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.
23 As a *,v file is parsed, the information pertaining to the file is
24 accumulated in memory, mostly in _RevisionData, _BranchData, and
25 _TagData objects. When parsing is complete, a final pass is made over
26 the data to create some final dependency links, collect statistics,
27 etc., then the _*Data objects are converted into CVSItem objects
28 (CVSRevision, CVSBranch, and CVSTag respectively) and the CVSItems are
29 dumped into databases.
31 During the data collection, persistent unique ids are allocated to
32 many types of objects: CVSFile, Symbol, and CVSItems. CVSItems are a
33 special case. CVSItem ids are unique across all CVSItem types, and
34 the ids are carried over from the corresponding data collection
37 _RevisionData -> CVSRevision
39 _BranchData -> CVSBranch
43 In a later pass it is possible to convert tags <-> branches. But even
44 if this occurs, the new branch or tag uses the same id as the old tag
52 from cvs2svn_lib
import config
53 from cvs2svn_lib
.common
import DB_OPEN_NEW
54 from cvs2svn_lib
.common
import warning_prefix
55 from cvs2svn_lib
.common
import error_prefix
56 from cvs2svn_lib
.common
import is_trunk_revision
57 from cvs2svn_lib
.common
import is_branch_revision_number
58 from cvs2svn_lib
.log
import logger
59 from cvs2svn_lib
.context
import Ctx
60 from cvs2svn_lib
.artifact_manager
import artifact_manager
61 from cvs2svn_lib
.cvs_path
import CVSFile
62 from cvs2svn_lib
.cvs_path
import CVSDirectory
63 from cvs2svn_lib
.symbol
import Symbol
64 from cvs2svn_lib
.symbol
import Trunk
65 from cvs2svn_lib
.cvs_item
import CVSRevision
66 from cvs2svn_lib
.cvs_item
import CVSBranch
67 from cvs2svn_lib
.cvs_item
import CVSTag
68 from cvs2svn_lib
.cvs_item
import cvs_revision_type_map
69 from cvs2svn_lib
.cvs_file_items
import VendorBranchError
70 from cvs2svn_lib
.cvs_file_items
import CVSFileItems
71 from cvs2svn_lib
.key_generator
import KeyGenerator
72 from cvs2svn_lib
.cvs_item_database
import NewCVSItemStore
73 from cvs2svn_lib
.symbol_statistics
import SymbolStatisticsCollector
74 from cvs2svn_lib
.metadata_database
import MetadataDatabase
75 from cvs2svn_lib
.metadata_database
import MetadataLogger
77 import cvs2svn_rcsparse
80 # A regular expression defining "valid" revision numbers (used to
81 # check that symbol definitions are reasonable).
82 _valid_revision_re
= re
.compile(r
'''
84 (?:\d+\.)+ # Digit groups with trailing dots
85 \d+ # And the last digit group.
89 _branch_revision_re
= re
.compile(r
'''
91 ((?:\d+\.\d+\.)+) # A nonzero even number of digit groups w/trailing dot
92 (?:0\.)? # CVS sticks an extra 0 here; RCS does not
93 (\d+) # And the last digit group
98 def is_same_line_of_development(rev1
, rev2
):
99 """Return True if rev1 and rev2 are on the same line of
100 development (i.e., both on trunk, or both on the same branch);
101 return False otherwise. Either rev1 or rev2 can be None, in
102 which case automatically return False."""
104 if rev1
is None or rev2
is None:
106 if is_trunk_revision(rev1
) and is_trunk_revision(rev2
):
107 # Trunk revisions have to be handled specially because the main
108 # trunk version number can be changed; e.g., from 1 to 2.
110 if rev1
[0:rev1
.rfind('.')] == rev2
[0:rev2
.rfind('.')]:
116 """We track the state of each revision so that in set_revision_info,
117 we can determine if our op is an add/change/delete. We can do this
118 because in set_revision_info, we'll have all of the _RevisionData
119 for a file at our fingertips, and we need to examine the state of
120 our prev_rev to determine if we're an add or a change. Without the
121 state of the prev_rev, we are unable to distinguish between an add
124 def __init__(self
, cvs_rev_id
, rev
, timestamp
, author
, state
):
125 # The id of this revision:
126 self
.cvs_rev_id
= cvs_rev_id
128 self
.timestamp
= timestamp
132 # If this is the first revision on a branch, then this is the
133 # branch_data of that branch; otherwise it is None.
134 self
.parent_branch_data
= None
136 # The revision number of the parent of this revision along the
137 # same line of development, if any. For the first revision R on a
138 # branch, we consider the revision from which R sprouted to be the
139 # 'parent'. If this is the root revision in the file's revision
140 # tree, then this field is None.
142 # Note that this revision can't be determined arithmetically (due
143 # to cvsadmin -o), which is why this field is necessary.
146 # The revision number of the primary child of this revision (the
147 # child along the same line of development), if any; otherwise,
151 # The _BranchData instances of branches that sprout from this
152 # revision, sorted in ascending order by branch number. It would
153 # be inconvenient to initialize it here because we would have to
154 # scan through all branches known by the _SymbolDataCollector to
155 # find the ones having us as the parent. Instead, this
156 # information is filled in by
157 # _FileDataCollector._resolve_dependencies() and sorted by
158 # _FileDataCollector._sort_branches().
