name, slang->sl_fname, FALSE) == FPC_SAME) break; if (slang == NULL) { /* Not loaded, try loading it now. The language name includes the * region name, the region is ignored otherwise. for int_wordlist * use an arbitrary name. */ if (round == 0) STRCPY(lang, "internal wordlist"); else { vim_strncpy(lang, gettail(spf_name), MAXWLEN); p = vim_strchr(lang, '.'); if (p != NULL) *p = NUL; /* truncate at ".encoding.add" */ } slang = spell_load_file(spf_name, lang, NULL, TRUE); /* If one of the languages has NOBREAK we assume the addition * files also have this. */ if (slang != NULL && nobreak) slang->sl_nobreak = TRUE; } if (slang != NULL && ga_grow(&ga, 1) == OK) { region_mask = REGION_ALL; if (use_region != NULL && !dont_use_region) { /* find region in sl_regions */ c = find_region(slang->sl_regions, use_region); if (c != REGION_ALL) region_mask = 1 << c; else if (*slang->sl_regions != NUL) /* This spell file is for other regions. */ region_mask = 0; } if (region_mask != 0) { LANGP_ENTRY(ga, ga.ga_len)->lp_slang = slang; LANGP_ENTRY(ga, ga.ga_len)->lp_sallang = NULL; LANGP_ENTRY(ga, ga.ga_len)->lp_replang = NULL; LANGP_ENTRY(ga, ga.ga_len)->lp_region = region_mask; ++ga.ga_len; use_midword(slang, wp); } } } /* Everything is fine, store the new b_langp value. */ ga_clear(&wp->w_s->b_langp); wp->w_s->b_langp = ga; /* For each language figure out what language to use for sound folding and * REP items. If the language doesn't support it itself use another one * with the same name. E.g. for "en-math" use "en". */ for (i = 0; i < ga.ga_len; ++i) { lp = LANGP_ENTRY(ga, i); /* sound folding */ if (lp->lp_slang->sl_sal.ga_len > 0) /* language does sound folding itself */ lp->lp_sallang = lp->lp_slang; else /* find first similar language that does sound folding */ for (j = 0; j < ga.ga_len; ++j) { lp2 = LANGP_ENTRY(ga, j); if (lp2->lp_slang->sl_sal.ga_len > 0 && STRNCMP(lp->lp_slang->sl_name, lp2->lp_slang->sl_name, 2) == 0) { lp->lp_sallang = lp2->lp_slang; break; } } /* REP items */ if (lp->lp_slang->sl_rep.ga_len > 0) /* language has REP items itself */ lp->lp_replang = lp->lp_slang; else /* find first similar language that has REP items */ for (j = 0; j < ga.ga_len; ++j) { lp2 = LANGP_ENTRY(ga, j); if (lp2->lp_slang->sl_rep.ga_len > 0 && STRNCMP(lp->lp_slang->sl_name, lp2->lp_slang->sl_name, 2) == 0) { lp->lp_replang = lp2->lp_slang; break; } } } theend: vim_free(spl_copy); recursive = FALSE; return ret_msg; } /* * Clear the midword characters for buffer "buf". */ static void clear_midword(wp) win_T *wp; { vim_memset(wp->w_s->b_spell_ismw, 0, 256); #ifdef FEAT_MBYTE vim_free(wp->w_s->b_spell_ismw_mb); wp->w_s->b_spell_ismw_mb = NULL; #endif } /* * Use the "sl_midword" field of language "lp" for buffer "buf". * They add up to any currently used midword characters. */ static void use_midword(lp, wp) slang_T *lp; win_T *wp; { char_u *p; if (lp->sl_midword == NULL) /* there aren't any */ return; for (p = lp->sl_midword; *p != NUL; ) #ifdef FEAT_MBYTE if (has_mbyte) { int c, l, n; char_u *bp; c = mb_ptr2char(p); l = (*mb_ptr2len)(p); if (c < 256 && l <= 2) wp->w_s->b_spell_ismw[c] = TRUE; else if (wp->w_s->b_spell_ismw_mb == NULL) /* First multi-byte char in "b_spell_ismw_mb". */ wp->w_s->b_spell_ismw_mb = vim_strnsave(p, l); else { /* Append multi-byte chars to "b_spell_ismw_mb". */ n = (int)STRLEN(wp->w_s->b_spell_ismw_mb); bp = vim_strnsave(wp->w_s->b_spell_ismw_mb, n + l); if (bp != NULL) { vim_free(wp->w_s->b_spell_ismw_mb); wp->w_s->b_spell_ismw_mb = bp; vim_strncpy(bp + n, p, l); } } p += l; } else #endif wp->w_s->b_spell_ismw[*p++] = TRUE; } /* * Find the region "region[2]" in "rp" (points to "sl_regions"). * Each region is simply stored as the two characters of it's name. * Returns the index if found (first is 0), REGION_ALL if not found. */ static int find_region(rp, region) char_u *rp; char_u *region; { int i; for (i = 0; ; i += 2) { if (rp[i] == NUL) return REGION_ALL; if (rp[i] == region[0] && rp[i + 1] == region[1]) break; } return i / 2; } /* * Return case type of word: * w word 0 * Word WF_ONECAP * W WORD WF_ALLCAP * WoRd wOrd WF_KEEPCAP */ static int captype(word, end) char_u *word; char_u *end; /* When NULL use up to NUL byte. */ { char_u *p; int c; int firstcap; int allcap; int past_second = FALSE; /* past second word char */ /* find first letter */ for (p = word; !spell_iswordp_nmw(p); mb_ptr_adv(p)) if (end == NULL ? *p == NUL : p >= end) return 0; /* only non-word characters, illegal word */ #ifdef FEAT_MBYTE if (has_mbyte) c = mb_ptr2char_adv(&p); else #endif c = *p++; firstcap = allcap = SPELL_ISUPPER(c); /* * Need to check all letters to find a word with mixed upper/lower. * But a word with an upper char only at start is a ONECAP. */ for ( ; end == NULL ? *p != NUL : p < end; mb_ptr_adv(p)) if (spell_iswordp_nmw(p)) { c = PTR2CHAR(p); if (!SPELL_ISUPPER(c)) { /* UUl -> KEEPCAP */ if (past_second && allcap) return WF_KEEPCAP; allcap = FALSE; } else if (!allcap) /* UlU -> KEEPCAP */ return WF_KEEPCAP; past_second = TRUE; } if (allcap) return WF_ALLCAP; if (firstcap) return WF_ONECAP; return 0; } /* * Like captype() but for a KEEPCAP word add ONECAP if the word starts with a * capital. So that make_case_word() can turn WOrd into Word. * Add ALLCAP for "WOrD". */ static int badword_captype(word, end) char_u *word; char_u *end; { int flags = captype(word, end); int c; int l, u; int first; char_u *p; if (flags & WF_KEEPCAP) { /* Count the number of UPPER and lower case letters. */ l = u = 0; first = FALSE; for (p = word; p < end; mb_ptr_adv(p)) { c = PTR2CHAR(p); if (SPELL_ISUPPER(c)) { ++u; if (p == word) first = TRUE; } else ++l; } /* If there are more UPPER than lower case letters suggest an * ALLCAP word. Otherwise, if the first letter is UPPER then * suggest ONECAP. Exception: "ALl" most likely should be "All", * require three upper case letters. */ if (u > l && u > 2) flags |= WF_ALLCAP; else if (first) flags |= WF_ONECAP; if (u >= 2 && l >= 2) /* maCARONI maCAroni */ flags |= WF_MIXCAP; } return flags; } # if defined(FEAT_MBYTE) || defined(EXITFREE) || defined(PROTO) /* * Free all languages. */ void spell_free_all() { slang_T *slang; buf_T *buf; char_u fname[MAXPATHL]; /* Go through all buffers and handle 'spelllang'. */ for (buf = firstbuf; buf != NULL; buf = buf->b_next) ga_clear(&buf->b_s.b_langp); while (first_lang != NULL) { slang = first_lang; first_lang = slang->sl_next; slang_free(slang); } if (int_wordlist != NULL) { /* Delete the internal wordlist and its .spl file */ mch_remove(int_wordlist); int_wordlist_spl(fname); mch_remove(fname); vim_free(int_wordlist); int_wordlist = NULL; } init_spell_chartab(); vim_free(repl_to); repl_to = NULL; vim_free(repl_from); repl_from = NULL; } # endif # if defined(FEAT_MBYTE) || defined(PROTO) /* * Clear all spelling tables and reload them. * Used after 'encoding' is set and when ":mkspell" was used. */ void spell_reload() { win_T *wp; /* Initialize the table for spell_iswordp(). */ init_spell_chartab(); /* Unload all allocated memory. */ spell_free_all(); /* Go through all buffers and handle 'spelllang'. */ for (wp = firstwin; wp != NULL; wp = wp->w_next) { /* Only load the wordlists when 'spelllang' is set and there is a * window for this buffer in which 'spell' is set. */ if (*wp->w_s->b_p_spl != NUL) { if (wp->w_p_spell) { (void)did_set_spelllang(wp); # ifdef FEAT_WINDOWS break; # endif } } } } # endif /* * Reload the spell file "fname" if it's loaded. */ static void spell_reload_one(fname, added_word) char_u *fname; int added_word; /* invoked through "zg" */ { slang_T *slang; int didit = FALSE; for (slang = first_lang; slang != NULL; slang = slang->sl_next) { if (fullpathcmp(fname, slang->sl_fname, FALSE) == FPC_SAME) { slang_clear(slang); if (spell_load_file(fname, NULL, slang, FALSE) == NULL) /* reloading failed, clear the language */ slang_clear(slang); redraw_all_later(SOME_VALID); didit = TRUE; } } /* When "zg" was used and the file wasn't loaded yet, should redo * 'spelllang' to load it now. */ if (added_word && !didit) did_set_spelllang(curwin); } /* * Functions for ":mkspell". */ #define MAXLINELEN 500 /* Maximum length in bytes of a line in a .aff and .dic file. */ /* * Main structure to store the contents of a ".aff" file. */ typedef struct afffile_S { char_u *af_enc; /* "SET", normalized, alloc'ed string or NULL */ int af_flagtype; /* AFT_CHAR, AFT_LONG, AFT_NUM or AFT_CAPLONG */ unsigned af_rare; /* RARE ID for rare word */ unsigned af_keepcase; /* KEEPCASE ID for keep-case word */ unsigned af_bad; /* BAD ID for banned word */ unsigned af_needaffix; /* NEEDAFFIX ID */ unsigned af_circumfix; /* CIRCUMFIX ID */ unsigned af_needcomp; /* NEEDCOMPOUND ID */ unsigned af_comproot; /* COMPOUNDROOT ID */ unsigned af_compforbid; /* COMPOUNDFORBIDFLAG ID */ unsigned af_comppermit; /* COMPOUNDPERMITFLAG ID */ unsigned af_nosuggest; /* NOSUGGEST ID */ int af_pfxpostpone; /* postpone prefixes without chop string and without flags */ hashtab_T af_pref; /* hashtable for prefixes, affheader_T */ hashtab_T af_suff; /* hashtable for suffixes, affheader_T */ hashtab_T af_comp; /* hashtable for compound flags, compitem_T */ } afffile_T; #define AFT_CHAR 0 /* flags are one character */ #define AFT_LONG 1 /* flags are two characters */ #define AFT_CAPLONG 2 /* flags are one or two characters */ #define AFT_NUM 3 /* flags are numbers, comma separated */ typedef struct affentry_S affentry_T; /* Affix entry from ".aff" file. Used for prefixes and suffixes. */ struct affentry_S { affentry_T *ae_next; /* next affix with same name/number */ char_u *ae_chop; /* text to chop off basic word (can be NULL) */ char_u *ae_add; /* text to add to basic word (can be NULL) */ char_u *ae_flags; /* flags on the affix (can be NULL) */ char_u *ae_cond; /* condition (NULL for ".") */ regprog_T *ae_prog; /* regexp program for ae_cond or NULL */ char ae_compforbid; /* COMPOUNDFORBIDFLAG found */ char ae_comppermit; /* COMPOUNDPERMITFLAG found */ }; #ifdef FEAT_MBYTE # define AH_KEY_LEN 17 /* 2 x 8 bytes + NUL */ #else # define AH_KEY_LEN 7 /* 6 digits + NUL */ #endif /* Affix header from ".aff" file. Used for af_pref and af_suff. */ typedef struct affheader_S { char_u ah_key[AH_KEY_LEN]; /* key for hashtab == name of affix */ unsigned ah_flag; /* affix name as number, uses "af_flagtype" */ int ah_newID; /* prefix ID after renumbering; 0 if not used */ int ah_combine; /* suffix may combine with prefix */ int ah_follows; /* another affix block should be following */ affentry_T *ah_first; /* first affix entry */ } affheader_T; #define HI2AH(hi) ((affheader_T *)(hi)->hi_key) /* Flag used in compound items. */ typedef struct compitem_S { char_u ci_key[AH_KEY_LEN]; /* key for hashtab == name of compound */ unsigned ci_flag; /* affix name as number, uses "af_flagtype" */ int ci_newID; /* affix ID after renumbering. */ } compitem_T; #define HI2CI(hi) ((compitem_T *)(hi)->hi_key) /* * Structure that is used to store the items in the word tree. This avoids * the need to keep track of each allocated thing, everything is freed all at * once after ":mkspell" is done. * Note: "sb_next" must be just before "sb_data" to make sure the alignment of * "sb_data" is correct for systems where pointers must be aligned on * pointer-size boundaries and sizeof(pointer) > sizeof(int) (e.g., Sparc). */ #define SBLOCKSIZE 16000 /* size of sb_data */ typedef struct sblock_S sblock_T; struct sblock_S { int sb_used; /* nr of bytes already in use */ sblock_T *sb_next; /* next block in list */ char_u sb_data[1]; /* data, actually longer */ }; /* * A node in the tree. */ typedef struct wordnode_S wordnode_T; struct wordnode_S { union /* shared to save space */ { char_u hashkey[6]; /* the hash key, only used while compressing */ int index; /* index in written nodes (valid after first round) */ } wn_u1; union /* shared to save space */ { wordnode_T *next; /* next node with same hash key */ wordnode_T *wnode; /* parent node that will write this node */ } wn_u2; wordnode_T *wn_child; /* child (next byte in word) */ wordnode_T *wn_sibling; /* next sibling (alternate byte in word, always sorted) */ int wn_refs; /* Nr. of references to this node. Only relevant for first node in a list of siblings, in following siblings it is always one. */ char_u wn_byte; /* Byte for this node. NUL for word end */ /* Info for when "wn_byte" is NUL. * In PREFIXTREE "wn_region" is used for the prefcondnr. * In the soundfolded word tree "wn_flags" has the MSW of the wordnr and * "wn_region" the LSW of the wordnr. */ char_u wn_affixID; /* supported/required prefix ID or 0 */ short_u wn_flags; /* WF_ flags */ short wn_region; /* region mask */ #ifdef SPELL_PRINTTREE int wn_nr; /* sequence nr for printing */ #endif }; #define WN_MASK 0xffff /* mask relevant bits of "wn_flags" */ #define HI2WN(hi) (wordnode_T *)((hi)->hi_key) /* * Info used while reading the spell files. */ typedef struct spellinfo_S { wordnode_T *si_foldroot; /* tree with case-folded words */ long si_foldwcount; /* nr of words in si_foldroot */ wordnode_T *si_keeproot; /* tree with keep-case words */ long si_keepwcount; /* nr of words in si_keeproot */ wordnode_T *si_prefroot; /* tree with postponed prefixes */ long si_sugtree; /* creating the soundfolding trie */ sblock_T *si_blocks; /* memory blocks used */ long si_blocks_cnt; /* memory blocks allocated */ int si_did_emsg; /* TRUE when ran out of memory */ long si_compress_cnt; /* words to add before lowering compression limit */ wordnode_T *si_first_free; /* List of nodes that have been freed during compression, linked by "wn_child" field. */ long si_free_count; /* number of nodes in si_first_free */ #ifdef SPELL_PRINTTREE int si_wordnode_nr; /* sequence nr for nodes */ #endif buf_T *si_spellbuf; /* buffer used to store soundfold word table */ int si_ascii; /* handling only ASCII words */ int si_add; /* addition file */ int si_clear_chartab; /* when TRUE clear char tables */ int si_region; /* region mask */ vimconv_T si_conv; /* for conversion to 'encoding' */ int si_memtot; /* runtime memory used */ int si_verbose; /* verbose messages */ int si_msg_count; /* number of words added since last message */ char_u *si_info; /* info text chars or NULL */ int si_region_count; /* number of regions supported (1 when there are no regions) */ char_u si_region_name[17]; /* region names; used only if * si_region_count > 1) */ garray_T si_rep; /* list of fromto_T entries from REP lines */ garray_T si_repsal; /* list of fromto_T entries from REPSAL lines */ garray_T si_sal; /* list of fromto_T entries from SAL lines */ char_u *si_sofofr; /* SOFOFROM text */ char_u *si_sofoto; /* SOFOTO text */ int si_nosugfile; /* NOSUGFILE item found */ int si_nosplitsugs; /* NOSPLITSUGS item found */ int si_followup; /* soundsalike: ? */ int si_collapse; /* soundsalike: ? */ hashtab_T si_commonwords; /* hashtable for common words */ time_t si_sugtime; /* timestamp for .sug file */ int si_rem_accents; /* soundsalike: remove accents */ garray_T si_map; /* MAP info concatenated */ char_u *si_midword; /* MIDWORD chars or NULL */ int si_compmax; /* max nr of words for compounding */ int si_compminlen; /* minimal length for compounding */ int si_compsylmax; /* max nr of syllables for compounding */ int si_compoptions; /* COMP_ flags */ garray_T si_comppat; /* CHECKCOMPOUNDPATTERN items, each stored as a string */ char_u *si_compflags; /* flags used for compounding */ char_u si_nobreak; /* NOBREAK */ char_u *si_syllable; /* syllable string */ garray_T si_prefcond; /* table with conditions for postponed * prefixes, each stored as a string */ int si_newprefID; /* current value for ah_newID */ int si_newcompID; /* current value for compound ID */ } spellinfo_T; static afffile_T *spell_read_aff __ARGS((spellinfo_T *spin, char_u *fname)); static int is_aff_rule __ARGS((char_u **items, int itemcnt, char *rulename, int mincount)); static void aff_process_flags __ARGS((afffile_T *affile, affentry_T *entry)); static int spell_info_item __ARGS((char_u *s)); static unsigned affitem2flag __ARGS((int flagtype, char_u *item, char_u *fname, int lnum)); static unsigned get_affitem __ARGS((int flagtype, char_u **pp)); static void process_compflags __ARGS((spellinfo_T *spin, afffile_T *aff, char_u *compflags)); static void check_renumber __ARGS((spellinfo_T *spin)); static int flag_in_afflist __ARGS((int flagtype, char_u *afflist, unsigned flag)); static void aff_check_number __ARGS((int spinval, int affval, char *name)); static void aff_check_string __ARGS((char_u *spinval, char_u *affval, char *name)); static int str_equal __ARGS((char_u *s1, char_u *s2)); static void add_fromto __ARGS((spellinfo_T *spin, garray_T *gap, char_u *from, char_u *to)); static int sal_to_bool __ARGS((char_u *s)); static int has_non_ascii __ARGS((char_u *s)); static void spell_free_aff __ARGS((afffile_T *aff)); static int spell_read_dic __ARGS((spellinfo_T *spin, char_u *fname, afffile_T *affile)); static int get_affix_flags __ARGS((afffile_T *affile, char_u *afflist)); static int get_pfxlist __ARGS((afffile_T *affile, char_u *afflist, char_u *store_afflist)); static void get_compflags __ARGS((afffile_T *affile, char_u *afflist, char_u *store_afflist)); static int store_aff_word __ARGS((spellinfo_T *spin, char_u *word, char_u *afflist, afffile_T *affile, hashtab_T *ht, hashtab_T *xht, int condit, int flags, char_u *pfxlist, int pfxlen)); static int spell_read_wordfile __ARGS((spellinfo_T *spin, char_u *fname)); static void *getroom __ARGS((spellinfo_T *spin, size_t len, int align)); static char_u *getroom_save __ARGS((spellinfo_T *spin, char_u *s)); static void free_blocks __ARGS((sblock_T *bl)); static wordnode_T *wordtree_alloc __ARGS((spellinfo_T *spin)); static int store_word __ARGS((spellinfo_T *spin, char_u *word, int flags, int region, char_u *pfxlist, int need_affix)); static int tree_add_word __ARGS((spellinfo_T *spin, char_u *word, wordnode_T *tree, int flags, int region, int affixID)); static wordnode_T *get_wordnode __ARGS((spellinfo_T *spin)); static int deref_wordnode __ARGS((spellinfo_T *spin, wordnode_T *node)); static void free_wordnode __ARGS((spellinfo_T *spin, wordnode_T *n)); static void wordtree_compress __ARGS((spellinfo_T *spin, wordnode_T *root)); static int node_compress __ARGS((spellinfo_T *spin, wordnode_T *node, hashtab_T *ht, int *tot)); static int node_equal __ARGS((wordnode_T *n1, wordnode_T *n2)); static int write_vim_spell __ARGS((spellinfo_T *spin, char_u *fname)); static void clear_node __ARGS((wordnode_T *node)); static int put_node __ARGS((FILE *fd, wordnode_T *node, int idx, int regionmask, int prefixtree)); static void spell_make_sugfile __ARGS((spellinfo_T *spin, char_u *wfname)); static int sug_filltree __ARGS((spellinfo_T *spin, slang_T *slang)); static int sug_maketable __ARGS((spellinfo_T *spin)); static int sug_filltable __ARGS((spellinfo_T *spin, wordnode_T *node, int startwordnr, garray_T *gap)); static int offset2bytes __ARGS((int nr, char_u *buf)); static int bytes2offset __ARGS((char_u **pp)); static void sug_write __ARGS((spellinfo_T *spin, char_u *fname)); static void mkspell __ARGS((int fcount, char_u **fnames, int ascii, int overwrite, int added_word)); static void spell_message __ARGS((spellinfo_T *spin, char_u *str)); static void init_spellfile __ARGS((void)); /* In the postponed prefixes tree wn_flags is used to store the WFP_ flags, * but it must be negative to indicate the prefix tree to tree_add_word(). * Use a negative number with the lower 8 bits zero. */ #define PFX_FLAGS -256 /* flags for "condit" argument of store_aff_word() */ #define CONDIT_COMB 1 /* affix must combine */ #define CONDIT_CFIX 2 /* affix must have CIRCUMFIX flag */ #define CONDIT_SUF 4 /* add a suffix for matching flags */ #define CONDIT_AFF 8 /* word already has an affix */ /* * Tunable parameters for when the tree is compressed. See 'mkspellmem'. */ static long compress_start = 30000; /* memory / SBLOCKSIZE */ static long compress_inc = 100; /* memory / SBLOCKSIZE */ static long compress_added = 500000; /* word count */ #ifdef SPELL_PRINTTREE /* * For debugging the tree code: print the current tree in a (more or less) * readable format, so that we can see what happens when adding a word and/or * compressing the tree. * Based on code from Olaf Seibert. */ #define PRINTLINESIZE 1000 #define PRINTWIDTH 6 #define PRINTSOME(l, depth, fmt, a1, a2) vim_snprintf(l + depth * PRINTWIDTH, \ PRINTLINESIZE - PRINTWIDTH * depth, fmt, a1, a2) static char line1[PRINTLINESIZE]; static char line2[PRINTLINESIZE]; static char line3[PRINTLINESIZE]; static void spell_clear_flags(wordnode_T *node) { wordnode_T *np; for (np = node; np != NULL; np = np->wn_sibling) { np->wn_u1.index = FALSE; spell_clear_flags(np->wn_child); } } static void spell_print_node(wordnode_T *node, int depth) { if (node->wn_u1.index) { /* Done this node before, print the reference. */ PRINTSOME(line1, depth, "(%d)", node->wn_nr, 0); PRINTSOME(line2, depth, " ", 0, 0); PRINTSOME(line3, depth, " ", 0, 0); msg(line1); msg(line2); msg(line3); } else { node->wn_u1.index = TRUE; if (node->wn_byte != NUL) { if (node->wn_child != NULL) PRINTSOME(line1, depth, " %c -> ", node->wn_byte, 0); else /* Cannot happen? */ PRINTSOME(line1, depth, " %c ???", node->wn_byte, 0); } else PRINTSOME(line1, depth, " $ ", 0, 0); PRINTSOME(line2, depth, "%d/%d ", node->wn_nr, node->wn_refs); if (node->wn_sibling != NULL) PRINTSOME(line3, depth, " | ", 0, 0); else PRINTSOME(line3, depth, " ", 0, 0); if (node->wn_byte == NUL) { msg(line1); msg(line2); msg(line3); } /* do the children */ if (node->wn_byte != NUL && node->wn_child != NULL) spell_print_node(node->wn_child, depth + 1); /* do the siblings */ if (node->wn_sibling != NULL) { /* get rid of all parent details except | */ STRCPY(line1, line3); STRCPY(line2, line3); spell_print_node(node->wn_sibling, depth); } } } static void spell_print_tree(wordnode_T *root) { if (root != NULL) { /* Clear the "wn_u1.index" fields, used to remember what has been * done. */ spell_clear_flags(root); /* Recursively print the tree. */ spell_print_node(root, 0); } } #endif /* SPELL_PRINTTREE */ /* * Read the affix file "fname". * Returns an afffile_T, NULL for complete failure. */ static afffile_T * spell_read_aff(spin, fname) spellinfo_T *spin; char_u *fname; { FILE *fd; afffile_T *aff; char_u rline[MAXLINELEN]; char_u *line; char_u *pc = NULL; #define MAXITEMCNT 30 char_u *(items[MAXITEMCNT]); int itemcnt; char_u *p; int lnum = 0; affheader_T *cur_aff = NULL; int did_postpone_prefix = FALSE; int aff_todo = 0; hashtab_T *tp; char_u *low = NULL; char_u *fol = NULL; char_u *upp = NULL; int do_rep; int do_repsal; int do_sal; int do_mapline; int found_map = FALSE; hashitem_T *hi; int l; int compminlen = 0; /* COMPOUNDMIN value */ int compsylmax = 0; /* COMPOUNDSYLMAX value */ int compoptions = 0; /* COMP_ flags */ int compmax = 0; /* COMPOUNDWORDMAX value */ char_u *compflags = NULL; /* COMPOUNDFLAG and COMPOUNDRULE concatenated */ char_u *midword = NULL; /* MIDWORD value */ char_u *syllable = NULL; /* SYLLABLE value */ char_u *sofofrom = NULL; /* SOFOFROM value */ char_u *sofoto = NULL; /* SOFOTO value */ /* * Open the file. */ fd = mch_fopen((char *)fname, "r"); if (fd == NULL) { EMSG2(_(e_notopen), fname); return