fat find: encoding fixesjed/python3-win32-compat

Use binary IO and byte strings in more places. Some formatting is more
cumbersome, but we really want to use raw bytes when interacting with Git
because that's what it provides us and that's what it expects. Decode when
writing to stdout because that is expected to be read by a human.

1 files changed, 13 insertions, 12 deletions

diff --git a/git-fat b/git-fat
index 8d0536c..d17b5e4 100755
--- a/git-fat
+++ b/git-fat
@@ -68,14 +68,14 @@ def difftreez_reader(input):
     :oldmode newmode oldsha1 newsha1 modflag\0filename\0:oldmode newmode ...
     """
     buffer = []
-    partial = ''
+    partial = b''
     while True:
         newread = input.read(BLOCK_SIZE)
         if not newread:
             break
         partial += newread
         while True:
-            head, sep, partial = partial.partition('\0')
+            head, sep, partial = partial.partition(b'\0')
             if not sep:
                 partial = head
                 break
@@ -387,7 +387,7 @@ class GitFat(object):
             This truncates to one hash per line.
             """
             for line in input:
-                output.write(line[:40] + '\n')
+                output.write(line[:40] + b'\n')
             output.close()
         revlist = subprocess.Popen(['git', 'rev-list', '--all', '--objects'], stdout=subprocess.PIPE, bufsize=-1)
         objcheck = subprocess.Popen(['git', 'cat-file', '--batch-check'], stdin=subprocess.PIPE, stdout=subprocess.PIPE, bufsize=-1)
@@ -397,7 +397,7 @@ class GitFat(object):
         # Build dict with the sizes of all large blobs
         for line in objcheck.stdout:
             objhash, blob, size = line.split()
-            if blob != 'blob':
+            if blob != b'blob':
                 continue
             size = int(size)
             numblobs += 1
@@ -421,6 +421,7 @@ class GitFat(object):
                                     stdin=revlist.stdout, stdout=subprocess.PIPE)
         for newblob, modflag, path in difftreez_reader(difftree.stdout):
             bsize = blobsizes.get(newblob)
+            path = path.decode(sys.getfilesystemencoding())
             if bsize:                     # We care about this blob
                 pathsizes[path].add(bsize)
         time1 = time.time()
@@ -433,7 +434,7 @@ class GitFat(object):
     def cmd_index_filter(self, args):
         # FIXME: Need input validation here
         manage_gitattributes = '--manage-gitattributes' in args
-        filelist = set(f.strip() for f in open(args[0]).readlines())
+        filelist = set(f.strip() for f in open(args[0], 'rb').readlines())
         lsfiles = subprocess.Popen(['git', 'ls-files', '-s'], stdout=subprocess.PIPE)
         updateindex = subprocess.Popen(['git', 'update-index', '--index-info'], stdin=subprocess.PIPE)
         for line in lsfiles.stdout:
@@ -446,7 +447,7 @@ class GitFat(object):
             # This file will contain the hash of the cleaned object
             hashfile = os.path.join(self.gitdir, b'fat', b'index-filter', blobhash)
             try:
-                cleanedobj = open(hashfile).read().rstrip()
+                cleanedobj = open(hashfile, 'wb').read().rstrip()
             except IOError:
                 catfile = subprocess.Popen(['git', 'cat-file', 'blob', blobhash], stdout=subprocess.PIPE)
                 hashobject = subprocess.Popen(['git', 'hash-object', '-w', '--stdin'], stdin=subprocess.PIPE, stdout=subprocess.PIPE)
@@ -460,19 +461,19 @@ class GitFat(object):
                 hashobject.wait()
                 filterclean.join()
                 mkdir_p(os.path.dirname(hashfile))
-                open(hashfile, 'w').write(cleanedobj + '\n')
-            updateindex.stdin.write('%s %s %s\t%s\n' % (mode, cleanedobj, stageno, filename))
+                open(hashfile, 'wb').write(cleanedobj + b'\n')
+            updateindex.stdin.write(mode + b' ' + cleanedobj + b' ' + stageno + b'\t' + filename + b'\n')
         if manage_gitattributes:
             try:
                 mode, blobsha1, stageno, filename = subprocess.check_output(['git', 'ls-files', '-s', '.gitattributes']).split()
                 gitattributes_lines = subprocess.check_output(['git', 'cat-file', 'blob', blobsha1]).splitlines()
             except ValueError:  # Nothing to unpack, thus no file
-                mode, stageno = '100644', '0'
+                mode, stageno = b'100644', b'0'
                 gitattributes_lines = []
-            gitattributes_extra = ['%s filter=fat -text' % line.split()[0] for line in filelist]
+            gitattributes_extra = [line.split()[0] + b' filter=fat -text' for line in filelist]
             hashobject = subprocess.Popen(['git', 'hash-object', '-w', '--stdin'], stdin=subprocess.PIPE, stdout=subprocess.PIPE)
-            stdout, stderr = hashobject.communicate('\n'.join(gitattributes_lines + gitattributes_extra) + '\n')
-            updateindex.stdin.write('%s %s %s\t%s\n' % (mode, stdout.strip(), stageno, '.gitattributes'))
+            stdout, stderr = hashobject.communicate(b'\n'.join(gitattributes_lines + gitattributes_extra) + b'\n')
+            updateindex.stdin.write(mode + b' ' + stdout.strip() + b' ' + stageno + b'\t.gitattributes\n')
         updateindex.stdin.close()
         lsfiles.wait()
         updateindex.wait()