moved all relevant files into the gitdb folder.

Submodule relinked to point to new github location, and moved as well

4 files changed, 360 insertions, 0 deletions

diff --git a/gitdb/test/performance/lib.py b/gitdb/test/performance/lib.py
new file mode 100644
index 0000000..45e0ca5
--- /dev/null
+++ b/gitdb/test/performance/lib.py
@@ -0,0 +1,50 @@
+"""Contains library functions"""
+import os
+from gitdb.test.lib import *
+import shutil
+import tempfile
+
+
+#{ Invvariants
+k_env_git_repo = "GITDB_TEST_GIT_REPO_BASE"
+#} END invariants
+
+
+#{ Utilities
+def resolve_or_fail(env_var):
+	""":return: resolved environment variable or raise EnvironmentError"""
+	try:
+		return os.environ[env_var]
+	except KeyError:
+		raise EnvironmentError("Please set the %r envrionment variable and retry" % env_var)
+	# END exception handling
+
+#} END utilities
+
+
+#{ Base Classes 
+
+class TestBigRepoR(TestBase):
+	"""TestCase providing access to readonly 'big' repositories using the following 
+	member variables:
+	
+	* gitrepopath
+	
+	 * read-only base path of the git source repository, i.e. .../git/.git"""
+	 
+	#{ Invariants
+	head_sha_2k = '235d521da60e4699e5bd59ac658b5b48bd76ddca'
+	head_sha_50 = '32347c375250fd470973a5d76185cac718955fd5'
+	#} END invariants 
+	
+	@classmethod
+	def setUpAll(cls):
+		try:
+			super(TestBigRepoR, cls).setUpAll()
+		except AttributeError:
+			pass
+		cls.gitrepopath = resolve_or_fail(k_env_git_repo)
+		assert cls.gitrepopath.endswith('.git')
+
+		
+#} END base classes
diff --git a/gitdb/test/performance/test_pack.py b/gitdb/test/performance/test_pack.py
new file mode 100644
index 0000000..32890dc
--- /dev/null
+++ b/gitdb/test/performance/test_pack.py
@@ -0,0 +1,86 @@
+"""Performance tests for object store"""
+from lib import (
+	TestBigRepoR 
+	)
+
+from gitdb.exc import UnsupportedOperation
+from gitdb.db.pack import PackedDB
+
+import sys
+import os
+from time import time
+import random
+
+class TestPackedDBPerformance(TestBigRepoR):
+	
+	def _test_pack_random_access(self):
+		pdb = PackedDB(os.path.join(self.gitrepopath, "objects/pack"))
+		
+		# sha lookup
+		st = time()
+		sha_list = list(pdb.sha_iter())
+		elapsed = time() - st
+		ns = len(sha_list)
+		print >> sys.stderr, "PDB: looked up %i shas by index in %f s ( %f shas/s )" % (ns, elapsed, ns / elapsed)
+		
+		# sha lookup: best-case and worst case access
+		pdb_pack_info = pdb._pack_info
+		# END shuffle shas
+		st = time()
+		for sha in sha_list:
+			pdb_pack_info(sha)
+		# END for each sha to look up
+		elapsed = time() - st
+		
+		# discard cache
+		del(pdb._entities)
+		pdb.entities()
+		print >> sys.stderr, "PDB: looked up %i sha in %i packs in %f s ( %f shas/s )" % (ns, len(pdb.entities()), elapsed, ns / elapsed)
+		# END for each random mode
+		
+		# query info and streams only
+		max_items = 10000			# can wait longer when testing memory
+		for pdb_fun in (pdb.info, pdb.stream):
+			st = time()
+			for sha in sha_list[:max_items]:
+				pdb_fun(sha)
+			elapsed = time() - st
+			print >> sys.stderr, "PDB: Obtained %i object %s by sha in %f s ( %f items/s )" % (max_items, pdb_fun.__name__.upper(), elapsed, max_items / elapsed)
+		# END for each function
+		
+		# retrieve stream and read all
+		max_items = 5000
+		pdb_stream = pdb.stream
+		total_size = 0
+		st = time()
+		for sha in sha_list[:max_items]:
+			stream = pdb_stream(sha)
+			stream.read()
+			total_size += stream.size
+		elapsed = time() - st
+		total_kib = total_size / 1000
+		print >> sys.stderr, "PDB: Obtained %i streams by sha and read all bytes totallying %i KiB ( %f KiB / s ) in %f s ( %f streams/s )" % (max_items, total_kib, total_kib/elapsed , elapsed, max_items / elapsed)
+		
+	def test_correctness(self):
+		pdb = PackedDB(os.path.join(self.gitrepopath, "objects/pack"))
+		# disabled for now as it used to work perfectly, checking big repositories takes a long time
+		print >> sys.stderr, "Endurance run: verify streaming of objects (crc and sha)"
+		for crc in range(2):