NULL; } vim_snprintf((char *)IObuff, IOSIZE, _("Reading affix file %s ..."), fname); spell_message(spin, IObuff); /* Only do REP lines when not done in another .aff file already. */ do_rep = spin->si_rep.ga_len == 0; /* Only do REPSAL lines when not done in another .aff file already. */ do_repsal = spin->si_repsal.ga_len == 0; /* Only do SAL lines when not done in another .aff file already. */ do_sal = spin->si_sal.ga_len == 0; /* Only do MAP lines when not done in another .aff file already. */ do_mapline = spin->si_map.ga_len == 0; /* * Allocate and init the afffile_T structure. */ aff = (afffile_T *)getroom(spin, sizeof(afffile_T), TRUE); if (aff == NULL) { fclose(fd); return NULL; } hash_init(&aff->af_pref); hash_init(&aff->af_suff); hash_init(&aff->af_comp); /* * Read all the lines in the file one by one. */ while (!vim_fgets(rline, MAXLINELEN, fd) && !got_int) { line_breakcheck(); ++lnum; /* Skip comment lines. */ if (*rline == '#') continue; /* Convert from "SET" to 'encoding' when needed. */ vim_free(pc); #ifdef FEAT_MBYTE if (spin->si_conv.vc_type != CONV_NONE) { pc = string_convert(&spin->si_conv, rline, NULL); if (pc == NULL) { smsg((char_u *)_("Conversion failure for word in %s line %d: %s"), fname, lnum, rline); continue; } line = pc; } else #endif { pc = NULL; line = rline; } /* Split the line up in white separated items. Put a NUL after each * item. */ itemcnt = 0; for (p = line; ; ) { while (*p != NUL && *p <= ' ') /* skip white space and CR/NL */ ++p; if (*p == NUL) break; if (itemcnt == MAXITEMCNT) /* too many items */ break; items[itemcnt++] = p; /* A few items have arbitrary text argument, don't split them. */ if (itemcnt == 2 && spell_info_item(items[0])) while (*p >= ' ' || *p == TAB) /* skip until CR/NL */ ++p; else while (*p > ' ') /* skip until white space or CR/NL */ ++p; if (*p == NUL) break; *p++ = NUL; } /* Handle non-empty lines. */ if (itemcnt > 0) { if (is_aff_rule(items, itemcnt, "SET", 2) && aff->af_enc == NULL) { #ifdef FEAT_MBYTE /* Setup for conversion from "ENC" to 'encoding'. */ aff->af_enc = enc_canonize(items[1]); if (aff->af_enc != NULL && !spin->si_ascii && convert_setup(&spin->si_conv, aff->af_enc, p_enc) == FAIL) smsg((char_u *)_("Conversion in %s not supported: from %s to %s"), fname, aff->af_enc, p_enc); spin->si_conv.vc_fail = TRUE; #else smsg((char_u *)_("Conversion in %s not supported"), fname); #endif } else if (is_aff_rule(items, itemcnt, "FLAG", 2) && aff->af_flagtype == AFT_CHAR) { if (STRCMP(items[1], "long") == 0) aff->af_flagtype = AFT_LONG; else if (STRCMP(items[1], "num") == 0) aff->af_flagtype = AFT_NUM; else if (STRCMP(items[1], "caplong") == 0) aff->af_flagtype = AFT_CAPLONG; else smsg((char_u *)_("Invalid value for FLAG in %s line %d: %s"), fname, lnum, items[1]); if (aff->af_rare != 0 || aff->af_keepcase != 0 || aff->af_bad != 0 || aff->af_needaffix != 0 || aff->af_circumfix != 0 || aff->af_needcomp != 0 || aff->af_comproot != 0 || aff->af_nosuggest != 0 || compflags != NULL || aff->af_suff.ht_used > 0 || aff->af_pref.ht_used > 0) smsg((char_u *)_("FLAG after using flags in %s line %d: %s"), fname, lnum, items[1]); } else if (spell_info_item(items[0])) { p = (char_u *)getroom(spin, (spin->si_info == NULL ? 0 : STRLEN(spin->si_info)) + STRLEN(items[0]) + STRLEN(items[1]) + 3, FALSE); if (p != NULL) { if (spin->si_info != NULL) { STRCPY(p, spin->si_info); STRCAT(p, "\n"); } STRCAT(p, items[0]); STRCAT(p, " "); STRCAT(p, items[1]); spin->si_info = p; } } else if (is_aff_rule(items, itemcnt, "MIDWORD", 2) && midword == NULL) { midword = getroom_save(spin, items[1]); } else if (is_aff_rule(items, itemcnt, "TRY", 2)) { /* ignored, we look in the tree for what chars may appear */ } /* TODO: remove "RAR" later */ else if ((is_aff_rule(items, itemcnt, "RAR", 2) || is_aff_rule(items, itemcnt, "RARE", 2)) && aff->af_rare == 0) { aff->af_rare = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } /* TODO: remove "KEP" later */ else if ((is_aff_rule(items, itemcnt, "KEP", 2) || is_aff_rule(items, itemcnt, "KEEPCASE", 2)) && aff->af_keepcase == 0) { aff->af_keepcase = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } else if ((is_aff_rule(items, itemcnt, "BAD", 2) || is_aff_rule(items, itemcnt, "FORBIDDENWORD", 2)) && aff->af_bad == 0) { aff->af_bad = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } else if (is_aff_rule(items, itemcnt, "NEEDAFFIX", 2) && aff->af_needaffix == 0) { aff->af_needaffix = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } else if (is_aff_rule(items, itemcnt, "CIRCUMFIX", 2) && aff->af_circumfix == 0) { aff->af_circumfix = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } else if (is_aff_rule(items, itemcnt, "NOSUGGEST", 2) && aff->af_nosuggest == 0) { aff->af_nosuggest = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } else if ((is_aff_rule(items, itemcnt, "NEEDCOMPOUND", 2) || is_aff_rule(items, itemcnt, "ONLYINCOMPOUND", 2)) && aff->af_needcomp == 0) { aff->af_needcomp = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } else if (is_aff_rule(items, itemcnt, "COMPOUNDROOT", 2) && aff->af_comproot == 0) { aff->af_comproot = affitem2flag(aff->af_flagtype, items[1], fname, lnum); } else if (is_aff_rule(items, itemcnt, "COMPOUNDFORBIDFLAG", 2) && aff->af_compforbid == 0) { aff->af_compforbid = affitem2flag(aff->af_flagtype, items[1], fname, lnum); if (aff->af_pref.ht_used > 0) smsg((char_u *)_("Defining COMPOUNDFORBIDFLAG after PFX item may give wrong results in %s line %d"), fname, lnum); } else if (is_aff_rule(items, itemcnt, "COMPOUNDPERMITFLAG", 2) && aff->af_comppermit == 0) { aff->af_comppermit = affitem2flag(aff->af_flagtype, items[1], fname, lnum); if (aff->af_pref.ht_used > 0) smsg((char_u *)_("Defining COMPOUNDPERMITFLAG after PFX item may give wrong results in %s line %d"), fname, lnum); } else if (is_aff_rule(items, itemcnt, "COMPOUNDFLAG", 2) && compflags == NULL) { /* Turn flag "c" into COMPOUNDRULE compatible string "c+", * "Na" into "Na+", "1234" into "1234+". */ p = getroom(spin, STRLEN(items[1]) + 2, FALSE); if (p != NULL) { STRCPY(p, items[1]); STRCAT(p, "+"); compflags = p; } } else if (is_aff_rule(items, itemcnt, "COMPOUNDRULES", 2)) { /* We don't use the count, but do check that it's a number and * not COMPOUNDRULE mistyped. */ if (atoi((char *)items[1]) == 0) smsg((char_u *)_("Wrong COMPOUNDRULES value in %s line %d: %s"), fname, lnum, items[1]); } else if (is_aff_rule(items, itemcnt, "COMPOUNDRULE", 2)) { /* Don't use the first rule if it is a number. */ if (compflags != NULL || *skipdigits(items[1]) != NUL) { /* Concatenate this string to previously defined ones, * using a slash to separate them. */ l = (int)STRLEN(items[1]) + 1; if (compflags != NULL) l += (int)STRLEN(compflags) + 1; p = getroom(spin, l, FALSE); if (p != NULL) { if (compflags != NULL) { STRCPY(p, compflags); STRCAT(p, "/"); } STRCAT(p, items[1]); compflags = p; } } } else if (is_aff_rule(items, itemcnt, "COMPOUNDWORDMAX", 2) && compmax == 0) { compmax = atoi((char *)items[1]); if (compmax == 0) smsg((char_u *)_("Wrong COMPOUNDWORDMAX value in %s line %d: %s"), fname, lnum, items[1]); } else if (is_aff_rule(items, itemcnt, "COMPOUNDMIN", 2) && compminlen == 0) { compminlen = atoi((char *)items[1]); if (compminlen == 0) smsg((char_u *)_("Wrong COMPOUNDMIN value in %s line %d: %s"), fname, lnum, items[1]); } else if (is_aff_rule(items, itemcnt, "COMPOUNDSYLMAX", 2) && compsylmax == 0) { compsylmax = atoi((char *)items[1]); if (compsylmax == 0) smsg((char_u *)_("Wrong COMPOUNDSYLMAX value in %s line %d: %s"), fname, lnum, items[1]); } else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDDUP", 1)) { compoptions |= COMP_CHECKDUP; } else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDREP", 1)) { compoptions |= COMP_CHECKREP; } else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDCASE", 1)) { compoptions |= COMP_CHECKCASE; } else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDTRIPLE", 1)) { compoptions |= COMP_CHECKTRIPLE; } else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDPATTERN", 2)) { if (atoi((char *)items[1]) == 0) smsg((char_u *)_("Wrong CHECKCOMPOUNDPATTERN value in %s line %d: %s"), fname, lnum, items[1]); } else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDPATTERN", 3)) { garray_T *gap = &spin->si_comppat; int i; /* Only add the couple if it isn't already there. */ for (i = 0; i < gap->ga_len - 1; i += 2) if (STRCMP(((char_u **)(gap->ga_data))[i], items[1]) == 0 && STRCMP(((char_u **)(gap->ga_data))[i + 1], items[2]) == 0) break; if (i >= gap->ga_len && ga_grow(gap, 2) == OK) { ((char_u **)(gap->ga_data))[gap->ga_len++] = getroom_save(spin, items[1]); ((char_u **)(gap->ga_data))[gap->ga_len++] = getroom_save(spin, items[2]); } } else if (is_aff_rule(items, itemcnt, "SYLLABLE", 2) && syllable == NULL) { syllable = getroom_save(spin, items[1]); } else if (is_aff_rule(items, itemcnt, "NOBREAK", 1)) { spin->si_nobreak = TRUE; } else if (is_aff_rule(items, itemcnt, "NOSPLITSUGS", 1)) { spin->si_nosplitsugs = TRUE; } else if (is_aff_rule(items, itemcnt, "NOSUGFILE", 1)) { spin->si_nosugfile = TRUE; } else if (is_aff_rule(items, itemcnt, "PFXPOSTPONE", 1)) { aff->af_pfxpostpone = TRUE; } else if ((STRCMP(items[0], "PFX") == 0 || STRCMP(items[0], "SFX") == 0) && aff_todo == 0 && itemcnt >= 4) { int lasti = 4; char_u key[AH_KEY_LEN]; if (*items[0] == 'P') tp = &aff->af_pref; else tp = &aff->af_suff; /* Myspell allows the same affix name to be used multiple * times. The affix files that do this have an undocumented * "S" flag on all but the last block, thus we check for that * and store it in ah_follows. */ vim_strncpy(key, items[1], AH_KEY_LEN - 1); hi = hash_find(tp, key); if (!HASHITEM_EMPTY(hi)) { cur_aff = HI2AH(hi); if (cur_aff->ah_combine != (*items[2] == 'Y')) smsg((char_u *)_("Different combining flag in continued affix block in %s line %d: %s"), fname, lnum, items[1]); if (!cur_aff->ah_follows) smsg((char_u *)_("Duplicate affix in %s line %d: %s"), fname, lnum, items[1]); } else { /* New affix letter. */ cur_aff = (affheader_T *)getroom(spin, sizeof(affheader_T), TRUE); if (cur_aff == NULL) break; cur_aff->ah_flag = affitem2flag(aff->af_flagtype, items[1], fname, lnum); if (cur_aff->ah_flag == 0 || STRLEN(items[1]) >= AH_KEY_LEN) break; if (cur_aff->ah_flag == aff->af_bad || cur_aff->ah_flag == aff->af_rare || cur_aff->ah_flag == aff->af_keepcase || cur_aff->ah_flag == aff->af_needaffix || cur_aff->ah_flag == aff->af_circumfix || cur_aff->ah_flag == aff->af_nosuggest || cur_aff->ah_flag == aff->af_needcomp || cur_aff->ah_flag == aff->af_comproot) smsg((char_u *)_("Affix also used for BAD/RARE/KEEPCASE/NEEDAFFIX/NEEDCOMPOUND/NOSUGGEST in %s line %d: %s"), fname, lnum, items[1]); STRCPY(cur_aff->ah_key, items[1]); hash_add(tp, cur_aff->ah_key); cur_aff->ah_combine = (*items[2] == 'Y'); } /* Check for the "S" flag, which apparently means that another * block with the same affix name is following. */ if (itemcnt > lasti && STRCMP(items[lasti], "S") == 0) { ++lasti; cur_aff->ah_follows = TRUE; } else cur_aff->ah_follows = FALSE; /* Myspell allows extra text after the item, but that might * mean mistakes go unnoticed. Require a comment-starter. */ if (itemcnt > lasti && *items[lasti] != '#') smsg((char_u *)_(e_afftrailing), fname, lnum, items[lasti]); if (STRCMP(items[2], "Y") != 0 && STRCMP(items[2], "N") != 0) smsg((char_u *)_("Expected Y or N in %s line %d: %s"), fname, lnum, items[2]); if (*items[0] == 'P' && aff->af_pfxpostpone) { if (cur_aff->ah_newID == 0) { /* Use a new number in the .spl file later, to be able * to handle multiple .aff files. */ check_renumber(spin); cur_aff->ah_newID = ++spin->si_newprefID; /* We only really use ah_newID if the prefix is * postponed. We know that only after handling all * the items. */ did_postpone_prefix = FALSE; } else /* Did use the ID in a previous block. */ did_postpone_prefix = TRUE; } aff_todo = atoi((char *)items[3]); } else if ((STRCMP(items[0], "PFX") == 0 || STRCMP(items[0], "SFX") == 0) && aff_todo > 0 && STRCMP(cur_aff->ah_key, items[1]) == 0 && itemcnt >= 5) { affentry_T *aff_entry; int upper = FALSE; int lasti = 5; /* Myspell allows extra text after the item, but that might * mean mistakes go unnoticed. Require a comment-starter. * Hunspell uses a "-" item. */ if (itemcnt > lasti && *items[lasti] != '#' && (STRCMP(items[lasti], "-") != 0 || itemcnt != lasti + 1)) smsg((char_u *)_(e_afftrailing), fname, lnum, items[lasti]); /* New item for an affix letter. */ --aff_todo; aff_entry = (affentry_T *)getroom(spin, sizeof(affentry_T), TRUE); if (aff_entry == NULL) break; if (STRCMP(items[2], "0") != 0) aff_entry->ae_chop = getroom_save(spin, items[2]); if (STRCMP(items[3], "0") != 0) { aff_entry->ae_add = getroom_save(spin, items[3]); /* Recognize flags on the affix: abcd/XYZ */ aff_entry->ae_flags = vim_strchr(aff_entry->ae_add, '/'); if (aff_entry->ae_flags != NULL) { *aff_entry->ae_flags++ = NUL; aff_process_flags(aff, aff_entry); } } /* Don't use an affix entry with non-ASCII characters when * "spin->si_ascii" is TRUE. */ if (!spin->si_ascii || !