« back to all changes in this revision
Viewing changes to bzrlib/tests/test_knit.py
- browse files at revision 2535.3.17
- compare with another revision
- compare with revision 5452
- download diff
- download tarball
- view history from revision 2535.3.17
- Committer: Andrew Bennetts
- Date: 2007-07-17 06:07:00 UTC
- mto: (2535.4.4 streaming-smart-fetch)
- mto: This revision was merged to the branch mainline in revision 2906.
- Revision ID: andrew.bennetts@canonical.com-20070717060700-blzcnto7drjgc7yc
[broken] Closer to a working Repository.fetch_revisions smart request.
added
removed
bzrlib/tests/test_knit.py
1092
1092
"""Get a data stream for an empty knit file."""
1093
1093
k1 = self.make_test_knit()
1094
1094
format, data_list, reader_callable = k1.get_data_stream([])
1095
self.assertEqual('knit-delta-plain', format)
1095
self.assertEqual('knit-plain', format)
1096
1096
self.assertEqual([], data_list)
1097
1097
content = reader_callable(None)
1098
1098
self.assertEqual('', content)
1114
1114
k1.add_lines(version_id, parents, split_lines(lines))
1115
1115
1116
1116
format, data_list, reader_callable = k1.get_data_stream(['text-a'])
1117
self.assertEqual('knit-delta-plain', format)
1117
self.assertEqual('knit-plain', format)
1118
1118
self.assertEqual(expected_data_list, data_list)
1119
1119
# There's only one record in the knit, so the content should be the
1120
1120
# entire knit data file's contents.
1143
1143
k1.add_lines(version_id, parents, split_lines(lines))
1144
1144
1145
1145
format, data_list, reader_callable = k1.get_data_stream(['text-m'])
1146
self.assertEqual('knit-delta-plain', format)
1146
self.assertEqual('knit-plain', format)
1147
1147
self.assertEqual(expected_data_list, data_list)
1148
1148
self.assertRecordContentEqual(k1, 'text-m', reader_callable(None))
1149
1149
1161
1161
]
1162
1162
1163
1163
format, data_list, reader_callable = k1.get_data_stream(['text-b'])
1164
self.assertEqual('knit-delta-plain', format)
1164
self.assertEqual('knit-plain', format)
1165
1165
self.assertEqual(expected_data_list, data_list)
1166
1166
self.assertRecordContentEqual(k1, 'text-b', reader_callable(None))
1167
1167
1190
1190
# case, they'll be in the order they were inserted into the knit.
1191
1191
format, data_list, reader_callable = k1.get_data_stream(
1192
1192
['text-d', 'text-b'])
1193
self.assertEqual('knit-delta-plain', format)
1193
self.assertEqual('knit-plain', format)
1194
1194
self.assertEqual(expected_data_list, data_list)
1195
1195
self.assertRecordContentEqual(k1, 'text-b', reader_callable(84))
1196
1196
self.assertRecordContentEqual(k1, 'text-d', reader_callable(84))
1228
1228
1229
1229
format, data_list, reader_callable = k1.get_data_stream(
1230
1230
['text-a', 'text-b', 'text-c', 'text-d', 'text-m'])
1231
self.assertEqual('knit-delta-plain', format)
1231
self.assertEqual('knit-plain', format)
1232
1232
self.assertEqual(expected_data_list, data_list)
1233
1233
for version_id, options, length, parents in expected_data_list:
1234
1234
bytes = reader_callable(length)
1243
1243
bytes = k1.get_stream_as_bytes(['text-a'])
1244
1244
data = bencode.bdecode(bytes)
1245
1245
format, record = data
1246
self.assertEqual('knit-delta-plain', format)
1246
self.assertEqual('knit-plain', format)
1247
1247
self.assertEqual(['text-a', ['fulltext'], []], record[:3])
1248
1248
self.assertRecordContentEqual(k1, 'text-a', record[3])
1249
1249
1278
1278
1279
1279
data = bencode.bdecode(bytes)
1280
1280
format = data.pop(0)
1281
self.assertEqual('knit-delta-plain', format)
1281
self.assertEqual('knit-plain', format)
1282
1282
1283
1283
for expected, actual in zip(expected_data_list, data):
1284
1284
expected_version = expected[0]
1436
1436
errors.KnitDataStreamIncompatible,
1437
1437
target.insert_data_stream, data_stream)
1438
1438
1439
def test_insert_data_stream_buffering_limit(self):
1440
"""insert_data_stream will batch the incoming records up to a certain
1441
size.