159 self
.branches_data
= []
161 # The revision numbers of the first commits on any branches on
162 # which commits occurred. This dependency is kept explicitly
163 # because otherwise a revision-only topological sort would miss
164 # the dependency that exists via branches_data.
165 self
.branches_revs_data
= []
167 # The _TagData instances of tags that are connected to this
171 # A token that may be set by a RevisionCollector, then used by
172 # RevisionReader to obtain the text again.
173 self
.revision_reader_token
= None
175 def get_first_on_branch_id(self
):
176 return self
.parent_branch_data
and self
.parent_branch_data
.id
180 """Collection area for information about a symbol in a single CVSFile.
182 SYMBOL is an instance of Symbol, undifferentiated as a Branch or a
183 Tag regardless of whether self is a _BranchData or a _TagData."""
185 def __init__(self
, id, symbol
):
186 """Initialize an object for SYMBOL."""
188 # The unique id that will be used for this particular symbol in
189 # this particular file. This same id will be used for the CVSItem
190 # that is derived from this instance.
193 # An instance of Symbol.
197 class _BranchData(_SymbolData
):
198 """Collection area for information about a Branch in a single CVSFile."""
200 def __init__(self
, id, symbol
, branch_number
):
201 _SymbolData
.__init
__(self
, id, symbol
)
203 # The branch number (e.g., '1.5.2') of this branch.
204 self
.branch_number
= branch_number
206 # The revision number of the revision from which this branch
207 # sprouts (e.g., '1.5').
208 self
.parent
= self
.branch_number
[:self
.branch_number
.rindex(".")]
210 # The revision number of the first commit on this branch, if any
211 # (e.g., '1.5.2.1'); otherwise, None.
215 class _TagData(_SymbolData
):
216 """Collection area for information about a Tag in a single CVSFile."""
218 def __init__(self
, id, symbol
, rev
):
219 _SymbolData
.__init
__(self
, id, symbol
)
221 # The revision number being tagged (e.g., '1.5.2.3').
225 class _SymbolDataCollector(object):
226 """Collect information about symbols in a single CVSFile."""
228 def __init__(self
, fdc
, cvs_file
):
230 self
.cvs_file
= cvs_file
232 self
.pdc
= self
.fdc
.pdc
233 self
.collect_data
= self
.fdc
.collect_data
235 # A list [(name, revision), ...] of symbols defined in the header
236 # of the file. The name has already been transformed using the
237 # symbol transform rules. If the symbol transform rules indicate
238 # that the symbol should be ignored, then it is never added to
239 # this list. This list is processed then deleted in
241 self
._symbol
_defs
= []
243 # A set containing the transformed names of symbols in this file
244 # (used to detect duplicates during processing of unlabeled
246 self
._defined
_symbols
= set()
248 # Map { branch_number : _BranchData }, where branch_number has an
249 # odd number of digits.
250 self
.branches_data
= { }
252 # Map { revision : [ tag_data ] }, where revision has an even
253 # number of digits, and the value is a list of _TagData objects
254 # for tags that apply to that revision.
257 def _add_branch(self
, name
, branch_number
):
258 """Record that BRANCH_NUMBER is the branch number for branch NAME,
259 and derive and record the revision from which NAME sprouts.
260 BRANCH_NUMBER is an RCS branch number with an odd number of
261 components, for example '1.7.2' (never '1.7.0.2'). Return the
262 _BranchData instance (which is usually newly-created)."""
264 branch_data
= self
.branches_data
.get(branch_number
)
266 if branch_data
is not None:
269 " branch '%s' already has name '%s',\n"
270 " cannot also have name '%s', ignoring the latter\n"
272 self
.cvs_file
.rcs_path
, branch_number
,
273 branch_data
.symbol
.name
, name
)
277 symbol
= self
.pdc
.get_symbol(name
)
278 branch_data
= _BranchData(
279 self
.collect_data
.item_key_generator
.gen_id(), symbol
, branch_number
281 self
.branches_data
[branch_number
] = branch_data
284 def _construct_distinct_name(self
, name
, original_name
):
285 """Construct a distinct symbol name from NAME.
287 If NAME is distinct, return it. If it is already used in this
288 file (as determined from its presence in self._defined_symbols),
289 construct and return a new name that is not already used."""
291 if name
not in self
._defined
_symbols
:
296 dup_name
= '%s-DUPLICATE-%d' % (name
, index
,)
297 if dup_name
not in self
._defined
_symbols
:
298 self
.collect_data
.record_fatal_error(
299 "Symbol name '%s' is already used in '%s'.\n"
300 "The unlabeled branch '%s' must be renamed using "
301 "--symbol-transform."
