~bzr-pqm/bzr/bzr.dev

# Copyright (C) 2005, 2006, 2007 Canonical Ltd

#

# This program is free software; you can redistribute it and/or modify

# it under the terms of the GNU General Public License as published by

# the Free Software Foundation; either version 2 of the License, or

# (at your option) any later version.

#

# This program is distributed in the hope that it will be useful,

# but WITHOUT ANY WARRANTY; without even the implied warranty of

# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

# GNU General Public License for more details.

#

# You should have received a copy of the GNU General Public License

# along with this program; if not, write to the Free Software

# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

"""Code to show logs of changes.

Various flavors of log can be produced:

* for one file, or the whole tree, and (not done yet) for

  files in a given directory

* in "verbose" mode with a description of what changed from one

  version to the next

* with file-ids and revision-ids shown

Logs are actually written out through an abstract LogFormatter

interface, which allows for different preferred formats.  Plugins can

register formats too.

Logs can be produced in either forward (oldest->newest) or reverse

(newest->oldest) order.

Logs can be filtered to show only revisions matching a particular

search string, or within a particular range of revisions.  The range

can be given as date/times, which are reduced to revisions before

calling in here.

In verbose mode we show a summary of what changed in each particular

revision.  Note that this is the delta for changes in that revision

relative to its left-most parent, not the delta relative to the last

logged revision.  So for example if you ask for a verbose log of

changes touching hello.c you will get a list of those revisions also

listing other things that were changed in the same revision, but not

all the changes since the previous revision that touched hello.c.

"""

import codecs

from itertools import (

    izip,

    )

import re

import sys

from warnings import (

    warn,

    )

from bzrlib.lazy_import import lazy_import

lazy_import(globals(), """

from bzrlib import (

    config,

    errors,

    repository as _mod_repository,

    revision as _mod_revision,

    revisionspec,

    trace,

    tsort,

    )

""")

from bzrlib import (

    registry,

    )

from bzrlib.osutils import (

    format_date,

    get_terminal_encoding,

    terminal_width,

    )

def find_touching_revisions(branch, file_id):

    """Yield a description of revisions which affect the file_id.

    Each returned element is (revno, revision_id, description)

    This is the list of revisions where the file is either added,

    modified, renamed or deleted.

    TODO: Perhaps some way to limit this to only particular revisions,

    or to traverse a non-mainline set of revisions?

    """

    last_ie = None

    last_path = None

    revno = 1

    for revision_id in branch.revision_history():

        this_inv = branch.repository.get_revision_inventory(revision_id)

        if file_id in this_inv:

            this_ie = this_inv[file_id]

            this_path = this_inv.id2path(file_id)

        else:

            this_ie = this_path = None

        # now we know how it was last time, and how it is in this revision.

        # are those two states effectively the same or not?

        if not this_ie and not last_ie:

            # not present in either

            pass

        elif this_ie and not last_ie:

            yield revno, revision_id, "added " + this_path

        elif not this_ie and last_ie:

            # deleted here

            yield revno, revision_id, "deleted " + last_path

        elif this_path != last_path:

            yield revno, revision_id, ("renamed %s => %s" % (last_path, this_path))

        elif (this_ie.text_size != last_ie.text_size

              or this_ie.text_sha1 != last_ie.text_sha1):

            yield revno, revision_id, "modified " + this_path

        last_ie = this_ie

        last_path = this_path

        revno += 1

def _enumerate_history(branch):

    rh = []

    revno = 1

    for rev_id in branch.revision_history():

        rh.append((revno, rev_id))

        revno += 1

    return rh

def show_log(branch,

             lf,

             specific_fileid=None,

             verbose=False,

             direction='reverse',

             start_revision=None,

             end_revision=None,

             search=None,

             limit=None):

    """Write out human-readable log of commits to this branch.

    lf

        LogFormatter object to show the output.

    specific_fileid

        If true, list only the commits affecting the specified

        file, rather than all commits.

    verbose

        If true show added/changed/deleted/renamed files.

    direction

        'reverse' (default) is latest to earliest;

