/brz/remove-bazaar

#!/usr/bin/env python

"""\

Read in a changeset output, and process it into a Changeset object.

"""

import os

import pprint

from cStringIO import StringIO

from bzrlib.tree import Tree

from bzrlib.trace import mutter, warning

from bzrlib.errors import BzrError

from bzrlib.xml5 import serializer_v5

from bzrlib.osutils import sha_file, sha_string

from bzrlib.revision import Revision

from bzrlib.inventory import Inventory, InventoryEntry

from common import decode, get_header, header_str

class BadChangeset(Exception): pass

class MalformedHeader(BadChangeset): pass

class MalformedPatches(BadChangeset): pass

class MalformedFooter(BadChangeset): pass

def _unescape(name):

    """Now we want to find the filename effected.

    Unfortunately the filename is written out as

    repr(filename), which means that it surrounds

    the name with quotes which may be single or double

    (single is preferred unless there is a single quote in

    the filename). And some characters will be escaped.

    TODO:   There has to be some pythonic way of undo-ing the

            representation of a string rather than using eval.

    """

    delimiter = name[0]

    if name[-1] != delimiter:

        raise BadChangeset('Could not properly parse the'

                ' filename: %r' % name)

    # We need to handle escaped hexadecimals too.

    return name[1:-1].replace('\"', '"').replace("\'", "'")

class RevisionInfo(object):

    """Gets filled out for each revision object that is read.

    """

    def __init__(self, revision_id):

        self.revision_id = revision_id

        self.sha1 = None

        self.committer = None

        self.date = None

        self.timestamp = None

        self.timezone = None

        self.inventory_sha1 = None

        self.parents = None

        self.parent_sha1s = {}

        self.message = None

    def __str__(self):

        return pprint.pformat(self.__dict__)

    def as_revision(self):

        rev = Revision(revision_id=self.revision_id,

            committer=self.committer,

            timestamp=float(self.timestamp),

            timezone=int(self.timezone),

            inventory_sha1=self.inventory_sha1,

            message='\n'.join(self.message))

        if self.parents:

            for parent in self.parents:

                revision_id, sha1 = parent.split()

                rev.parent_ids.append(revision_id)

                self.parent_sha1s[revision_id] = sha1

        return rev

class ChangesetInfo(object):

    """This contains the meta information. Stuff that allows you to

    recreate the revision or inventory XML.

    """

    def __init__(self):

        self.committer = None

        self.date = None

        self.message = None

        self.base = None

        self.base_sha1 = None

        # A list of RevisionInfo objects

        self.revisions = []

        self.actions = []

        # The next entries are created during complete_info() and

        # other post-read functions.

        # A list of real Revision objects

        self.real_revisions = []

        self.timestamp = None

        self.timezone = None

    def __str__(self):

        return pprint.pformat(self.__dict__)

    def complete_info(self):

        """This makes sure that all information is properly

        split up, based on the assumptions that can be made

        when information is missing.

        """

        from common import unpack_highres_date

        # Put in all of the guessable information.

        if not self.timestamp and self.date:

            self.timestamp, self.timezone = unpack_highres_date(self.date)

        self.real_revisions = []

        for rev in self.revisions:

            if rev.timestamp is None:

                if rev.date is not None:

                    rev.timestamp, rev.timezone = \

                            unpack_highres_date(rev.date)

                else:

                    rev.timestamp = self.timestamp

                    rev.timezone = self.timezone

            if rev.message is None and self.message:

                rev.message = self.message

            if rev.committer is None and self.committer:

                rev.committer = self.committer

            self.real_revisions.append(rev.as_revision())

        if self.base is None:

            # When we don't have a base, then the real base

            # is the first parent of the first revision listed

            rev = self.real_revisions[0]

            if len(rev.parent_ids) == 0:

                # There is no base listed, and

                # the lowest revision doesn't have a parent

                # so this is probably against the empty tree

                # and thus base truly is None

                self.base = None

                self.base_sha1 = None

            else:

                self.base = rev.parent_ids[0]

                # In general, if self.base is None, self.base_sha1 should

                # also be None

                raise NotImplementedError('Cannot determine self.base_sha1, need some sort of branch')

    def _get_target(self):

        """Return the target revision."""

        if len(self.real_revisions) > 0:

            return self.real_revisions[-1].revision_id

        elif len(self.revisions) > 0:

            return self.revisions[-1].revision_id

        return None

    target = property(_get_target, doc='The target revision id')

class ChangesetReader(object):

    """This class reads in a changeset from a file, and returns

    a Changeset object, which can then be applied against a tree.

    """

    def __init__(self, from_file):

        """Read in the changeset from the file.

        :param from_file: A file-like object (must have iterator support).