302 % (name
, self
.cvs_file
.rcs_path
, original_name
,)
306 def _add_unlabeled_branch(self
, branch_number
):
307 original_name
= "unlabeled-" + branch_number
308 name
= self
.transform_symbol(original_name
, branch_number
)
310 self
.collect_data
.record_fatal_error(
311 "The unlabeled branch '%s' in '%s' contains commits.\n"
312 "It may not be ignored via a symbol transform. (Use --exclude "
314 % (original_name
, self
.cvs_file
.rcs_path
,)
316 # Retain the original name to allow the conversion to continue:
319 distinct_name
= self
._construct
_distinct
_name
(name
, original_name
)
320 self
._defined
_symbols
.add(distinct_name
)
321 return self
._add
_branch
(distinct_name
, branch_number
)
323 def _add_tag(self
, name
, revision
):
324 """Record that tag NAME refers to the specified REVISION."""
326 symbol
= self
.pdc
.get_symbol(name
)
328 self
.collect_data
.item_key_generator
.gen_id(), symbol
, revision
330 self
.tags_data
.setdefault(revision
, []).append(tag_data
)
333 def transform_symbol(self
, name
, revision
):
334 """Transform a symbol according to the project's symbol transforms.
336 Transform the symbol with the original name NAME and canonicalized
337 revision number REVISION. Return the new symbol name or None if
338 the symbol should be ignored entirely.
340 Log the results of the symbol transform if necessary."""
343 # Apply any user-defined symbol transforms to the symbol name:
344 name
= self
.cvs_file
.project
.transform_symbol(
345 self
.cvs_file
, name
, revision
350 self
.pdc
.log_symbol_transform(old_name
, None)
352 " symbol '%s'=%s ignored in %s"
353 % (old_name
, revision
, self
.cvs_file
.rcs_path
,)
357 self
.pdc
.log_symbol_transform(old_name
, name
)
359 " symbol '%s'=%s transformed to '%s' in %s"
360 % (old_name
, revision
, name
, self
.cvs_file
.rcs_path
,)
365 def define_symbol(self
, name
, revision
):
366 """Record a symbol definition for later processing."""
368 # Canonicalize the revision number:
369 revision
= _branch_revision_re
.sub(r
'\1\2', revision
)
371 # Apply any user-defined symbol transforms to the symbol name:
372 name
= self
.transform_symbol(name
, revision
)
375 # Verify that the revision number is valid:
376 if _valid_revision_re
.match(revision
):
377 # The revision number is valid; record it for later processing:
378 self
._symbol
_defs
.append( (name
, revision
) )
382 ' branch %r references invalid revision %s\n'
383 ' and will be ignored.'
384 % (self
.cvs_file
.rcs_path
, name
, revision
,)
387 def _eliminate_trivial_duplicate_defs(self
, symbol_defs
):
388 """Iterate through SYMBOL_DEFS, Removing identical duplicate definitions.
390 Duplicate definitions of symbol names have been seen in the wild,
391 and they can also happen when --symbol-transform is used. If a
392 symbol is defined to the same revision number repeatedly, then
393 ignore all but the last definition."""
395 # Make a copy, since we have to iterate through the definitions
397 symbol_defs
= list(symbol_defs
)
399 # A map { (name, revision) : [index,...] } of the indexes where
400 # symbol definitions name=revision were found:
401 known_definitions
= {}
402 for (i
, symbol_def
) in enumerate(symbol_defs
):
403 known_definitions
.setdefault(symbol_def
, []).append(i
)
405 # A set of the indexes of entries that have to be removed from
408 for ((name
, revision
), indexes
) in known_definitions
.iteritems():
412 " symbol %s:%s defined multiple times; ignoring duplicates\n"
413 % (self
.cvs_file
.rcs_path
, name
, revision
,)
415 dup_indexes
.update(indexes
[:-1])
417 for (i
, symbol_def
) in enumerate(symbol_defs
):
418 if i
not in dup_indexes
:
421 def _process_duplicate_defs(self
, symbol_defs
):
422 """Iterate through SYMBOL_DEFS, processing duplicate names.
424 Duplicate definitions of symbol names have been seen in the wild,
425 and they can also happen when --symbol-transform is used. If a
426 symbol is defined multiple times, then it is a fatal error. This
427 method should be called after _eliminate_trivial_duplicate_defs()."""
429 # Make a copy, since we have to access multiple times:
430 symbol_defs
= list(symbol_defs
)
432 # A map {name : [index,...]} mapping the names of symbols to a
433 # list of their definitions' indexes in symbol_defs:
435 for (i
, (name
, revision
)) in enumerate(symbol_defs
):
436 known_symbols
.setdefault(name
, []).append(i
)
438 known_symbols
= known_symbols
.items()
441 for (name
, indexes
) in known_symbols
:
443 # This symbol was defined multiple times.
444 self
.collect_data
.record_fatal_error(
445 "Multiple definitions of the symbol '%s' in '%s': %s" % (
446 name
, self
.cvs_file
.rcs_path
,
447 ' '.join([symbol_defs
[i
][1] for i
in indexes
]),
450 # Ignore all but the last definition for now, to allow the
451 # conversion to proceed:
452 dup_indexes
.update(indexes
[:-1])
454 for (i
, symbol_def
) in enumerate(symbol_defs
):
455 if i
not in dup_indexes
:
458 def _process_symbol(self
, name
, revision
):
459 """Process a symbol called NAME, which is associated with REVISON.
