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
54 from cvs2svn_lib
import config
55 from cvs2svn_lib
.common
import DB_OPEN_NEW
56 from cvs2svn_lib
.common
import warning_prefix
57 from cvs2svn_lib
.common
import error_prefix
58 from cvs2svn_lib
.common
import is_trunk_revision
59 from cvs2svn_lib
.log
import Log
60 from cvs2svn_lib
.context
import Ctx
61 from cvs2svn_lib
.artifact_manager
import artifact_manager
62 from cvs2svn_lib
.cvs_path
import CVSPath
63 from cvs2svn_lib
.cvs_path
import CVSFile
64 from cvs2svn_lib
.cvs_path
import CVSDirectory
65 from cvs2svn_lib
.symbol
import Symbol
66 from cvs2svn_lib
.symbol
import Trunk
67 from cvs2svn_lib
.cvs_item
import CVSRevision
68 from cvs2svn_lib
.cvs_item
import CVSBranch
69 from cvs2svn_lib
.cvs_item
import CVSTag
70 from cvs2svn_lib
.cvs_item
import cvs_revision_type_map
71 from cvs2svn_lib
.cvs_file_items
import VendorBranchError
72 from cvs2svn_lib
.cvs_file_items
import CVSFileItems
73 from cvs2svn_lib
.key_generator
import KeyGenerator
74 from cvs2svn_lib
.cvs_item_database
import NewCVSItemStore
75 from cvs2svn_lib
.symbol_statistics
import SymbolStatisticsCollector
76 from cvs2svn_lib
.metadata_database
import MetadataDatabase
77 from cvs2svn_lib
.metadata_database
import MetadataLogger
78 from cvs2svn_lib
.repository_walker
import walk_repository
80 import cvs2svn_rcsparse
83 # A regular expression defining "valid" revision numbers (used to
84 # check that symbol definitions are reasonable).
85 _valid_revision_re
= re
.compile(r
'''
87 (?:\d+\.)+ # Digit groups with trailing dots
88 \d+ # And the last digit group.
92 _branch_revision_re
= re
.compile(r
'''
94 ((?:\d+\.\d+\.)+) # A nonzero even number of digit groups w/trailing dot
95 (?:0\.)? # CVS sticks an extra 0 here; RCS does not
96 (\d+) # And the last digit group
101 def is_branch_revision_number(rev
):
102 """Return True iff REV is a branch revision number.
104 REV is a CVS revision number in canonical form (i.e., with zeros
105 removed). Return True iff it refers to a whole branch, as opposed
106 to a single revision."""
108 return rev
.count('.') % 2 == 0
111 def is_same_line_of_development(rev1
, rev2
):
112 """Return True if rev1 and rev2 are on the same line of
113 development (i.e., both on trunk, or both on the same branch);
114 return False otherwise. Either rev1 or rev2 can be None, in
115 which case automatically return False."""
117 if rev1
is None or rev2
is None:
119 if rev1
.count('.') == 1 and rev2
.count('.') == 1:
121 if rev1
[0:rev1
.rfind('.')] == rev2
[0:rev2
.rfind('.')]:
127 """We track the state of each revision so that in set_revision_info,
128 we can determine if our op is an add/change/delete. We can do this
129 because in set_revision_info, we'll have all of the _RevisionData
130 for a file at our fingertips, and we need to examine the state of
131 our prev_rev to determine if we're an add or a change. Without the
132 state of the prev_rev, we are unable to distinguish between an add
135 def __init__(self
, cvs_rev_id
, rev
, timestamp
, author
, state
):
136 # The id of this revision:
137 self
.cvs_rev_id
= cvs_rev_id
139 self
.timestamp
= timestamp
141 self
.original_timestamp
= timestamp
144 # If this is the first revision on a branch, then this is the
145 # branch_data of that branch; otherwise it is None.
146 self
.parent_branch_data
= None
148 # The revision number of the parent of this revision along the
149 # same line of development, if any. For the first revision R on a
150 # branch, we consider the revision from which R sprouted to be the
151 # 'parent'. If this is the root revision in the file's revision
152 # tree, then this field is None.
154 # Note that this revision can't be determined arithmetically (due
155 # to cvsadmin -o), which is why this field is necessary.
158 # The revision number of the primary child of this revision (the
159 # child along the same line of development), if any; otherwise,
163 # The _BranchData instances of branches that sprout from this
164 # revision, sorted in ascending order by branch number. It would
165 # be inconvenient to initialize it here because we would have to
166 # scan through all branches known by the _SymbolDataCollector to
167 # find the ones having us as the parent. Instead, this
168 # information is filled in by
169 # _FileDataCollector._resolve_dependencies() and sorted by
170 # _FileDataCollector._sort_branches().
171 self
.branches_data
= []
173 # The revision numbers of the first commits on any branches on
174 # which commits occurred. This dependency is kept explicitly
175 # because otherwise a revision-only topological sort would miss
176 # the dependency that exists via branches_data.
177 self
.branches_revs_data
= []
179 # The _TagData instances of tags that are connected to this
183 # A token that may be set by a RevisionCollector, then used by
184 # RevisionReader to obtain the text again.
