# revlogdeltas.py - Logic around delta computation for revlog

#

# Copyright 2005-2007 Matt Mackall <mpm@selenic.com>

# Copyright 2018 Octobus <contact@octobus.net>

#

# This software may be used and distributed according to the terms of the

# GNU General Public License version 2 or any later version.

"""Helper class to compute deltas stored inside revlogs"""

from __future__ import absolute_import

import collections

import struct

# import stuff from node for others to import from revlog

from ..node import (

    nullrev,

)

from ..i18n import _

from .constants import (

    REVIDX_ISCENSORED,

    REVIDX_RAWTEXT_CHANGING_FLAGS,

)

from ..thirdparty import (

    attr,

)

from .. import (

    error,

    mdiff,

)

# maximum <delta-chain-data>/<revision-text-length> ratio

LIMIT_DELTA2TEXT = 2

class _testrevlog(object):

    """minimalist fake revlog to use in doctests"""

        """data is an list of revision payload boundaries"""

        self._data = data

        self._srdensitythreshold = density

        self._srmingapsize = mingap

    def start(self, rev):

        if rev == 0:

            return 0

        return self._data[rev - 1]

    def end(self, rev):

        return self._data[rev]

    def length(self, rev):

        return self.end(rev) - self.start(rev)

    def __len__(self):

        return len(self._data)

def slicechunk(revlog, revs, targetsize=None):

    """slice revs to reduce the amount of unrelated data to be read from disk.

    ``revs`` is sliced into groups that should be read in one time.

    Assume that revs are sorted.

    The initial chunk is sliced until the overall density (payload/chunks-span

    ratio) is above `revlog._srdensitythreshold`. No gap smaller than

    `revlog._srmingapsize` is skipped.

    If `targetsize` is set, no chunk larger than `targetsize` will be yield.

    For consistency with other slicing choice, this limit won't go lower than

    `revlog._srmingapsize`.

    If individual revisions chunk are larger than this limit, they will still

    be raised individually.

    >>> revlog = _testrevlog([

    ...  5,  #00 (5)

    ...  10, #01 (5)

    ...  12, #02 (2)

    ...  12, #03 (empty)

    ...  27, #04 (15)

    ...  31, #05 (4)

    ...  31, #06 (empty)

    ...  42, #07 (11)

    ...  47, #08 (5)

    ...  47, #09 (empty)

    ...  48, #10 (1)

    ...  51, #11 (3)

    ...  74, #12 (23)

    ...  85, #13 (11)

    ...  86, #14 (1)

    ...  91, #15 (5)

    ... ])

    >>> list(slicechunk(revlog, list(range(16))))

    [[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]]

    >>> list(slicechunk(revlog, [0, 15]))

    [[0], [15]]

    >>> list(slicechunk(revlog, [0, 11, 15]))

    [[0], [11], [15]]

    >>> list(slicechunk(revlog, [0, 11, 13, 15]))

    [[0], [11, 13, 15]]

    >>> list(slicechunk(revlog, [1, 2, 3, 5, 8, 10, 11, 14]))

    [[1, 2], [5, 8, 10, 11], [14]]

    Slicing with a maximum chunk size

    >>> list(slicechunk(revlog, [0, 11, 13, 15], targetsize=15))

    [[0], [11], [13], [15]]

    >>> list(slicechunk(revlog, [0, 11, 13, 15], targetsize=20))

    [[0], [11], [13, 15]]

    """

    if targetsize is not None:

        targetsize = max(targetsize, revlog._srmingapsize)

    # targetsize should not be specified when evaluating delta candidates:

    # * targetsize is used to ensure we stay within specification when reading,

    for chunk in _slicechunktodensity(revlog, revs,

                                      revlog._srdensitythreshold,

                                      revlog._srmingapsize):

        for subchunk in _slicechunktosize(revlog, chunk, targetsize):

            yield subchunk

def _slicechunktosize(revlog, revs, targetsize=None):

    """slice revs to match the target size

    This is intended to be used on chunk that density slicing selected by that

    are still too large compared to the read garantee of revlog. This might

    happens when "minimal gap size" interrupted the slicing or when chain are

    built in a way that create large blocks next to each other.