461 REVISION is a canonical revision number with zeros removed, for
462 example: '1.7', '1.7.2', or '1.1.1' or '1.1.1.1'. NAME is a
463 transformed branch or tag name."""
465 # Add symbol to our records:
466 if is_branch_revision_number(revision
):
467 self
._add
_branch
(name
, revision
)
469 self
._add
_tag
(name
, revision
)
471 def process_symbols(self
):
472 """Process the symbol definitions from SELF._symbol_defs."""
474 symbol_defs
= self
._symbol
_defs
475 del self
._symbol
_defs
477 symbol_defs
= self
._eliminate
_trivial
_duplicate
_defs
(symbol_defs
)
478 symbol_defs
= self
._process
_duplicate
_defs
(symbol_defs
)
480 for (name
, revision
) in symbol_defs
:
481 self
._defined
_symbols
.add(name
)
482 self
._process
_symbol
(name
, revision
)
485 def rev_to_branch_number(revision
):
486 """Return the branch_number of the branch on which REVISION lies.
488 REVISION is a branch revision number with an even number of
489 components; for example '1.7.2.1' (never '1.7.2' nor '1.7.0.2').
490 The return value is the branch number (for example, '1.7.2').
491 Return none iff REVISION is a trunk revision such as '1.2'."""
493 if is_trunk_revision(revision
):
495 return revision
[:revision
.rindex(".")]
497 def rev_to_branch_data(self
, revision
):
498 """Return the branch_data of the branch on which REVISION lies.
500 REVISION must be a branch revision number with an even number of
501 components; for example '1.7.2.1' (never '1.7.2' nor '1.7.0.2').
502 Raise KeyError iff REVISION is unknown."""
504 assert not is_trunk_revision(revision
)
506 return self
.branches_data
[self
.rev_to_branch_number(revision
)]
508 def rev_to_lod(self
, revision
):
509 """Return the line of development on which REVISION lies.
511 REVISION must be a revision number with an even number of
512 components. Raise KeyError iff REVISION is unknown."""
514 if is_trunk_revision(revision
):
515 return self
.pdc
.trunk
517 return self
.rev_to_branch_data(revision
).symbol
520 class _FileDataCollector(cvs2svn_rcsparse
.Sink
):
521 """Class responsible for collecting RCS data for a particular file.
523 Any collected data that need to be remembered are stored into the
524 referenced CollectData instance."""
526 def __init__(self
, pdc
, cvs_file
):
527 """Create an object that is prepared to receive data for CVS_FILE.
528 CVS_FILE is a CVSFile instance. COLLECT_DATA is used to store the
529 information collected about the file."""
532 self
.cvs_file
= cvs_file
534 self
.collect_data
= self
.pdc
.collect_data
535 self
.project
= self
.cvs_file
.project
537 # A place to store information about the symbols in this file:
538 self
.sdc
= _SymbolDataCollector(self
, self
.cvs_file
)
540 # { revision : _RevisionData instance }
543 # Lists [ (parent, child) ] of revision number pairs indicating
544 # that revision child depends on revision parent along the main
545 # line of development.
546 self
._primary
_dependencies
= []
548 # If set, this is an RCS branch number -- rcsparse calls this the
549 # "principal branch", but CVS and RCS refer to it as the "default
550 # branch", so that's what we call it, even though the rcsparse API
551 # setter method is still 'set_principal_branch'.
552 self
.default_branch
= None
554 # True iff revision 1.1 of the file appears to have been imported
555 # (as opposed to added normally).
556 self
._file
_imported
= False
558 def _get_rev_id(self
, revision
):
561 return self
._rev
_data
[revision
].cvs_rev_id
563 def set_principal_branch(self
, branch
):
564 """This is a callback method declared in Sink."""
566 if branch
.find('.') == -1:
567 # This just sets the default branch to trunk. Normally this
568 # shouldn't occur, but it has been seen in at least one CVS
569 # repository. Just ignore it.
572 m
= _branch_revision_re
.match(branch
)
574 self
.collect_data
.record_fatal_error(
575 'The default branch %s in file %r is not a valid branch number'
576 % (branch
, self
.cvs_file
.rcs_path
,)
580 branch
= m
.group(1) + m
.group(2)
581 if branch
.count('.') != 2:
582 # We don't know how to deal with a non-top-level default
583 # branch (what does CVS do?). So if this case is detected,
585 self
.collect_data
.record_fatal_error(
586 'The default branch %s in file %r is not a top-level branch'
587 % (branch
, self
.cvs_file
.rcs_path
,)
591 self
.default_branch
= branch
593 def define_tag(self
, name
, revision
):
594 """Remember the symbol name and revision, but don't process them yet.
596 This is a callback method declared in Sink."""
598 self
.sdc
.define_symbol(name
, revision
)
600 def set_expansion(self
, mode
):
601 """This is a callback method declared in Sink."""
603 self
.cvs_file
.mode
= mode
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 logger
.error('File %r contains duplicate definitions of revision %s.'