+			count = 0
+			st = time()
+			for entity in pdb.entities():
+				pack_verify = entity.is_valid_stream
+				sha_by_index = entity.index().sha
+				for index in xrange(entity.index().size()):
+					try:
+						assert pack_verify(sha_by_index(index), use_crc=crc)
+						count += 1
+					except UnsupportedOperation:
+						pass
+					# END ignore old indices
+				# END for each index
+			# END for each entity
+			elapsed = time() - st
+			print >> sys.stderr, "PDB: verified %i objects (crc=%i) in %f s ( %f objects/s )" % (count, crc, elapsed, count / elapsed)
+		# END for each verify mode
+		
diff --git a/gitdb/test/performance/test_pack_streaming.py b/gitdb/test/performance/test_pack_streaming.py
new file mode 100644
index 0000000..4d47cdf
--- /dev/null
+++ b/gitdb/test/performance/test_pack_streaming.py
@@ -0,0 +1,33 @@
+"""Specific test for pack streams only"""
+from lib import (
+	TestBigRepoR 
+	)
+
+from gitdb.db.pack import PackedDB
+
+import os
+import sys
+from time import time
+
+class TestPackStreamingPerformance(TestBigRepoR):
+	
+	def test_stream_reading(self):
+		pdb = PackedDB(os.path.join(self.gitrepopath, "objects/pack"))
+		
+		# streaming only, meant for --with-profile runs
+		ni = 5000
+		count = 0
+		pdb_stream = pdb.stream
+		total_size = 0
+		st = time()
+		for sha in pdb.sha_iter():
+			if count == ni:
+				break
+			stream = pdb_stream(sha)
+			stream.read()
+			total_size += stream.size
+			count += 1
+		elapsed = time() - st
+		total_kib = total_size / 1000
+		print >> sys.stderr, "PDB Streaming: Got %i streams by sha and read all bytes totallying %i KiB ( %f KiB / s ) in %f s ( %f streams/s )" % (ni, total_kib, total_kib/elapsed , elapsed, ni / elapsed)
+		
diff --git a/gitdb/test/performance/test_stream.py b/gitdb/test/performance/test_stream.py
new file mode 100644
index 0000000..1afc1a1
--- /dev/null
+++ b/gitdb/test/performance/test_stream.py
@@ -0,0 +1,191 @@
+"""Performance data streaming performance"""
+from lib import TestBigRepoR
+from gitdb.db import *
+from gitdb.base import *
+from gitdb.stream import *
+from gitdb.util import (
+							pool,
+							bin_to_hex
+						)
+from gitdb.typ import str_blob_type
+from gitdb.fun import chunk_size
+
+from async import ( 
+	IteratorReader, 
+	ChannelThreadTask,
+	)
+
+from cStringIO import StringIO
+from time import time
+import os
+import sys
+import stat
+import subprocess
+
+
+from lib import (
+	TestBigRepoR,
+	make_memory_file,
+	with_rw_directory
+	)
+
+
+#{ Utilities
+def read_chunked_stream(stream):
+	total = 0
+	while True:
+		chunk = stream.read(chunk_size)
+		total += len(chunk)
+		if len(chunk) < chunk_size:
+			break
+	# END read stream loop
+	assert total == stream.size
+	return stream
+	
+	
+class TestStreamReader(ChannelThreadTask):
+	"""Expects input streams and reads them in chunks. It will read one at a time, 
+	requireing a queue chunk of size 1"""
+	def __init__(self, *args):
+		super(TestStreamReader, self).__init__(*args)
+		self.fun = read_chunked_stream
+		self.max_chunksize = 1
+	
+
+#} END utilities
+
+class TestObjDBPerformance(TestBigRepoR):
+	
+	large_data_size_bytes = 1000*1000*50		# some MiB should do it
+	moderate_data_size_bytes = 1000*1000*1		# just 1 MiB
+	
+	@with_rw_directory
+	def test_large_data_streaming(self, path):
+		ldb = LooseObjectDB(path)
+		string_ios = list()			# list of streams we previously created
+		
+		# serial mode 
+		for randomize in range(2):
+			desc = (randomize and 'random ') or ''
+			print >> sys.stderr, "Creating %s data ..." % desc
+			st = time()
+			size, stream = make_memory_file(self.large_data_size_bytes, randomize)
+			elapsed = time() - st
+			print >> sys.stderr, "Done (in %f s)" % elapsed
+			string_ios.append(stream)
+			
+			# writing - due to the compression it will seem faster than it is 
+			st = time()
+			sha = ldb.store(IStream('blob', size, stream)).binsha
+			elapsed_add = time() - st
+			assert ldb.has_object(sha)
+			db_file = ldb.readable_db_object_path(bin_to_hex(sha))