    >>> revlog = _testrevlog([

    ...  3,  #0 (3)

    ...  5,  #1 (2)

    ...  6,  #2 (1)

    ...  8,  #3 (2)

    ...  8,  #4 (empty)

    ...  11, #5 (3)

    ...  12, #6 (1)

    ...  13, #7 (1)

    ...  14, #8 (1)

    ... ])

    Cases where chunk is already small enough

    >>> list(_slicechunktosize(revlog, [0], 3))

    [[0]]

    >>> list(_slicechunktosize(revlog, [6, 7], 3))

    [[6, 7]]

    >>> list(_slicechunktosize(revlog, [0], None))

    [[0]]

    >>> list(_slicechunktosize(revlog, [6, 7], None))

    [[6, 7]]

    cases where we need actual slicing

    >>> list(_slicechunktosize(revlog, [0, 1], 3))

    [[0], [1]]

    >>> list(_slicechunktosize(revlog, [1, 3], 3))

    [[1], [3]]

    >>> list(_slicechunktosize(revlog, [1, 2, 3], 3))

    [[1, 2], [3]]

    >>> list(_slicechunktosize(revlog, [3, 5], 3))

    [[3], [5]]

    >>> list(_slicechunktosize(revlog, [3, 4, 5], 3))

    [[3], [5]]

    >>> list(_slicechunktosize(revlog, [5, 6, 7, 8], 3))

    [[5], [6, 7, 8]]

    >>> list(_slicechunktosize(revlog, [0, 1, 2, 3, 4, 5, 6, 7, 8], 3))

    [[0], [1, 2], [3], [5], [6, 7, 8]]

    Case with too large individual chunk (must return valid chunk)

    >>> list(_slicechunktosize(revlog, [0, 1], 2))

    [[0], [1]]

    >>> list(_slicechunktosize(revlog, [1, 3], 1))

    [[1], [3]]

    >>> list(_slicechunktosize(revlog, [3, 4, 5], 2))

    [[3], [5]]

    """

    assert targetsize is None or 0 <= targetsize

    if targetsize is None or segmentspan(revlog, revs) <= targetsize:

        yield revs

        return

    startrevidx = 0

    startdata = revlog.start(revs[0])

    endrevidx = 0

    iterrevs = enumerate(revs)

    next(iterrevs) # skip first rev.

    for idx, r in iterrevs:

        span = revlog.end(r) - startdata

        if span <= targetsize:

            endrevidx = idx

        else:

            chunk = _trimchunk(revlog, revs, startrevidx, endrevidx + 1)

            if chunk:

                yield chunk

            startrevidx = idx

            startdata = revlog.start(r)

            endrevidx = idx

    yield _trimchunk(revlog, revs, startrevidx)

def _slicechunktodensity(revlog, revs, targetdensity=0.5,

                         mingapsize=0):

    """slice revs to reduce the amount of unrelated data to be read from disk.

    ``revs`` is sliced into groups that should be read in one time.

    Assume that revs are sorted.

    The initial chunk is sliced until the overall density (payload/chunks-span

    ratio) is above `targetdensity`. No gap smaller than `mingapsize` is

    skipped.

    >>> revlog = _testrevlog([

    ...  5,  #00 (5)

    ...  10, #01 (5)

    ...  12, #02 (2)

    ...  12, #03 (empty)

    ...  27, #04 (15)

    ...  31, #05 (4)

    ...  31, #06 (empty)

    ...  42, #07 (11)

    ...  47, #08 (5)

    ...  47, #09 (empty)

    ...  48, #10 (1)

    ...  51, #11 (3)

    ...  74, #12 (23)

    ...  85, #13 (11)

    ...  86, #14 (1)

    ...  91, #15 (5)

    ... ])

