/brz/remove-bazaar

# groupcompress, a bzr plugin providing new compression logic.

# Copyright (C) 2008 Canonical Limited.

# 

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

# it under the terms of the GNU General Public License version 2 as published

# by the Free Software Foundation.

# 

# 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., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301 USA

# 

"""Core compression logic for compressing streams of related files."""

from itertools import izip

from cStringIO import StringIO

import zlib

from bzrlib import (

    annotate,

    debug,

    diff,

    errors,

    graph as _mod_graph,

    osutils,

    pack,

    patiencediff,

    )

from bzrlib.graph import Graph

from bzrlib.knit import _DirectPackAccess

from bzrlib.osutils import (

    contains_whitespace,

    contains_linebreaks,

    sha_string,

    sha_strings,

    split_lines,

    )

from bzrlib.btree_index import BTreeBuilder

from bzrlib.lru_cache import LRUSizeCache

from bzrlib.plugins.groupcompress import equivalence_table

from bzrlib.tsort import topo_sort

from bzrlib.versionedfile import (

    adapter_registry,

    AbsentContentFactory,

    ChunkedContentFactory,

    FulltextContentFactory,

    VersionedFiles,

    )

def parse(line_list):

    result = []

    lines = iter(line_list)

    next = lines.next

    label_line = lines.next()

    sha1_line = lines.next()

    if (not label_line.startswith('label: ') or

        not sha1_line.startswith('sha1: ')):

        raise AssertionError("bad text record %r" % lines)

    label = tuple(label_line[7:-1].split('\x00'))

    sha1 = sha1_line[6:-1]

    for header in lines:

        op = header[0]

        numbers = header[2:]

        numbers = [int(n) for n in header[2:].split(',')]

        if op == 'c':

            result.append((op, numbers[0], numbers[1], None))

        else:

            contents = [next() for i in xrange(numbers[0])]

            result.append((op, None, numbers[0], contents))

    return label, sha1, result

def apply_delta(basis, delta):

    """Apply delta to this object to become new_version_id."""

    lines = []

    last_offset = 0

    # eq ranges occur where gaps occur

    # start, end refer to offsets in basis

    for op, start, count, delta_lines in delta:

        if op == 'c':

            lines.append(basis[start:start+count])

        else:

            lines.extend(delta_lines)

    trim_encoding_newline(lines)

    return lines

def trim_encoding_newline(lines):

    if lines[-1] == '\n':

        del lines[-1]

    else:

        lines[-1] = lines[-1][:-1]

class GroupCompressor(object):

    """Produce a serialised group of compressed texts.

    It contains code very similar to SequenceMatcher because of having a similar

    task. However some key differences apply:

     - there is no junk, we want a minimal edit not a human readable diff.

     - we don't filter very common lines (because we don't know where a good

       range will start, and after the first text we want to be emitting minmal

       edits only.

     - we chain the left side, not the right side

     - we incrementally update the adjacency matrix as new lines are provided.

     - we look for matches in all of the left side, so the routine which does

       the analagous task of find_longest_match does not need to filter on the

       left side.

    """

    _equivalence_table_class = equivalence_table.EquivalenceTable

    def __init__(self, delta=True):

        """Create a GroupCompressor.

        :paeam delta: If False, do not compress records.

        """

        self._delta = delta

        self.line_offsets = []

        self.endpoint = 0

        self.input_bytes = 0

        self.line_locations = self._equivalence_table_class([])

        self.lines = self.line_locations.lines

        self.labels_deltas = {}

    def get_matching_blocks(self, lines):

        """Return an the ranges in lines which match self.lines.

        :param lines: lines to compress

        :return: A list of (old_start, new_start, length) tuples which reflect

            a region in self.lines that is present in lines.  The last element

            of the list is always (old_len, new_len, 0) to provide a end point

            for generating instructions from the matching blocks list.