        """

        object.__init__(self)

        self.from_file = from_file

        self._next_line = None

        self.info = ChangesetInfo()

        # We put the actual inventory ids in the footer, so that the patch

        # is easier to read for humans.

        # Unfortunately, that means we need to read everything before we

        # can create a proper changeset.

        self._read()

        self._validate()

    def _read(self):

        self._read_header()

        self._read_patches()

        self._read_footer()

    def _validate(self):

        """Make sure that the information read in makes sense

        and passes appropriate checksums.

        """

        # Fill in all the missing blanks for the revisions

        # and generate the real_revisions list.

        self.info.complete_info()

        self._validate_revisions()

    def _validate_revisions(self):

        """Make sure all revision entries match their checksum."""

        # This is a mapping from each revision id to it's sha hash

        rev_to_sha1 = {}

        for rev, rev_info in zip(self.info.real_revisions, self.info.revisions):

            assert rev.revision_id == rev_info.revision_id

            sio = StringIO()

            serializer_v5.write_revision(rev, sio)

            sio.seek(0)

            sha1 = sha_file(sio)

            if sha1 != rev_info.sha1:

                raise BzrError('Revision checksum mismatch.'

                    ' For revision_id {%s} supplied sha1 (%s) != measured (%s)'

                    % (rev.revision_id, rev_info.sha1, sha1))

            if rev_to_sha1.has_key(rev.revision_id):

                raise BzrError('Revision {%s} given twice in the list'

                        % (rev.revision_id))

            rev_to_sha1[rev.revision_id] = sha1

        # Now that we've checked all the sha1 sums, we can make sure that

        # at least for the small list we have, all of the references are

        # valid.

        for rev in self.info.real_revisions:

            for p_id in rev.parent_ids:

                if p_id in rev_to_sha1:

                    if parent.revision_sha1 != rev_to_sha1[parent.revision_id]:

                        raise BzrError('Parent revision checksum mismatch.'

                                ' A parent was referenced with an'

                                ' incorrect checksum'

                                ': {%r} %s != %s' % (parent.revision_id,

                                            parent.revision_sha1,

                                            rev_to_sha1[parent.revision_id]))

    def _validate_references_from_branch(self, branch):

        """Now that we have a branch which should have some of the

        revisions we care about, go through and validate all of them

        that we can.

        """

        rev_to_sha = {}

        inv_to_sha = {}

        def add_sha(d, revision_id, sha1):

            if revision_id is None:

                if sha1 is not None:

                    raise BzrError('A Null revision should always'

                        'have a null sha1 hash')

                return

            if revision_id in d:

                # This really should have been validated as part

                # of _validate_revisions but lets do it again

                if sha1 != d[revision_id]:

                    raise BzrError('** Revision %r referenced with 2 different'

                            ' sha hashes %s != %s' % (revision_id,

                                sha1, d[revision_id]))

            else:

                d[revision_id] = sha1

        add_sha(rev_to_sha, self.info.base, self.info.base_sha1)

        # All of the contained revisions were checked

        # in _validate_revisions

        checked = {}

        for rev_info in self.info.revisions:

            checked[rev_info.revision_id] = True

            add_sha(rev_to_sha, rev_info.revision_id, rev_info.sha1)

        for (rev, rev_info) in zip(self.info.real_revisions, self.info.revisions):

            add_sha(inv_to_sha, rev_info.revision_id, rev_info.inventory_sha1)

            for p_id, sha1 in rev_info.parent_sha1s.iteritems():

                add_sha(rev_to_sha, p_id, sha1)

        count = 0

        missing = {}

        for revision_id, sha1 in rev_to_sha.iteritems():

            if branch.has_revision(revision_id):

                local_sha1 = branch.get_revision_sha1(revision_id)

                if sha1 != local_sha1:

                    raise BzrError('sha1 mismatch. For revision id {%s}' 

                            'local: %s, cset: %s' % (revision_id, local_sha1, sha1))

                else:

                    count += 1

            elif revision_id not in checked:

                missing[revision_id] = sha1

        for inv_id, sha1 in inv_to_sha.iteritems():

            if branch.has_revision(inv_id):

                # TODO: Currently branch.get_inventory_sha1() just returns the value

                # that is stored in the revision text. Which is *really* bogus, because

                # that means we aren't validating the actual text, just that we wrote 

                # and read the string. But for now, what the hell.

                local_sha1 = branch.get_inventory_sha1(inv_id)

                if sha1 != local_sha1:

                    raise BzrError('sha1 mismatch. For inventory id {%s}' 

                            'local: %s, cset: %s' % (inv_id, local_sha1, sha1))

                else:

                    count += 1

        if len(missing) > 0:

            # I don't know if this is an error yet

            warning('Not all revision hashes could be validated.'