        'forward' is earliest to latest.

    start_revision

        If not None, only show revisions >= start_revision

    end_revision

        If not None, only show revisions <= end_revision

    search

        If not None, only show revisions with matching commit messages

    limit

        If not None or 0, only show limit revisions

    """

    branch.lock_read()

    try:

        if getattr(lf, 'begin_log', None):

            lf.begin_log()

        _show_log(branch, lf, specific_fileid, verbose, direction,

                  start_revision, end_revision, search, limit)

        if getattr(lf, 'end_log', None):

            lf.end_log()

    finally:

        branch.unlock()

def _show_log(branch,

             lf,

             specific_fileid=None,

             verbose=False,

             direction='reverse',

             start_revision=None,

             end_revision=None,

             search=None,

             limit=None):

    """Worker function for show_log - see show_log."""

    if not isinstance(lf, LogFormatter):

        warn("not a LogFormatter instance: %r" % lf)

    if specific_fileid:

        trace.mutter('get log for file_id %r', specific_fileid)

    generate_merge_revisions = getattr(lf, 'supports_merge_revisions', False)

    allow_single_merge_revision = getattr(lf,

        'supports_single_merge_revision', False)

    view_revisions = calculate_view_revisions(branch, start_revision,

                                              end_revision, direction,

                                              specific_fileid,

                                              generate_merge_revisions,

                                              allow_single_merge_revision)

    rev_tag_dict = {}

    generate_tags = getattr(lf, 'supports_tags', False)

    if generate_tags:

        if branch.supports_tags():

            rev_tag_dict = branch.tags.get_reverse_tag_dict()

    generate_delta = verbose and getattr(lf, 'supports_delta', False)

    # now we just print all the revisions

    log_count = 0

    revision_iterator = make_log_rev_iterator(branch, view_revisions,

        generate_delta, search)

    for revs in revision_iterator:

        for (rev_id, revno, merge_depth), rev, delta in revs:

            lr = LogRevision(rev, revno, merge_depth, delta,

                             rev_tag_dict.get(rev_id))

            lf.log_revision(lr)

            if limit:

                log_count += 1

                if log_count >= limit:

                    return

def calculate_view_revisions(branch, start_revision, end_revision, direction,

                             specific_fileid, generate_merge_revisions,

                             allow_single_merge_revision):

    if (not generate_merge_revisions and start_revision is end_revision is

        None and direction == 'reverse' and specific_fileid is None):

        return _linear_view_revisions(branch)

    mainline_revs, rev_nos, start_rev_id, end_rev_id = \

        _get_mainline_revs(branch, start_revision, end_revision)

    if not mainline_revs:

        return []

    if direction == 'reverse':

        start_rev_id, end_rev_id = end_rev_id, start_rev_id

    generate_single_revision = False

    if ((not generate_merge_revisions)

        and ((start_rev_id and (start_rev_id not in rev_nos))

            or (end_rev_id and (end_rev_id not in rev_nos)))):

        generate_single_revision = ((start_rev_id == end_rev_id)

            and allow_single_merge_revision)

        if not generate_single_revision:

            raise errors.BzrCommandError('Selected log formatter only supports'

                ' mainline revisions.')

        generate_merge_revisions = generate_single_revision

    view_revs_iter = get_view_revisions(mainline_revs, rev_nos, branch,

                          direction, include_merges=generate_merge_revisions)

    view_revisions = _filter_revision_range(list(view_revs_iter),

                                            start_rev_id,

                                            end_rev_id)

    if view_revisions and generate_single_revision:

        view_revisions = view_revisions[0:1]

    if specific_fileid:

        view_revisions = _filter_revisions_touching_file_id(branch,

                                                         specific_fileid,

                                                         view_revisions,

                                                         direction)

    # rebase merge_depth - unless there are no revisions or 

    # either the first or last revision have merge_depth = 0.

    if view_revisions and view_revisions[0][2] and view_revisions[-1][2]:

        min_depth = min([d for r,n,d in view_revisions])

        if min_depth != 0:

            view_revisions = [(r,n,d-min_depth) for r,n,d in view_revisions]

    return view_revisions

def _linear_view_revisions(branch):

    start_revno, start_revision_id = branch.last_revision_info()

    repo = branch.repository

    revision_ids = repo.iter_reverse_revision_history(start_revision_id)

    for num, revision_id in enumerate(revision_ids):

        yield revision_id, str(start_revno - num), 0

def make_log_rev_iterator(branch, view_revisions, generate_delta, search):

    """Create a revision iterator for log.