185 self
.revision_reader_token
= None
187 def get_first_on_branch_id(self
):
188 return self
.parent_branch_data
and self
.parent_branch_data
.id
192 """Collection area for information about a symbol in a single CVSFile.
194 SYMBOL is an instance of Symbol, undifferentiated as a Branch or a
195 Tag regardless of whether self is a _BranchData or a _TagData."""
197 def __init__(self
, id, symbol
):
198 """Initialize an object for SYMBOL."""
200 # The unique id that will be used for this particular symbol in
201 # this particular file. This same id will be used for the CVSItem
202 # that is derived from this instance.
205 # An instance of Symbol.
209 class _BranchData(_SymbolData
):
210 """Collection area for information about a Branch in a single CVSFile."""
212 def __init__(self
, id, symbol
, branch_number
):
213 _SymbolData
.__init
__(self
, id, symbol
)
215 # The branch number (e.g., '1.5.2') of this branch.
216 self
.branch_number
= branch_number
218 # The revision number of the revision from which this branch
219 # sprouts (e.g., '1.5').
220 self
.parent
= self
.branch_number
[:self
.branch_number
.rindex(".")]
222 # The revision number of the first commit on this branch, if any
223 # (e.g., '1.5.2.1'); otherwise, None.
227 class _TagData(_SymbolData
):
228 """Collection area for information about a Tag in a single CVSFile."""
230 def __init__(self
, id, symbol
, rev
):
231 _SymbolData
.__init
__(self
, id, symbol
)
233 # The revision number being tagged (e.g., '1.5.2.3').
237 class _SymbolDataCollector(object):
238 """Collect information about symbols in a single CVSFile."""
240 def __init__(self
, fdc
, cvs_file
):
242 self
.cvs_file
= cvs_file
244 self
.pdc
= self
.fdc
.pdc
245 self
.collect_data
= self
.fdc
.collect_data
247 # A list [(name, revision), ...] of symbols defined in the header
248 # of the file. The name has already been transformed using the
249 # symbol transform rules. If the symbol transform rules indicate
250 # that the symbol should be ignored, then it is never added to
251 # this list. This list is processed then deleted in
253 self
._symbol
_defs
= []
255 # A set containing the transformed names of symbols in this file
256 # (used to detect duplicates during processing of unlabeled
258 self
._defined
_symbols
= set()
260 # Map { branch_number : _BranchData }, where branch_number has an
261 # odd number of digits.
262 self
.branches_data
= { }
264 # Map { revision : [ tag_data ] }, where revision has an even
265 # number of digits, and the value is a list of _TagData objects
266 # for tags that apply to that revision.
269 def _add_branch(self
, name
, branch_number
):
270 """Record that BRANCH_NUMBER is the branch number for branch NAME,
271 and derive and record the revision from which NAME sprouts.
272 BRANCH_NUMBER is an RCS branch number with an odd number of
273 components, for example '1.7.2' (never '1.7.0.2'). Return the
274 _BranchData instance (which is usually newly-created)."""
276 branch_data
= self
.branches_data
.get(branch_number
)
278 if branch_data
is not None:
281 " branch '%s' already has name '%s',\n"
282 " cannot also have name '%s', ignoring the latter\n"
284 self
.cvs_file
.filename
, branch_number
,
285 branch_data
.symbol
.name
, name
)
289 symbol
= self
.pdc
.get_symbol(name
)
290 branch_data
= _BranchData(
291 self
.collect_data
.item_key_generator
.gen_id(), symbol
, branch_number
293 self
.branches_data
[branch_number
] = branch_data
296 def _construct_distinct_name(self
, name
, original_name
):
297 """Construct a distinct symbol name from NAME.
299 If NAME is distinct, return it. If it is already used in this
300 file (as determined from its presence in self._defined_symbols),
301 construct and return a new name that is not already used."""
303 if name
not in self
._defined
_symbols
:
308 dup_name
= '%s-DUPLICATE-%d' % (name
, index
,)
309 if dup_name
not in self
._defined
_symbols
:
310 self
.collect_data
.record_fatal_error(
311 "Symbol name '%s' is already used in '%s'.\n"
312 "The unlabeled branch '%s' must be renamed using "
313 "--symbol-transform."
314 % (name
, self
.cvs_file
.filename
, original_name
,)
318 def _add_unlabeled_branch(self
, branch_number
):
319 original_name
= "unlabeled-" + branch_number
320 name
= self
.transform_symbol(original_name
, branch_number
)
322 self
.collect_data
.record_fatal_error(
323 "The unlabeled branch '%s' in '%s' contains commits.\n"
324 "It may not be ignored via a symbol transform. (Use --exclude "
326 % (original_name
, self
.cvs_file
.filename
,)
328 # Retain the original name to allow the conversion to continue:
331 distinct_name
= self
._construct
_distinct
_name
(name
, original_name
)
332 self
._defined
_symbols
.add(distinct_name
)
333 return self
._add
_branch
(distinct_name
, branch_number
)
335 def _add_tag(self
, name
, revision
):
336 """Record that tag NAME refers to the specified REVISION."""