                    ' Unable validate %d hashes' % len(missing))

        mutter('Verified %d sha hashes for the changeset.' % count)

    def _validate_inventory(self, inv):

        """At this point we should have generated the ChangesetTree,

        so build up an inventory, and make sure the hashes match.

        """

        assert inv is not None

        # Now we should have a complete inventory entry.

        sio = StringIO()

        serializer_v5.write_inventory(inv, sio)

        sio.seek(0)

        sha1 = sha_file(sio)

        # Target revision is the last entry in the real_revisions list

        rev = self.info.real_revisions[-1]

        if sha1 != rev.inventory_sha1:

            open(',,bogus-inv', 'wb').write(sio.getvalue())

            raise BzrError('Inventory sha hash mismatch.')

    def get_changeset(self, branch):

        """Return the meta information, and a Changeset tree which can

        be used to populate the local stores and working tree, respectively.

        """

        self._validate_references_from_branch(branch)

        cset_tree = ChangesetTree(branch.revision_tree(self.info.base))

        self._update_tree(cset_tree)

        inv = cset_tree.inventory

        self._validate_inventory(inv)

        return self.info, cset_tree

    def _next(self):

        """yield the next line, but secretly

        keep 1 extra line for peeking.

        """

        for line in self.from_file:

            last = self._next_line

            self._next_line = line

            if last is not None:

                #mutter('yielding line: %r' % last)

                yield last

        last = self._next_line

        self._next_line = None

        #mutter('yielding line: %r' % last)

        yield last

    def _read_header(self):

        """Read the bzr header"""

        header = get_header()

        found = False

        for line in self._next():

            if found:

                # not all mailers will keep trailing whitespace

                if line == '#\n':

                    line = '# \n'

                if (not line.startswith('# ') or not line.endswith('\n')

                        or decode(line[2:-1]) != header[0]):

                    raise MalformedHeader('Found a header, but it'

                        ' was improperly formatted')

                header.pop(0) # We read this line.

                if not header:

                    break # We found everything.

            elif (line.startswith('#') and line.endswith('\n')):

                line = decode(line[1:-1].strip())

                if line[:len(header_str)] == header_str:

                    if line == header[0]:

                        found = True

                    else:

                        raise MalformedHeader('Found what looks like'

                                ' a header, but did not match')

                    header.pop(0)

        else:

            raise MalformedHeader('Did not find an opening header')

        for line in self._next():

            # The bzr header is terminated with a blank line

            # which does not start with '#'

            if line == '\n':

                break

            self._handle_next(line)

    def _read_next_entry(self, line, indent=1):

        """Read in a key-value pair

        """

        if not line.startswith('#'):

            raise MalformedHeader('Bzr header did not start with #')

        line = decode(line[1:-1]) # Remove the '#' and '\n'

        if line[:indent] == ' '*indent:

            line = line[indent:]

        if not line:

            return None, None# Ignore blank lines

        loc = line.find(': ')

        if loc != -1:

            key = line[:loc]

            value = line[loc+2:]

            if not value:

                value = self._read_many(indent=indent+3)

        elif line[-1:] == ':':

            key = line[:-1]

            value = self._read_many(indent=indent+3)

        else:

            raise MalformedHeader('While looking for key: value pairs,'

                    ' did not find the colon %r' % (line))

        key = key.replace(' ', '_')

        #mutter('found %s: %s' % (key, value))

        return key, value

    def _handle_next(self, line):

        key, value = self._read_next_entry(line, indent=1)

        if key is None:

            return

        if key == 'revision':

            self._read_revision(value)

        elif hasattr(self.info, key):

            if getattr(self.info, key) is None:

                setattr(self.info, key, value)

            else:

                raise MalformedHeader('Duplicated Key: %s' % key)

        else:

            # What do we do with a key we don't recognize

            raise MalformedHeader('Unknown Key: %s' % key)

    def _read_many(self, indent):

        """If a line ends with no entry, that means that it should be

        followed with multiple lines of values.

        This detects the end of the list, because it will be a line that

        does not start properly indented.

        """

        values = []

        start = '#' + (' '*indent)

        if self._next_line is None or self._next_line[:len(start)] != start:

            return values

        for line in self._next():

            values.append(decode(line[len(start):-1]))

            if self._next_line is None or self._next_line[:len(start)] != start:

                break

        return values

    def _read_one_patch(self):

        """Read in one patch, return the complete patch, along with

        the next line.