(has_non_ascii(aff_entry->ae_chop) || has_non_ascii(aff_entry->ae_add))) { aff_entry->ae_next = cur_aff->ah_first; cur_aff->ah_first = aff_entry; if (STRCMP(items[4], ".") != 0) { char_u buf[MAXLINELEN]; aff_entry->ae_cond = getroom_save(spin, items[4]); if (*items[0] == 'P') sprintf((char *)buf, "^%s", items[4]); else sprintf((char *)buf, "%s$", items[4]); aff_entry->ae_prog = vim_regcomp(buf, RE_MAGIC + RE_STRING + RE_STRICT); if (aff_entry->ae_prog == NULL) smsg((char_u *)_("Broken condition in %s line %d: %s"), fname, lnum, items[4]); } /* For postponed prefixes we need an entry in si_prefcond * for the condition. Use an existing one if possible. * Can't be done for an affix with flags, ignoring * COMPOUNDFORBIDFLAG and COMPOUNDPERMITFLAG. */ if (*items[0] == 'P' && aff->af_pfxpostpone && aff_entry->ae_flags == NULL) { /* When the chop string is one lower-case letter and * the add string ends in the upper-case letter we set * the "upper" flag, clear "ae_chop" and remove the * letters from "ae_add". The condition must either * be empty or start with the same letter. */ if (aff_entry->ae_chop != NULL && aff_entry->ae_add != NULL #ifdef FEAT_MBYTE && aff_entry->ae_chop[(*mb_ptr2len)( aff_entry->ae_chop)] == NUL #else && aff_entry->ae_chop[1] == NUL #endif ) { int c, c_up; c = PTR2CHAR(aff_entry->ae_chop); c_up = SPELL_TOUPPER(c); if (c_up != c && (aff_entry->ae_cond == NULL || PTR2CHAR(aff_entry->ae_cond) == c)) { p = aff_entry->ae_add + STRLEN(aff_entry->ae_add); mb_ptr_back(aff_entry->ae_add, p); if (PTR2CHAR(p) == c_up) { upper = TRUE; aff_entry->ae_chop = NULL; *p = NUL; /* The condition is matched with the * actual word, thus must check for the * upper-case letter. */ if (aff_entry->ae_cond != NULL) { char_u buf[MAXLINELEN]; #ifdef FEAT_MBYTE if (has_mbyte) { onecap_copy(items[4], buf, TRUE); aff_entry->ae_cond = getroom_save( spin, buf); } else #endif *aff_entry->ae_cond = c_up; if (aff_entry->ae_cond != NULL) { sprintf((char *)buf, "^%s", aff_entry->ae_cond); vim_free(aff_entry->ae_prog); aff_entry->ae_prog = vim_regcomp( buf, RE_MAGIC + RE_STRING); } } } } } if (aff_entry->ae_chop == NULL && aff_entry->ae_flags == NULL) { int idx; char_u **pp; int n; /* Find a previously used condition. */ for (idx = spin->si_prefcond.ga_len - 1; idx >= 0; --idx) { p = ((char_u **)spin->si_prefcond.ga_data)[idx]; if (str_equal(p, aff_entry->ae_cond)) break; } if (idx < 0 && ga_grow(&spin->si_prefcond, 1) == OK) { /* Not found, add a new condition. */ idx = spin->si_prefcond.ga_len++; pp = ((char_u **)spin->si_prefcond.ga_data) + idx; if (aff_entry->ae_cond == NULL) *pp = NULL; else *pp = getroom_save(spin, aff_entry->ae_cond); } /* Add the prefix to the prefix tree. */ if (aff_entry->ae_add == NULL) p = (char_u *)""; else p = aff_entry->ae_add; /* PFX_FLAGS is a negative number, so that * tree_add_word() knows this is the prefix tree. */ n = PFX_FLAGS; if (!cur_aff->ah_combine) n |= WFP_NC; if (upper) n |= WFP_UP; if (aff_entry->ae_comppermit) n |= WFP_COMPPERMIT; if (aff_entry->ae_compforbid) n |= WFP_COMPFORBID; tree_add_word(spin, p, spin->si_prefroot, n, idx, cur_aff->ah_newID); did_postpone_prefix = TRUE; } /* Didn't actually use ah_newID, backup si_newprefID. */ if (aff_todo == 0 && !did_postpone_prefix) { --spin->si_newprefID; cur_aff->ah_newID = 0; } } } } else if (is_aff_rule(items, itemcnt, "FOL", 2) && fol == NULL) { fol = vim_strsave(items[1]); } else if (is_aff_rule(items, itemcnt, "LOW", 2) && low == NULL) { low = vim_strsave(items[1]); } else if (is_aff_rule(items, itemcnt, "UPP", 2) && upp == NULL) { upp = vim_strsave(items[1]); } else if (is_aff_rule(items, itemcnt, "REP", 2) || is_aff_rule(items, itemcnt, "REPSAL", 2)) { /* Ignore REP/REPSAL count */; if (!isdigit(*items[1])) smsg((char_u *)_("Expected REP(SAL) count in %s line %d"), fname, lnum); } else if ((STRCMP(items[0], "REP") == 0 || STRCMP(items[0], "REPSAL") == 0) && itemcnt >= 3) { /* REP/REPSAL item */ /* Myspell ignores extra arguments, we require it starts with * # to detect mistakes. */ if (itemcnt > 3 && items[3][0] != '#') smsg((char_u *)_(e_afftrailing), fname, lnum, items[3]); if (items[0][3] == 'S' ? do_repsal : do_rep) { /* Replace underscore with space (can't include a space * directly). */ for (p = items[1]; *p != NUL; mb_ptr_adv(p)) if (*p == '_') *p = ' '; for (p = items[2]; *p != NUL; mb_ptr_adv(p)) if (*p == '_') *p = ' '; add_fromto(spin, items[0][3] == 'S' ? &spin->si_repsal : &spin->si_rep, items[1], items[2]); } } else if (is_aff_rule(items, itemcnt, "MAP", 2)) { /* MAP item or count */ if (!found_map) { /* First line contains the count. */ found_map = TRUE; if (!isdigit(*items[1])) smsg((char_u *)_("Expected MAP count in %s line %d"), fname, lnum); } else if (do_mapline) { int c; /* Check that every character appears only once. */ for (p = items[1]; *p != NUL; ) { #ifdef FEAT_MBYTE c = mb_ptr2char_adv(&p); #else c = *p++; #endif if ((spin->si_map.ga_len > 0 && vim_strchr(spin->si_map.ga_data, c) != NULL) || vim_strchr(p, c) != NULL) smsg((char_u *)_("Duplicate character in MAP in %s line %d"), fname, lnum); } /* We simply concatenate all the MAP strings, separated by * slashes. */ ga_concat(&spin->si_map, items[1]); ga_append(&spin->si_map, '/'); } } /* Accept "SAL from to" and "SAL from to #comment". */ else if (is_aff_rule(items, itemcnt, "SAL", 3)) { if (do_sal) { /* SAL item (sounds-a-like) * Either one of the known keys or a from-to pair. */ if (STRCMP(items[1], "followup") == 0) spin->si_followup = sal_to_bool(items[2]); else if (STRCMP(items[1], "collapse_result") == 0) spin->si_collapse = sal_to_bool(items[2]); else if (STRCMP(items[1], "remove_accents") == 0) spin->si_rem_accents = sal_to_bool(items[2]); else /* when "to" is "_" it means empty */ add_fromto(spin, &spin->si_sal, items[1], STRCMP(items[2], "_") == 0 ? (char_u *)"" : items[2]); } } else if (is_aff_rule(items, itemcnt, "SOFOFROM", 2) && sofofrom == NULL) { sofofrom = getroom_save(spin, items[1]); } else if (is_aff_rule(items, itemcnt, "SOFOTO", 2) && sofoto == NULL) { sofoto = getroom_save(spin, items[1]); } else if (STRCMP(items[0], "COMMON") == 0) { int i; for (i = 1; i < itemcnt; ++i) { if (HASHITEM_EMPTY(hash_find(&spin->si_commonwords, items[i]))) { p = vim_strsave(items[i]); if (p == NULL) break; hash_add(&spin->si_commonwords, p); } } } else smsg((char_u *)_("Unrecognized or duplicate item in %s line %d: %s"), fname, lnum, items[0]); } } if (fol != NULL || low != NULL || upp != NULL) { if (spin->si_clear_chartab) { /* Clear the char type tables, don't want to use any of the * currently used spell properties. */ init_spell_chartab(); spin->si_clear_chartab = FALSE; } /* * Don't write a word table for an ASCII file, so that we don't check * for conflicts with a word table that matches 'encoding'. * Don't write one for utf-8 either, we use utf_*() and * mb_get_class(), the list of chars in the file will be incomplete. */ if (!spin->si_ascii #ifdef FEAT_MBYTE && !enc_utf8 #endif ) { if (fol == NULL || low == NULL || upp == NULL) smsg((char_u *)_("Missing FOL/LOW/UPP line in %s"), fname); else (void)set_spell_chartab(fol, low, upp); } vim_free(fol); vim_free(low); vim_free(upp); } /* Use compound specifications of the .aff file for the spell info. */ if (compmax != 0) { aff_check_number(spin->si_compmax, compmax, "COMPOUNDWORDMAX"); spin->si_compmax = compmax; } if (compminlen != 0) { aff_check_number(spin->si_compminlen, compminlen, "COMPOUNDMIN"); spin->si_compminlen = compminlen; } if (compsylmax != 0) { if (syllable == NULL) smsg((char_u *)_("COMPOUNDSYLMAX used without SYLLABLE")); aff_check_number(spin->si_compsylmax, compsylmax, "COMPOUNDSYLMAX"); spin->si_compsylmax = compsylmax; } if (compoptions != 0) { aff_check_number(spin->si_compoptions, compoptions, "COMPOUND options"); spin->si_compoptions |= compoptions; } if (compflags != NULL) process_compflags(spin, aff, compflags); /* Check that we didn't use too many renumbered flags. */ if (spin->si_newcompID < spin->si_newprefID) { if (spin->si_newcompID == 127 || spin->si_newcompID == 255) MSG(_("Too many postponed prefixes")); else if (spin->si_newprefID == 0 || spin->si_newprefID == 127) MSG(_("Too many compound flags")); else MSG(_("Too many postponed prefixes and/or compound flags")); } if (syllable != NULL) { aff_check_string(spin->si_syllable, syllable, "SYLLABLE"); spin->si_syllable = syllable; } if (sofofrom != NULL || sofoto != NULL) { if (sofofrom == NULL || sofoto == NULL) smsg((char_u *)_("Missing SOFO%s line in %s"), sofofrom == NULL ? "FROM" : "TO", fname); else if (spin->si_sal.ga_len > 0) smsg((char_u *)_("Both SAL and SOFO lines in %s"), fname); else { aff_check_string(spin->si_sofofr, sofofrom, "SOFOFROM"); aff_check_string(spin->si_sofoto, sofoto, "SOFOTO"); spin->si_sofofr = sofofrom; spin->si_sofoto = sofoto; } } if (midword != NULL) { aff_check_string(spin->si_midword, midword, "MIDWORD"); spin->si_midword = midword; } vim_free(pc); fclose(fd); return aff; } /* * Return TRUE when items[0] equals "rulename", there are "mincount" items or * a comment is following after item "mincount". */ static int is_aff_rule(items, itemcnt, rulename, mincount) char_u **items; int itemcnt; char *rulename; int mincount; { return (STRCMP(items[0], rulename) == 0 && (itemcnt == mincount || (itemcnt > mincount && items[mincount][0] == '#'))); } /* * For affix "entry" move COMPOUNDFORBIDFLAG and COMPOUNDPERMITFLAG from * ae_flags to ae_comppermit and ae_compforbid. */ static void aff_process_flags(affile, entry) afffile_T *affile; affentry_T *entry; { char_u *p; char_u *prevp; unsigned flag; if (entry->ae_flags != NULL && (affile->af_compforbid != 0 || affile->af_comppermit != 0)) { for (p = entry->ae_flags; *p != NUL; ) { prevp = p; flag = get_affitem(affile->af_flagtype, &p); if (flag == affile->af_comppermit || flag == affile->af_compforbid) { STRMOVE(prevp, p); p = prevp; if (flag == affile->af_comppermit) entry->ae_comppermit = TRUE; else entry->ae_compforbid = TRUE; } if (affile->af_flagtype == AFT_NUM && *p == ',') ++p; } if (*entry->ae_flags == NUL) entry->ae_flags = NULL; /* nothing left */ } } /* * Return TRUE if "s" is the name of an info item in the affix file. */ static int