338 symbol
= self
.pdc
.get_symbol(name
)
340 self
.collect_data
.item_key_generator
.gen_id(), symbol
, revision
342 self
.tags_data
.setdefault(revision
, []).append(tag_data
)
345 def transform_symbol(self
, name
, revision
):
346 """Transform a symbol according to the project's symbol transforms.
348 Transform the symbol with the original name NAME and canonicalized
349 revision number REVISION. Return the new symbol name or None if
350 the symbol should be ignored entirely.
352 Log the results of the symbol transform if necessary."""
355 # Apply any user-defined symbol transforms to the symbol name:
356 name
= self
.cvs_file
.project
.transform_symbol(
357 self
.cvs_file
, name
, revision
362 self
.pdc
.log_symbol_transform(old_name
, None)
364 " symbol '%s'=%s ignored in %s"
365 % (old_name
, revision
, self
.cvs_file
.filename
,)
369 self
.pdc
.log_symbol_transform(old_name
, name
)
371 " symbol '%s'=%s transformed to '%s' in %s"
372 % (old_name
, revision
, name
, self
.cvs_file
.filename
,)
377 def define_symbol(self
, name
, revision
):
378 """Record a symbol definition for later processing."""
380 # Canonicalize the revision number:
381 revision
= _branch_revision_re
.sub(r
'\1\2', revision
)
383 # Apply any user-defined symbol transforms to the symbol name:
384 name
= self
.transform_symbol(name
, revision
)
387 # Verify that the revision number is valid:
388 if _valid_revision_re
.match(revision
):
389 # The revision number is valid; record it for later processing:
390 self
._symbol
_defs
.append( (name
, revision
) )
394 ' branch %r references invalid revision %s\n'
395 ' and will be ignored.'
396 % (self
.cvs_file
.filename
, name
, revision
,)
399 def _eliminate_trivial_duplicate_defs(self
, symbol_defs
):
400 """Iterate through SYMBOL_DEFS, Removing identical duplicate definitions.
402 Duplicate definitions of symbol names have been seen in the wild,
403 and they can also happen when --symbol-transform is used. If a
404 symbol is defined to the same revision number repeatedly, then
405 ignore all but the last definition."""
407 # Make a copy, since we have to iterate through the definitions
409 symbol_defs
= list(symbol_defs
)
411 # A map { (name, revision) : [index,...] } of the indexes where
412 # symbol definitions name=revision were found:
413 known_definitions
= {}
414 for (i
, symbol_def
) in enumerate(symbol_defs
):
415 known_definitions
.setdefault(symbol_def
, []).append(i
)
417 # A set of the indexes of entries that have to be removed from
420 for ((name
, revision
), indexes
) in known_definitions
.iteritems():
424 " symbol %s:%s defined multiple times; ignoring duplicates\n"
425 % (self
.cvs_file
.filename
, name
, revision
,)
427 dup_indexes
.update(indexes
[:-1])
429 for (i
, symbol_def
) in enumerate(symbol_defs
):
430 if i
not in dup_indexes
:
433 def _process_duplicate_defs(self
, symbol_defs
):
434 """Iterate through SYMBOL_DEFS, processing duplicate names.
436 Duplicate definitions of symbol names have been seen in the wild,
437 and they can also happen when --symbol-transform is used. If a
438 symbol is defined multiple times, then it is a fatal error. This
439 method should be called after _eliminate_trivial_duplicate_defs()."""
441 # Make a copy, since we have to access multiple times:
442 symbol_defs
= list(symbol_defs
)
444 # A map {name : [index,...]} mapping the names of symbols to a
445 # list of their definitions' indexes in symbol_defs:
447 for (i
, (name
, revision
)) in enumerate(symbol_defs
):
448 known_symbols
.setdefault(name
, []).append(i
)
450 known_symbols
= known_symbols
.items()
453 for (name
, indexes
) in known_symbols
:
455 # This symbol was defined multiple times.
456 self
.collect_data
.record_fatal_error(
457 "Multiple definitions of the symbol '%s' in '%s': %s" % (
458 name
, self
.cvs_file
.filename
,
459 ' '.join([symbol_defs
[i
][1] for i
in indexes
]),
462 # Ignore all but the last definition for now, to allow the
463 # conversion to proceed:
464 dup_indexes
.update(indexes
[:-1])
466 for (i
, symbol_def
) in enumerate(symbol_defs
):
467 if i
not in dup_indexes
:
470 def _process_symbol(self
, name
, revision
):
471 """Process a symbol called NAME, which is associated with REVISON.
473 REVISION is a canonical revision number with zeros removed, for
474 example: '1.7', '1.7.2', or '1.1.1' or '1.1.1.1'. NAME is a
475 transformed branch or tag name."""
477 # Add symbol to our records:
478 if is_branch_revision_number(revision
):
479 self
._add
_branch
(name
, revision
)
481 self
._add
_tag
(name
, revision
)
483 def process_symbols(self
):
484 """Process the symbol definitions from SELF._symbol_defs."""
