1 # (Be in -*- python -*- mode.)
3 # ====================================================================
4 # Copyright (c) 2000-2009 CollabNet. All rights reserved.
6 # This software is licensed as described in the file COPYING, which
7 # you should have received as part of this distribution. The terms
8 # are also available at http://subversion.tigris.org/license-1.html.
9 # If newer versions of this license are posted there, you may use a
10 # newer version instead, at your option.
12 # This software consists of voluntary contributions made by many
13 # individuals. For exact contribution history, see the revision
14 # history and logs, available at http://cvs2svn.tigris.org/.
15 # ====================================================================
17 """Data collection classes.
19 This module contains the code used to collect data from the CVS
20 repository. It parses *,v files, recording all useful information
21 except for the actual file contents (though even the file contents
22 might be recorded by the RevisionRecorder if one is configured).
24 As a *,v file is parsed, the information pertaining to the file is
25 accumulated in memory, mostly in _RevisionData, _BranchData, and
26 _TagData objects. When parsing is complete, a final pass is made over
27 the data to create some final dependency links, collect statistics,
28 etc., then the _*Data objects are converted into CVSItem objects
29 (CVSRevision, CVSBranch, and CVSTag respectively) and the CVSItems are
30 dumped into databases.
32 During the data collection, persistent unique ids are allocated to
33 many types of objects: CVSFile, Symbol, and CVSItems. CVSItems are a
34 special case. CVSItem ids are unique across all CVSItem types, and
35 the ids are carried over from the corresponding data collection
38 _RevisionData -> CVSRevision
40 _BranchData -> CVSBranch
44 In a later pass it is possible to convert tags <-> branches. But even
45 if this occurs, the new branch or tag uses the same id as the old tag
55 from cvs2svn_lib
import config
56 from cvs2svn_lib
.common
import DB_OPEN_NEW
57 from cvs2svn_lib
.common
import warning_prefix
58 from cvs2svn_lib
.common
import error_prefix
59 from cvs2svn_lib
.common
import is_trunk_revision
60 from cvs2svn_lib
.log
import Log
61 from cvs2svn_lib
.context
import Ctx
62 from cvs2svn_lib
.artifact_manager
import artifact_manager
63 from cvs2svn_lib
.cvs_path
import CVSPath
64 from cvs2svn_lib
.cvs_path
import CVSFile
65 from cvs2svn_lib
.cvs_path
import CVSDirectory
66 from cvs2svn_lib
.symbol
import Symbol
67 from cvs2svn_lib
.symbol
import Trunk
68 from cvs2svn_lib
.cvs_item
import CVSRevision
69 from cvs2svn_lib
.cvs_item
import CVSBranch
70 from cvs2svn_lib
.cvs_item
import CVSTag
71 from cvs2svn_lib
.cvs_item
import cvs_revision_type_map
72 from cvs2svn_lib
.cvs_file_items
import VendorBranchError
73 from cvs2svn_lib
.cvs_file_items
import CVSFileItems
74 from cvs2svn_lib
.key_generator
import KeyGenerator
75 from cvs2svn_lib
.cvs_item_database
import NewCVSItemStore
76 from cvs2svn_lib
.symbol_statistics
import SymbolStatisticsCollector
77 from cvs2svn_lib
.metadata_database
import MetadataDatabase
78 from cvs2svn_lib
.metadata_database
import MetadataLogger
79 from cvs2svn_lib
.repository_walker
import walk_repository
81 import cvs2svn_rcsparse
84 # A regular expression defining "valid" revision numbers (used to
85 # check that symbol definitions are reasonable).
86 _valid_revision_re
= re
.compile(r
'''
88 (?:\d+\.)+ # Digit groups with trailing dots
89 \d+ # And the last digit group.
93 _branch_revision_re
= re
.compile(r
'''
95 ((?:\d+\.\d+\.)+) # A nonzero even number of digit groups w/trailing dot
96 (?:0\.)? # CVS sticks an extra 0 here; RCS does not
97 (\d+) # And the last digit group
102 def is_branch_revision_number(rev
):
103 """Return True iff REV is a branch revision number.
105 REV is a CVS revision number in canonical form (i.e., with zeros
106 removed). Return True iff it refers to a whole branch, as opposed
107 to a single revision."""
109 return rev
.count('.') % 2 == 0
112 def is_same_line_of_development(rev1
, rev2
):
113 """Return True if rev1 and rev2 are on the same line of
114 development (i.e., both on trunk, or both on the same branch);
115 return False otherwise. Either rev1 or rev2 can be None, in
116 which case automatically return False."""
118 if rev1
is None or rev2
is None:
120 if rev1
.count('.') == 1 and rev2
.count('.') == 1:
122 if rev1
[0:rev1
.rfind('.')] == rev2
[0:rev2
.rfind('.')]:
128 """We track the state of each revision so that in set_revision_info,
129 we can determine if our op is an add/change/delete. We can do this
130 because in set_revision_info, we'll have all of the _RevisionData
131 for a file at our fingertips, and we need to examine the state of
132 our prev_rev to determine if we're an add or a change. Without the
133 state of the prev_rev, we are unable to distinguish between an add
136 def __init__(self
, cvs_rev_id
, rev
, timestamp
, author
, state
):
137 # The id of this revision:
138 self
.cvs_rev_id
= cvs_rev_id
140 self
.timestamp
= timestamp
142 self
.original_timestamp
= timestamp
145 # If this is the first revision on a branch, then this is the
146 # branch_data of that branch; otherwise it is None.