        """

        result = []

        pos = 0

        line_locations = self.line_locations

        line_locations.set_right_lines(lines)

        # We either copy a range (while there are reusable lines) or we 

        # insert new lines. To find reusable lines we traverse 

        locations = None

        max_pos = len(lines)

        max_time = 0.0

        max_info = None

        result_append = result.append

        while pos < max_pos:

            block, pos, locations = _get_longest_match(line_locations, pos,

                                                       max_pos, locations)

            if block is not None:

                result_append(block)

        result_append((len(self.lines), len(lines), 0))

        return result

    def compress(self, key, lines, expected_sha):

        """Compress lines with label key.

        :param key: A key tuple. It is stored in the output

            for identification of the text during decompression. If the last

            element is 'None' it is replaced with the sha1 of the text -

            e.g. sha1:xxxxxxx.

        :param lines: The lines to be compressed. Must be split

            on \n, with the \n preserved.'

        :param expected_sha: If non-None, the sha the lines are blieved to

            have. During compression the sha is calculated; a mismatch will

            cause an error.

        :return: The sha1 of lines, and the number of bytes accumulated in

            the group output so far.

        """

        sha1 = sha_strings(lines)

        if key[-1] is None:

            key = key[:-1] + ('sha1:' + sha1,)

        label = '\x00'.join(key)

        # setup good encoding for trailing \n support.

        if not lines or lines[-1].endswith('\n'):

            lines.append('\n')

        else:

            lines[-1] = lines[-1] + '\n'

        new_lines = []

        new_lines.append('label: %s\n' % label)

        new_lines.append('sha1: %s\n' % sha1)

        index_lines = [False, False]

        pos = 0

        range_len = 0

        range_start = 0

        flush_range = self.flush_range

        copy_ends = None

        blocks = self.get_matching_blocks(lines)

        current_pos = 0

        # We either copy a range (while there are reusable lines) or we 

        # insert new lines. To find reusable lines we traverse 

        for old_start, new_start, range_len in blocks:

            if new_start != current_pos:

                # non-matching region

                flush_range(current_pos, None, new_start - current_pos,

                    lines, new_lines, index_lines)

            current_pos = new_start + range_len

            if not range_len:

                continue

            flush_range(new_start, old_start, range_len, lines,

                new_lines, index_lines)

        delta_start = (self.endpoint, len(self.lines))

        self.output_lines(new_lines, index_lines)

        trim_encoding_newline(lines)

        self.input_bytes += sum(map(len, lines))

        delta_end = (self.endpoint, len(self.lines))

        self.labels_deltas[key] = (delta_start, delta_end)

        return sha1, self.endpoint

    def extract(self, key):

        """Extract a key previously added to the compressor.

        :param key: The key to extract.

        :return: An iterable over bytes and the sha1.

        """

        delta_details = self.labels_deltas[key]

        delta_lines = self.lines[delta_details[0][1]:delta_details[1][1]]

        label, sha1, delta = parse(delta_lines)

        if label != key:

            raise AssertionError("wrong key: %r, wanted %r" % (label, key))

        # Perhaps we want to keep the line offsets too in memory at least?

        lines = apply_delta(''.join(self.lines), delta)

        sha1 = sha_strings(lines)

        return lines, sha1

    def flush_range(self, range_start, copy_start, range_len, lines, new_lines, index_lines):

        insert_instruction = "i,%d\n" % range_len

        if copy_start is not None:

            # range stops, flush and start a new copy range

            stop_byte = self.line_offsets[copy_start + range_len - 1]

            if copy_start == 0:

                start_byte = 0

            else:

                start_byte = self.line_offsets[copy_start - 1]

            bytes = stop_byte - start_byte

            copy_control_instruction = "c,%d,%d\n" % (start_byte, bytes)

            if (bytes + len(insert_instruction) >

                len(copy_control_instruction)):

                new_lines.append(copy_control_instruction)

                index_lines.append(False)

                return

        # not copying, or inserting is shorter than copying, so insert.

        new_lines.append(insert_instruction)

        new_lines.extend(lines[range_start:range_start+range_len])

        index_lines.append(False)

        index_lines.extend([copy_start is None]*range_len)

    def output_lines(self, new_lines, index_lines):

        """Output some lines.