486 symbol_defs
= self
._symbol
_defs
487 del self
._symbol
_defs
489 symbol_defs
= self
._eliminate
_trivial
_duplicate
_defs
(symbol_defs
)
490 symbol_defs
= self
._process
_duplicate
_defs
(symbol_defs
)
492 for (name
, revision
) in symbol_defs
:
493 self
._defined
_symbols
.add(name
)
494 self
._process
_symbol
(name
, revision
)
497 def rev_to_branch_number(revision
):
498 """Return the branch_number of the branch on which REVISION lies.
500 REVISION is 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 The return value is the branch number (for example, '1.7.2').
503 Return none iff REVISION is a trunk revision such as '1.2'."""
505 if is_trunk_revision(revision
):
507 return revision
[:revision
.rindex(".")]
509 def rev_to_branch_data(self
, revision
):
510 """Return the branch_data of the branch on which REVISION lies.
512 REVISION must be a branch revision number with an even number of
513 components; for example '1.7.2.1' (never '1.7.2' nor '1.7.0.2').
514 Raise KeyError iff REVISION is unknown."""
516 assert not is_trunk_revision(revision
)
518 return self
.branches_data
[self
.rev_to_branch_number(revision
)]
520 def rev_to_lod(self
, revision
):
521 """Return the line of development on which REVISION lies.
523 REVISION must be a revision number with an even number of
524 components. Raise KeyError iff REVISION is unknown."""
526 if is_trunk_revision(revision
):
527 return self
.pdc
.trunk
529 return self
.rev_to_branch_data(revision
).symbol
532 class _FileDataCollector(cvs2svn_rcsparse
.Sink
):
533 """Class responsible for collecting RCS data for a particular file.
535 Any collected data that need to be remembered are stored into the
536 referenced CollectData instance."""
538 def __init__(self
, pdc
, cvs_file
):
539 """Create an object that is prepared to receive data for CVS_FILE.
540 CVS_FILE is a CVSFile instance. COLLECT_DATA is used to store the
541 information collected about the file."""
544 self
.cvs_file
= cvs_file
546 self
.collect_data
= self
.pdc
.collect_data
547 self
.project
= self
.cvs_file
.project
549 # A place to store information about the symbols in this file:
550 self
.sdc
= _SymbolDataCollector(self
, self
.cvs_file
)
552 # { revision : _RevisionData instance }
555 # Lists [ (parent, child) ] of revision number pairs indicating
556 # that revision child depends on revision parent along the main
557 # line of development.
558 self
._primary
_dependencies
= []
560 # If set, this is an RCS branch number -- rcsparse calls this the
561 # "principal branch", but CVS and RCS refer to it as the "default
562 # branch", so that's what we call it, even though the rcsparse API
563 # setter method is still 'set_principal_branch'.
564 self
.default_branch
= None
566 # True iff revision 1.1 of the file appears to have been imported
567 # (as opposed to added normally).
568 self
._file
_imported
= False
570 def _get_rev_id(self
, revision
):
573 return self
._rev
_data
[revision
].cvs_rev_id
575 def set_principal_branch(self
, branch
):
576 """This is a callback method declared in Sink."""
578 if branch
.find('.') == -1:
579 # This just sets the default branch to trunk. Normally this
580 # shouldn't occur, but it has been seen in at least one CVS
581 # repository. Just ignore it.
584 self
.default_branch
= branch
586 def set_expansion(self
, mode
):
587 """This is a callback method declared in Sink."""
589 self
.cvs_file
.mode
= mode
591 def set_description(self
, description
):
592 """This is a callback method declared in Sink."""
594 self
.cvs_file
.description
= description
596 def define_tag(self
, name
, revision
):
597 """Remember the symbol name and revision, but don't process them yet.
599 This is a callback method declared in Sink."""
601 self
.sdc
.define_symbol(name
, revision
)
603 def admin_completed(self
):
604 """This is a callback method declared in Sink."""
606 self
.sdc
.process_symbols()
608 def define_revision(self
, revision
, timestamp
, author
, state
,
610 """This is a callback method declared in Sink."""
612 for branch
in branches
:
614 branch_data
= self
.sdc
.rev_to_branch_data(branch
)
616 # Normally we learn about the branches from the branch names
617 # and numbers parsed from the symbolic name header. But this
618 # must have been an unlabeled branch that slipped through the
619 # net. Generate a name for it and create a _BranchData record
621 branch_data
= self
.sdc
._add
_unlabeled
_branch
(
622 self
.sdc
.rev_to_branch_number(branch
))
624 assert branch_data
.child
is None
625 branch_data
.child
= branch
627 if revision
in self
._rev
_data
:
628 # This revision has already been seen.
629 Log().error('File %r contains duplicate definitions of revision %s.'
630 % (self
.cvs_file
.filename
, revision
,))
633 # Record basic information about the revision:
634 rev_data
= _RevisionData(
635 self
.collect_data
.item_key_generator
.gen_id(),
636 revision
, int(timestamp
), author
, state
)
637 self
._rev
_data
[revision
] = rev_data
639 # When on trunk, the RCS 'next' revision number points to what
640 # humans might consider to be the 'previous' revision number. For
641 # example, 1.3's RCS 'next' is 1.2.
643 # However, on a branch, the RCS 'next' revision number really does
644 # point to what humans would consider to be the 'next' revision
645 # number. For example, 1.1.2.1's RCS 'next' would be 1.1.2.2.
