16
16
# You should have received a copy of the GNU General Public License
17
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
20
"""Versioned text file storage api."""
22
22
from copy import copy
23
23
from cStringIO import StringIO
25
26
from zlib import adler32
27
28
from bzrlib.lazy_import import lazy_import
457
469
if isinstance(version_ids, basestring):
458
470
version_ids = [version_ids]
459
471
raise NotImplementedError(self.get_ancestry)
461
473
def get_ancestry_with_ghosts(self, version_ids):
462
474
"""Return a list of all ancestors of given version(s). This
463
475
will not include the null revision.
465
477
Must raise RevisionNotPresent if any of the given versions are
466
478
not present in file history.
468
480
Ghosts that are known about will be included in ancestry list,
469
481
but are not explicitly marked.
471
483
raise NotImplementedError(self.get_ancestry_with_ghosts)
473
485
def get_parent_map(self, version_ids):
474
486
"""Get a map of the parents of version_ids.
538
550
unchanged Alive in both a and b (possibly created in both)
539
551
new-a Created in a
540
552
new-b Created in b
541
ghost-a Killed in a, unborn in b
553
ghost-a Killed in a, unborn in b
542
554
ghost-b Killed in b, unborn in a
543
555
irrelevant Not in either revision
545
557
raise NotImplementedError(VersionedFile.plan_merge)
547
559
def weave_merge(self, plan, a_marker=TextMerge.A_MARKER,
548
560
b_marker=TextMerge.B_MARKER):
549
561
return PlanWeaveMerge(plan, a_marker, b_marker).merge_lines()[0]
825
838
raise NotImplementedError(self.add_lines)
840
def _add_text(self, key, parents, text, nostore_sha=None, random_id=False):
841
"""Add a text to the store.
843
This is a private function for use by CommitBuilder.
845
:param key: The key tuple of the text to add. If the last element is
846
None, a CHK string will be generated during the addition.
847
:param parents: The parents key tuples of the text to add.
848
:param text: A string containing the text to be committed.
849
:param nostore_sha: Raise ExistingContent and do not add the lines to
850
the versioned file if the digest of the lines matches this.
851
:param random_id: If True a random id has been selected rather than
852
an id determined by some deterministic process such as a converter
853
from a foreign VCS. When True the backend may choose not to check
854
for uniqueness of the resulting key within the versioned file, so
855
this should only be done when the result is expected to be unique
857
:param check_content: If True, the lines supplied are verified to be
858
bytestrings that are correctly formed lines.
859
:return: The text sha1, the number of bytes in the text, and an opaque
860
representation of the inserted version which can be provided
861
back to future _add_text calls in the parent_texts dictionary.
863
# The default implementation just thunks over to .add_lines(),
864
# inefficient, but it works.
865
return self.add_lines(key, parents, osutils.split_lines(text),
866
nostore_sha=nostore_sha,
827
870
def add_mpdiffs(self, records):
828
871
"""Add mpdiffs to this VersionedFile.
871
914
raise NotImplementedError(self.annotate)
873
916
def check(self, progress_bar=None):
874
"""Check this object for integrity."""
917
"""Check this object for integrity.
919
:param progress_bar: A progress bar to output as the check progresses.
920
:param keys: Specific keys within the VersionedFiles to check. When
921
this parameter is not None, check() becomes a generator as per
922
get_record_stream. The difference to get_record_stream is that
923
more or deeper checks will be performed.
924
:return: None, or if keys was supplied a generator as per
875
927
raise NotImplementedError(self.check)
878
930
def check_not_reserved_id(version_id):
879
931
revision.check_not_reserved_id(version_id)
933
def clear_cache(self):
934
"""Clear whatever caches this VersionedFile holds.
936
This is generally called after an operation has been performed, when we
937
don't expect to be using this versioned file again soon.
881
940
def _check_lines_not_unicode(self, lines):
882
941
"""Check that lines being added to a versioned file are not unicode."""
883
942
for line in lines:
890
949
if '\n' in line[:-1]:
891
950
raise errors.BzrBadParameterContainsNewline("lines")
952
def get_known_graph_ancestry(self, keys):
953
"""Get a KnownGraph instance with the ancestry of keys."""
954
# most basic implementation is a loop around get_parent_map
958
this_parent_map = self.get_parent_map(pending)
959
parent_map.update(this_parent_map)
961
map(pending.update, this_parent_map.itervalues())
962
pending = pending.difference(parent_map)
963
kg = _mod_graph.KnownGraph(parent_map)
893
966
def get_parent_map(self, keys):
894
967
"""Get a map of the parents of keys.