        :param new_lines: The lines to output.

        :param index_lines: A boolean flag for each line - when True, index

            that line.

        """

        # indexed_newlines = [idx for idx, val in enumerate(index_lines)

        #                          if val and new_lines[idx] == '\n']

        # if indexed_newlines:

        #     import pdb; pdb.set_trace()

        endpoint = self.endpoint

        self.line_locations.extend_lines(new_lines, index_lines)

        for line in new_lines:

            endpoint += len(line)

            self.line_offsets.append(endpoint)

        self.endpoint = endpoint

    def ratio(self):

        """Return the overall compression ratio."""

        return float(self.input_bytes) / float(self.endpoint)

def make_pack_factory(graph, delta, keylength):

    """Create a factory for creating a pack based groupcompress.

    This is only functional enough to run interface tests, it doesn't try to

    provide a full pack environment.

    :param graph: Store a graph.

    :param delta: Delta compress contents.

    :param keylength: How long should keys be.

    """

    def factory(transport):

        parents = graph or delta

        ref_length = 0

        if graph:

            ref_length += 1

        graph_index = BTreeBuilder(reference_lists=ref_length,

            key_elements=keylength)

        stream = transport.open_write_stream('newpack')

        writer = pack.ContainerWriter(stream.write)

        writer.begin()

        index = _GCGraphIndex(graph_index, lambda:True, parents=parents,

            add_callback=graph_index.add_nodes)

        access = _DirectPackAccess({})

        access.set_writer(writer, graph_index, (transport, 'newpack'))

        result = GroupCompressVersionedFiles(index, access, delta)

        result.stream = stream

        result.writer = writer

        return result

    return factory

def cleanup_pack_group(versioned_files):

    versioned_files.writer.end()

    versioned_files.stream.close()

class GroupCompressVersionedFiles(VersionedFiles):

    """A group-compress based VersionedFiles implementation."""

    def __init__(self, index, access, delta=True):

        """Create a GroupCompressVersionedFiles object.

        :param index: The index object storing access and graph data.

        :param access: The access object storing raw data.

        :param delta: Whether to delta compress or just entropy compress.

        """

        self._index = index

        self._access = access

        self._delta = delta

        self._unadded_refs = {}

        self._group_cache = LRUSizeCache(max_size=50*1024*1024)

    def add_lines(self, key, parents, lines, parent_texts=None,

        left_matching_blocks=None, nostore_sha=None, random_id=False,

        check_content=True):

        """Add a text to the store.

        :param key: The key tuple of the text to add.

        :param parents: The parents key tuples of the text to add.

        :param lines: A list of lines. Each line must be a bytestring. And all

            of them except the last must be terminated with \n and contain no

            other \n's. The last line may either contain no \n's or a single

            terminating \n. If the lines list does meet this constraint the add

            routine may error or may succeed - but you will be unable to read

            the data back accurately. (Checking the lines have been split

            correctly is expensive and extremely unlikely to catch bugs so it

            is not done at runtime unless check_content is True.)

        :param parent_texts: An optional dictionary containing the opaque 

            representations of some or all of the parents of version_id to

            allow delta optimisations.  VERY IMPORTANT: the texts must be those

            returned by add_lines or data corruption can be caused.

        :param left_matching_blocks: a hint about which areas are common

            between the text and its left-hand-parent.  The format is

            the SequenceMatcher.get_matching_blocks format.

        :param nostore_sha: Raise ExistingContent and do not add the lines to

            the versioned file if the digest of the lines matches this.

        :param random_id: If True a random id has been selected rather than

            an id determined by some deterministic process such as a converter

            from a foreign VCS. When True the backend may choose not to check

            for uniqueness of the resulting key within the versioned file, so

            this should only be done when the result is expected to be unique

            anyway.