647 # In other words, in RCS, 'next' always means "where to find the next
648 # deltatext that you need this revision to retrieve.
650 # That said, we don't *want* RCS's behavior here, so we determine
651 # whether we're on trunk or a branch and set the dependencies
654 if is_trunk_revision(revision
):
655 self
._primary
_dependencies
.append( (next
, revision
,) )
657 self
._primary
_dependencies
.append( (revision
, next
,) )
659 def _resolve_primary_dependencies(self
):
660 """Resolve the dependencies listed in self._primary_dependencies."""
662 for (parent
, child
,) in self
._primary
_dependencies
:
663 parent_data
= self
._rev
_data
[parent
]
664 assert parent_data
.child
is None
665 parent_data
.child
= child
667 child_data
= self
._rev
_data
[child
]
668 assert child_data
.parent
is None
669 child_data
.parent
= parent
671 def _resolve_branch_dependencies(self
):
672 """Resolve dependencies involving branches."""
674 for branch_data
in self
.sdc
.branches_data
.values():
675 # The branch_data's parent has the branch as a child regardless
676 # of whether the branch had any subsequent commits:
678 parent_data
= self
._rev
_data
[branch_data
.parent
]
682 ' branch %r references non-existing revision %s\n'
683 ' and will be ignored.'
684 % (self
.cvs_file
.filename
, branch_data
.symbol
.name
,
685 branch_data
.parent
,))
686 del self
.sdc
.branches_data
[branch_data
.branch_number
]
688 parent_data
.branches_data
.append(branch_data
)
690 # If the branch has a child (i.e., something was committed on
691 # the branch), then we store a reference to the branch_data
692 # there, define the child's parent to be the branch's parent,
693 # and list the child in the branch parent's branches_revs_data:
694 if branch_data
.child
is not None:
695 child_data
= self
._rev
_data
[branch_data
.child
]
696 assert child_data
.parent_branch_data
is None
697 child_data
.parent_branch_data
= branch_data
698 assert child_data
.parent
is None
699 child_data
.parent
= branch_data
.parent
700 parent_data
.branches_revs_data
.append(branch_data
.child
)
702 def _sort_branches(self
):
703 """Sort the branches sprouting from each revision in creation order.
705 Creation order is taken to be the reverse of the order that they
706 are listed in the symbols part of the RCS file. (If a branch is
707 created then deleted, a later branch can be assigned the recycled
708 branch number; therefore branch numbers are not an indication of
711 for rev_data
in self
._rev
_data
.values():
712 rev_data
.branches_data
.sort(lambda a
, b
: - cmp(a
.id, b
.id))
714 def _resolve_tag_dependencies(self
):
715 """Resolve dependencies involving tags."""
717 for (rev
, tag_data_list
) in self
.sdc
.tags_data
.items():
719 parent_data
= self
._rev
_data
[rev
]
723 ' the following tag(s) reference non-existing revision %s\n'
724 ' and will be ignored:\n'
726 self
.cvs_file
.filename
, rev
,
727 ', '.join([repr(tag_data
.symbol
.name
)
728 for tag_data
in tag_data_list
]),))
729 del self
.sdc
.tags_data
[rev
]
731 for tag_data
in tag_data_list
:
732 assert tag_data
.rev
== rev
733 # The tag_data's rev has the tag as a child:
734 parent_data
.tags_data
.append(tag_data
)
736 def _determine_operation(self
, rev_data
):
737 prev_rev_data
= self
._rev
_data
.get(rev_data
.parent
)
738 return cvs_revision_type_map
[(
739 rev_data
.state
!= 'dead',
740 prev_rev_data
is not None and prev_rev_data
.state
!= 'dead',
743 def _get_cvs_revision(self
, rev_data
):
744 """Create and return a CVSRevision for REV_DATA."""
748 for branch_data
in rev_data
.branches_data
751 branch_commit_ids
= [
752 self
._get
_rev
_id
(rev
)
753 for rev
in rev_data
.branches_revs_data
758 for tag_data
in rev_data
.tags_data
761 revision_type
= self
._determine
_operation
(rev_data
)
763 return revision_type(
764 self
._get
_rev
_id
(rev_data
.rev
), self
.cvs_file
,
765 rev_data
.timestamp
, None,
766 self
._get
_rev
_id
(rev_data
.parent
),
767 self
._get
_rev
_id
(rev_data
.child
),
770 self
.sdc
.rev_to_lod(rev_data
.rev
),
771 rev_data
.get_first_on_branch_id(),
773 tag_ids
, branch_ids
, branch_commit_ids
,
774 rev_data
.revision_reader_token
777 def _get_cvs_revisions(self
):
778 """Generate the CVSRevisions present in this file."""
780 for rev_data
in self
._rev
_data
.itervalues():
781 yield self
._get
_cvs
_revision
(rev_data
)
783 def _get_cvs_branches(self
):
784 """Generate the CVSBranches present in this file."""