    :param branch: The branch being logged.

    :param view_revisions: The revisions being viewed.

    :param generate_delta: Whether to generate a delta for each revision.

    :param search: A user text search string.

    :return: An iterator over lists of ((rev_id, revno, merge_depth), rev,

        delta).

    """

    # Convert view_revisions into (view, None, None) groups to fit with

    # the standard interface here.

    if type(view_revisions) == list:

        # A single batch conversion is faster than many incremental ones.

        # As we have all the data, do a batch conversion.

        nones = [None] * len(view_revisions)

        log_rev_iterator = iter([zip(view_revisions, nones, nones)])

    else:

        def _convert():

            for view in view_revisions:

                yield (view, None, None)

        log_rev_iterator = iter([_convert()])

    for adapter in log_adapters:

        log_rev_iterator = adapter(branch, generate_delta, search,

            log_rev_iterator)

    return log_rev_iterator

def _make_search_filter(branch, generate_delta, search, log_rev_iterator):

    """Create a filtered iterator of log_rev_iterator matching on a regex.

    :param branch: The branch being logged.

    :param generate_delta: Whether to generate a delta for each revision.

    :param search: A user text search string.

    :param log_rev_iterator: An input iterator containing all revisions that

        could be displayed, in lists.

    :return: An iterator over lists of ((rev_id, revno, merge_depth), rev,

        delta).

    """

    if search is None:

        return log_rev_iterator

    # Compile the search now to get early errors.

    searchRE = re.compile(search, re.IGNORECASE)

    return _filter_message_re(searchRE, log_rev_iterator)

def _filter_message_re(searchRE, log_rev_iterator):

    for revs in log_rev_iterator:

        new_revs = []

        for (rev_id, revno, merge_depth), rev, delta in revs:

            if searchRE.search(rev.message):

                new_revs.append(((rev_id, revno, merge_depth), rev, delta))

        yield new_revs

def _make_delta_filter(branch, generate_delta, search, log_rev_iterator):

    """Add revision deltas to a log iterator if needed.

    :param branch: The branch being logged.

    :param generate_delta: Whether to generate a delta for each revision.

    :param search: A user text search string.

    :param log_rev_iterator: An input iterator containing all revisions that

        could be displayed, in lists.

    :return: An iterator over lists of ((rev_id, revno, merge_depth), rev,

        delta).

    """

    if not generate_delta:

        return log_rev_iterator

    return _generate_deltas(branch.repository, log_rev_iterator)

def _generate_deltas(repository, log_rev_iterator):

    """Create deltas for each batch of revisions in log_rev_iterator."""

    for revs in log_rev_iterator:

        revisions = [rev[1] for rev in revs]

        deltas = repository.get_deltas_for_revisions(revisions)

        revs = [(rev[0], rev[1], delta) for rev, delta in izip(revs, deltas)]

        yield revs

def _make_revision_objects(branch, generate_delta, search, log_rev_iterator):

    """Extract revision objects from the repository

    :param branch: The branch being logged.

    :param generate_delta: Whether to generate a delta for each revision.

    :param search: A user text search string.

    :param log_rev_iterator: An input iterator containing all revisions that

        could be displayed, in lists.

    :return: An iterator over lists of ((rev_id, revno, merge_depth), rev,

        delta).

    """

    repository = branch.repository

    for revs in log_rev_iterator:

        # r = revision_id, n = revno, d = merge depth

        revision_ids = [view[0] for view, _, _ in revs]

        revisions = repository.get_revisions(revision_ids)

        revs = [(rev[0], revision, rev[2]) for rev, revision in

            izip(revs, revisions)]

        yield revs

def _make_batch_filter(branch, generate_delta, search, log_rev_iterator):

    """Group up a single large batch into smaller ones.