        :param check_content: If True, the lines supplied are verified to be

            bytestrings that are correctly formed lines.

        :return: The text sha1, the number of bytes in the text, and an opaque

                 representation of the inserted version which can be provided

                 back to future add_lines calls in the parent_texts dictionary.

        """

        self._index._check_write_ok()

        self._check_add(key, lines, random_id, check_content)

        if parents is None:

            # The caller might pass None if there is no graph data, but kndx

            # indexes can't directly store that, so we give them

            # an empty tuple instead.

            parents = ()

        # double handling for now. Make it work until then.

        length = sum(map(len, lines))

        record = ChunkedContentFactory(key, parents, None, lines)

        sha1 = list(self._insert_record_stream([record], random_id=random_id))[0]

        return sha1, length, None

    def annotate(self, key):

        """See VersionedFiles.annotate."""

        graph = Graph(self)

        parent_map = self.get_parent_map([key])

        if not parent_map:

            raise errors.RevisionNotPresent(key, self)

        if parent_map[key] is not None:

            search = graph._make_breadth_first_searcher([key])

            keys = set()

            while True:

                try:

                    present, ghosts = search.next_with_ghosts()

                except StopIteration:

                    break

                keys.update(present)

            parent_map = self.get_parent_map(keys)

        else:

            keys = [key]

            parent_map = {key:()}

        head_cache = _mod_graph.FrozenHeadsCache(graph)

        parent_cache = {}

        reannotate = annotate.reannotate

        for record in self.get_record_stream(keys, 'topological', True):

            key = record.key

            chunks = osutils.chunks_to_lines(record.get_bytes_as('chunked'))

            parent_lines = [parent_cache[parent] for parent in parent_map[key]]

            parent_cache[key] = list(

                reannotate(parent_lines, fulltext, key, None, head_cache))

        return parent_cache[key]

    def check(self, progress_bar=None):

        """See VersionedFiles.check()."""

        keys = self.keys()

        for record in self.get_record_stream(keys, 'unordered', True):

            record.get_bytes_as('fulltext')

    def _check_add(self, key, lines, random_id, check_content):

        """check that version_id and lines are safe to add."""

        version_id = key[-1]

        if version_id is not None:

            if contains_whitespace(version_id):

                raise InvalidRevisionId(version_id, self)

        self.check_not_reserved_id(version_id)

        # TODO: If random_id==False and the key is already present, we should

        # probably check that the existing content is identical to what is

        # being inserted, and otherwise raise an exception.  This would make

        # the bundle code simpler.

        if check_content:

            self._check_lines_not_unicode(lines)

            self._check_lines_are_lines(lines)

    def get_parent_map(self, keys):

        """Get a map of the parents of keys.

        :param keys: The keys to look up parents for.

        :return: A mapping from keys to parents. Absent keys are absent from

            the mapping.

        """

        result = {}

        sources = [self._index]

        source_results = []

        missing = set(keys)

        for source in sources:

            if not missing:

                break

            new_result = source.get_parent_map(missing)

            source_results.append(new_result)

            result.update(new_result)

            missing.difference_update(set(new_result))

        if self._unadded_refs:

            for key in missing:

                if key in self._unadded_refs:

                    result[key] = self._unadded_refs[key]

        return result

    def get_record_stream(self, keys, ordering, include_delta_closure):

        """Get a stream of records for keys.

        :param keys: The keys to include.

        :param ordering: Either 'unordered' or 'topological'. A topologically

            sorted stream has compression parents strictly before their

            children.

        :param include_delta_closure: If True then the closure across any

            compression parents will be included (in the opaque data).

        :return: An iterator of ContentFactory objects, each of which is only

            valid until the iterator is advanced.