786 for branch_data
in self
.sdc
.branches_data
.values():
788 branch_data
.id, self
.cvs_file
, branch_data
.symbol
,
789 branch_data
.branch_number
,
790 self
.sdc
.rev_to_lod(branch_data
.parent
),
791 self
._get
_rev
_id
(branch_data
.parent
),
792 self
._get
_rev
_id
(branch_data
.child
),
796 def _get_cvs_tags(self
):
797 """Generate the CVSTags present in this file."""
799 for tags_data
in self
.sdc
.tags_data
.values():
800 for tag_data
in tags_data
:
802 tag_data
.id, self
.cvs_file
, tag_data
.symbol
,
803 self
.sdc
.rev_to_lod(tag_data
.rev
),
804 self
._get
_rev
_id
(tag_data
.rev
),
808 def tree_completed(self
):
809 """The revision tree has been parsed.
811 Analyze it for consistency and connect some loose ends.
813 This is a callback method declared in Sink."""
815 self
._resolve
_primary
_dependencies
()
816 self
._resolve
_branch
_dependencies
()
817 self
._sort
_branches
()
818 self
._resolve
_tag
_dependencies
()
820 # Compute the preliminary CVSFileItems for this file:
822 cvs_items
.extend(self
._get
_cvs
_revisions
())
823 cvs_items
.extend(self
._get
_cvs
_branches
())
824 cvs_items
.extend(self
._get
_cvs
_tags
())
825 self
._cvs
_file
_items
= CVSFileItems(
826 self
.cvs_file
, self
.pdc
.trunk
, cvs_items
829 self
._cvs
_file
_items
.check_link_consistency()
831 def set_revision_info(self
, revision
, log
, text
):
832 """This is a callback method declared in Sink."""
834 rev_data
= self
._rev
_data
[revision
]
835 cvs_rev
= self
._cvs
_file
_items
[rev_data
.cvs_rev_id
]
837 if cvs_rev
.metadata_id
is not None:
838 # Users have reported problems with repositories in which the
839 # deltatext block for revision 1.1 appears twice. It is not
840 # known whether this results from a CVS/RCS bug, or from botched
841 # hand-editing of the repository. In any case, empirically, cvs
842 # and rcs both use the first version when checking out data, so
843 # that's what we will do. (For the record: "cvs log" fails on
844 # such a file; "rlog" prints the log message from the first
845 # block and ignores the second one.)
848 " Deltatext block for revision %s appeared twice;\n"
849 " ignoring the second occurrence.\n"
850 % (warning_prefix
, self
.cvs_file
.filename
, revision
,)
854 if is_trunk_revision(revision
):
857 branch_name
= self
.sdc
.rev_to_branch_data(revision
).symbol
.name
859 cvs_rev
.metadata_id
= self
.collect_data
.metadata_logger
.store(
860 self
.project
, branch_name
, rev_data
.author
, log
862 cvs_rev
.deltatext_exists
= bool(text
)
864 # If this is revision 1.1, determine whether the file appears to
865 # have been created via 'cvs add' instead of 'cvs import'. The
866 # test is that the log message CVS uses for 1.1 in imports is
867 # "Initial revision\n" with no period. (This fact helps determine
868 # whether this file might have had a default branch in the past.)
869 if revision
== '1.1':
870 self
._file
_imported
= (log
== 'Initial revision\n')
872 def parse_completed(self
):
873 """Finish the processing of this file.
875 This is a callback method declared in Sink."""
877 # Make sure that there was an info section for each revision:
878 for cvs_item
in self
._cvs
_file
_items
.values():
879 if isinstance(cvs_item
, CVSRevision
) and cvs_item
.metadata_id
is None:
880 self
.collect_data
.record_fatal_error(
881 '%r has no deltatext section for revision %s'
882 % (self
.cvs_file
.filename
, cvs_item
.rev
,)
885 def _process_ntdbrs(self
):
886 """Fix up any non-trunk default branch revisions (if present).
888 If a non-trunk default branch is determined to have existed, yield
889 the _RevisionData.ids for all revisions that were once non-trunk
890 default revisions, in dependency order.
892 There are two cases to handle:
894 One case is simple. The RCS file lists a default branch
895 explicitly in its header, such as '1.1.1'. In this case, we know
896 that every revision on the vendor branch is to be treated as head
897 of trunk at that point in time.
899 But there's also a degenerate case. The RCS file does not
900 currently have a default branch, yet we can deduce that for some
901 period in the past it probably *did* have one. For example, the
902 file has vendor revisions 1.1.1.1 -> 1.1.1.96, all of which are
903 dated before 1.2, and then it has 1.1.1.97 -> 1.1.1.100 dated
904 after 1.2. In this case, we should record 1.1.1.96 as the last
905 vendor revision to have been the head of the default branch.
907 If any non-trunk default branch revisions are found:
909 - Set their ntdbr members to True.
911 - Connect the last one with revision 1.2.
913 - Remove revision 1.1 if it is not needed.