147 self
.parent_branch_data
= None
149 # The revision number of the parent of this revision along the
150 # same line of development, if any. For the first revision R on a
151 # branch, we consider the revision from which R sprouted to be the
152 # 'parent'. If this is the root revision in the file's revision
153 # tree, then this field is None.
155 # Note that this revision can't be determined arithmetically (due
156 # to cvsadmin -o), which is why this field is necessary.
159 # The revision number of the primary child of this revision (the
160 # child along the same line of development), if any; otherwise,
164 # The _BranchData instances of branches that sprout from this
165 # revision, sorted in ascending order by branch number. It would
166 # be inconvenient to initialize it here because we would have to
167 # scan through all branches known by the _SymbolDataCollector to
168 # find the ones having us as the parent. Instead, this
169 # information is filled in by
170 # _FileDataCollector._resolve_dependencies() and sorted by
171 # _FileDataCollector._sort_branches().
172 self
.branches_data
= []
174 # The revision numbers of the first commits on any branches on
175 # which commits occurred. This dependency is kept explicitly
176 # because otherwise a revision-only topological sort would miss
177 # the dependency that exists via branches_data.
178 self
.branches_revs_data
= []
180 # The _TagData instances of tags that are connected to this
184 # A token that may be returned from
185 # RevisionRecorder.record_text(). It can be used by
186 # RevisionReader to obtain the text again.
187 self
.revision_recorder_token
= None
189 def get_first_on_branch_id(self
):
190 return self
.parent_branch_data
and self
.parent_branch_data
.id
194 """Collection area for information about a symbol in a single CVSFile.
196 SYMBOL is an instance of Symbol, undifferentiated as a Branch or a
197 Tag regardless of whether self is a _BranchData or a _TagData."""
199 def __init__(self
, id, symbol
):
200 """Initialize an object for SYMBOL."""
202 # The unique id that will be used for this particular symbol in
203 # this particular file. This same id will be used for the CVSItem
204 # that is derived from this instance.
207 # An instance of Symbol.
211 class _BranchData(_SymbolData
):
212 """Collection area for information about a Branch in a single CVSFile."""
214 def __init__(self
, id, symbol
, branch_number
):
215 _SymbolData
.__init
__(self
, id, symbol
)
217 # The branch number (e.g., '1.5.2') of this branch.
218 self
.branch_number
= branch_number
220 # The revision number of the revision from which this branch
221 # sprouts (e.g., '1.5').
222 self
.parent
= self
.branch_number
[:self
.branch_number
.rindex(".")]
224 # The revision number of the first commit on this branch, if any
225 # (e.g., '1.5.2.1'); otherwise, None.
229 class _TagData(_SymbolData
):
230 """Collection area for information about a Tag in a single CVSFile."""
232 def __init__(self
, id, symbol
, rev
):
233 _SymbolData
.__init
__(self
, id, symbol
)
235 # The revision number being tagged (e.g., '1.5.2.3').
239 class _SymbolDataCollector(object):
240 """Collect information about symbols in a single CVSFile."""
242 def __init__(self
, fdc
, cvs_file
):
244 self
.cvs_file
= cvs_file
246 self
.pdc
= self
.fdc
.pdc
247 self
.collect_data
= self
.fdc
.collect_data
249 # A list [(name, revision), ...] of symbols defined in the header
250 # of the file. The name has already been transformed using the
251 # symbol transform rules. If the symbol transform rules indicate
252 # that the symbol should be ignored, then it is never added to
253 # this list. This list is processed then deleted in
255 self
._symbol
_defs
= []
257 # A set containing the transformed names of symbols in this file
258 # (used to detect duplicates during processing of unlabeled
260 self
._defined
_symbols
= set()
262 # Map { branch_number : _BranchData }, where branch_number has an
263 # odd number of digits.
264 self
.branches_data
= { }
266 # Map { revision : [ tag_data ] }, where revision has an even
267 # number of digits, and the value is a list of _TagData objects
268 # for tags that apply to that revision.
271 def _add_branch(self
, name
, branch_number
):
272 """Record that BRANCH_NUMBER is the branch number for branch NAME,
273 and derive and record the revision from which NAME sprouts.
274 BRANCH_NUMBER is an RCS branch number with an odd number of
275 components, for example '1.7.2' (never '1.7.0.2'). Return the
276 _BranchData instance (which is usually newly-created)."""
278 branch_data
= self
.branches_data
.get(branch_number
)
280 if branch_data
is not None:
283 " branch '%s' already has name '%s',\n"
284 " cannot also have name '%s', ignoring the latter\n"
286 self
.cvs_file
.filename
, branch_number
,
287 branch_data
.symbol
.name
, name
)
291 symbol
= self
.pdc
.get_symbol(name
)
292 branch_data
= _BranchData(
293 self
.collect_data
.item_key_generator
.gen_id(), symbol
, branch_number
295 self
.branches_data
[branch_number
] = branch_data
298 def _construct_distinct_name(self
, name
, original_name
):
299 """Construct a distinct symbol name from NAME.
301 If NAME is distinct, return it. If it is already used in this
302 file (as determined from its presence in self._defined_symbols),
303 construct and return a new name that is not already used."""
305 if name
not in self
._defined
_symbols
:
310 dup_name
= '%s-DUPLICATE-%d' % (name
, index
,)
311 if dup_name
not in self
._defined
_symbols
:
312 self
.collect_data
.record_fatal_error(
313 "Symbol name '%s' is already used in '%s'.\n"
314 "The unlabeled branch '%s' must be renamed using "
315 "--symbol-transform."