926
999
has_key = index._has_key_from_parent_map
1001
def get_missing_compression_parent_keys(self):
1002
"""Return an iterable of keys of missing compression parents.
1004
Check this after calling insert_record_stream to find out if there are
1005
any missing compression parents. If there are, the records that
1006
depend on them are not able to be inserted safely. The precise
1007
behaviour depends on the concrete VersionedFiles class in use.
1009
Classes that do not support this will raise NotImplementedError.
1011
raise NotImplementedError(self.get_missing_compression_parent_keys)
928
1013
def insert_record_stream(self, stream):
929
1014
"""Insert a record stream into this container.
931
:param stream: A stream of records to insert.
1016
:param stream: A stream of records to insert.
933
1018
:seealso VersionedFile.get_record_stream:
1075
1160
result.append((prefix + (origin,), line))
1078
def check(self, progress_bar=None):
1163
def get_annotator(self):
1164
return annotate.Annotator(self)
1166
def check(self, progress_bar=None, keys=None):
1079
1167
"""See VersionedFiles.check()."""
1168
# XXX: This is over-enthusiastic but as we only thunk for Weaves today
1169
# this is tolerable. Ideally we'd pass keys down to check() and
1170
# have the older VersiondFile interface updated too.
1080
1171
for prefix, vf in self._iter_all_components():
1173
if keys is not None:
1174
return self.get_record_stream(keys, 'unordered', True)
1083
1176
def get_parent_map(self, keys):
1084
1177
"""Get a map of the parents of keys.
1386
1479
elif state == 'conflicted-b':
1387
1480
ch_b = ch_a = True
1388
1481
lines_b.append(line)
1482
elif state == 'killed-both':
1483
# This counts as a change, even though there is no associated
1390
1487
if state not in ('irrelevant', 'ghost-a', 'ghost-b',
1391
'killed-base', 'killed-both'):
1392
1489
raise AssertionError(state)
1393
1490
for struct in outstanding_struct():
1493
def base_from_plan(self):
1494
"""Construct a BASE file from the plan text."""
1496
for state, line in self.plan:
1497
if state in ('killed-a', 'killed-b', 'killed-both', 'unchanged'):
1498
# If unchanged, then this line is straight from base. If a or b
1499
# or both killed the line, then it *used* to be in base.
1500
base_lines.append(line)
1502
if state not in ('killed-base', 'irrelevant',
1503
'ghost-a', 'ghost-b',
1505
'conflicted-a', 'conflicted-b'):
1506
# killed-base, irrelevant means it doesn't apply
1507
# ghost-a/ghost-b are harder to say for sure, but they
1508
# aren't in the 'inc_c' which means they aren't in the
1509
# shared base of a & b. So we don't include them. And
1510
# obviously if the line is newly inserted, it isn't in base
1512
# If 'conflicted-a' or b, then it is new vs one base, but
1513
# old versus another base. However, if we make it present
1514
# in the base, it will be deleted from the target, and it
1515
# seems better to get a line doubled in the merge result,
1516
# rather than have it deleted entirely.
1517
# Example, each node is the 'text' at that point:
1525
# There was a criss-cross conflict merge. Both sides
1526
# include the other, but put themselves first.
1527
# Weave marks this as a 'clean' merge, picking OTHER over
1528
# THIS. (Though the details depend on order inserted into
1530
# LCA generates a plan:
1531
# [('unchanged', M),
1532
# ('conflicted-b', b),
1534
# ('conflicted-a', b),
1536
# If you mark 'conflicted-*' as part of BASE, then a 3-way
1537
# merge tool will cleanly generate "MaN" (as BASE vs THIS
1538
# removes one 'b', and BASE vs OTHER removes the other)
1539
# If you include neither, 3-way creates a clean "MbabN" as
1540
# THIS adds one 'b', and OTHER does too.
1541
# It seems that having the line 2 times is better than
1542
# having it omitted. (Easier to manually delete than notice
1543
# it needs to be added.)
1544
raise AssertionError('Unknown state: %s' % (state,))
1397
1548
class WeaveMerge(PlanWeaveMerge):
1398
1549
"""Weave merge that takes a VersionedFile and two versions as its input."""