        :return: action, lines, do_continue

        """

        #mutter('_read_one_patch: %r' % self._next_line)

        # Peek and see if there are no patches

        if self._next_line is None or self._next_line.startswith('#'):

            return None, [], False

        first = True

        lines = []

        for line in self._next():

            if first:

                if not line.startswith('==='):

                    raise MalformedPatches('The first line of all patches'

                        ' should be a bzr meta line "==="'

                        ': %r' % line)

                action = decode(line[4:-1])

            if self._next_line is not None and self._next_line.startswith('==='):

                return action, lines, True

            elif self._next_line is None or self._next_line.startswith('#'):

                return action, lines, False

            if first:

                first = False

            else:

                lines.append(line)

        return action, lines, False

    def _read_patches(self):

        do_continue = True

        while do_continue:

            action, lines, do_continue = self._read_one_patch()

            if action is not None:

                self.info.actions.append((action, lines))

    def _read_revision(self, revision_id):

        """Revision entries have extra information associated.

        """

        rev_info = RevisionInfo(revision_id)

        start = '#    '

        for line in self._next():

            key,value = self._read_next_entry(line, indent=4)

            #if key is None:

            #    continue

            if hasattr(rev_info, key):

                if getattr(rev_info, key) is None:

                    setattr(rev_info, key, value)

                else:

                    raise MalformedHeader('Duplicated Key: %s' % key)

            else:

                # What do we do with a key we don't recognize

                raise MalformedHeader('Unknown Key: %s' % key)

            if self._next_line is None or self._next_line[:len(start)] != start:

                break

        self.info.revisions.append(rev_info)

    def _read_footer(self):

        """Read the rest of the meta information.

        :param first_line:  The previous step iterates past what it

                            can handle. That extra line is given here.

        """

        for line in self._next():

            self._handle_next(line)

            if self._next_line is None or self._next_line.startswith('#'):

                break

    def _update_tree(self, cset_tree):

        """This fills out a ChangesetTree based on the information

        that was read in.

        :param cset_tree: A ChangesetTree to update with the new information.

        """

        def get_rev_id(info, file_id, kind):

            if info is not None:

                if not info.starts_with('last-changed:'):

                    raise BzrError("Last changed revision should start with 'last-changed:'"

                        ': %r' % info)

                revision_id = decode(info[13:])

            elif cset_tree._last_changed_revision_ids.has_key(file_id):

                return cset_tree._last_changed_revision_ids[file_id]

            else:

                revision_id = self.info.revisions[-1].revision_id

            cset_tree.note_last_changed(file_id, revision_id)

            return revision_id

        def renamed(kind, extra, lines):

            info = extra.split(' // ')

            if len(info) < 2:

                raise BzrError('renamed action lines need both a from and to'

                        ': %r' % extra)

            old_path = info[0]

            if info[1].startswith('=> '):

                new_path = info[1][3:]

            else:

                new_path = info[1]

            file_id = cset_tree.path2id(old_path)

            # print '%r %r %r' % (old_path, new_path, file_id)

            if len(info) > 2:

                revision = get_rev_id(info[2], file_id, kind)

            else:

                revision = get_rev_id(None, file_id, kind)

            cset_tree.note_rename(old_path, new_path)

            if lines:

                cset_tree.note_patch(new_path, ''.join(lines))

        def removed(kind, extra, lines):

            info = extra.split(' // ')

            if len(info) > 1:

                # TODO: in the future we might allow file ids to be

                # given for removed entries

                raise BzrError('removed action lines should only have the path'

                        ': %r' % extra)

            path = info[0]

            cset_tree.note_deletion(path)

        def added(kind, extra, lines):

            info = extra.split(' // ')

            if len(info) <= 1:

                raise BzrError('add action lines require the path and file id'

                        ': %r' % extra)

            elif len(info) > 3:

                raise BzrError('add action lines have fewer than 3 entries.'

                        ': %r' % extra)

            path = info[0]

            if not info[1].startswith('file-id:'):

                raise BzrError('The file-id should follow the path for an add'

                        ': %r' % extra)

            file_id = info[1][8:]

            cset_tree.note_id(file_id, path, kind)

            if kind == 'directory':

                return

            if len(info) > 2:

                revision = get_rev_id(info[2], file_id, kind)

            else:

                revision = get_rev_id(None, file_id, kind)

            cset_tree.note_patch(path, ''.join(lines))

        def modified(kind, extra, lines):

            info = extra.split(' // ')

            if len(info) < 1:

                raise BzrError('modified action lines have at least'

                        'the path in them: %r' % extra)

            path = info[0]

            file_id = cset_tree.path2id(path)

            if len(info) > 1:

                revision = get_rev_id(info[1], file_id, kind)

            else:

                revision = get_rev_id(None, file_id, kind)

            cset_tree.note_patch(path, ''.join(lines))

        valid_actions = {

            'renamed':renamed,

            'removed':removed,

            'added':added,

            'modified':modified

        }

        for action_line, lines in self.info.actions:

            first = action_line.find(' ')

            if first == -1:

                raise BzrError('Bogus action line'

                        ' (no opening space): %r' % action_line)

            second = action_line.find(' ', first+1)