spell_info_item(s) char_u *s; { return STRCMP(s, "NAME") == 0 || STRCMP(s, "HOME") == 0 || STRCMP(s, "VERSION") == 0 || STRCMP(s, "AUTHOR") == 0 || STRCMP(s, "EMAIL") == 0 || STRCMP(s, "COPYRIGHT") == 0; } /* * Turn an affix flag name into a number, according to the FLAG type. * returns zero for failure. */ static unsigned affitem2flag(flagtype, item, fname, lnum) int flagtype; char_u *item; char_u *fname; int lnum; { unsigned res; char_u *p = item; res = get_affitem(flagtype, &p); if (res == 0) { if (flagtype == AFT_NUM) smsg((char_u *)_("Flag is not a number in %s line %d: %s"), fname, lnum, item); else smsg((char_u *)_("Illegal flag in %s line %d: %s"), fname, lnum, item); } if (*p != NUL) { smsg((char_u *)_(e_affname), fname, lnum, item); return 0; } return res; } /* * Get one affix name from "*pp" and advance the pointer. * Returns zero for an error, still advances the pointer then. */ static unsigned get_affitem(flagtype, pp) int flagtype; char_u **pp; { int res; if (flagtype == AFT_NUM) { if (!VIM_ISDIGIT(**pp)) { ++*pp; /* always advance, avoid getting stuck */ return 0; } res = getdigits(pp); } else { #ifdef FEAT_MBYTE res = mb_ptr2char_adv(pp); #else res = *(*pp)++; #endif if (flagtype == AFT_LONG || (flagtype == AFT_CAPLONG && res >= 'A' && res <= 'Z')) { if (**pp == NUL) return 0; #ifdef FEAT_MBYTE res = mb_ptr2char_adv(pp) + (res << 16); #else res = *(*pp)++ + (res << 16); #endif } } return res; } /* * Process the "compflags" string used in an affix file and append it to * spin->si_compflags. * The processing involves changing the affix names to ID numbers, so that * they fit in one byte. */ static void process_compflags(spin, aff, compflags) spellinfo_T *spin; afffile_T *aff; char_u *compflags; { char_u *p; char_u *prevp; unsigned flag; compitem_T *ci; int id; int len; char_u *tp; char_u key[AH_KEY_LEN]; hashitem_T *hi; /* Make room for the old and the new compflags, concatenated with a / in * between. Processing it makes it shorter, but we don't know by how * much, thus allocate the maximum. */ len = (int)STRLEN(compflags) + 1; if (spin->si_compflags != NULL) len += (int)STRLEN(spin->si_compflags) + 1; p = getroom(spin, len, FALSE); if (p == NULL) return; if (spin->si_compflags != NULL) { STRCPY(p, spin->si_compflags); STRCAT(p, "/"); } spin->si_compflags = p; tp = p + STRLEN(p); for (p = compflags; *p != NUL; ) { if (vim_strchr((char_u *)"/?*+[]", *p) != NULL) /* Copy non-flag characters directly. */ *tp++ = *p++; else { /* First get the flag number, also checks validity. */ prevp = p; flag = get_affitem(aff->af_flagtype, &p); if (flag != 0) { /* Find the flag in the hashtable. If it was used before, use * the existing ID. Otherwise add a new entry. */ vim_strncpy(key, prevp, p - prevp); hi = hash_find(&aff->af_comp, key); if (!HASHITEM_EMPTY(hi)) id = HI2CI(hi)->ci_newID; else { ci = (compitem_T *)getroom(spin, sizeof(compitem_T), TRUE); if (ci == NULL) break; STRCPY(ci->ci_key, key); ci->ci_flag = flag; /* Avoid using a flag ID that has a special meaning in a * regexp (also inside []). */ do { check_renumber(spin); id = spin->si_newcompID--; } while (vim_strchr((char_u *)"/?*+[]\\-^", id) != NULL); ci->ci_newID = id; hash_add(&aff->af_comp, ci->ci_key); } *tp++ = id; } if (aff->af_flagtype == AFT_NUM && *p == ',') ++p; } } *tp = NUL; } /* * Check that the new IDs for postponed affixes and compounding don't overrun * each other. We have almost 255 available, but start at 0-127 to avoid * using two bytes for utf-8. When the 0-127 range is used up go to 128-255. * When that is used up an error message is given. */ static void check_renumber(spin) spellinfo_T *spin; { if (spin->si_newprefID == spin->si_newcompID && spin->si_newcompID < 128) { spin->si_newprefID = 127; spin->si_newcompID = 255; } } /* * Return TRUE if flag "flag" appears in affix list "afflist". */ static int flag_in_afflist(flagtype, afflist, flag) int flagtype; char_u *afflist; unsigned flag; { char_u *p; unsigned n; switch (flagtype) { case AFT_CHAR: return vim_strchr(afflist, flag) != NULL; case AFT_CAPLONG: case AFT_LONG: for (p = afflist; *p != NUL; ) { #ifdef FEAT_MBYTE n = mb_ptr2char_adv(&p); #else n = *p++; #endif if ((flagtype == AFT_LONG || (n >= 'A' && n <= 'Z')) && *p != NUL) #ifdef FEAT_MBYTE n = mb_ptr2char_adv(&p) + (n << 16); #else n = *p++ + (n << 16); #endif if (n == flag) return TRUE; } break; case AFT_NUM: for (p = afflist; *p != NUL; ) { n = getdigits(&p); if (n == flag) return TRUE; if (*p != NUL) /* skip over comma */ ++p; } break; } return FALSE; } /* * Give a warning when "spinval" and "affval" numbers are set and not the same. */ static void aff_check_number(spinval, affval, name) int spinval; int affval; char *name; { if (spinval != 0 && spinval != affval) smsg((char_u *)_("%s value differs from what is used in another .aff file"), name); } /* * Give a warning when "spinval" and "affval" strings are set and not the same. */ static void aff_check_string(spinval, affval, name) char_u *spinval; char_u *affval; char *name; { if (spinval != NULL && STRCMP(spinval, affval) != 0) smsg((char_u *)_("%s value differs from what is used in another .aff file"), name); } /* * Return TRUE if strings "s1" and "s2" are equal. Also consider both being * NULL as equal. */ static int str_equal(s1, s2) char_u *s1; char_u *s2; { if (s1 == NULL || s2 == NULL) return s1 == s2; return STRCMP(s1, s2) == 0; } /* * Add a from-to item to "gap". Used for REP and SAL items. * They are stored case-folded. */ static void add_fromto(spin, gap, from, to) spellinfo_T *spin; garray_T *gap; char_u *from; char_u *to; { fromto_T *ftp; char_u word[MAXWLEN]; if (ga_grow(gap, 1) == OK) { ftp = ((fromto_T *)gap->ga_data) + gap->ga_len; (void)spell_casefold(from, (int)STRLEN(from), word, MAXWLEN); ftp->ft_from = getroom_save(spin, word); (void)spell_casefold(to, (int)STRLEN(to), word, MAXWLEN); ftp->ft_to = getroom_save(spin, word); ++gap->ga_len; } } /* * Convert a boolean argument in a SAL line to TRUE or FALSE; */ static int sal_to_bool(s) char_u *s; { return STRCMP(s, "1") == 0 || STRCMP(s, "true") == 0; } /* * Return TRUE if string "s" contains a non-ASCII character (128 or higher). * When "s" is NULL FALSE is returned. */ static int has_non_ascii(s) char_u *s; { char_u *p; if (s != NULL) for (p = s; *p != NUL; ++p) if (*p >= 128) return TRUE; return FALSE; } /* * Free the structure filled by spell_read_aff(). */ static void spell_free_aff(aff) afffile_T *aff; { hashtab_T *ht; hashitem_T *hi; int todo; affheader_T *ah; affentry_T *ae; vim_free(aff->af_enc); /* All this trouble to free the "ae_prog" items... */ for (ht = &aff->af_pref; ; ht = &aff->af_suff) { todo = (int)ht->ht_used; for (hi = ht->ht_array; todo > 0; ++hi) { if (!HASHITEM_EMPTY(hi)) { --todo; ah = HI2AH(hi); for (ae = ah->ah_first; ae != NULL; ae = ae->ae_next) vim_free(ae->ae_prog); } } if (ht == &aff->af_suff) break; } hash_clear(&aff->af_pref); hash_clear(&aff->af_suff); hash_clear(&aff->af_comp); } /* * Read dictionary file "fname". * Returns OK or FAIL; */ static int spell_read_dic(spin, fname, affile) spellinfo_T *spin; char_u *fname; afffile_T *affile; { hashtab_T ht; char_u line[MAXLINELEN]; char_u *p; char_u *afflist; char_u store_afflist[MAXWLEN]; int pfxlen; int need_affix; char_u *dw; char_u *pc; char_u *w; int l; hash_T hash; hashitem_T *hi; FILE *fd; int lnum = 1; int non_ascii = 0; int retval = OK; char_u message[MAXLINELEN + MAXWLEN]; int flags; int duplicate = 0; /* * Open the file. */ fd = mch_fopen((char *)fname, "r"); if (fd == NULL) { EMSG2(_(e_notopen), fname); return FAIL; } /* The hashtable is only used to detect duplicated words. */ hash_init(&ht); vim_snprintf((char *)IObuff, IOSIZE, _("Reading dictionary file %s ..."), fname); spell_message(spin, IObuff); /* start with a message for the first line */ spin->si_msg_count = 999999; /* Read and ignore the first line: word count. */ (void)vim_fgets(line, MAXLINELEN, fd); if (!vim_isdigit(*skipwhite(line))) EMSG2(_("E760: No word count in %s"), fname); /* * Read all the lines in the file one by one. * The words are converted to 'encoding' here, before being added to * the hashtable. */ while (!vim_fgets(line, MAXLINELEN, fd) && !got_int) { line_breakcheck(); ++lnum; if (line[0] == '#' || line[0] == '/') continue; /* comment line */ /* Remove CR, LF and white space from the end. White space halfway * the word is kept to allow e.g., "et al.". */ l = (int)STRLEN(line); while (l > 0 && line[l - 1] <= ' ') --l; if (l == 0) continue; /* empty line */ line[l] = NUL; #ifdef FEAT_MBYTE /* Convert from "SET" to 'encoding' when needed. */ if (spin->si_conv.vc_type != CONV_NONE) { pc = string_convert(&spin->si_conv, line, NULL); if (pc == NULL) { smsg((char_u *)_("Conversion failure for word in %s line %d: %s"), fname, lnum, line); continue; } w = pc; } else #endif { pc = NULL; w = line; } /* Truncate the word at the "/", set "afflist" to what follows. * Replace "\/" by "/" and "\\" by "\". */ afflist = NULL; for (p = w; *p != NUL; mb_ptr_adv(p)) { if (*p == '\\' && (p[1] == '\\' || p[1] == '/')) STRMOVE(p, p + 1); else if (*p == '/') { *p = NUL; afflist = p + 1; break; } } /* Skip non-ASCII words when "spin->si_ascii" is TRUE. */ if (spin->si_ascii && has_non_ascii(w)) { ++non_ascii; vim_free(pc); continue; } /* This takes time, print a message every 10000 words. */ if (spin->si_verbose && spin->si_msg_count > 10000) { spin->si_msg_count = 0; vim_snprintf((char *)message, sizeof(message), _("line %6d, word %6d - %s"), lnum, spin->si_foldwcount + spin->si_keepwcount, w); msg_start(); msg_puts_long_attr(message, 0); msg_clr_eos(); msg_didout = FALSE; msg_col = 0; out_flush(); } /* Store the word in the hashtable to be able to find duplicates. */ dw = (char_u *)getroom_save(spin, w); if (dw == NULL) { retval = FAIL; vim_free(pc); break; } hash = hash_hash(dw); hi = hash_lookup(&ht, dw, hash); if (!HASHITEM_EMPTY(hi)) { if (p_verbose > 0) smsg((char_u *)_("Duplicate word in %s line %d: %s"), fname, lnum, dw); else if (duplicate == 0) smsg((char_u *)_("First duplicate word in %s line %d: %s"), fname, lnum, dw); ++duplicate; } else hash_add_item(&ht, hi, dw, hash); flags = 0; store_afflist[0] = NUL; pfxlen = 0; need_affix = FALSE; if (afflist != NULL) { /* Extract flags from the affix list. */ flags |= get_affix_flags(affile, afflist); if (affile->af_needaffix != 0 && flag_in_afflist( affile->af_flagtype, afflist, affile->af_needaffix)) need_affix = TRUE; if (affile->af_pfxpostpone) /* Need to store the list of prefix IDs with the word. */ pfxlen = get_pfxlist(affile, afflist, store_afflist); if (spin->si_compflags != NULL) /* Need to store the list of compound flags with the word. * Concatenate them to the list of prefix IDs. */ get_compflags(affile, afflist, store_afflist + pfxlen); } /* Add the word to the word tree(s). */ if (store_word(spin, dw, flags, spin->si_region, store_afflist, need_affix) == FAIL) retval = FAIL; if (afflist != NULL) { /* Find all matching suffixes and add the resulting words. * Additionally do matching prefixes that combine. */ if (store_aff_word(spin, dw, afflist, affile, &affile->af_suff, &affile->af_pref, CONDIT_SUF, flags, store_afflist, pfxlen) == FAIL) retval = FAIL; /* Find all matching prefixes and add the resulting words. */ if (store_aff_word(spin, dw, afflist, affile, &affile->af_pref, NULL, CONDIT_SUF, flags, store_afflist, pfxlen) == FAIL) retval = FAIL; } vim_free(pc); } if (duplicate > 0) smsg((char_u *)_("%d duplicate word(s) in %s"), duplicate, fname); if (spin->si_ascii && non_ascii > 0) smsg((char_u *)_("Ignored %d word(s) with non-ASCII characters in %s"), non_ascii, fname); hash_clear(&ht); fclose(fd); return retval; } /* * Check for affix flags in "afflist" that are turned into word flags. * Return WF_ flags. */ static int get_affix_flags(affile, afflist) afffile_T *affile; char_u *afflist; { int flags = 0; if (affile->af_keepcase != 0 && flag_in_afflist( affile->af_flagtype, afflist, affile->af_keepcase)) flags |= WF_KEEPCAP | WF_FIXCAP; if (affile->af_rare != 0 && flag_in_afflist( affile->af_flagtype, afflist, affile->af_rare)) flags |= WF_RARE; if (affile->af_bad != 0 && flag_in_afflist( affile->af_flagtype, afflist, affile->af_bad)) flags |= WF_BANNED; if (affile->af_needcomp != 0 && flag_in_afflist( affile->af_flagtype, afflist, affile->af_needcomp)) flags |= WF_NEEDCOMP; if (affile->af_comproot != 0 && flag_in_afflist( affile->af_flagtype, afflist, affile->af_comproot)) flags |= WF_COMPROOT; if (affile->af_nosuggest != 0 && flag_in_afflist( affile->af_flagtype, afflist, affile->af_nosuggest)) flags |= WF_NOSUGGEST; return flags; } /* * Get the list of prefix IDs from the affix list "afflist". * Used for PFXPOSTPONE. * Put the resulting flags in "store_afflist[MAXWLEN]" with a terminating NUL * and return the number of affixes. */ static int get_pfxlist(affile, afflist, store_afflist) afffile_T *affile; char_u *afflist; char_u *store_afflist; { char_u *p; char_u *prevp; int cnt = 0; int id; char_u key[AH_KEY_LEN]; hashitem_T *hi; for (p = afflist; *p != NUL; ) { prevp = p; if (get_affitem(affile->af_flagtype, &p) != 0) { /* A flag is a postponed prefix flag if it appears in "af_pref" * and it's ID is not zero. */ vim_strncpy(key, prevp, p - prevp); hi = hash_find(&affile->af_pref, key); if (!HASHITEM_EMPTY(hi)) { id = HI2AH(hi)->ah_newID; if (id != 0) store_afflist[cnt++] = id; } } if (affile->af_flagtype == AFT_NUM && *p == ',') ++p; } store_afflist[cnt] = NUL; return cnt; } /* * Get the list of compound IDs from the affix list "afflist" that are used * for compound words. * Puts the flags in "store_afflist[]". */ static void get_compflags(affile, afflist, store_afflist) afffile_T *affile; char_u *afflist; char_u *store_afflist; { char_u *p; char_u *prevp; int cnt = 0; char_u key[AH_KEY_LEN]; hashitem_T *hi; for (p = afflist; *p != NUL; ) { prevp = p; if (get_affitem(affile->af_flagtype, &p) != 0) { /* A flag is a compound flag if it appears in "af_comp". */ vim_strncpy(key, prevp, p - prevp); hi = hash_find(&affile->af_comp, key); if (!HASHITEM_EMPTY(hi)) store_afflist[cnt++] = HI2CI(hi)->ci_newID; } if (affile->af_flagtype == AFT_NUM && *p == ',') ++p; } store_afflist[cnt] = NUL; } /* * Apply affixes to a word and store the resulting words. * "ht" is the hashtable with affentry_T that need to be applied, either * prefixes or suffixes. * "xht", when not NULL, is the prefix hashtable, to be used additionally on * the resulting words for combining affixes. * * Returns FAIL when out of memory. */ static int store_aff_word(spin, word, afflist, affile, ht, xht, condit, flags, pfxlist, pfxlen) spellinfo_T *spin; /* spell info */ char_u *word; /* basic word start */ char_u *afflist; /* list of names of supported affixes */ afffile_T *affile; hashtab_T *ht; hashtab_T *xht; int condit; /* CONDIT_SUF et al. */ int flags; /* flags for the word */ char_u *pfxlist; /* list of prefix IDs */ int pfxlen; /* nr of flags in "pfxlist" for prefixes, rest * is compound flags */ { int todo; hashitem_T *hi; affheader_T *ah; affentry_T *ae; regmatch_T regmatch; char_u newword[MAXWLEN]; int retval = OK; int i, j; char_u *p; int use_flags; char_u *use_pfxlist; int use_pfxlen; int need_affix; char_u store_afflist[MAXWLEN]; char_u pfx_pfxlist[MAXWLEN]; size_t wordlen = STRLEN(word); int use_condit; todo = (int)ht->ht_used; for (hi = ht->ht_array; todo > 0 && retval == OK; ++hi) { if (!HASHITEM_EMPTY(hi)) { --todo; ah = HI2AH(hi); /* Check that the affix combines, if required, and that the word * supports this affix. */ if (((condit & CONDIT_COMB) == 0 || ah->ah_combine) && flag_in_afflist(affile->af_flagtype, afflist, ah->ah_flag)) { /* Loop over all affix entries with this name. */ for (ae = ah->ah_first; ae != NULL; ae = ae->ae_next) { /* Check the condition. It's not logical to match case * here, but it is required for compatibility with * Myspell. * Another requirement from Myspell is that the chop * string is shorter than the word itself. * For prefixes, when "PFXPOSTPONE" was used, only do * prefixes with a chop string and/or flags. * When a previously added affix had CIRCUMFIX this one * must have it too, if it had not then this one must not * have one either. */ regmatch.regprog = ae->ae_prog; regmatch.rm_ic = FALSE; if ((xht != NULL || !affile->af_pfxpostpone || ae->ae_chop != NULL || ae->ae_flags != NULL) && (ae->ae_chop == NULL || STRLEN(ae->ae_chop) < wordlen) && (ae->ae_prog == NULL || vim_regexec(®match, word, (colnr_T)0)) && (((condit & CONDIT_CFIX) == 0) == ((condit & CONDIT_AFF) == 0 || ae->ae_flags == NULL || !flag_in_afflist(affile->af_flagtype, ae->ae_flags, affile->af_circumfix)))) { /* Match. Remove the chop and add the affix. */ if (xht == NULL) { /* prefix: chop/add at the start of the word */ if (ae->ae_add == NULL) *newword = NUL; else vim_strncpy(newword, ae->ae_add, MAXWLEN - 1); p = word; if (ae->ae_chop != NULL) { /* Skip chop string. */ #ifdef FEAT_MBYTE if (has_mbyte) { i = mb_charlen(ae->ae_chop); for ( ; i > 0; --i) mb_ptr_adv(p); } else #endif p += STRLEN(ae->ae_chop); } STRCAT(newword, p); } else { /* suffix: chop/add at the end of the word */ vim_strncpy(newword, word, MAXWLEN - 1); if (ae->ae_chop != NULL) { /* Remove chop string. */ p = newword + STRLEN(newword); i = (int)MB_CHARLEN(ae->ae_chop); for ( ; i > 0; --i) mb_ptr_back(newword, p); *p = NUL; } if (ae->ae_add != NULL) STRCAT(newword, ae->ae_add); } use_flags = flags; use_pfxlist = pfxlist; use_pfxlen = pfxlen; need_affix = FALSE; use_condit = condit | CONDIT_COMB | CONDIT_AFF; if (ae->ae_flags != NULL) { /* Extract flags from the affix list. */ use_flags |= get_affix_flags(affile, ae->ae_flags); if (affile->af_needaffix != 0 && flag_in_afflist( affile->af_flagtype, ae->ae_flags, affile->af_needaffix)) need_affix = TRUE; /* When there is a CIRCUMFIX flag the other affix * must also have it and we don't add the word * with one affix. */ if (affile->af_circumfix != 0 && flag_in_afflist( affile->af_flagtype, ae->ae_flags, affile->af_circumfix)) { use_condit |= CONDIT_CFIX; if ((condit & CONDIT_CFIX) == 0) need_affix = TRUE; } if (affile->af_pfxpostpone || spin->si_compflags != NULL) { if (affile->af_pfxpostpone) /* Get prefix IDS from the affix list. */ use_pfxlen = get_pfxlist(affile, ae->ae_flags, store_afflist); else use_pfxlen = 0; use_pfxlist = store_afflist; /* Combine the prefix IDs. Avoid adding the * same ID twice. */ for (i = 0; i < pfxlen; ++i) { for (j = 0; j < use_pfxlen; ++j) if (pfxlist[i] == use_pfxlist[j]) break; if (j == use_pfxlen) use_pfxlist[use_pfxlen++] = pfxlist[i]; } if (spin->si_compflags != NULL) /* Get compound IDS from the affix list. */ get_compflags(affile, ae->ae_flags, use_pfxlist + use_pfxlen); /* Combine the list of compound flags. * Concatenate them to the prefix IDs list. * Avoid adding the same ID twice. */ for (i = pfxlen; pfxlist[i] != NUL; ++i) { for (j = use_pfxlen; use_pfxlist[j] != NUL; ++j) if (pfxlist[i] == use_pfxlist[j]) break; if (use_pfxlist[j] == NUL) { use_pfxlist[j++] = pfxlist[i]; use_pfxlist[j] = NUL; } } } } /* Obey a "COMPOUNDFORBIDFLAG" of the affix: don't * use the compound flags. */ if (use_pfxlist != NULL && ae->ae_compforbid) { vim_strncpy(pfx_pfxlist, use_pfxlist, use_pfxlen); use_pfxlist = pfx_pfxlist; } /* When there are postponed prefixes... */ if (spin->si_prefroot != NULL && spin->si_prefroot->wn_sibling != NULL) { /* ... add a flag to indicate an affix was used. */ use_flags |= WF_HAS_AFF; /* ... don't use a prefix list if combining * affixes is not allowed. But do use the * compound flags after them. */ if (!ah->ah_combine && use_pfxlist != NULL) use_pfxlist += use_pfxlen; } /* When compounding is supported and there is no * "COMPOUNDPERMITFLAG" then forbid compounding on the * side where the affix is applied. */ if (spin->si_compflags != NULL && !ae->ae_comppermit) { if (xht != NULL) use_flags |= WF_NOCOMPAFT; else use_flags |= WF_NOCOMPBEF; } /* Store the modified word. */ if (store_word(spin, newword, use_flags, spin->si_region, use_pfxlist, need_affix) == FAIL) retval = FAIL; /* When added a prefix or a first suffix and the affix * has flags may add a(nother) suffix. RECURSIVE! */ if ((condit & CONDIT_SUF) && ae->ae_flags != NULL) if (store_aff_word(spin, newword, ae->ae_flags, affile, &affile->af_suff, xht, use_condit & (xht == NULL ? ~0 : ~CONDIT_SUF), use_flags, use_pfxlist, pfxlen) == FAIL) retval = FAIL; /* When added a suffix and combining is allowed also * try adding a prefix additionally. Both for the * word flags and for the affix flags. RECURSIVE! */ if (xht != NULL && ah->ah_combine) { if (store_aff_word(spin, newword, afflist, affile, xht, NULL, use_condit, use_flags, use_pfxlist, pfxlen) == FAIL || (ae->ae_flags != NULL && store_aff_word(spin, newword, ae->ae_flags, affile, xht, NULL, use_condit, use_flags, use_pfxlist, pfxlen) == FAIL)) retval = FAIL; } } } } } } return retval; } /* * Read a file with a list of words. */ static int spell_read_wordfile(spin, fname) spellinfo_T *spin; char_u *fname; { FILE *fd; long lnum = 0; char_u rline[MAXLINELEN]; char_u *line; char_u *pc = NULL; char_u *p; int l; int retval = OK; int did_word = FALSE; int non_ascii = 0; int flags; int regionmask; /* * Open the file. */ fd = mch_fopen((char *)fname, "r"); if (fd == NULL) { EMSG2(_(e_notopen), fname); return FAIL; } vim_snprintf((char *)IObuff, IOSIZE, _("Reading word file %s ..."), fname); spell_message(spin, IObuff); /* * Read all the lines in the file one by one. */ while (!vim_fgets(rline, MAXLINELEN, fd) && !got_int) { line_breakcheck(); ++lnum; /* Skip comment lines. */ if (*rline == '#') continue; /* Remove CR, LF and white space from the end. */ l = (int)STRLEN(rline); while (l > 0 && rline[l - 1] <= ' ') --l; if (l == 0) continue; /* empty or blank line */ rline[l] = NUL; /* Convert from "/encoding={encoding}" to 'encoding' when needed. */ vim_free(pc); #ifdef FEAT_MBYTE if (spin->si_conv.vc_type != CONV_NONE) { pc = string_convert(&spin->si_conv, rline, NULL); if (pc == NULL) { smsg((char_u *)_("Conversion failure for word in %s line %d: %s"), fname, lnum, rline); continue; } line = pc; } else #endif { pc = NULL; line = rline; } if (*line == '/') { ++line; if (STRNCMP(line, "encoding=", 9) == 0) { if (spin->si_conv.vc_type != CONV_NONE) smsg((char_u *)_("Duplicate /encoding= line ignored in %s line %d: %s"), fname, lnum, line - 1); else if (did_word) smsg((char_u *)_("/encoding= line after word ignored in %s line %d: %s"), fname, lnum, line - 1); else { #ifdef FEAT_MBYTE char_u *enc; /* Setup for conversion to 'encoding'. */ line += 9; enc = enc_canonize(line); if (enc != NULL && !spin->si_ascii && convert_setup(&spin->si_conv, enc, p_enc) == FAIL) smsg((char_u *)_("Conversion in %s not supported: from %s to %s"), fname, line, p_enc); vim_free(enc); spin->si_conv.vc_fail = TRUE; #else smsg((char_u *)_("Conversion in %s not supported"), fname); #endif } continue; } if (STRNCMP(line, "regions=", 8) == 0) { if (spin->si_region_count > 1) smsg((char_u *)_("Duplicate /regions= line ignored in %s line %d: %s"), fname, lnum, line); else { line += 8; if (STRLEN(line) > 16) smsg((char_u *)_("Too many regions in %s line %d: %s"), fname, lnum, line); else { spin->si_region_count = (int)STRLEN(line) / 2; STRCPY(spin->si_region_name, line); /* Adjust