918 if self
.default_branch
:
919 vendor_cvs_branch_id
= self
.sdc
.branches_data
[self
.default_branch
].id
920 vendor_lod_items
= self
._cvs
_file
_items
.get_lod_items(
921 self
._cvs
_file
_items
[vendor_cvs_branch_id
]
923 if not self
._cvs
_file
_items
.process_live_ntdb(vendor_lod_items
):
925 elif self
._file
_imported
:
926 vendor_branch_data
= self
.sdc
.branches_data
.get('1.1.1')
927 if vendor_branch_data
is None:
930 vendor_lod_items
= self
._cvs
_file
_items
.get_lod_items(
931 self
._cvs
_file
_items
[vendor_branch_data
.id]
933 if not self
._cvs
_file
_items
.process_historical_ntdb(
939 except VendorBranchError
, e
:
940 self
.collect_data
.record_fatal_error(str(e
))
943 if self
._file
_imported
:
944 self
._cvs
_file
_items
.imported_remove_1_1(vendor_lod_items
)
946 self
._cvs
_file
_items
.check_link_consistency()
948 def get_cvs_file_items(self
):
949 """Finish up and return a CVSFileItems instance for this file.
951 This method must only be called once."""
953 self
._process
_ntdbrs
()
955 # Break a circular reference loop, allowing the memory for self
956 # and sdc to be freed.
959 return self
._cvs
_file
_items
962 class _ProjectDataCollector
:
963 def __init__(self
, collect_data
, project
):
964 self
.collect_data
= collect_data
965 self
.project
= project
968 # The Trunk LineOfDevelopment object for this project:
970 self
.collect_data
.symbol_key_generator
.gen_id(), self
.project
972 self
.project
.trunk_id
= self
.trunk
.id
974 # This causes a record for self.trunk to spring into existence:
975 self
.collect_data
.register_trunk(self
.trunk
)
977 # A map { name -> Symbol } for all known symbols in this project.
978 # The symbols listed here are undifferentiated into Branches and
979 # Tags because the same name might appear as a branch in one file
980 # and a tag in another.
983 # A map { (old_name, new_name) : count } indicating how many files
984 # were affected by each each symbol name transformation:
985 self
.symbol_transform_counts
= {}
987 def get_symbol(self
, name
):
988 """Return the Symbol object for the symbol named NAME in this project.
990 If such a symbol does not yet exist, allocate a new symbol_id,
991 create a Symbol instance, store it in self.symbols, and return it."""
993 symbol
= self
.symbols
.get(name
)
996 self
.collect_data
.symbol_key_generator
.gen_id(),
998 self
.symbols
[name
] = symbol
1001 def log_symbol_transform(self
, old_name
, new_name
):
1002 """Record that OLD_NAME was transformed to NEW_NAME in one file.
1004 This information is used to generated a statistical summary of
1005 symbol transforms."""
1008 self
.symbol_transform_counts
[old_name
, new_name
] += 1
1010 self
.symbol_transform_counts
[old_name
, new_name
] = 1
1012 def summarize_symbol_transforms(self
):
1013 if self
.symbol_transform_counts
and Log().is_on(Log
.NORMAL
):
1015 log
.normal('Summary of symbol transforms:')
1016 transforms
= self
.symbol_transform_counts
.items()
1018 for ((old_name
, new_name
), count
) in transforms
:
1019 if new_name
is None:
1020 log
.normal(' "%s" ignored in %d files' % (old_name
, count
,))
1023 ' "%s" transformed to "%s" in %d files'
1024 % (old_name
, new_name
, count
,)
1027 def process_file(self
, cvs_file
):
1028 Log().normal(cvs_file
.filename
)
1029 fdc
= _FileDataCollector(self
, cvs_file
)
1031 cvs2svn_rcsparse
.parse(open(cvs_file
.filename
, 'rb'), fdc
)
1032 except (cvs2svn_rcsparse
.common
.RCSParseError
, ValueError, RuntimeError):
1033 self
.collect_data
.record_fatal_error(
1034 "%r is not a valid ,v file" % (cvs_file
.filename
,)
1036 # Abort the processing of this file, but let the pass continue
1040 Log().warn("Exception occurred while parsing %s" % cvs_file
.filename
)
1045 return fdc
.get_cvs_file_items()
1049 """Repository for data collected by parsing the CVS repository files.
1051 This class manages the databases into which information collected
1052 from the CVS repository is stored. The data are stored into this
1053 class by _FileDataCollector instances, one of which is created for
1054 each file to be parsed."""
1056 def __init__(self
, stats_keeper
):
1057 self
._cvs
_item
_store
= NewCVSItemStore(
1058 artifact_manager
.get_temp_file(config
.CVS_ITEMS_STORE
))
1059 self
.metadata_db
= MetadataDatabase(
1060 artifact_manager
.get_temp_file(config
.METADATA_STORE
),
1061 artifact_manager
.get_temp_file(config
.METADATA_INDEX_TABLE
),
1064 self
.metadata_logger
= MetadataLogger(self
.metadata_db
)
1065 self
.fatal_errors
= []
1067 self
.symbol_stats
= SymbolStatisticsCollector()
1068 self
.stats_keeper
= stats_keeper
1070 # Key generator for CVSFiles:
1071 self
.file_key_generator
= KeyGenerator()
1073 # Key generator for CVSItems:
1074 self
.item_key_generator
= KeyGenerator()
1076 # Key generator for Symbols:
1077 self
.symbol_key_generator
= KeyGenerator()
1079 def record_fatal_error(self
, err
):
1080 """Record that fatal error ERR was found.