632 % (self
.cvs_file
.rcs_path
, 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 tree_completed(self
):
662 """The revision tree has been parsed.
664 Analyze it for consistency and connect some loose ends.
666 This is a callback method declared in Sink."""
668 self
._resolve
_primary
_dependencies
()
669 self
._resolve
_branch
_dependencies
()
670 self
._sort
_branches
()
671 self
._resolve
_tag
_dependencies
()
673 # Compute the preliminary CVSFileItems for this file:
675 cvs_items
.extend(self
._get
_cvs
_revisions
())
676 cvs_items
.extend(self
._get
_cvs
_branches
())
677 cvs_items
.extend(self
._get
_cvs
_tags
())
678 self
._cvs
_file
_items
= CVSFileItems(
679 self
.cvs_file
, self
.pdc
.trunk
, cvs_items
682 self
._cvs
_file
_items
.check_link_consistency()
684 def _resolve_primary_dependencies(self
):
685 """Resolve the dependencies listed in self._primary_dependencies."""
687 for (parent
, child
,) in self
._primary
_dependencies
:
688 parent_data
= self
._rev
_data
[parent
]
689 assert parent_data
.child
is None
690 parent_data
.child
= child
692 child_data
= self
._rev
_data
[child
]
693 assert child_data
.parent
is None
694 child_data
.parent
= parent
696 def _resolve_branch_dependencies(self
):
697 """Resolve dependencies involving branches."""
699 for branch_data
in self
.sdc
.branches_data
.values():
700 # The branch_data's parent has the branch as a child regardless
701 # of whether the branch had any subsequent commits:
703 parent_data
= self
._rev
_data
[branch_data
.parent
]
707 ' branch %r references non-existing revision %s\n'
708 ' and will be ignored.'
709 % (self
.cvs_file
.rcs_path
, branch_data
.symbol
.name
,
710 branch_data
.parent
,))
711 del self
.sdc
.branches_data
[branch_data
.branch_number
]
713 parent_data
.branches_data
.append(branch_data
)
715 # If the branch has a child (i.e., something was committed on
716 # the branch), then we store a reference to the branch_data
717 # there, define the child's parent to be the branch's parent,
718 # and list the child in the branch parent's branches_revs_data:
719 if branch_data
.child
is not None:
720 child_data
= self
._rev
_data
[branch_data
.child
]
721 assert child_data
.parent_branch_data
is None
722 child_data
.parent_branch_data
= branch_data
723 assert child_data
.parent
is None
724 child_data
.parent
= branch_data
.parent
725 parent_data
.branches_revs_data
.append(branch_data
.child
)
727 def _sort_branches(self
):
728 """Sort the branches sprouting from each revision in creation order.
730 Creation order is taken to be the reverse of the order that they
731 are listed in the symbols part of the RCS file. (If a branch is
732 created then deleted, a later branch can be assigned the recycled
733 branch number; therefore branch numbers are not an indication of
736 for rev_data
in self
._rev
_data
.values():
737 rev_data
.branches_data
.sort(lambda a
, b
: - cmp(a
.id, b
.id))
739 def _resolve_tag_dependencies(self
):
740 """Resolve dependencies involving tags."""
742 for (rev
, tag_data_list
) in self
.sdc
.tags_data
.items():
744 parent_data
= self
._rev
_data
[rev
]
748 ' the following tag(s) reference non-existing revision %s\n'
749 ' and will be ignored:\n'
751 self
.cvs_file
.rcs_path
, rev
,
752 ', '.join([repr(tag_data
.symbol
.name
)
753 for tag_data
in tag_data_list
]),))
754 del self
.sdc
.tags_data
[rev
]
756 for tag_data
in tag_data_list
:
757 assert tag_data
.rev
== rev
758 # The tag_data's rev has the tag as a child:
759 parent_data
.tags_data
.append(tag_data
)
761 def _get_cvs_branches(self
):
762 """Generate the CVSBranches present in this file."""
764 for branch_data
in self
.sdc
.branches_data
.values():
766 branch_data
.id, self
.cvs_file
, branch_data
.symbol
,
767 branch_data
.branch_number
,
768 self
.sdc
.rev_to_lod(branch_data
.parent
),
769 self
._get
_rev
_id
(branch_data
.parent
),
770 self
._get
_rev
_id
(branch_data
.child
),
774 def _get_cvs_tags(self
):
775 """Generate the CVSTags present in this file."""
777 for tags_data
in self
.sdc
.tags_data
.values():
778 for tag_data
in tags_data
:
780 tag_data
.id, self
.cvs_file
, tag_data
.symbol
,
781 self
.sdc
.rev_to_lod(tag_data
.rev
),
782 self
._get
_rev
_id
(tag_data
.rev
),
786 def set_description(self
, description
):
787 """This is a callback method declared in Sink."""
789 self
.cvs_file
.description
= description
790 self
.cvs_file
.determine_file_properties(Ctx().file_property_setters
)
792 def set_revision_info(self
, revision
, log
, text
):
793 """This is a callback method declared in Sink."""