the mask for a word valid in all regions. */ spin->si_region = (1 << spin->si_region_count) - 1; } } continue; } smsg((char_u *)_("/ line ignored in %s line %d: %s"), fname, lnum, line - 1); continue; } flags = 0; regionmask = spin->si_region; /* Check for flags and region after a slash. */ p = vim_strchr(line, '/'); if (p != NULL) { *p++ = NUL; while (*p != NUL) { if (*p == '=') /* keep-case word */ flags |= WF_KEEPCAP | WF_FIXCAP; else if (*p == '!') /* Bad, bad, wicked word. */ flags |= WF_BANNED; else if (*p == '?') /* Rare word. */ flags |= WF_RARE; else if (VIM_ISDIGIT(*p)) /* region number(s) */ { if ((flags & WF_REGION) == 0) /* first one */ regionmask = 0; flags |= WF_REGION; l = *p - '0'; if (l > spin->si_region_count) { smsg((char_u *)_("Invalid region nr in %s line %d: %s"), fname, lnum, p); break; } regionmask |= 1 << (l - 1); } else { smsg((char_u *)_("Unrecognized flags in %s line %d: %s"), fname, lnum, p); break; } ++p; } } /* Skip non-ASCII words when "spin->si_ascii" is TRUE. */ if (spin->si_ascii && has_non_ascii(line)) { ++non_ascii; continue; } /* Normal word: store it. */ if (store_word(spin, line, flags, regionmask, NULL, FALSE) == FAIL) { retval = FAIL; break; } did_word = TRUE; } vim_free(pc); fclose(fd); if (spin->si_ascii && non_ascii > 0) { vim_snprintf((char *)IObuff, IOSIZE, _("Ignored %d words with non-ASCII characters"), non_ascii); spell_message(spin, IObuff); } return retval; } /* * Get part of an sblock_T, "len" bytes long. * This avoids calling free() for every little struct we use (and keeping * track of them). * The memory is cleared to all zeros. * Returns NULL when out of memory. */ static void * getroom(spin, len, align) spellinfo_T *spin; size_t len; /* length needed */ int align; /* align for pointer */ { char_u *p; sblock_T *bl = spin->si_blocks; if (align && bl != NULL) /* Round size up for alignment. On some systems structures need to be * aligned to the size of a pointer (e.g., SPARC). */ bl->sb_used = (bl->sb_used + sizeof(char *) - 1) & ~(sizeof(char *) - 1); if (bl == NULL || bl->sb_used + len > SBLOCKSIZE) { if (len >= SBLOCKSIZE) bl = NULL; else /* Allocate a block of memory. It is not freed until much later. */ bl = (sblock_T *)alloc_clear( (unsigned)(sizeof(sblock_T) + SBLOCKSIZE)); if (bl == NULL) { if (!spin->si_did_emsg) { EMSG(_("E845: Insufficient memory, word list will be incomplete")); spin->si_did_emsg = TRUE; } return NULL; } bl->sb_next = spin->si_blocks; spin->si_blocks = bl; bl->sb_used = 0; ++spin->si_blocks_cnt; } p = bl->sb_data + bl->sb_used; bl->sb_used += (int)len; return p; } /* * Make a copy of a string into memory allocated with getroom(). * Returns NULL when out of memory. */ static char_u * getroom_save(spin, s) spellinfo_T *spin; char_u *s; { char_u *sc; sc = (char_u *)getroom(spin, STRLEN(s) + 1, FALSE); if (sc != NULL) STRCPY(sc, s); return sc; } /* * Free the list of allocated sblock_T. */ static void free_blocks(bl) sblock_T *bl; { sblock_T *next; while (bl != NULL) { next = bl->sb_next; vim_free(bl); bl = next; } } /* * Allocate the root of a word tree. * Returns NULL when out of memory. */ static wordnode_T * wordtree_alloc(spin) spellinfo_T *spin; { return (wordnode_T *)getroom(spin, sizeof(wordnode_T), TRUE); } /* * Store a word in the tree(s). * Always store it in the case-folded tree. For a keep-case word this is * useful when the word can also be used with all caps (no WF_FIXCAP flag) and * used to find suggestions. * For a keep-case word also store it in the keep-case tree. * When "pfxlist" is not NULL store the word for each postponed prefix ID and * compound flag. */ static int store_word(spin, word, flags, region, pfxlist, need_affix) spellinfo_T *spin; char_u *word; int flags; /* extra flags, WF_BANNED */ int region; /* supported region(s) */ char_u *pfxlist; /* list of prefix IDs or NULL */ int need_affix; /* only store word with affix ID */ { int len = (int)STRLEN(word); int ct = captype(word, word + len); char_u foldword[MAXWLEN]; int res = OK; char_u *p; (void)spell_casefold(word, len, foldword, MAXWLEN); for (p = pfxlist; res == OK; ++p) { if (!need_affix || (p != NULL && *p != NUL)) res = tree_add_word(spin, foldword, spin->si_foldroot, ct | flags, region, p == NULL ? 0 : *p); if (p == NULL || *p == NUL) break; } ++spin->si_foldwcount; if (res == OK && (ct == WF_KEEPCAP || (flags & WF_KEEPCAP))) { for (p = pfxlist; res == OK; ++p) { if (!need_affix || (p != NULL && *p != NUL)) res = tree_add_word(spin, word, spin->si_keeproot, flags, region, p == NULL ? 0 : *p); if (p == NULL || *p == NUL) break; } ++spin->si_keepwcount; } return res; } /* * Add word "word" to a word tree at "root". * When "flags" < 0 we are adding to the prefix tree where "flags" is used for * "rare" and "region" is the condition nr. * Returns FAIL when out of memory. */ static int tree_add_word(spin, word, root, flags, region, affixID) spellinfo_T *spin; char_u *word; wordnode_T *root; int flags; int region; int affixID; { wordnode_T *node = root; wordnode_T *np; wordnode_T *copyp, **copyprev; wordnode_T **prev = NULL; int i; /* Add each byte of the word to the tree, including the NUL at the end. */ for (i = 0; ; ++i) { /* When there is more than one reference to this node we need to make * a copy, so that we can modify it. Copy the whole list of siblings * (we don't optimize for a partly shared list of siblings). */ if (node != NULL && node->wn_refs > 1) { --node->wn_refs; copyprev = prev; for (copyp = node; copyp != NULL; copyp = copyp->wn_sibling) { /* Allocate a new node and copy the info. */ np = get_wordnode(spin); if (np == NULL) return FAIL; np->wn_child = copyp->wn_child; if (np->wn_child != NULL) ++np->wn_child->wn_refs; /* child gets extra ref */ np->wn_byte = copyp->wn_byte; if (np->wn_byte == NUL) { np->wn_flags = copyp->wn_flags; np->wn_region = copyp->wn_region; np->wn_affixID = copyp->wn_affixID; } /* Link the new node in the list, there will be one ref. */ np->wn_refs = 1; if (copyprev != NULL) *copyprev = np; copyprev = &np->wn_sibling; /* Let "node" point to the head of the copied list. */ if (copyp == node) node = np; } } /* Look for the sibling that has the same character. They are sorted * on byte value, thus stop searching when a sibling is found with a * higher byte value. For zero bytes (end of word) the sorting is * done on flags and then on affixID. */ while (node != NULL && (node->wn_byte < word[i] || (node->wn_byte == NUL && (flags < 0 ? node->wn_affixID < (unsigned)affixID : (node->wn_flags < (unsigned)(flags & WN_MASK) || (node->wn_flags == (flags & WN_MASK) && (spin->si_sugtree ? (node->wn_region & 0xffff) < region : node->wn_affixID < (unsigned)affixID))))))) { prev = &node->wn_sibling; node = *prev; } if (node == NULL || node->wn_byte != word[i] || (word[i] == NUL && (flags < 0 || spin->si_sugtree || node->wn_flags != (flags & WN_MASK) || node->wn_affixID != affixID))) { /* Allocate a new node. */ np = get_wordnode(spin); if (np == NULL) return FAIL; np->wn_byte = word[i]; /* If "node" is NULL this is a new child or the end of the sibling * list: ref count is one. Otherwise use ref count of sibling and * make ref count of sibling one (matters when inserting in front * of the list of siblings). */ if (node == NULL) np->wn_refs = 1; else { np->wn_refs = node->wn_refs; node->wn_refs = 1; } if (prev != NULL) *prev = np; np->wn_sibling = node; node = np; } if (word[i] == NUL) { node->wn_flags = flags; node->wn_region |= region; node->wn_affixID = affixID; break; } prev = &node->wn_child; node = *prev; } #ifdef SPELL_PRINTTREE smsg("Added \"%s\"", word); spell_print_tree(root->wn_sibling); #endif /* count nr of words added since last message */ ++spin->si_msg_count; if (spin->si_compress_cnt > 1) { if (--spin->si_compress_cnt == 1) /* Did enough words to lower the block count limit. */ spin->si_blocks_cnt += compress_inc; } /* * When we have allocated lots of memory we need to compress the word tree * to free up some room. But compression is slow, and we might actually * need that room, thus only compress in the following situations: * 1. When not compressed before (si_compress_cnt == 0): when using * "compress_start" blocks. * 2. When compressed before and used "compress_inc" blocks before * adding "compress_added" words (si_compress_cnt > 1). * 3. When compressed before, added "compress_added" words * (si_compress_cnt == 1) and the number of free nodes drops below the * maximum word length. */ #ifndef SPELL_PRINTTREE if (spin->si_compress_cnt == 1 ? spin->si_free_count < MAXWLEN : spin->si_blocks_cnt >= compress_start) #endif { /* Decrement the block counter. The effect is that we compress again * when the freed up room has been used and another "compress_inc" * blocks have been allocated. Unless "compress_added" words have * been added, then the limit is put back again. */ spin->si_blocks_cnt -= compress_inc; spin->si_compress_cnt = compress_added; if (spin->si_verbose) { msg_start(); msg_puts((char_u *)_(msg_compressing)); msg_clr_eos(); msg_didout = FALSE; msg_col = 0; out_flush(); } /* Compress both trees. Either they both have many nodes, which makes * compression useful, or one of them is small, which means * compression goes fast. But when filling the souldfold word tree * there is no keep-case tree. */ wordtree_compress(spin, spin->si_foldroot); if (affixID >= 0) wordtree_compress(spin, spin->si_keeproot); } return OK; } /* * Check the 'mkspellmem' option. Return FAIL if it's wrong. * Sets "sps_flags". */ int spell_check_msm() { char_u *p = p_msm; long start = 0; long incr = 0; long added = 0; if (!VIM_ISDIGIT(*p)) return FAIL; /* block count = (value * 1024) / SBLOCKSIZE (but avoid overflow)*/ start = (getdigits(&p) * 10) / (SBLOCKSIZE / 102); if (*p != ',') return FAIL; ++p; if (!VIM_ISDIGIT(*p)) return FAIL; incr = (getdigits(&p) * 102) / (SBLOCKSIZE / 10); if (*p != ',') return FAIL; ++p; if (!VIM_ISDIGIT(*p)) return FAIL; added = getdigits(&p) * 1024; if (*p != NUL) return FAIL; if (start == 0 || incr == 0 || added == 0 || incr > start) return FAIL; compress_start = start; compress_inc = incr; compress_added = added; return OK; } /* * Get a wordnode_T, either from the list of previously freed nodes or * allocate a new one. * Returns NULL when out of memory. */ static wordnode_T * get_wordnode(spin) spellinfo_T *spin; { wordnode_T *n; if (spin->si_first_free == NULL) n = (wordnode_T *)getroom(spin, sizeof(wordnode_T), TRUE); else { n = spin->si_first_free; spin->si_first_free = n->wn_child; vim_memset(n, 0, sizeof(wordnode_T)); --spin->si_free_count; } #ifdef SPELL_PRINTTREE if (n != NULL) n->wn_nr = ++spin->si_wordnode_nr; #endif return n; } /* * Decrement the reference count on a node (which is the head of a list of * siblings). If the reference count becomes zero free the node and its * siblings. * Returns the number of nodes actually freed. */ static int deref_wordnode(spin, node) spellinfo_T *spin; wordnode_T *node; { wordnode_T *np; int cnt = 0; if (--node->wn_refs == 0) { for (np = node; np != NULL; np = np->wn_sibling) { if (np->wn_child != NULL) cnt += deref_wordnode(spin, np->wn_child); free_wordnode(spin, np); ++cnt; } ++cnt; /* length field */ } return cnt; } /* * Free a wordnode_T for re-use later. * Only the "wn_child" field becomes invalid. */ static void free_wordnode(spin, n) spellinfo_T *spin; wordnode_T *n; { n->wn_child = spin->si_first_free; spin->si_first_free = n; ++spin->si_free_count; } /* * Compress a tree: find tails that are identical and can be shared. */ static void wordtree_compress(spin, root) spellinfo_T *spin; wordnode_T *root; { hashtab_T ht; int n; int tot = 0; int