    :param branch: The branch being logged.

    :param generate_delta: Whether to generate a delta for each revision.

    :param search: A user text search string.

    :param log_rev_iterator: An input iterator containing all revisions that

        could be displayed, in lists.

    :return: An iterator over lists of ((rev_id, revno, merge_depth), rev, delta).

    """

    repository = branch.repository

    num = 9

    for batch in log_rev_iterator:

        batch = iter(batch)

        while True:

            step = [detail for _, detail in zip(range(num), batch)]

            if len(step) == 0:

                break

            yield step

            num = min(int(num * 1.5), 200)

def _get_mainline_revs(branch, start_revision, end_revision):

    """Get the mainline revisions from the branch.

    Generates the list of mainline revisions for the branch.

    :param  branch: The branch containing the revisions. 

    :param  start_revision: The first revision to be logged.

            For backwards compatibility this may be a mainline integer revno,

            but for merge revision support a RevisionInfo is expected.

    :param  end_revision: The last revision to be logged.

            For backwards compatibility this may be a mainline integer revno,

            but for merge revision support a RevisionInfo is expected.

    :return: A (mainline_revs, rev_nos, start_rev_id, end_rev_id) tuple.

    """

    branch_revno, branch_last_revision = branch.last_revision_info()

    if branch_revno == 0:

        return None, None, None, None

    # For mainline generation, map start_revision and end_revision to 

    # mainline revnos. If the revision is not on the mainline choose the 

    # appropriate extreme of the mainline instead - the extra will be 

    # filtered later.

    # Also map the revisions to rev_ids, to be used in the later filtering

    # stage.

    start_rev_id = None 

    if start_revision is None:

        start_revno = 1

    else:

        if isinstance(start_revision, revisionspec.RevisionInfo):

            start_rev_id = start_revision.rev_id

            start_revno = start_revision.revno or 1

        else:

            branch.check_real_revno(start_revision)

            start_revno = start_revision

    end_rev_id = None

    if end_revision is None:

        end_revno = branch_revno

    else:

        if isinstance(end_revision, revisionspec.RevisionInfo):

            end_rev_id = end_revision.rev_id

            end_revno = end_revision.revno or branch_revno

        else:

            branch.check_real_revno(end_revision)

            end_revno = end_revision

    if ((start_rev_id == _mod_revision.NULL_REVISION)

        or (end_rev_id == _mod_revision.NULL_REVISION)):

        raise errors.BzrCommandError('Logging revision 0 is invalid.')

    if start_revno > end_revno:

        raise errors.BzrCommandError("Start revision must be older than "

                                     "the end revision.")

    if end_revno < start_revno:

        return None, None, None, None

    cur_revno = branch_revno

    rev_nos = {}

    mainline_revs = []

    for revision_id in branch.repository.iter_reverse_revision_history(

                        branch_last_revision):

        if cur_revno < start_revno:

            # We have gone far enough, but we always add 1 more revision

            rev_nos[revision_id] = cur_revno

            mainline_revs.append(revision_id)

            break

        if cur_revno <= end_revno:

            rev_nos[revision_id] = cur_revno

            mainline_revs.append(revision_id)

        cur_revno -= 1

    else:

        # We walked off the edge of all revisions, so we add a 'None' marker

        mainline_revs.append(None)

    mainline_revs.reverse()

    # override the mainline to look like the revision history.

    return mainline_revs, rev_nos, start_rev_id, end_rev_id

def _filter_revision_range(view_revisions, start_rev_id, end_rev_id):

    """Filter view_revisions based on revision ranges.

    :param view_revisions: A list of (revision_id, dotted_revno, merge_depth) 

            tuples to be filtered.

    :param start_rev_id: If not NONE specifies the first revision to be logged.

            If NONE then all revisions up to the end_rev_id are logged.

    :param end_rev_id: If not NONE specifies the last revision to be logged.