316 % (name
, self
.cvs_file
.filename
, original_name
,)
320 def _add_unlabeled_branch(self
, branch_number
):
321 original_name
= "unlabeled-" + branch_number
322 name
= self
.transform_symbol(original_name
, branch_number
)
324 self
.collect_data
.record_fatal_error(
325 "The unlabeled branch '%s' in '%s' contains commits.\n"
326 "It may not be ignored via a symbol transform. (Use --exclude "
328 % (original_name
, self
.cvs_file
.filename
,)
330 # Retain the original name to allow the conversion to continue:
333 distinct_name
= self
._construct
_distinct
_name
(name
, original_name
)
334 self
._defined
_symbols
.add(distinct_name
)
335 return self
._add
_branch
(distinct_name
, branch_number
)
337 def _add_tag(self
, name
, revision
):
338 """Record that tag NAME refers to the specified REVISION."""
340 symbol
= self
.pdc
.get_symbol(name
)
342 self
.collect_data
.item_key_generator
.gen_id(), symbol
, revision
344 self
.tags_data
.setdefault(revision
, []).append(tag_data
)
347 def transform_symbol(self
, name
, revision
):
348 """Transform a symbol according to the project's symbol transforms.
350 Transform the symbol with the original name NAME and canonicalized
351 revision number REVISION. Return the new symbol name or None if
352 the symbol should be ignored entirely.
354 Log the results of the symbol transform if necessary."""
357 # Apply any user-defined symbol transforms to the symbol name:
358 name
= self
.cvs_file
.project
.transform_symbol(
359 self
.cvs_file
, name
, revision
364 self
.pdc
.log_symbol_transform(old_name
, None)
366 " symbol '%s'=%s ignored in %s"
367 % (old_name
, revision
, self
.cvs_file
.filename
,)
371 self
.pdc
.log_symbol_transform(old_name
, name
)
373 " symbol '%s'=%s transformed to '%s' in %s"
374 % (old_name
, revision
, name
, self
.cvs_file
.filename
,)
379 def define_symbol(self
, name
, revision
):
380 """Record a symbol definition for later processing."""
382 # Canonicalize the revision number:
383 revision
= _branch_revision_re
.sub(r
'\1\2', revision
)
385 # Apply any user-defined symbol transforms to the symbol name:
386 name
= self
.transform_symbol(name
, revision
)
389 # Verify that the revision number is valid:
390 if _valid_revision_re
.match(revision
):
391 # The revision number is valid; record it for later processing:
392 self
._symbol
_defs
.append( (name
, revision
) )
396 ' branch %r references invalid revision %s\n'
397 ' and will be ignored.'
398 % (self
.cvs_file
.filename
, name
, revision
,)
401 def _eliminate_trivial_duplicate_defs(self
, symbol_defs
):
402 """Iterate through SYMBOL_DEFS, Removing identical duplicate definitions.
404 Duplicate definitions of symbol names have been seen in the wild,
405 and they can also happen when --symbol-transform is used. If a
406 symbol is defined to the same revision number repeatedly, then
407 ignore all but the last definition."""
409 # Make a copy, since we have to iterate through the definitions
411 symbol_defs
= list(symbol_defs
)
413 # A map { (name, revision) : [index,...] } of the indexes where
414 # symbol definitions name=revision were found:
415 known_definitions
= {}
416 for (i
, symbol_def
) in enumerate(symbol_defs
):
417 known_definitions
.setdefault(symbol_def
, []).append(i
)
419 # A set of the indexes of entries that have to be removed from
422 for ((name
, revision
), indexes
) in known_definitions
.iteritems():
426 " symbol %s:%s defined multiple times; ignoring duplicates\n"
427 % (self
.cvs_file
.filename
, name
, revision
,)
429 dup_indexes
.update(indexes
[:-1])
431 for (i
, symbol_def
) in enumerate(symbol_defs
):
432 if i
not in dup_indexes
:
435 def _process_duplicate_defs(self
, symbol_defs
):
436 """Iterate through SYMBOL_DEFS, processing duplicate names.
438 Duplicate definitions of symbol names have been seen in the wild,
439 and they can also happen when --symbol-transform is used. If a
440 symbol is defined multiple times, then it is a fatal error. This
441 method should be called after _eliminate_trivial_duplicate_defs()."""
443 # Make a copy, since we have to access multiple times:
444 symbol_defs
= list(symbol_defs
)
446 # A map {name : [index,...]} mapping the names of symbols to a
447 # list of their definitions' indexes in symbol_defs:
449 for (i
, (name
, revision
)) in enumerate(symbol_defs
):
450 known_symbols
.setdefault(name
, []).append(i
)
452 known_symbols
= known_symbols
.items()
455 for (name
, indexes
) in known_symbols
:
457 # This symbol was defined multiple times.
458 self
.collect_data
.record_fatal_error(
459 "Multiple definitions of the symbol '%s' in '%s': %s" % (
460 name
, self
.cvs_file
.filename
,
461 ' '.join([symbol_defs
[i
][1] for i
in indexes
]),
464 # Ignore all but the last definition for now, to allow the
465 # conversion to proceed:
466 dup_indexes
.update(indexes
[:-1])
468 for (i
, symbol_def
) in enumerate(symbol_defs
):
469 if i
not in dup_indexes
:
472 def _process_symbol(self
, name
, revision
):
473 """Process a symbol called NAME, which is associated with REVISON.
