diff options
Diffstat (limited to 'git/test/performance/test_streams.py')
| -rw-r--r-- | git/test/performance/test_streams.py | 44 |
1 files changed, 22 insertions, 22 deletions
diff --git a/git/test/performance/test_streams.py b/git/test/performance/test_streams.py index 93e88841..cac53a06 100644 --- a/git/test/performance/test_streams.py +++ b/git/test/performance/test_streams.py @@ -18,16 +18,16 @@ from lib import ( class TestObjDBPerformance(TestBigRepoR): - + large_data_size_bytes = 1000*1000*10 # some MiB should do it moderate_data_size_bytes = 1000*1000*1 # just 1 MiB - + @with_rw_repo('HEAD', bare=True) def test_large_data_streaming(self, rwrepo): # TODO: This part overlaps with the same file in gitdb.test.performance.test_stream # It should be shared if possible ldb = LooseObjectDB(os.path.join(rwrepo.git_dir, 'objects')) - + for randomize in range(2): desc = (randomize and 'random ') or '' print >> sys.stderr, "Creating %s data ..." % desc @@ -35,7 +35,7 @@ class TestObjDBPerformance(TestBigRepoR): size, stream = make_memory_file(self.large_data_size_bytes, randomize) elapsed = time() - st print >> sys.stderr, "Done (in %f s)" % elapsed - + # writing - due to the compression it will seem faster than it is st = time() binsha = ldb.store(IStream('blob', size, stream)).binsha @@ -43,22 +43,22 @@ class TestObjDBPerformance(TestBigRepoR): assert ldb.has_object(binsha) db_file = ldb.readable_db_object_path(bin_to_hex(binsha)) 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(binsha) 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() @@ -71,21 +71,21 @@ class TestObjDBPerformance(TestBigRepoR): 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 git has something to do os.remove(db_file) - + # VS. CGIT ########## # CGIT ! Can using the cgit programs be faster ? proc = rwrepo.git.hash_object('-w', '--stdin', as_process=True, istream=subprocess.PIPE) - + # write file - pump everything in at once to be a fast as possible data = stream.getvalue() # cache it st = time() @@ -96,15 +96,15 @@ class TestObjDBPerformance(TestBigRepoR): gelapsed_add = time() - st del(data) assert gitsha == bin_to_hex(binsha) # we do it the same way, right ? - + # as its the same sha, we reuse our path fsize_kib = os.path.getsize(db_file) / 1000 print >> sys.stderr, "Added %i KiB (filesize = %i KiB) of %s data to using git-hash-object in %f s ( %f Write KiB / s)" % (size_kib, fsize_kib, desc, gelapsed_add, size_kib / gelapsed_add) - + # compare ... print >> sys.stderr, "Git-Python is %f %% faster than git when adding big %s files" % (100.0 - (elapsed_add / gelapsed_add) * 100, desc) - - + + # read all st = time() s, t, size, data = rwrepo.git.get_object_data(gitsha) @@ -113,8 +113,8 @@ class TestObjDBPerformance(TestBigRepoR): # compare print >> sys.stderr, "Git-Python is %f %% faster than git when reading big %sfiles" % (100.0 - (elapsed_readall / gelapsed_readall) * 100, desc) - - + + # read chunks st = time() s, t, size, stream = rwrepo.git.stream_object_data(gitsha) @@ -125,7 +125,7 @@ class TestObjDBPerformance(TestBigRepoR): # END read stream gelapsed_readchunks = time() - st print >> sys.stderr, "Read %i KiB of %s data in %i KiB chunks from git-cat-file in %f s ( %f Read KiB / s)" % (size_kib, desc, cs_kib, gelapsed_readchunks, size_kib / gelapsed_readchunks) - + # compare print >> sys.stderr, "Git-Python is %f %% faster than git when reading big %s files in chunks" % (100.0 - (elapsed_readchunks / gelapsed_readchunks) * 100, desc) # END for each randomization factor |