1400
def __init__(self, versionedfile, ver_a, ver_b,
1551
def __init__(self, versionedfile, ver_a, ver_b,
1401
1552
a_marker=PlanWeaveMerge.A_MARKER, b_marker=PlanWeaveMerge.B_MARKER):
1402
1553
plan = versionedfile.plan_merge(ver_a, ver_b)
1403
1554
PlanWeaveMerge.__init__(self, plan, a_marker, b_marker)
1406
1557
class VirtualVersionedFiles(VersionedFiles):
1407
"""Dummy implementation for VersionedFiles that uses other functions for
1558
"""Dummy implementation for VersionedFiles that uses other functions for
1408
1559
obtaining fulltexts and parent maps.
1410
This is always on the bottom of the stack and uses string keys
1561
This is always on the bottom of the stack and uses string keys
1411
1562
(rather than tuples) internally.
1469
1620
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1470
1621
for i, (key,) in enumerate(keys):
1471
1622
if pb is not None:
1472
pb.update("iterating texts", i, len(keys))
1623
pb.update("Finding changed lines", i, len(keys))
1473
1624
for l in self._get_lines(key):
1628
def network_bytes_to_kind_and_offset(network_bytes):
1629
"""Strip of a record kind from the front of network_bytes.
1631
:param network_bytes: The bytes of a record.
1632
:return: A tuple (storage_kind, offset_of_remaining_bytes)
1634
line_end = network_bytes.find('\n')
1635
storage_kind = network_bytes[:line_end]
1636
return storage_kind, line_end + 1
1639
class NetworkRecordStream(object):
1640
"""A record_stream which reconstitures a serialised stream."""
1642
def __init__(self, bytes_iterator):
1643
"""Create a NetworkRecordStream.
1645
:param bytes_iterator: An iterator of bytes. Each item in this
1646
iterator should have been obtained from a record_streams'
1647
record.get_bytes_as(record.storage_kind) call.
1649
self._bytes_iterator = bytes_iterator
1650
self._kind_factory = {
1651
'fulltext': fulltext_network_to_record,
1652
'groupcompress-block': groupcompress.network_block_to_records,
1653
'knit-ft-gz': knit.knit_network_to_record,
1654
'knit-delta-gz': knit.knit_network_to_record,
1655
'knit-annotated-ft-gz': knit.knit_network_to_record,
1656
'knit-annotated-delta-gz': knit.knit_network_to_record,
1657
'knit-delta-closure': knit.knit_delta_closure_to_records,
1663
:return: An iterator as per VersionedFiles.get_record_stream().
1665
for bytes in self._bytes_iterator:
1666
storage_kind, line_end = network_bytes_to_kind_and_offset(bytes)
1667
for record in self._kind_factory[storage_kind](
1668
storage_kind, bytes, line_end):
1672
def fulltext_network_to_record(kind, bytes, line_end):
1673
"""Convert a network fulltext record to record."""
1674
meta_len, = struct.unpack('!L', bytes[line_end:line_end+4])
1675
record_meta = bytes[line_end+4:line_end+4+meta_len]
1676
key, parents = bencode.bdecode_as_tuple(record_meta)
1677
if parents == 'nil':
1679
fulltext = bytes[line_end+4+meta_len:]
1680
return [FulltextContentFactory(key, parents, None, fulltext)]
1683
def _length_prefix(bytes):
1684
return struct.pack('!L', len(bytes))
1687
def record_to_fulltext_bytes(record):
1688
if record.parents is None:
1691
parents = record.parents
1692
record_meta = bencode.bencode((record.key, parents))
1693
record_content = record.get_bytes_as('fulltext')
1694
return "fulltext\n%s%s%s" % (
1695
_length_prefix(record_meta), record_meta, record_content)
1698
def sort_groupcompress(parent_map):
1699
"""Sort and group the keys in parent_map into groupcompress order.
1701
groupcompress is defined (currently) as reverse-topological order, grouped
1704
:return: A sorted-list of keys
1706
# gc-optimal ordering is approximately reverse topological,
1707
# properly grouped by file-id.
1709
for item in parent_map.iteritems():
1711
if isinstance(key, str) or len(key) == 1:
1716
per_prefix_map[prefix].append(item)
1718
per_prefix_map[prefix] = [item]
1721
for prefix in sorted(per_prefix_map):
1722
present_keys.extend(reversed(tsort.topo_sort(per_prefix_map[prefix])))
added
removed
bzrlib/versionedfile.py