795 rev_data
= self
._rev
_data
[revision
]
796 cvs_rev
= self
._cvs
_file
_items
[rev_data
.cvs_rev_id
]
798 if cvs_rev
.metadata_id
is not None:
799 # Users have reported problems with repositories in which the
800 # deltatext block for revision 1.1 appears twice. It is not
801 # known whether this results from a CVS/RCS bug, or from botched
802 # hand-editing of the repository. In any case, empirically, cvs
803 # and rcs both use the first version when checking out data, so
804 # that's what we will do. (For the record: "cvs log" fails on
805 # such a file; "rlog" prints the log message from the first
806 # block and ignores the second one.)
809 " Deltatext block for revision %s appeared twice;\n"
810 " ignoring the second occurrence.\n"
811 % (warning_prefix
, self
.cvs_file
.rcs_path
, revision
,)
815 if is_trunk_revision(revision
):
818 branch_name
= self
.sdc
.rev_to_branch_data(revision
).symbol
.name
820 cvs_rev
.metadata_id
= self
.collect_data
.metadata_logger
.store(
821 self
.project
, branch_name
, rev_data
.author
, log
823 cvs_rev
.deltatext_exists
= bool(text
)
825 # If this is revision 1.1, determine whether the file appears to
826 # have been created via 'cvs add' instead of 'cvs import'. The
827 # test is that the log message CVS uses for 1.1 in imports is
828 # "Initial revision\n" with no period. (This fact helps determine
829 # whether this file might have had a default branch in the past.)
830 if revision
== '1.1':
831 self
._file
_imported
= (log
== 'Initial revision\n')
833 def parse_completed(self
):
834 """Finish the processing of this file.
836 This is a callback method declared in Sink."""
838 # Make sure that there was an info section for each revision:
839 for cvs_item
in self
._cvs
_file
_items
.values():
840 if isinstance(cvs_item
, CVSRevision
) and cvs_item
.metadata_id
is None:
841 self
.collect_data
.record_fatal_error(
842 '%r has no deltatext section for revision %s'
843 % (self
.cvs_file
.rcs_path
, cvs_item
.rev
,)
846 def _determine_operation(self
, rev_data
):
847 prev_rev_data
= self
._rev
_data
.get(rev_data
.parent
)
848 return cvs_revision_type_map
[(
849 rev_data
.state
!= 'dead',
850 prev_rev_data
is not None and prev_rev_data
.state
!= 'dead',
853 def _get_cvs_revisions(self
):
854 """Generate the CVSRevisions present in this file."""
856 for rev_data
in self
._rev
_data
.itervalues():
857 yield self
._get
_cvs
_revision
(rev_data
)
859 def _get_cvs_revision(self
, rev_data
):
860 """Create and return a CVSRevision for REV_DATA."""
864 for branch_data
in rev_data
.branches_data
867 branch_commit_ids
= [
868 self
._get
_rev
_id
(rev
)
869 for rev
in rev_data
.branches_revs_data
874 for tag_data
in rev_data
.tags_data
877 revision_type
= self
._determine
_operation
(rev_data
)
879 return revision_type(
880 self
._get
_rev
_id
(rev_data
.rev
), self
.cvs_file
,
881 rev_data
.timestamp
, None,
882 self
._get
_rev
_id
(rev_data
.parent
),
883 self
._get
_rev
_id
(rev_data
.child
),
886 self
.sdc
.rev_to_lod(rev_data
.rev
),
887 rev_data
.get_first_on_branch_id(),
889 tag_ids
, branch_ids
, branch_commit_ids
,
890 rev_data
.revision_reader_token
893 def get_cvs_file_items(self
):
894 """Finish up and return a CVSFileItems instance for this file.
896 This method must only be called once."""
898 self
._process
_ntdbrs
()
900 # Break a circular reference loop, allowing the memory for self
901 # and sdc to be freed.
904 return self
._cvs
_file
_items
906 def _process_ntdbrs(self
):
907 """Fix up any non-trunk default branch revisions (if present).
909 If a non-trunk default branch is determined to have existed, yield
910 the _RevisionData.ids for all revisions that were once non-trunk
911 default revisions, in dependency order.
913 There are two cases to handle:
915 One case is simple. The RCS file lists a default branch
916 explicitly in its header, such as '1.1.1'. In this case, we know
917 that every revision on the vendor branch is to be treated as head
918 of trunk at that point in time.
920 But there's also a degenerate case. The RCS file does not
921 currently have a default branch, yet we can deduce that for some
922 period in the past it probably *did* have one. For example, the
923 file has vendor revisions 1.1.1.1 -> 1.1.1.96, all of which are
924 dated before 1.2, and then it has 1.1.1.97 -> 1.1.1.100 dated
925 after 1.2. In this case, we should record 1.1.1.96 as the last
926 vendor revision to have been the head of the default branch.
928 If any non-trunk default branch revisions are found:
930 - Set their ntdbr members to True.
932 - Connect the last one with revision 1.2.
934 - Remove revision 1.1 if it is not needed.
939 if self
.default_branch
:
941 vendor_cvs_branch_id
= self
.sdc
.branches_data
[self
.default_branch
].id
945 ' vendor branch %r is not present in file and will be ignored.'