475 REVISION is a canonical revision number with zeros removed, for
476 example: '1.7', '1.7.2', or '1.1.1' or '1.1.1.1'. NAME is a
477 transformed branch or tag name."""
479 # Add symbol to our records:
480 if is_branch_revision_number(revision
):
481 self
._add
_branch
(name
, revision
)
483 self
._add
_tag
(name
, revision
)
485 def process_symbols(self
):
486 """Process the symbol definitions from SELF._symbol_defs."""
488 symbol_defs
= self
._symbol
_defs
489 del self
._symbol
_defs
491 symbol_defs
= self
._eliminate
_trivial
_duplicate
_defs
(symbol_defs
)
492 symbol_defs
= self
._process
_duplicate
_defs
(symbol_defs
)
494 for (name
, revision
) in symbol_defs
:
495 self
._defined
_symbols
.add(name
)
496 self
._process
_symbol
(name
, revision
)
499 def rev_to_branch_number(revision
):
500 """Return the branch_number of the branch on which REVISION lies.
502 REVISION is a branch revision number with an even number of
503 components; for example '1.7.2.1' (never '1.7.2' nor '1.7.0.2').
504 The return value is the branch number (for example, '1.7.2').
505 Return none iff REVISION is a trunk revision such as '1.2'."""
507 if is_trunk_revision(revision
):
509 return revision
[:revision
.rindex(".")]
511 def rev_to_branch_data(self
, revision
):
512 """Return the branch_data of the branch on which REVISION lies.
514 REVISION must be a branch revision number with an even number of
515 components; for example '1.7.2.1' (never '1.7.2' nor '1.7.0.2').
516 Raise KeyError iff REVISION is unknown."""
518 assert not is_trunk_revision(revision
)
520 return self
.branches_data
[self
.rev_to_branch_number(revision
)]
522 def rev_to_lod(self
, revision
):
523 """Return the line of development on which REVISION lies.
525 REVISION must be a revision number with an even number of
526 components. Raise KeyError iff REVISION is unknown."""
528 if is_trunk_revision(revision
):
529 return self
.pdc
.trunk
531 return self
.rev_to_branch_data(revision
).symbol
534 class _FileDataCollector(cvs2svn_rcsparse
.Sink
):
535 """Class responsible for collecting RCS data for a particular file.
537 Any collected data that need to be remembered are stored into the
538 referenced CollectData instance."""
540 def __init__(self
, pdc
, cvs_file
):
541 """Create an object that is prepared to receive data for CVS_FILE.
542 CVS_FILE is a CVSFile instance. COLLECT_DATA is used to store the
543 information collected about the file."""
546 self
.cvs_file
= cvs_file
548 self
.collect_data
= self
.pdc
.collect_data
549 self
.project
= self
.cvs_file
.project
551 # A place to store information about the symbols in this file:
552 self
.sdc
= _SymbolDataCollector(self
, self
.cvs_file
)
554 # { revision : _RevisionData instance }
557 # Lists [ (parent, child) ] of revision number pairs indicating
558 # that revision child depends on revision parent along the main
559 # line of development.
560 self
._primary
_dependencies
= []
562 # If set, this is an RCS branch number -- rcsparse calls this the
563 # "principal branch", but CVS and RCS refer to it as the "default
564 # branch", so that's what we call it, even though the rcsparse API
565 # setter method is still 'set_principal_branch'.
566 self
.default_branch
= None
568 # True iff revision 1.1 of the file appears to have been imported
569 # (as opposed to added normally).
570 self
._file
_imported
= False
572 def _get_rev_id(self
, revision
):
575 return self
._rev
_data
[revision
].cvs_rev_id
577 def set_principal_branch(self
, branch
):
578 """This is a callback method declared in Sink."""
580 if branch
.find('.') == -1:
581 # This just sets the default branch to trunk. Normally this
582 # shouldn't occur, but it has been seen in at least one CVS
583 # repository. Just ignore it.
586 self
.default_branch
= branch
588 def set_expansion(self
, mode
):
589 """This is a callback method declared in Sink."""
591 self
.cvs_file
.mode
= mode
593 def set_description(self
, description
):
594 """This is a callback method declared in Sink."""
596 self
.cvs_file
.description
= description
598 def define_tag(self
, name
, revision
):
599 """Remember the symbol name and revision, but don't process them yet.
601 This is a callback method declared in Sink."""
603 self
.sdc
.define_symbol(name
, revision
)
605 def admin_completed(self
):
606 """This is a callback method declared in Sink."""
608 self
.sdc
.process_symbols()
610 def define_revision(self
, revision
, timestamp
, author
, state
,
612 """This is a callback method declared in Sink."""
614 for branch
in branches
:
616 branch_data
= self
.sdc
.rev_to_branch_data(branch
)
618 # Normally we learn about the branches from the branch names
619 # and numbers parsed from the symbolic name header. But this
620 # must have been an unlabeled branch that slipped through the
621 # net. Generate a name for it and create a _BranchData record
623 branch_data
= self
.sdc
._add
_unlabeled
_branch
(
624 self
.sdc
.rev_to_branch_number(branch
))
626 assert branch_data
.child
is None
627 branch_data
.child
= branch
629 if revision
in self
._rev
_data
:
630 # This revision has already been seen.