+			fsize_kib = os.path.getsize(db_file) / 1000
+			
+			
+			size_kib = size / 1000
+			print >> sys.stderr, "Added %i KiB (filesize = %i KiB) of %s data to loose odb in %f s ( %f Write KiB / s)" % (size_kib, fsize_kib, desc, elapsed_add, size_kib / elapsed_add)
+			
+			# reading all at once
+			st = time()
+			ostream = ldb.stream(sha)
+			shadata = ostream.read()
+			elapsed_readall = time() - st
+			
+			stream.seek(0)
+			assert shadata == stream.getvalue()
+			print >> sys.stderr, "Read %i KiB of %s data at once from loose odb in %f s ( %f Read KiB / s)" % (size_kib, desc, elapsed_readall, size_kib / elapsed_readall)
+			
+			
+			# reading in chunks of 1 MiB
+			cs = 512*1000
+			chunks = list()
+			st = time()
+			ostream = ldb.stream(sha)
+			while True:
+				data = ostream.read(cs)
+				chunks.append(data)
+				if len(data) < cs:
+					break
+			# END read in chunks
+			elapsed_readchunks = time() - st
+			
+			stream.seek(0)
+			assert ''.join(chunks) == stream.getvalue()
+			
+			cs_kib = cs / 1000
+			print >> sys.stderr, "Read %i KiB of %s data in %i KiB chunks from loose odb in %f s ( %f Read KiB / s)" % (size_kib, desc, cs_kib, elapsed_readchunks, size_kib / elapsed_readchunks)
+			
+			# del db file so we keep something to do
+			os.remove(db_file)
+		# END for each randomization factor
+		
+		
+		# multi-threaded mode
+		# want two, should be supported by most of todays cpus
+		pool.set_size(2)
+		total_kib = 0
+		nsios = len(string_ios)
+		for stream in string_ios:
+			stream.seek(0)
+			total_kib += len(stream.getvalue()) / 1000
+		# END rewind
+		
+		def istream_iter():
+			for stream in string_ios:
+				stream.seek(0)
+				yield IStream(str_blob_type, len(stream.getvalue()), stream)
+			# END for each stream
+		# END util
+		
+		# write multiple objects at once, involving concurrent compression
+		reader = IteratorReader(istream_iter())
+		istream_reader = ldb.store_async(reader)
+		istream_reader.task().max_chunksize = 1
+		
+		st = time()
+		istreams = istream_reader.read(nsios)
+		assert len(istreams) == nsios
+		elapsed = time() - st
+		
+		print >> sys.stderr, "Threads(%i): Compressed %i KiB of data in loose odb in %f s ( %f Write KiB / s)" % (pool.size(), total_kib, elapsed, total_kib / elapsed)
+		
+		# decompress multiple at once, by reading them
+		# chunk size is not important as the stream will not really be decompressed
+		
+		# until its read
+		istream_reader = IteratorReader(iter([ i.binsha for i in istreams ]))
+		ostream_reader = ldb.stream_async(istream_reader)
+		
+		chunk_task = TestStreamReader(ostream_reader, "chunker", None)
+		output_reader = pool.add_task(chunk_task)
+		output_reader.task().max_chunksize = 1
+		
+		st = time()
+		assert len(output_reader.read(nsios)) == nsios
+		elapsed = time() - st
+		
+		print >> sys.stderr, "Threads(%i): Decompressed %i KiB of data in loose odb in %f s ( %f Read KiB / s)" % (pool.size(), total_kib, elapsed, total_kib / elapsed)
+		
+		# store the files, and read them back. For the reading, we use a task 
+		# as well which is chunked into one item per task. Reading all will
+		# very quickly result in two threads handling two bytestreams of 
+		# chained compression/decompression streams
+		reader = IteratorReader(istream_iter())
+		istream_reader = ldb.store_async(reader)
+		istream_reader.task().max_chunksize = 1
+		
+		istream_to_sha = lambda items: [ i.binsha for i in items ]
+		istream_reader.set_post_cb(istream_to_sha)
+		
+		ostream_reader = ldb.stream_async(istream_reader)
+		
+		chunk_task = TestStreamReader(ostream_reader, "chunker", None)
+		output_reader = pool.add_task(chunk_task)
+		output_reader.max_chunksize = 1
+		
+		st = time()
+		assert len(output_reader.read(nsios)) == nsios
+		elapsed = time() - st
+		
+		print >> sys.stderr, "Threads(%i): Compressed and decompressed and read %i KiB of data in loose odb in %f s ( %f Combined KiB / s)" % (pool.size(), total_kib, elapsed, total_kib / elapsed)