946 % (warning_prefix
, self
.cvs_file
.rcs_path
, self
.default_branch
,)
948 self
.default_branch
= None
950 vendor_lod_items
= self
._cvs
_file
_items
.get_lod_items(
951 self
._cvs
_file
_items
[vendor_cvs_branch_id
]
953 if not self
._cvs
_file
_items
.process_live_ntdb(vendor_lod_items
):
955 elif self
._file
_imported
:
956 vendor_branch_data
= self
.sdc
.branches_data
.get('1.1.1')
957 if vendor_branch_data
is None:
960 vendor_lod_items
= self
._cvs
_file
_items
.get_lod_items(
961 self
._cvs
_file
_items
[vendor_branch_data
.id]
963 if not self
._cvs
_file
_items
.process_historical_ntdb(
969 except VendorBranchError
, e
:
970 self
.collect_data
.record_fatal_error(str(e
))
973 if self
._file
_imported
:
974 self
._cvs
_file
_items
.imported_remove_1_1(vendor_lod_items
)
976 self
._cvs
_file
_items
.check_link_consistency()
979 class _ProjectDataCollector
:
980 def __init__(self
, collect_data
, project
):
981 self
.collect_data
= collect_data
982 self
.project
= project
985 # The Trunk LineOfDevelopment object for this project:
987 self
.collect_data
.symbol_key_generator
.gen_id(), self
.project
989 self
.project
.trunk_id
= self
.trunk
.id
991 # This causes a record for self.trunk to spring into existence:
992 self
.collect_data
.register_trunk(self
.trunk
)
994 # A map { name -> Symbol } for all known symbols in this project.
995 # The symbols listed here are undifferentiated into Branches and
996 # Tags because the same name might appear as a branch in one file
997 # and a tag in another.
1000 # A map { (old_name, new_name) : count } indicating how many files
1001 # were affected by each each symbol name transformation:
1002 self
.symbol_transform_counts
= {}
1004 def get_symbol(self
, name
):
1005 """Return the Symbol object for the symbol named NAME in this project.
1007 If such a symbol does not yet exist, allocate a new symbol_id,
1008 create a Symbol instance, store it in self.symbols, and return it."""
1010 symbol
= self
.symbols
.get(name
)
1013 self
.collect_data
.symbol_key_generator
.gen_id(),
1015 self
.symbols
[name
] = symbol
1018 def log_symbol_transform(self
, old_name
, new_name
):
1019 """Record that OLD_NAME was transformed to NEW_NAME in one file.
1021 This information is used to generated a statistical summary of
1022 symbol transforms."""
1025 self
.symbol_transform_counts
[old_name
, new_name
] += 1
1027 self
.symbol_transform_counts
[old_name
, new_name
] = 1
1029 def summarize_symbol_transforms(self
):
1030 if self
.symbol_transform_counts
and logger
.is_on(logger
.NORMAL
):
1031 logger
.normal('Summary of symbol transforms:')
1032 transforms
= self
.symbol_transform_counts
.items()
1034 for ((old_name
, new_name
), count
) in transforms
:
1035 if new_name
is None:
1036 logger
.normal(' "%s" ignored in %d files' % (old_name
, count
,))
1039 ' "%s" transformed to "%s" in %d files'
1040 % (old_name
, new_name
, count
,)
1043 def process_file(self
, cvs_file
):
1044 logger
.normal(cvs_file
.rcs_path
)
1045 fdc
= _FileDataCollector(self
, cvs_file
)
1047 cvs2svn_rcsparse
.parse(open(cvs_file
.rcs_path
, 'rb'), fdc
)
1048 except (cvs2svn_rcsparse
.common
.RCSParseError
, RuntimeError):
1049 self
.collect_data
.record_fatal_error(
1050 "%r is not a valid ,v file" % (cvs_file
.rcs_path
,)
1052 # Abort the processing of this file, but let the pass continue
1055 except ValueError, e
:
1056 self
.collect_data
.record_fatal_error(
1057 "%r is not a valid ,v file (%s)" % (cvs_file
.rcs_path
, str(e
),)
1059 # Abort the processing of this file, but let the pass continue
1063 logger
.warn("Exception occurred while parsing %s" % cvs_file
.rcs_path
)
1068 return fdc
.get_cvs_file_items()
1072 """Repository for data collected by parsing the CVS repository files.
1074 This class manages the databases into which information collected
1075 from the CVS repository is stored. The data are stored into this
1076 class by _FileDataCollector instances, one of which is created for
1077 each file to be parsed."""
1079 def __init__(self
, stats_keeper
):
1080 self
._cvs
_item
_store
= NewCVSItemStore(
1081 artifact_manager
.get_temp_file(config
.CVS_ITEMS_STORE
))
1082 self
.metadata_db
= MetadataDatabase(
1083 artifact_manager
.get_temp_file(config
.METADATA_STORE
),
1084 artifact_manager
.get_temp_file(config
.METADATA_INDEX_TABLE
),
1087 self
.metadata_logger
= MetadataLogger(self
.metadata_db
)
1088 self
.fatal_errors
= []
1090 self
.symbol_stats
= SymbolStatisticsCollector()
1091 self
.stats_keeper
= stats_keeper
1093 # Key generator for CVSItems:
1094 self
.item_key_generator
= KeyGenerator()
1096 # Key generator for Symbols:
1097 self
.symbol_key_generator
= KeyGenerator()
1099 def record_fatal_error(self
, err
):
1100 """Record that fatal error ERR was found.