631 Log().error('File %r contains duplicate definitions of revision %s.'
632 % (self
.cvs_file
.filename
, revision
,))
635 # Record basic information about the revision:
636 rev_data
= _RevisionData(
637 self
.collect_data
.item_key_generator
.gen_id(),
638 revision
, int(timestamp
), author
, state
)
639 self
._rev
_data
[revision
] = rev_data
641 # When on trunk, the RCS 'next' revision number points to what
642 # humans might consider to be the 'previous' revision number. For
643 # example, 1.3's RCS 'next' is 1.2.
645 # However, on a branch, the RCS 'next' revision number really does
646 # point to what humans would consider to be the 'next' revision
647 # number. For example, 1.1.2.1's RCS 'next' would be 1.1.2.2.
649 # In other words, in RCS, 'next' always means "where to find the next
650 # deltatext that you need this revision to retrieve.
652 # That said, we don't *want* RCS's behavior here, so we determine
653 # whether we're on trunk or a branch and set the dependencies
656 if is_trunk_revision(revision
):
657 self
._primary
_dependencies
.append( (next
, revision
,) )
659 self
._primary
_dependencies
.append( (revision
, next
,) )
661 def _resolve_primary_dependencies(self
):
662 """Resolve the dependencies listed in self._primary_dependencies."""
664 for (parent
, child
,) in self
._primary
_dependencies
:
665 parent_data
= self
._rev
_data
[parent
]
666 assert parent_data
.child
is None
667 parent_data
.child
= child
669 child_data
= self
._rev
_data
[child
]
670 assert child_data
.parent
is None
671 child_data
.parent
= parent
673 def _resolve_branch_dependencies(self
):
674 """Resolve dependencies involving branches."""
676 for branch_data
in self
.sdc
.branches_data
.values():
677 # The branch_data's parent has the branch as a child regardless
678 # of whether the branch had any subsequent commits:
680 parent_data
= self
._rev
_data
[branch_data
.parent
]
684 ' branch %r references non-existing revision %s\n'
685 ' and will be ignored.'
686 % (self
.cvs_file
.filename
, branch_data
.symbol
.name
,
687 branch_data
.parent
,))
688 del self
.sdc
.branches_data
[branch_data
.branch_number
]
690 parent_data
.branches_data
.append(branch_data
)
692 # If the branch has a child (i.e., something was committed on
693 # the branch), then we store a reference to the branch_data
694 # there, define the child's parent to be the branch's parent,
695 # and list the child in the branch parent's branches_revs_data:
696 if branch_data
.child
is not None:
697 child_data
= self
._rev
_data
[branch_data
.child
]
698 assert child_data
.parent_branch_data
is None
699 child_data
.parent_branch_data
= branch_data
700 assert child_data
.parent
is None
701 child_data
.parent
= branch_data
.parent
702 parent_data
.branches_revs_data
.append(branch_data
.child
)
704 def _sort_branches(self
):
705 """Sort the branches sprouting from each revision in creation order.
707 Creation order is taken to be the reverse of the order that they
708 are listed in the symbols part of the RCS file. (If a branch is
709 created then deleted, a later branch can be assigned the recycled
710 branch number; therefore branch numbers are not an indication of
713 for rev_data
in self
._rev
_data
.values():
714 rev_data
.branches_data
.sort(lambda a
, b
: - cmp(a
.id, b
.id))
716 def _resolve_tag_dependencies(self
):
717 """Resolve dependencies involving tags."""
719 for (rev
, tag_data_list
) in self
.sdc
.tags_data
.items():
721 parent_data
= self
._rev
_data
[rev
]
725 ' the following tag(s) reference non-existing revision %s\n'
726 ' and will be ignored:\n'
728 self
.cvs_file
.filename
, rev
,
729 ', '.join([repr(tag_data
.symbol
.name
)
730 for tag_data
in tag_data_list
]),))
731 del self
.sdc
.tags_data
[rev
]
733 for tag_data
in tag_data_list
:
734 assert tag_data
.rev
== rev
735 # The tag_data's rev has the tag as a child:
736 parent_data
.tags_data
.append(tag_data
)
738 def _determine_operation(self
, rev_data
):
739 prev_rev_data
= self
._rev
_data
.get(rev_data
.parent
)
740 return cvs_revision_type_map
[(
741 rev_data
.state
!= 'dead',
742 prev_rev_data
is not None and prev_rev_data
.state
!= 'dead',
745 def _get_cvs_revision(self
, rev_data
):
746 """Create and return a CVSRevision for REV_DATA."""
750 for branch_data
in rev_data
.branches_data
753 branch_commit_ids
= [
754 self
._get
_rev
_id
(rev
)
755 for rev
in rev_data
.branches_revs_data
760 for tag_data
in rev_data
.tags_data
763 revision_type
= self
._determine
_operation
(rev_data
)
765 return revision_type(
766 self
._get
_rev
_id
(rev_data
.rev
), self
.cvs_file
,
767 rev_data
.timestamp
, None,
768 self
._get
_rev
_id
(rev_data
.parent
),
769 self
._get
_rev
_id
(rev_data
.child
),
772 self
.sdc
.rev_to_lod(rev_data
.rev
),
773 rev_data
.get_first_on_branch_id(),
775 tag_ids
, branch_ids
, branch_commit_ids
,
776 rev_data
.revision_recorder_token
)
778 def _get_cvs_revisions(self
):
779 """Generate the CVSRevisions present in this file."""