perc; /* Skip the root itself, it's not actually used. The first sibling is the * start of the tree. */ if (root->wn_sibling != NULL) { hash_init(&ht); n = node_compress(spin, root->wn_sibling, &ht, &tot); #ifndef SPELL_PRINTTREE if (spin->si_verbose || p_verbose > 2) #endif { if (tot > 1000000) perc = (tot - n) / (tot / 100); else if (tot == 0) perc = 0; else perc = (tot - n) * 100 / tot; vim_snprintf((char *)IObuff, IOSIZE, _("Compressed %d of %d nodes; %d (%d%%) remaining"), n, tot, tot - n, perc); spell_message(spin, IObuff); } #ifdef SPELL_PRINTTREE spell_print_tree(root->wn_sibling); #endif hash_clear(&ht); } } /* * Compress a node, its siblings and its children, depth first. * Returns the number of compressed nodes. */ static int node_compress(spin, node, ht, tot) spellinfo_T *spin; wordnode_T *node; hashtab_T *ht; int *tot; /* total count of nodes before compressing, incremented while going through the tree */ { wordnode_T *np; wordnode_T *tp; wordnode_T *child; hash_T hash; hashitem_T *hi; int len = 0; unsigned nr, n; int compressed = 0; /* * Go through the list of siblings. Compress each child and then try * finding an identical child to replace it. * Note that with "child" we mean not just the node that is pointed to, * but the whole list of siblings of which the child node is the first. */ for (np = node; np != NULL && !got_int; np = np->wn_sibling) { ++len; if ((child = np->wn_child) != NULL) { /* Compress the child first. This fills hashkey. */ compressed += node_compress(spin, child, ht, tot); /* Try to find an identical child. */ hash = hash_hash(child->wn_u1.hashkey); hi = hash_lookup(ht, child->wn_u1.hashkey, hash); if (!HASHITEM_EMPTY(hi)) { /* There are children we encountered before with a hash value * identical to the current child. Now check if there is one * that is really identical. */ for (tp = HI2WN(hi); tp != NULL; tp = tp->wn_u2.next) if (node_equal(child, tp)) { /* Found one! Now use that child in place of the * current one. This means the current child and all * its siblings is unlinked from the tree. */ ++tp->wn_refs; compressed += deref_wordnode(spin, child); np->wn_child = tp; break; } if (tp == NULL) { /* No other child with this hash value equals the child of * the node, add it to the linked list after the first * item. */ tp = HI2WN(hi); child->wn_u2.next = tp->wn_u2.next; tp->wn_u2.next = child; } } else /* No other child has this hash value, add it to the * hashtable. */ hash_add_item(ht, hi, child->wn_u1.hashkey, hash); } } *tot += len + 1; /* add one for the node that stores the length */ /* * Make a hash key for the node and its siblings, so that we can quickly * find a lookalike node. This must be done after compressing the sibling * list, otherwise the hash key would become invalid by the compression. */ node->wn_u1.hashkey[0] = len; nr = 0; for (np = node; np != NULL; np = np->wn_sibling) { if (np->wn_byte == NUL) /* end node: use wn_flags, wn_region and wn_affixID */ n = np->wn_flags + (np->wn_region << 8) + (np->wn_affixID << 16); else /* byte node: use the byte value and the child pointer */ n = (unsigned)(np->wn_byte + ((long_u)np->wn_child << 8)); nr = nr * 101 + n; } /* Avoid NUL bytes, it terminates the hash key. */ n = nr & 0xff; node->wn_u1.hashkey[1] = n == 0 ? 1 : n; n = (nr >> 8) & 0xff; node->wn_u1.hashkey[2] = n == 0 ? 1 : n; n = (nr >> 16) & 0xff; node->wn_u1.hashkey[3] = n == 0 ? 1 : n; n = (nr >> 24) & 0xff; node->wn_u1.hashkey[4] = n == 0 ? 1 : n; node->wn_u1.hashkey[5] = NUL; /* Check for CTRL-C pressed now and then. */ fast_breakcheck(); return compressed; } /* * Return TRUE when two nodes have identical siblings and children. */ static int node_equal(n1, n2) wordnode_T *n1; wordnode_T *n2; { wordnode_T *p1; wordnode_T *p2; for (p1 = n1, p2 = n2; p1 != NULL && p2 != NULL; p1 = p1->wn_sibling, p2 = p2->wn_sibling) if (p1->wn_byte != p2->wn_byte || (p1->wn_byte == NUL ? (p1->wn_flags != p2->wn_flags || p1->wn_region != p2->wn_region || p1->wn_affixID != p2->wn_affixID) : (p1->wn_child != p2->wn_child))) break; return p1 == NULL && p2 == NULL; } static int #ifdef __BORLANDC__ _RTLENTRYF #endif rep_compare __ARGS((const void *s1, const void *s2)); /* * Function given to qsort() to sort the REP items on "from" string. */ static int #ifdef __BORLANDC__ _RTLENTRYF #endif rep_compare(s1, s2) const void *s1; const void *s2; { fromto_T *p1 = (fromto_T *)s1; fromto_T *p2 = (fromto_T *)s2; return STRCMP(p1->ft_from, p2->ft_from); } /* * Write the Vim .spl file "fname". * Return FAIL or OK; */ static int write_vim_spell(spin, fname) spellinfo_T *spin; char_u *fname; { FILE *fd; int regionmask; int round; wordnode_T *tree; int nodecount; int i; int l; garray_T *gap; fromto_T *ftp; char_u *p; int rr; int retval = OK; size_t fwv = 1; /* collect return value of fwrite() to avoid warnings from picky compiler */ fd = mch_fopen((char *)fname, "w"); if (fd == NULL) { EMSG2(_(e_notopen), fname); return FAIL; } /*

: */ /* */ fwv &= fwrite(VIMSPELLMAGIC, VIMSPELLMAGICL, (size_t)1, fd); if (fwv != (size_t)1) /* Catch first write error, don't try writing more. */ goto theend; putc(VIMSPELLVERSION, fd); /* */ /* * :

... */ /* SN_INFO: */ if (spin->si_info != NULL) { putc(SN_INFO, fd); /* */ putc(0, fd); /* */ i = (int)STRLEN(spin->si_info); put_bytes(fd, (long_u)i, 4); /* */ fwv &= fwrite(spin->si_info, (size_t)i, (size_t)1, fd); /* */ } /* SN_REGION: ... * Write the region names only if there is more than one. */ if (spin->si_region_count > 1) { putc(SN_REGION, fd); /* */ putc(SNF_REQUIRED, fd); /* */ l = spin->si_region_count * 2; put_bytes(fd, (long_u)l, 4); /* */ fwv &= fwrite(spin->si_region_name, (size_t)l, (size_t)1, fd); /* ... */ regionmask = (1 << spin->si_region_count) - 1; } else regionmask = 0; /* SN_CHARFLAGS: * * The table with character flags and the table for case folding. * This makes sure the same characters are recognized as word characters * when generating an when using a spell file. * Skip this for ASCII, the table may conflict with the one used for * 'encoding'. * Also skip this for an .add.spl file, the main spell file must contain * the table (avoids that it conflicts). File is shorter too. */ if (!spin->si_ascii && !spin->si_add) { char_u folchars[128 * 8]; int flags; putc(SN_CHARFLAGS, fd); /* */ putc(SNF_REQUIRED, fd); /* */ /* Form the string first, we need to know its length. */ l = 0; for (i = 128; i < 256; ++i) { #ifdef FEAT_MBYTE if (has_mbyte) l += mb_char2bytes(spelltab.st_fold[i], folchars + l); else #endif folchars[l++] = spelltab.st_fold[i]; } put_bytes(fd, (long_u)(1 + 128 + 2 + l), 4); /* */ fputc(128, fd); /* */ for (i = 128; i < 256; ++i) { flags = 0; if (spelltab.st_isw[i]) flags |= CF_WORD; if (spelltab.st_isu[i]) flags |= CF_UPPER; fputc(flags, fd); /* */ } put_bytes(fd, (long_u)l, 2); /* */ fwv &= fwrite(folchars, (size_t)l, (size_t)1, fd); /* */ } /* SN_MIDWORD: */ if (spin->si_midword != NULL) { putc(SN_MIDWORD, fd); /* */ putc(SNF_REQUIRED, fd); /* */ i = (int)STRLEN(spin->si_midword); put_bytes(fd, (long_u)i, 4); /* */ fwv &= fwrite(spin->si_midword, (size_t)i, (size_t)1, fd); /* */ } /* SN_PREFCOND: ... */ if (spin->si_prefcond.ga_len > 0) { putc(SN_PREFCOND, fd); /* */ putc(SNF_REQUIRED, fd); /* */ l = write_spell_prefcond(NULL, &spin->si_prefcond); put_bytes(fd, (long_u)l, 4); /* */ write_spell_prefcond(fd, &spin->si_prefcond); } /* SN_REP: ... * SN_SAL: ... * SN_REPSAL: ... */ /* round 1: SN_REP section * round 2: SN_SAL section (unless SN_SOFO is used) * round 3: SN_REPSAL section */ for (round = 1; round <= 3; ++round) { if (round == 1) gap = &spin->si_rep; else if (round == 2) { /* Don't write SN_SAL when using a SN_SOFO section */ if (spin->si_sofofr != NULL && spin->si_sofoto != NULL) continue; gap = &spin->si_sal; } else gap = &spin->si_repsal; /* Don't write the section if there are no items. */ if (gap->ga_len == 0) continue; /* Sort the REP/REPSAL items. */ if (round != 2) qsort(gap->ga_data, (size_t)gap->ga_len, sizeof(fromto_T), rep_compare); i = round == 1 ? SN_REP : (round == 2 ? SN_SAL : SN_REPSAL); putc(i, fd); /* */ /* This is for making suggestions, section is not required. */ putc(0, fd); /* */ /* Compute the length of what follows. */ l = 2; /* count or */ for (i = 0; i < gap->ga_len; ++i) { ftp = &((fromto_T *)gap->ga_data)[i]; l += 1 + (int)STRLEN(ftp->ft_from); /* count <*fromlen> and <*from> */ l += 1 + (int)STRLEN(ftp->ft_to); /* count <*tolen> and <*to> */ } if (round == 2) ++l; /* count */ put_bytes(fd, (long_u)l, 4); /* */ if (round == 2) { i = 0; if (spin->si_followup) i |= SAL_F0LLOWUP; if (spin->si_collapse) i |= SAL_COLLAPSE; if (spin->si_rem_accents) i |= SAL_REM_ACCENTS; putc(i, fd); /* */ } put_bytes(fd, (long_u)gap->ga_len, 2); /* or */ for (i = 0; i < gap->ga_len; ++i) { /* : */ /* : */ ftp = &((fromto_T *)gap->ga_data)[i]; for (rr = 1; rr <= 2; ++rr) { p = rr == 1 ? ftp->ft_from : ftp->ft_to; l = (int)STRLEN(p); putc(l, fd); if (l > 0) fwv &= fwrite(p, l, (size_t)1, fd); } } } /* SN_SOFO: * This is for making suggestions, section is not required. */ if (spin->si_sofofr != NULL && spin->si_sofoto != NULL) { putc(SN_SOFO, fd); /* */ putc(0, fd); /* */ l = (int)STRLEN(spin->si_sofofr); put_bytes(fd, (long_u)(l + STRLEN(spin->si_sofoto) + 4), 4); /* */ put_bytes(fd, (long_u)l, 2); /* */ fwv &= fwrite(spin->si_sofofr, l, (size_t)1, fd); /* */ l = (int)STRLEN(spin->si_sofoto); put_bytes(fd, (long_u)l, 2); /* */ fwv &= fwrite(spin->si_sofoto, l, (size_t)1, fd); /* */ } /* SN_WORDS: ... * This is for making suggestions, section is not required. */ if (spin->si_commonwords.ht_used > 0) { putc(SN_WORDS, fd); /* */ putc(0, fd); /* */ /* round 1: count the bytes * round 2: write the bytes */ for (round = 1; round <= 2; ++round) { int todo; int len = 0; hashitem_T *hi; todo = (int)spin->si_commonwords.ht_used; for (hi = spin->si_commonwords.ht_array; todo > 0; ++hi) if (!HASHITEM_EMPTY(hi)) { l = (int)STRLEN(hi->hi_key) + 1; len += l; if (round == 2) /* */ fwv &= fwrite(hi->hi_key, (size_t)l, (size_t)1, fd); --todo; } if (round == 1) put_bytes(fd, (long_u)len, 4); /* */ } } /* SN_MAP: * This is for making suggestions, section is not required. */ if (spin->si_map.ga_len > 0) { putc(SN_MAP, fd); /* */ putc(0, fd); /* */ l = spin->si_map.ga_len; put_bytes(fd, (long_u)l, 4); /* */ fwv &= fwrite(spin->si_map.ga_data, (size_t)l, (size_t)1, fd); /* */ } /* SN_SUGFILE: * This is used to notify that a .sug file may be available and at the * same time allows for checking that a .sug file that is found matches * with this .spl file. That's because the word numbers must be exactly * right. */ if (!spin->si_nosugfile && (spin->si_sal.ga_len > 0 || (spin->si_sofofr != NULL && spin->si_sofoto != NULL))) { putc(SN_SUGFILE, fd); /* */ putc(0, fd); /* */ put_bytes(fd, (long_u)8, 4); /* */ /* Set si_sugtime and write it to the file. */ spin->si_sugtime = time(NULL); put_time(fd, spin->si_sugtime); /* */ } /* SN_NOSPLITSUGS: nothing * This is used to notify that no suggestions with word splits are to be * made. */ if (spin->si_nosplitsugs) { putc(SN_NOSPLITSUGS, fd); /* */ putc(0, fd); /* */ put_bytes(fd, (long_u)0, 4); /* */ } /* SN_COMPOUND: compound info. * We don't mark it required, when not supported all compound words will * be bad words. */ if (spin->si_compflags != NULL) { putc(SN_COMPOUND, fd); /* */ putc(0, fd); /* */ l = (int)STRLEN(spin->si_compflags); for (i = 0; i < spin->si_comppat.ga_len; ++i) l += (int)STRLEN(((char_u **)(spin->si_comppat.ga_data))[i]) + 1; put_bytes(fd, (long_u)(l + 7), 4); /* */ putc(spin->si_compmax, fd); /* */ putc(spin->si_compminlen, fd); /* */ putc(spin->si_compsylmax, fd); /*