            If NONE then all revisions up to the end of the log are logged.

    :return: The filtered view_revisions.

    """

    if start_rev_id or end_rev_id: 

        revision_ids = [r for r, n, d in view_revisions]

        if start_rev_id:

            start_index = revision_ids.index(start_rev_id)

        else:

            start_index = 0

        if start_rev_id == end_rev_id:

            end_index = start_index

        else:

            if end_rev_id:

                end_index = revision_ids.index(end_rev_id)

            else:

                end_index = len(view_revisions) - 1

        # To include the revisions merged into the last revision, 

        # extend end_rev_id down to, but not including, the next rev

        # with the same or lesser merge_depth

        end_merge_depth = view_revisions[end_index][2]

        try:

            for index in xrange(end_index+1, len(view_revisions)+1):

                if view_revisions[index][2] <= end_merge_depth:

                    end_index = index - 1

                    break

        except IndexError:

            # if the search falls off the end then log to the end as well

            end_index = len(view_revisions) - 1

        view_revisions = view_revisions[start_index:end_index+1]

    return view_revisions

def _filter_revisions_touching_file_id(branch, file_id, view_revisions,

                                       direction):

    r"""Return the list of revision ids which touch a given file id.

    The function filters view_revisions and returns a subset.

    This includes the revisions which directly change the file id,

    and the revisions which merge these changes. So if the

    revision graph is::

        A-.

        |\ \

        B C E

        |/ /

        D |

        |\|

        | F

        |/

        G

    And 'C' changes a file, then both C and D will be returned. F will not be

    returned even though it brings the changes to C into the branch starting

    with E. (Note that if we were using F as the tip instead of G, then we

    would see C, D, F.)

    This will also be restricted based on a subset of the mainline.

    :param branch: The branch where we can get text revision information.

    :param file_id: Filter out revisions that do not touch file_id.

    :param view_revisions: A list of (revision_id, dotted_revno, merge_depth)

        tuples. This is the list of revisions which will be filtered. It is

        assumed that view_revisions is in merge_sort order (either forward or

        reverse).

    :param direction: The direction of view_revisions.  See also

        reverse_by_depth, and get_view_revisions

    :return: A list of (revision_id, dotted_revno, merge_depth) tuples.

    """

    # Lookup all possible text keys to determine which ones actually modified

    # the file.

    text_keys = [(file_id, rev_id) for rev_id, revno, depth in view_revisions]

    # Looking up keys in batches of 1000 can cut the time in half, as well as

    # memory consumption. GraphIndex *does* like to look for a few keys in

    # parallel, it just doesn't like looking for *lots* of keys in parallel.

    # TODO: This code needs to be re-evaluated periodically as we tune the

    #       indexing layer. We might consider passing in hints as to the known

    #       access pattern (sparse/clustered, high success rate/low success

    #       rate). This particular access is clustered with a low success rate.

    get_parent_map = branch.repository.texts.get_parent_map

    modified_text_revisions = set()

    chunk_size = 1000

    for start in xrange(0, len(text_keys), chunk_size):

        next_keys = text_keys[start:start + chunk_size]

        # Only keep the revision_id portion of the key

        modified_text_revisions.update(

            [k[1] for k in get_parent_map(next_keys)])

    del text_keys, next_keys

    result = []

    if direction == 'forward':

        # TODO: The algorithm for finding 'merges' of file changes expects

        #       'reverse' order (the default from 'merge_sort()'). Instead of

        #       forcing this, we could just use the reverse_by_depth order.

        view_revisions = reverse_by_depth(view_revisions)

    # Track what revisions will merge the current revision, replace entries

    # with 'None' when they have been added to result

    current_merge_stack = [None]

    for info in view_revisions:

        rev_id, revno, depth = info

        if depth == len(current_merge_stack):

            current_merge_stack.append(info)

        else:

            del current_merge_stack[depth + 1:]

            current_merge_stack[-1] = info

        if rev_id in modified_text_revisions:

            # This needs to be logged, along with the extra revisions

            for idx in xrange(len(current_merge_stack)):

                node = current_merge_stack[idx]

                if node is not None:

                    result.append(node)

                    current_merge_stack[idx] = None

    if direction == 'forward':

        result = reverse_by_depth(result)

    return result

def get_view_revisions(mainline_revs, rev_nos, branch, direction,

                       include_merges=True):

    """Produce an iterator of revisions to show

    :return: an iterator of (revision_id, revno, merge_depth)

    (if there is no revno for a revision, None is supplied)

    """

    if include_merges is False:

        revision_ids = mainline_revs[1:]

        if direction == 'reverse':

            revision_ids.reverse()

        for revision_id in revision_ids:

            yield revision_id, str(rev_nos[revision_id]), 0

        return

    graph = branch.repository.get_graph()

    # This asks for all mainline revisions, which means we only have to spider

    # sideways, rather than depth history. That said, its still size-of-history

    # and should be addressed.

    # mainline_revisions always includes an extra revision at the beginning, so

    # don't request it.

    parent_map = dict(((key, value) for key, value in

        graph.iter_ancestry(mainline_revs[1:]) if value is not None))

    # filter out ghosts; merge_sort errors on ghosts.

    rev_graph = _mod_repository._strip_NULL_ghosts(parent_map)

    merge_sorted_revisions = tsort.merge_sort(

        rev_graph,

        mainline_revs[-1],

        mainline_revs,

        generate_revno=True)

    if direction == 'forward':

        # forward means oldest first.

        merge_sorted_revisions = reverse_by_depth(merge_sorted_revisions)

    elif direction != 'reverse':

        raise ValueError('invalid direction %r' % direction)

    for sequence, rev_id, merge_depth, revno, end_of_merge in merge_sorted_revisions:

        yield rev_id, '.'.join(map(str, revno)), merge_depth

def reverse_by_depth(merge_sorted_revisions, _depth=0):

    """Reverse revisions by depth.

    Revisions with a different depth are sorted as a group with the previous

    revision of that depth.  There may be no topological justification for this,

    but it looks much nicer.

    """

    zd_revisions = []

    for val in merge_sorted_revisions:

        if val[2] == _depth:

            zd_revisions.append([val])

        else:

            zd_revisions[-1].append(val)

    for revisions in zd_revisions:

        if len(revisions) > 1:

            revisions[1:] = reverse_by_depth(revisions[1:], _depth + 1)

    zd_revisions.reverse()

    result = []

    for chunk in zd_revisions:

        result.extend(chunk)

    return result

class LogRevision(object):

    """A revision to be logged (by LogFormatter.log_revision).

    A simple wrapper for the attributes of a revision to be logged.

    The attributes may or may not be populated, as determined by the 

    logging options and the log formatter capabilities.

    """

    def __init__(self, rev=None, revno=None, merge_depth=0, delta=None,

                 tags=None):

        self.rev = rev

        self.revno = revno

        self.merge_depth = merge_depth

        self.delta = delta

        self.tags = tags

class LogFormatter(object):

    """Abstract class to display log messages.

    At a minimum, a derived class must implement the log_revision method.

    If the LogFormatter needs to be informed of the beginning or end of

    a log it should implement the begin_log and/or end_log hook methods.

    A LogFormatter should define the following supports_XXX flags 

    to indicate which LogRevision attributes it supports:

    - supports_delta must be True if this log formatter supports delta.

        Otherwise the delta attribute may not be populated.

    - supports_merge_revisions must be True if this log formatter supports 

        merge revisions.  If not, and if supports_single_merge_revisions is

        also not True, then only mainline revisions will be passed to the 

        formatter.

    - supports_single_merge_revision must be True if this log formatter

        supports logging only a single merge revision.  This flag is

        only relevant if supports_merge_revisions is not True.

    - supports_tags must be True if this log formatter supports tags.

        Otherwise the tags attribute may not be populated.

    Plugins can register functions to show custom revision properties using

    the properties_handler_registry. The registered function

    must respect the following interface description:

        def my_show_properties(properties_dict):

            # code that returns a dict {'name':'value'} of the properties 

            # to be shown

    """