781 for rev_data
in self
._rev
_data
.itervalues():
782 yield self
._get
_cvs
_revision
(rev_data
)
784 def _get_cvs_branches(self
):
785 """Generate the CVSBranches present in this file."""
787 for branch_data
in self
.sdc
.branches_data
.values():
789 branch_data
.id, self
.cvs_file
, branch_data
.symbol
,
790 branch_data
.branch_number
,
791 self
.sdc
.rev_to_lod(branch_data
.parent
),
792 self
._get
_rev
_id
(branch_data
.parent
),
793 self
._get
_rev
_id
(branch_data
.child
),
797 def _get_cvs_tags(self
):
798 """Generate the CVSTags present in this file."""
800 for tags_data
in self
.sdc
.tags_data
.values():
801 for tag_data
in tags_data
:
803 tag_data
.id, self
.cvs_file
, tag_data
.symbol
,
804 self
.sdc
.rev_to_lod(tag_data
.rev
),
805 self
._get
_rev
_id
(tag_data
.rev
),
809 def tree_completed(self
):
810 """The revision tree has been parsed.
812 Analyze it for consistency and connect some loose ends.
814 This is a callback method declared in Sink."""
816 self
._resolve
_primary
_dependencies
()
817 self
._resolve
_branch
_dependencies
()
818 self
._sort
_branches
()
819 self
._resolve
_tag
_dependencies
()
821 # Compute the preliminary CVSFileItems for this file:
823 cvs_items
.extend(self
._get
_cvs
_revisions
())
824 cvs_items
.extend(self
._get
_cvs
_branches
())
825 cvs_items
.extend(self
._get
_cvs
_tags
())
826 self
._cvs
_file
_items
= CVSFileItems(
827 self
.cvs_file
, self
.pdc
.trunk
, cvs_items
830 self
._cvs
_file
_items
.check_link_consistency()
832 # Warm up the revision recorder:
833 self
.collect_data
.revision_recorder
.start_file(self
.cvs_file
)
835 def set_revision_info(self
, revision
, log
, text
):
836 """This is a callback method declared in Sink."""
838 rev_data
= self
._rev
_data
[revision
]
839 cvs_rev
= self
._cvs
_file
_items
[rev_data
.cvs_rev_id
]
841 if cvs_rev
.metadata_id
is not None:
842 # Users have reported problems with repositories in which the
843 # deltatext block for revision 1.1 appears twice. It is not
844 # known whether this results from a CVS/RCS bug, or from botched
845 # hand-editing of the repository. In any case, empirically, cvs
846 # and rcs both use the first version when checking out data, so
847 # that's what we will do. (For the record: "cvs log" fails on
848 # such a file; "rlog" prints the log message from the first
849 # block and ignores the second one.)
852 " Deltatext block for revision %s appeared twice;\n"
853 " ignoring the second occurrence.\n"
854 % (warning_prefix
, self
.cvs_file
.filename
, revision
,)
858 if is_trunk_revision(revision
):
861 branch_name
= self
.sdc
.rev_to_branch_data(revision
).symbol
.name
863 cvs_rev
.metadata_id
= self
.collect_data
.metadata_logger
.store(
864 self
.project
, branch_name
, rev_data
.author
, log
866 cvs_rev
.deltatext_exists
= bool(text
)
868 # If this is revision 1.1, determine whether the file appears to
869 # have been created via 'cvs add' instead of 'cvs import'. The
870 # test is that the log message CVS uses for 1.1 in imports is
871 # "Initial revision\n" with no period. (This fact helps determine
872 # whether this file might have had a default branch in the past.)
873 if revision
== '1.1':
874 self
._file
_imported
= (log
== 'Initial revision\n')
876 cvs_rev
.revision_recorder_token
= \
877 self
.collect_data
.revision_recorder
.record_text(cvs_rev
, log
, text
)
879 def parse_completed(self
):
880 """Finish the processing of this file.
882 This is a callback method declared in Sink."""
884 # Make sure that there was an info section for each revision:
885 for cvs_item
in self
._cvs
_file
_items
.values():
886 if isinstance(cvs_item
, CVSRevision
) and cvs_item
.metadata_id
is None:
887 self
.collect_data
.record_fatal_error(
888 '%r has no deltatext section for revision %s'
889 % (self
.cvs_file
.filename
, cvs_item
.rev
,)
892 def _process_ntdbrs(self
):
893 """Fix up any non-trunk default branch revisions (if present).
895 If a non-trunk default branch is determined to have existed, yield
896 the _RevisionData.ids for all revisions that were once non-trunk
897 default revisions, in dependency order.
899 There are two cases to handle:
901 One case is simple. The RCS file lists a default branch
902 explicitly in its header, such as '1.1.1'. In this case, we know
903 that every revision on the vendor branch is to be treated as head
904 of trunk at that point in time.
906 But there's also a degenerate case. The RCS file does not
907 currently have a default branch, yet we can deduce that for some
908 period in the past it probably *did* have one. For example, the
909 file has vendor revisions 1.1.1.1 -> 1.1.1.96, all of which are
910 dated before 1.2, and then it has 1.1.1.97 -> 1.1.1.100 dated
911 after 1.2. In this case, we should record 1.1.1.96 as the last
912 vendor revision to have been the head of the default branch.
914 If any non-trunk default branch revisions are found:
916 - Set their ntdbr members to True.
918 - Connect the last one with revision 1.2.
920 - Remove revision 1.1 if it is not needed.