1082 ERR is a string (without trailing newline) describing the error.
1083 Output the error to stderr immediately, and record a copy to be
1084 output again in a summary at the end of CollectRevsPass."""
1086 err
= '%s: %s' % (error_prefix
, err
,)
1087 Log().error(err
+ '\n')
1088 self
.fatal_errors
.append(err
)
1090 def add_cvs_directory(self
, cvs_directory
):
1091 """Record CVS_DIRECTORY."""
1093 Ctx()._cvs
_path
_db
.log_path(cvs_directory
)
1095 def add_cvs_file_items(self
, cvs_file_items
):
1096 """Record the information from CVS_FILE_ITEMS.
1098 Store the CVSFile to _cvs_path_db under its persistent id, store
1099 the CVSItems, and record the CVSItems to self.stats_keeper."""
1101 Ctx()._cvs
_path
_db
.log_path(cvs_file_items
.cvs_file
)
1102 self
._cvs
_item
_store
.add(cvs_file_items
)
1104 self
.stats_keeper
.record_cvs_file(cvs_file_items
.cvs_file
)
1105 for cvs_item
in cvs_file_items
.values():
1106 self
.stats_keeper
.record_cvs_item(cvs_item
)
1108 def register_trunk(self
, trunk
):
1109 """Create a symbol statistics record for the specified trunk LOD."""
1111 # This causes a record to spring into existence:
1112 self
.symbol_stats
[trunk
]
1114 def _process_cvs_file_items(self
, cvs_file_items
):
1115 """Process the CVSFileItems from one CVSFile."""
1117 # Remove an initial delete on trunk if it is not needed:
1118 cvs_file_items
.remove_unneeded_initial_trunk_delete(self
.metadata_db
)
1120 # Remove initial branch deletes that are not needed:
1121 cvs_file_items
.remove_initial_branch_deletes(self
.metadata_db
)
1123 # If this is a --trunk-only conversion, discard all branches and
1124 # tags, then draft any non-trunk default branch revisions to
1126 if Ctx().trunk_only
:
1127 cvs_file_items
.exclude_non_trunk()
1129 cvs_file_items
.check_link_consistency()
1131 self
.add_cvs_file_items(cvs_file_items
)
1132 self
.symbol_stats
.register(cvs_file_items
)
1134 def process_project(self
, project
):
1135 Ctx()._projects
[project
.id] = project
1137 pdc
= _ProjectDataCollector(self
, project
)
1139 found_rcs_file
= False
1140 for cvs_path
in walk_repository(
1141 project
, self
.file_key_generator
, self
.record_fatal_error
1143 if isinstance(cvs_path
, CVSDirectory
):
1144 self
.add_cvs_directory(cvs_path
)
1146 cvs_file_items
= pdc
.process_file(cvs_path
)
1147 self
._process
_cvs
_file
_items
(cvs_file_items
)
1148 found_rcs_file
= True
1150 if not found_rcs_file
:
1151 self
.record_fatal_error(
1152 'No RCS files found under %r!\n'
1153 'Are you absolutely certain you are pointing cvs2svn\n'
1154 'at a CVS repository?\n'
1155 % (project
.project_cvs_repos_path
,)
1158 pdc
.summarize_symbol_transforms()
1160 self
.num_files
+= pdc
.num_files
1161 Log().verbose('Processed', self
.num_files
, 'files')
1163 def _register_empty_subdirectories(self
):
1164 """Set the CVSDirectory.empty_subdirectory_id members."""
1168 for path
in Ctx()._cvs
_path
_db
.itervalues()
1169 if isinstance(path
, CVSDirectory
)
1171 for path
in Ctx()._cvs
_path
_db
.itervalues():
1172 if isinstance(path
, CVSFile
):
1173 directory
= path
.parent_directory
1174 while directory
is not None and directory
in directories
:
1175 directories
.remove(directory
)
1176 directory
= directory
.parent_directory
1177 for directory
in directories
:
1178 if directory
.parent_directory
is not None:
1179 directory
.parent_directory
.empty_subdirectory_ids
.append(directory
.id)
1181 def _set_cvs_path_ordinals(self
):
1182 cvs_files
= list(Ctx()._cvs
_path
_db
.itervalues())
1183 cvs_files
.sort(CVSPath
.slow_compare
)
1184 for (i
, cvs_file
) in enumerate(cvs_files
):
1185 cvs_file
.ordinal
= i
1188 """Close the data structures associated with this instance.
1190 Return a list of fatal errors encountered while processing input.
1191 Each list entry is a string describing one fatal error."""
1193 self
.symbol_stats
.purge_ghost_symbols()
1194 self
.symbol_stats
.close()
1195 self
.symbol_stats
= None
1196 self
.metadata_logger
= None
1197 self
.metadata_db
.close()
1198 self
.metadata_db
= None
1199 self
._cvs
_item
_store
.close()
1200 self
._cvs
_item
_store
= None
1201 self
._register
_empty
_subdirectories
()
1202 self
._set
_cvs
_path
_ordinals
()
1203 retval
= self
.fatal_errors
1204 self
.fatal_errors
= None