1102 ERR is a string (without trailing newline) describing the error.
1103 Output the error to stderr immediately, and record a copy to be
1104 output again in a summary at the end of CollectRevsPass."""
1106 err
= '%s: %s' % (error_prefix
, err
,)
1107 logger
.error(err
+ '\n')
1108 self
.fatal_errors
.append(err
)
1110 def add_cvs_directory(self
, cvs_directory
):
1111 """Record CVS_DIRECTORY."""
1113 Ctx()._cvs
_path
_db
.log_path(cvs_directory
)
1115 def add_cvs_file_items(self
, cvs_file_items
):
1116 """Record the information from CVS_FILE_ITEMS.
1118 Store the CVSFile to _cvs_path_db under its persistent id, store
1119 the CVSItems, and record the CVSItems to self.stats_keeper."""
1121 Ctx()._cvs
_path
_db
.log_path(cvs_file_items
.cvs_file
)
1122 self
._cvs
_item
_store
.add(cvs_file_items
)
1124 self
.stats_keeper
.record_cvs_file(cvs_file_items
.cvs_file
)
1125 for cvs_item
in cvs_file_items
.values():
1126 self
.stats_keeper
.record_cvs_item(cvs_item
)
1128 def register_trunk(self
, trunk
):
1129 """Create a symbol statistics record for the specified trunk LOD."""
1131 # This causes a record to spring into existence:
1132 self
.symbol_stats
[trunk
]
1134 def _process_cvs_file_items(self
, cvs_file_items
):
1135 """Process the CVSFileItems from one CVSFile."""
1137 # Remove an initial delete on trunk if it is not needed:
1138 cvs_file_items
.remove_unneeded_initial_trunk_delete(self
.metadata_db
)
1140 # Remove initial branch deletes that are not needed:
1141 cvs_file_items
.remove_initial_branch_deletes(self
.metadata_db
)
1143 # If this is a --trunk-only conversion, discard all branches and
1144 # tags, then draft any non-trunk default branch revisions to
1146 if Ctx().trunk_only
:
1147 cvs_file_items
.exclude_non_trunk()
1149 cvs_file_items
.check_link_consistency()
1151 self
.add_cvs_file_items(cvs_file_items
)
1152 self
.symbol_stats
.register(cvs_file_items
)
1154 def process_project(self
, project
, cvs_paths
):
1155 pdc
= _ProjectDataCollector(self
, project
)
1157 found_rcs_file
= False
1158 for cvs_path
in cvs_paths
:
1159 if isinstance(cvs_path
, CVSDirectory
):
1160 self
.add_cvs_directory(cvs_path
)
1162 cvs_file_items
= pdc
.process_file(cvs_path
)
1163 self
._process
_cvs
_file
_items
(cvs_file_items
)
1164 found_rcs_file
= True
1166 if not found_rcs_file
:
1167 self
.record_fatal_error(
1168 'No RCS files found under %r!\n'
1169 'Are you absolutely certain you are pointing cvs2svn\n'
1170 'at a CVS repository?\n'
1171 % (project
.project_cvs_repos_path
,)
1174 pdc
.summarize_symbol_transforms()
1176 self
.num_files
+= pdc
.num_files
1177 logger
.verbose('Processed', self
.num_files
, 'files')
1179 def _register_empty_subdirectories(self
):
1180 """Set the CVSDirectory.empty_subdirectory_id members."""
1184 for path
in Ctx()._cvs
_path
_db
.itervalues()
1185 if isinstance(path
, CVSDirectory
)
1187 for path
in Ctx()._cvs
_path
_db
.itervalues():
1188 if isinstance(path
, CVSFile
):
1189 directory
= path
.parent_directory
1190 while directory
is not None and directory
in directories
:
1191 directories
.remove(directory
)
1192 directory
= directory
.parent_directory
1193 for directory
in directories
:
1194 if directory
.parent_directory
is not None:
1195 directory
.parent_directory
.empty_subdirectory_ids
.append(directory
.id)
1198 """Close the data structures associated with this instance.
1200 Return a list of fatal errors encountered while processing input.
1201 Each list entry is a string describing one fatal error."""
1203 self
.symbol_stats
.purge_ghost_symbols()
1204 self
.symbol_stats
.close()
1205 self
.symbol_stats
= None
1206 self
.metadata_logger
= None
1207 self
.metadata_db
.close()
1208 self
.metadata_db
= None
1209 self
._cvs
_item
_store
.close()
1210 self
._cvs
_item
_store
= None
1211 self
._register
_empty
_subdirectories
()
1212 retval
= self
.fatal_errors
1213 self
.fatal_errors
= None