1442
1443
This isn't testing correctness in the way other tests in this file do,
1444
just a performance/resource-use characteristic.
1445
"""
1446
target = self.make_test_knit(name='target')
1447
# Instrument target. We want to log the size of writes, and not
1448
# actually perform the insert because we aren't using real data.
1449
add_raw_records_calls = []
1450
def fake_add_raw_records(records, bytes):
1451
add_raw_records_calls.append(len(bytes))
1452
target._add_raw_records = fake_add_raw_records
1453
1454
data_stream = (
1455
target.get_format_signature(),
1456
[('v1', [], 30, []), ('v2', [], 30, []), ('v3', [], 30, [])],
1457
StringIO('x' * 90).read
1458
)
1459
1460
# Insert 3 records of size 30, when bufsize is 64. No individual write
1461
# should exceed 64, so in this case we expect [60, 30] (i.e. the first
1462
# two records will be read and written in one go).
1463
target.insert_data_stream(data_stream, buffer_size=64)
1464
self.assertEqual([60, 30], add_raw_records_calls)
1465
1466
def test_insert_data_stream_buffering_large_records(self):
1467
"""insert_data_stream's batching copes with records larger than the
1468
buffer size.
1469
"""
1470
target = self.make_test_knit(name='target')
1471
# Instrument target. We want to log the size of writes, and not
1472
# actually perform the insert because we aren't using real data.
1473
add_raw_records_calls = []
1474
def fake_add_raw_records(records, bytes):
1475
add_raw_records_calls.append(len(bytes))
1476
target._add_raw_records = fake_add_raw_records
1477
1478
data_stream = (
1479
target.get_format_signature(),
1480
[('v1', [], 100, []), ('v2', [], 100, [])],
1481
StringIO('x' * 200).read
1482
)
1483
1484
# Insert 1 record of size 100, when the buffer_size is much smaller than
1485
# that. Note that _add_raw_records is never called with no records,
1486
# i.e. if the buffer is empty, then flushing it does not trigger an
1487
# empty write.
1488
target.insert_data_stream(data_stream, buffer_size=20)
1489
self.assertEqual([100, 100], add_raw_records_calls)
1490
1491
def test_insert_data_stream_buffering_flushed_by_known_record(self):
1492
"""insert_data_stream's flushes its buffers (if any) when it needs to do
1493
consistency checks on a record from the stream.
1494
"""
1495
target = self.make_test_knit(name='target')
1496
# Insert a record real record.
1497
target.add_lines('v1', [], split_lines(TEXT_1))
1498
# Now instrument target. We want to log the size of writes, and not
1499
# actually perform the insert because we aren't using real data.
1500
add_raw_records_calls = []
1501
def fake_add_raw_records(records, bytes):
1502
add_raw_records_calls.append(len(bytes))
1503
target._add_raw_records = fake_add_raw_records
1504
1505
# Create a file with a superficially valid knit header, gzip it.
1506
sio = StringIO()
1507
gzip_file = gzip.GzipFile(mode='wb', fileobj=sio)
1508
gzip_file.write('xx v1 yy %s\n' % target.get_sha1('v1'))
1509
gzip_file.close()
1510
sio.seek(0)
1511
data_stream = (
1512
target.get_format_signature(),
1513
[('v1', [], len(sio.getvalue()), [])],
1514
sio.read,
1515
)
1516
1517
# Nothing is written; the buffer had nothing to flush.
1518
target.insert_data_stream(data_stream)
1519
self.assertEqual([], add_raw_records_calls)