        """

        # keys might be a generator

        keys = set(keys)

        if not keys:

            return

        if not self._index.has_graph:

            # Cannot topological order when no graph has been stored.

            ordering = 'unordered'

        # Cheap: iterate

        locations = self._index.get_build_details(keys)

        if ordering == 'topological':

            # would be better to not globally sort initially but instead

            # start with one key, recurse to its oldest parent, then grab

            # everything in the same group, etc.

            parent_map = dict((key, details[2]) for key, details in

                locations.iteritems())

            local = frozenset(keys).intersection(set(self._unadded_refs))

            for key in local:

                parent_map[key] = self._unadded_refs[key]

                locations[key] = None

            present_keys = topo_sort(parent_map)

            # Now group by source:

        else:

            present_keys = locations.keys()

            local = frozenset(keys).intersection(set(self._unadded_refs))

            for key in local:

                present_keys.append(key)

                locations[key] = None

        absent_keys = keys.difference(set(locations))

        for key in absent_keys:

            yield AbsentContentFactory(key)

        for key in present_keys:

            if key in self._unadded_refs:

                lines, sha1 = self._compressor.extract(key)

                parents = self._unadded_refs[key]

            else:

                index_memo, _, parents, (method, _) = locations[key]

                read_memo = index_memo[0:3]

                # get the group:

                try:

                    plain = self._group_cache[read_memo]

                except KeyError:

                    # read the group

                    zdata = self._access.get_raw_records([read_memo]).next()

                    # decompress - whole thing - this is not a bug, as it

                    # permits caching. We might want to store the partially

                    # decompresed group and decompress object, so that recent

                    # texts are not penalised by big groups.

                    decomp = zlib.decompressobj()

                    plain = decomp.decompress(zdata) #, index_memo[4])

                    self._group_cache[read_memo] = plain

                # cheapo debugging:

                # print len(zdata), len(plain)

                # parse - requires split_lines, better to have byte offsets

                # here (but not by much - we only split the region for the

                # recipe, and we often want to end up with lines anyway.

                delta_lines = split_lines(plain[index_memo[3]:index_memo[4]])

                label, sha1, delta = parse(delta_lines)

                if label != key:

                    raise AssertionError("wrong key: %r, wanted %r" % (label, key))

                lines = apply_delta(plain, delta)

            yield ChunkedContentFactory(key, parents, sha1, lines)

    def get_sha1s(self, keys):

        """See VersionedFiles.get_sha1s()."""

        result = {}

        for record in self.get_record_stream(keys, 'unordered', True):

            if record.sha1 != None:

                result[record.key] = record.sha1

            else:

                if record.storage_kind != 'absent':

                    result[record.key] == sha_string(record.get_bytes_as(

                        'fulltext'))

        return result

    def insert_record_stream(self, stream):

        """Insert a record stream into this container.

        :param stream: A stream of records to insert. 

        :return: None

        :seealso VersionedFiles.get_record_stream:

        """

        for _ in self._insert_record_stream(stream):

            pass

    def _insert_record_stream(self, stream, random_id=False):

        """Internal core to insert a record stream into this container.

        This helper function has a different interface than insert_record_stream

        to allow add_lines to be minimal, but still return the needed data.

        :param stream: A stream of records to insert. 

        :return: An iterator over the sha1 of the inserted records.

        :seealso insert_record_stream:

        :seealso add_lines:

        """

        def get_adapter(adapter_key):

            try:

                return adapters[adapter_key]

            except KeyError:

                adapter_factory = adapter_registry.get(adapter_key)

                adapter = adapter_factory(self)

                adapters[adapter_key] = adapter

                return adapter

        adapters = {}

        # This will go up to fulltexts for gc to gc fetching, which isn't

        # ideal.

        self._compressor = GroupCompressor(self._delta)

        self._unadded_refs = {}

        keys_to_add = []

        basis_end = 0

        groups = 1

        def flush():

            compressed = zlib.compress(''.join(self._compressor.lines))

            index, start, length = self._access.add_raw_records(

                [(None, len(compressed))], compressed)[0]

            nodes = []

            for key, reads, refs in keys_to_add:

                nodes.append((key, "%d %d %s" % (start, length, reads), refs))

            self._index.add_records(nodes, random_id=random_id)

        for record in stream:

            # Raise an error when a record is missing.

            if record.storage_kind == 'absent':

                raise errors.RevisionNotPresent([record.key], self)

            elif record.storage_kind in ('chunked', 'fulltext'):

                lines = osutils.chunks_to_lines(record.get_bytes_as('chunked'))

            else:

                adapter_key = record.storage_kind, 'fulltext'

                adapter = get_adapter(adapter_key)

                bytes = adapter.get_bytes(record,

                    record.get_bytes_as(record.storage_kind))

                lines = osutils.split_lines(bytes)

            found_sha1, end_point = self._compressor.compress(record.key,

                lines, record.sha1)

            if record.key[-1] is None:

                key = record.key[:-1] + ('sha1:' + found_sha1,)

            else:

                key = record.key

            self._unadded_refs[key] = record.parents

            yield found_sha1

            keys_to_add.append((key, '%d %d' % (basis_end, end_point),

                (record.parents,)))

            basis_end = end_point

            if basis_end > 1024 * 1024 * 20:

                flush()

                self._compressor = GroupCompressor(self._delta)

                self._unadded_refs = {}

                keys_to_add = []

                basis_end = 0

                groups += 1

        if len(keys_to_add):

            flush()

        self._compressor = None

        self._unadded_refs = {}

    def iter_lines_added_or_present_in_keys(self, keys, pb=None):

        """Iterate over the lines in the versioned files from keys.

        This may return lines from other keys. Each item the returned

        iterator yields is a tuple of a line and a text version that that line

        is present in (not introduced in).

        Ordering of results is in whatever order is most suitable for the

        underlying storage format.

        If a progress bar is supplied, it may be used to indicate progress.

        The caller is responsible for cleaning up progress bars (because this

        is an iterator).

        NOTES:

         * Lines are normalised by the underlying store: they will all have \n

           terminators.

         * Lines are returned in arbitrary order.

        :return: An iterator over (line, key).

        """

        if pb is None:

            pb = progress.DummyProgress()

        keys = set(keys)

        total = len(keys)

        # we don't care about inclusions, the caller cares.

        # but we need to setup a list of records to visit.

        # we need key, position, length

        for key_idx, record in enumerate(self.get_record_stream(keys,

            'unordered', True)):

            # XXX: todo - optimise to use less than full texts.

            key = record.key

            pb.update('Walking content.', key_idx, total)

            if record.storage_kind == 'absent':

                raise errors.RevisionNotPresent(record.key, self)

            lines = split_lines(record.get_bytes_as('fulltext'))

            for line in lines:

                yield line, key

        pb.update('Walking content.', total, total)

    def keys(self):

        """See VersionedFiles.keys."""

        if 'evil' in debug.debug_flags:

            trace.mutter_callsite(2, "keys scales with size of history")

        sources = [self._index]

        result = set()

        for source in sources:

            result.update(source.keys())

        return result

class _GCGraphIndex(object):

    """Mapper from GroupCompressVersionedFiles needs into GraphIndex storage."""

    def __init__(self, graph_index, is_locked, parents=True,

        add_callback=None):

        """Construct a _GCGraphIndex on a graph_index.

        :param graph_index: An implementation of bzrlib.index.GraphIndex.

        :param is_locked: A callback to check whether the object should answer

            queries.

        :param parents: If True, record knits parents, if not do not record 

            parents.

        :param add_callback: If not None, allow additions to the index and call

            this callback with a list of added GraphIndex nodes:

            [(node, value, node_refs), ...]

        :param is_locked: A callback, returns True if the index is locked and

            thus usable.

        """

        self._add_callback = add_callback

        self._graph_index = graph_index