925 if self
.default_branch
:
926 vendor_cvs_branch_id
= self
.sdc
.branches_data
[self
.default_branch
].id
927 vendor_lod_items
= self
._cvs
_file
_items
.get_lod_items(
928 self
._cvs
_file
_items
[vendor_cvs_branch_id
]
930 if not self
._cvs
_file
_items
.process_live_ntdb(vendor_lod_items
):
932 elif self
._file
_imported
:
933 vendor_branch_data
= self
.sdc
.branches_data
.get('1.1.1')
934 if vendor_branch_data
is None:
937 vendor_lod_items
= self
._cvs
_file
_items
.get_lod_items(
938 self
._cvs
_file
_items
[vendor_branch_data
.id]
940 if not self
._cvs
_file
_items
.process_historical_ntdb(
946 except VendorBranchError
, e
:
947 self
.collect_data
.record_fatal_error(str(e
))
950 if self
._file
_imported
:
951 self
._cvs
_file
_items
.imported_remove_1_1(vendor_lod_items
)
953 self
._cvs
_file
_items
.check_link_consistency()
955 def get_cvs_file_items(self
):
956 """Finish up and return a CVSFileItems instance for this file.
958 This method must only be called once."""
960 self
._process
_ntdbrs
()
962 # Break a circular reference loop, allowing the memory for self
963 # and sdc to be freed.
966 return self
._cvs
_file
_items
969 class _ProjectDataCollector
:
970 def __init__(self
, collect_data
, project
):
971 self
.collect_data
= collect_data
972 self
.project
= project
975 # The Trunk LineOfDevelopment object for this project:
977 self
.collect_data
.symbol_key_generator
.gen_id(), self
.project
979 self
.project
.trunk_id
= self
.trunk
.id
981 # This causes a record for self.trunk to spring into existence:
982 self
.collect_data
.register_trunk(self
.trunk
)
984 # A map { name -> Symbol } for all known symbols in this project.
985 # The symbols listed here are undifferentiated into Branches and
986 # Tags because the same name might appear as a branch in one file
987 # and a tag in another.
990 # A map { (old_name, new_name) : count } indicating how many files
991 # were affected by each each symbol name transformation:
992 self
.symbol_transform_counts
= {}
994 def get_symbol(self
, name
):
995 """Return the Symbol object for the symbol named NAME in this project.
997 If such a symbol does not yet exist, allocate a new symbol_id,
998 create a Symbol instance, store it in self.symbols, and return it."""
1000 symbol
= self
.symbols
.get(name
)
1003 self
.collect_data
.symbol_key_generator
.gen_id(),
1005 self
.symbols
[name
] = symbol
1008 def log_symbol_transform(self
, old_name
, new_name
):
1009 """Record that OLD_NAME was transformed to NEW_NAME in one file.
1011 This information is used to generated a statistical summary of
1012 symbol transforms."""
1015 self
.symbol_transform_counts
[old_name
, new_name
] += 1
1017 self
.symbol_transform_counts
[old_name
, new_name
] = 1
1019 def summarize_symbol_transforms(self
):
1020 if self
.symbol_transform_counts
and Log().is_on(Log
.NORMAL
):
1022 log
.normal('Summary of symbol transforms:')
1023 transforms
= self
.symbol_transform_counts
.items()
1025 for ((old_name
, new_name
), count
) in transforms
:
1026 if new_name
is None:
1027 log
.normal(' "%s" ignored in %d files' % (old_name
, count
,))
1030 ' "%s" transformed to "%s" in %d files'
1031 % (old_name
, new_name
, count
,)
1034 def process_file(self
, cvs_file
):
1035 Log().normal(cvs_file
.filename
)
1036 fdc
= _FileDataCollector(self
, cvs_file
)
1038 cvs2svn_rcsparse
.parse(open(cvs_file
.filename
, 'rb'), fdc
)
1039 except (cvs2svn_rcsparse
.common
.RCSParseError
, ValueError, RuntimeError):
1040 self
.collect_data
.record_fatal_error(
1041 "%r is not a valid ,v file" % (cvs_file
.filename
,)
1043 # Abort the processing of this file, but let the pass continue
1047 Log().warn("Exception occurred while parsing %s" % cvs_file
.filename
)
1052 return fdc
.get_cvs_file_items()
1056 """Repository for data collected by parsing the CVS repository files.
1058 This class manages the databases into which information collected
1059 from the CVS repository is stored. The data are stored into this
1060 class by _FileDataCollector instances, one of which is created for
1061 each file to be parsed."""
1063 def __init__(self
, revision_recorder
, stats_keeper
):
1064 self
.revision_recorder
= revision_recorder
1065 self
._cvs
_item
_store
= NewCVSItemStore(
1066 artifact_manager
.get_temp_file(config
.CVS_ITEMS_STORE
))
1067 self
.metadata_db
= MetadataDatabase(
1068 artifact_manager
.get_temp_file(config
.METADATA_STORE
),
1069 artifact_manager
.get_temp_file(config
.METADATA_INDEX_TABLE
),
1072 self
.metadata_logger
= MetadataLogger(self
.metadata_db
)
1073 self
.fatal_errors
= []
1075 self
.symbol_stats
= SymbolStatisticsCollector()
1076 self
.stats_keeper
= stats_keeper
1078 # Key generator for CVSFiles:
1079 self
.file_key_generator
= KeyGenerator()
1081 # Key generator for CVSItems:
1082 self
.item_key_generator
= KeyGenerator()
1084 # Key generator for Symbols:
1085 self
.symbol_key_generator
= KeyGenerator()
1087 self
.revision_recorder
.start()
1089 def record_fatal_error(self
, err
):
1090 """Record that fatal error ERR was found.