1520
1439
# def test_insert_data_stream_buffering_limit(self):
1440
# """insert_data_stream will batch the incoming records up to a certain
1441
# size.
1442
#
1443
# This isn't testing correctness in the way other tests in this file do,
1444
# just a performance/resource-use characteristic.
1445
# """
1446
# target = self.make_test_knit(name='target')
1447
# # Instrument target. We want to log the size of writes, and not
1448
# # actually perform the insert because we aren't using real data.
1449
# add_raw_records_calls = []
1450
# def fake_add_raw_records(records, bytes):
1451
# add_raw_records_calls.append(len(bytes))
1452
# target._add_raw_records = fake_add_raw_records
1453
#
1454
# data_stream = (
1455
# target.get_format_signature(),
1456
# [('v1', [], 30, []), ('v2', [], 30, []), ('v3', [], 30, [])],
1457
# StringIO('x' * 90).read
1458
# )
1459
#
1460
# # Insert 3 records of size 30, when bufsize is 64. No individual write
1461
# # should exceed 64, so in this case we expect [60, 30] (i.e. the first
1462
# # two records will be read and written in one go).
1463
# target.insert_data_stream(data_stream, buffer_size=64)
1464
# self.assertEqual([60, 30], add_raw_records_calls)
1465
#
1466
# def test_insert_data_stream_buffering_large_records(self):
1467
# """insert_data_stream's batching copes with records larger than the
1468
# buffer size.
1469
# """
1470
# target = self.make_test_knit(name='target')
1471
# # Instrument target. We want to log the size of writes, and not
1472
# # actually perform the insert because we aren't using real data.
1473
# add_raw_records_calls = []
1474
# def fake_add_raw_records(records, bytes):
1475
# add_raw_records_calls.append(len(bytes))
1476
# target._add_raw_records = fake_add_raw_records
1477
#
1478
# data_stream = (
1479
# target.get_format_signature(),
1480
# [('v1', [], 100, []), ('v2', [], 100, [])],
1481
# StringIO('x' * 200).read
1482
# )
1483
#
1484
# # Insert 1 record of size 100, when the buffer_size is much smaller than
1485
# # that. Note that _add_raw_records is never called with no records,
1486
# # i.e. if the buffer is empty, then flushing it does not trigger an
1487
# # empty write.
1488
# target.insert_data_stream(data_stream, buffer_size=20)
1489
# self.assertEqual([100, 100], add_raw_records_calls)
1490
#
1491
# def test_insert_data_stream_buffering_flushed_by_known_record(self):
1492
# """insert_data_stream's flushes its buffers (if any) when it needs to do
1493
# consistency checks on a record from the stream.
1494
# """
1495
# target = self.make_test_knit(name='target')
1496
# # Insert a record real record.
1497
# target.add_lines('v1', [], split_lines(TEXT_1))
1498
# # Now instrument target. We want to log the size of writes, and not
1499
# # actually perform the insert because we aren't using real data.
1500
# add_raw_records_calls = []
1501
# def fake_add_raw_records(records, bytes):
1502
# add_raw_records_calls.append(len(bytes))
1503
# target._add_raw_records = fake_add_raw_records
1504
#
1505
# # Create a file with a superficially valid knit header, gzip it.
1506
# sio = StringIO()
1507
# gzip_file = gzip.GzipFile(mode='wb', fileobj=sio)
1508
# gzip_file.write('xx v1 yy %s\n' % target.get_sha1('v1'))
1509
# gzip_file.close()
1510
# sio.seek(0)
1511
# data_stream = (
1512
# target.get_format_signature(),
1513
# [('v1', [], len(sio.getvalue()), [])],
1514
# sio.read,
1515
# )
1516
#
1517
# # Nothing is written; the buffer had nothing to flush.
1518
# target.insert_data_stream(data_stream)
1519
# self.assertEqual([], add_raw_records_calls)
1520
#
1521
1521
# * test that a stream of "already present version, then new version"
1522
1522
# inserts correctly.
1523
1523
Loggerhead is a web-based interface for Breezy
Version: 2.0.1