1092 ERR is a string (without trailing newline) describing the error.
1093 Output the error to stderr immediately, and record a copy to be
1094 output again in a summary at the end of CollectRevsPass."""
1096 err
= '%s: %s' % (error_prefix
, err
,)
1097 Log().error(err
+ '\n')
1098 self
.fatal_errors
.append(err
)
1100 def add_cvs_directory(self
, cvs_directory
):
1101 """Record CVS_DIRECTORY."""
1103 Ctx()._cvs
_path
_db
.log_path(cvs_directory
)
1105 def add_cvs_file_items(self
, cvs_file_items
):
1106 """Record the information from CVS_FILE_ITEMS.
1108 Store the CVSFile to _cvs_path_db under its persistent id, store
1109 the CVSItems, and record the CVSItems to self.stats_keeper."""
1111 Ctx()._cvs
_path
_db
.log_path(cvs_file_items
.cvs_file
)
1112 self
._cvs
_item
_store
.add(cvs_file_items
)
1114 self
.stats_keeper
.record_cvs_file(cvs_file_items
.cvs_file
)
1115 for cvs_item
in cvs_file_items
.values():
1116 self
.stats_keeper
.record_cvs_item(cvs_item
)
1118 def register_trunk(self
, trunk
):
1119 """Create a symbol statistics record for the specified trunk LOD."""
1121 # This causes a record to spring into existence:
1122 self
.symbol_stats
[trunk
]
1124 def _process_cvs_file_items(self
, cvs_file_items
):
1125 """Process the CVSFileItems from one CVSFile."""
1127 # Remove an initial delete on trunk if it is not needed:
1128 cvs_file_items
.remove_unneeded_initial_trunk_delete(self
.metadata_db
)
1130 # Remove initial branch deletes that are not needed:
1131 cvs_file_items
.remove_initial_branch_deletes(self
.metadata_db
)
1133 # If this is a --trunk-only conversion, discard all branches and
1134 # tags, then draft any non-trunk default branch revisions to
1136 if Ctx().trunk_only
:
1137 cvs_file_items
.exclude_non_trunk()
1139 cvs_file_items
.check_link_consistency()
1141 self
.revision_recorder
.finish_file(cvs_file_items
)
1142 self
.add_cvs_file_items(cvs_file_items
)
1143 self
.symbol_stats
.register(cvs_file_items
)
1145 def process_project(self
, project
):
1146 Ctx()._projects
[project
.id] = project
1148 pdc
= _ProjectDataCollector(self
, project
)
1150 found_rcs_file
= False
1151 for cvs_path
in walk_repository(
1152 project
, self
.file_key_generator
, self
.record_fatal_error
1154 if isinstance(cvs_path
, CVSDirectory
):
1155 self
.add_cvs_directory(cvs_path
)
1157 cvs_file_items
= pdc
.process_file(cvs_path
)
1158 self
._process
_cvs
_file
_items
(cvs_file_items
)
1159 found_rcs_file
= True
1161 if not found_rcs_file
:
1162 self
.record_fatal_error(
1163 'No RCS files found under %r!\n'
1164 'Are you absolutely certain you are pointing cvs2svn\n'
1165 'at a CVS repository?\n'
1166 % (project
.project_cvs_repos_path
,)
1169 pdc
.summarize_symbol_transforms()
1171 self
.num_files
+= pdc
.num_files
1172 Log().verbose('Processed', self
.num_files
, 'files')
1174 def _register_empty_subdirectories(self
):
1175 """Set the CVSDirectory.empty_subdirectory_id members."""
1179 for path
in Ctx()._cvs
_path
_db
.itervalues()
1180 if isinstance(path
, CVSDirectory
)
1182 for path
in Ctx()._cvs
_path
_db
.itervalues():
1183 if isinstance(path
, CVSFile
):
1184 directory
= path
.parent_directory
1185 while directory
is not None and directory
in directories
:
1186 directories
.remove(directory
)
1187 directory
= directory
.parent_directory
1188 for directory
in directories
:
1189 if directory
.parent_directory
is not None:
1190 directory
.parent_directory
.empty_subdirectory_ids
.append(directory
.id)
1192 def _set_cvs_path_ordinals(self
):
1193 cvs_files
= list(Ctx()._cvs
_path
_db
.itervalues())
1194 cvs_files
.sort(CVSPath
.slow_compare
)
1195 for (i
, cvs_file
) in enumerate(cvs_files
):
1196 cvs_file
.ordinal
= i
1199 """Close the data structures associated with this instance.
1201 Return a list of fatal errors encountered while processing input.
1202 Each list entry is a string describing one fatal error."""
1204 self
.revision_recorder
.finish()
1205 self
.symbol_stats
.purge_ghost_symbols()
1206 self
.symbol_stats
.close()
1207 self
.symbol_stats
= None
1208 self
.metadata_logger
= None
1209 self
.metadata_db
.close()
1210 self
.metadata_db
= None
1211 self
._cvs
_item
_store
.close()
1212 self
._cvs
_item
_store
= None
1213 self
._register
_empty
_subdirectories
()
1214 self
._set
_cvs
_path
_ordinals
()
1215 self
.revision_recorder
= None
1216 retval
= self
.fatal_errors
1217 self
.fatal_errors
= None