    >>> list(_slicechunktodensity(revlog, list(range(16))))

    [[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]]

    >>> list(_slicechunktodensity(revlog, [0, 15]))

    [[0], [15]]

    >>> list(_slicechunktodensity(revlog, [0, 11, 15]))

    [[0], [11], [15]]

    >>> list(_slicechunktodensity(revlog, [0, 11, 13, 15]))

    [[0], [11, 13, 15]]

    >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14]))

    [[1, 2], [5, 8, 10, 11], [14]]

    >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14],

    ...                           mingapsize=20))

    [[1, 2, 3, 5, 8, 10, 11], [14]]

    >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14],

    ...                           targetdensity=0.95))

    [[1, 2], [5], [8, 10, 11], [14]]

    >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14],

    ...                           targetdensity=0.95, mingapsize=12))

    [[1, 2], [5, 8, 10, 11], [14]]

    """

    start = revlog.start

    length = revlog.length

    if len(revs) <= 1:

        yield revs

        return

    deltachainspan = segmentspan(revlog, revs)

    if deltachainspan < mingapsize:

        yield revs

        return

    readdata = deltachainspan

    chainpayload = sum(length(r) for r in revs)

    if deltachainspan:

        density = chainpayload / float(deltachainspan)

    else:

        density = 1.0

    if density >= targetdensity:

        yield revs

        return

    # Store the gaps in a heap to have them sorted by decreasing size

    gaps = []

    prevend = None

    for i, rev in enumerate(revs):

        revstart = start(rev)

        revlen = length(rev)

        # Skip empty revisions to form larger holes

        if revlen == 0:

            continue

        if prevend is not None:

            gapsize = revstart - prevend

            # only consider holes that are large enough

            if gapsize > mingapsize:

                gaps.append((gapsize, i))

        prevend = revstart + revlen

    # sort the gaps to pop them from largest to small

    gaps.sort()

    # Collect the indices of the largest holes until the density is acceptable

    selected = []

    while gaps and density < targetdensity:

        gapsize, gapidx = gaps.pop()

        selected.append(gapidx)

        # the gap sizes are stored as negatives to be sorted decreasingly

        # by the heap

        readdata -= gapsize

        if readdata > 0:

            density = chainpayload / float(readdata)

        else:

            density = 1.0

    selected.sort()

    # Cut the revs at collected indices

    previdx = 0

    for idx in selected:

        chunk = _trimchunk(revlog, revs, previdx, idx)

        if chunk:

            yield chunk

        previdx = idx

    chunk = _trimchunk(revlog, revs, previdx)

    if chunk:

        yield chunk

def _trimchunk(revlog, revs, startidx, endidx=None):

    """returns revs[startidx:endidx] without empty trailing revs

    Doctest Setup

    >>> revlog = _testrevlog([

    ...  5,  #0

    ...  10, #1

    ...  12, #2

    ...  12, #3 (empty)

    ...  17, #4

    ...  21, #5

    ...  21, #6 (empty)

    ... ])

    Contiguous cases:

    >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0)

    [0, 1, 2, 3, 4, 5]

    >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0, 5)

    [0, 1, 2, 3, 4]

    >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0, 4)

    [0, 1, 2]

    >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 2, 4)

    [2]

    >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 3)

    [3, 4, 5]

    >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 3, 5)

    [3, 4]

    Discontiguous cases:

    >>> _trimchunk(revlog, [1, 3, 5, 6], 0)

    [1, 3, 5]

    >>> _trimchunk(revlog, [1, 3, 5, 6], 0, 2)

    [1]

    >>> _trimchunk(revlog, [1, 3, 5, 6], 1, 3)

    [3, 5]

    >>> _trimchunk(revlog, [1, 3, 5, 6], 1)

    [3, 5]

    """

    length = revlog.length

    if endidx is None:

        endidx = len(revs)

    # If we have a non-emtpy delta candidate, there are nothing to trim

    if revs[endidx - 1] < len(revlog):

        # Trim empty revs at the end, except the very first revision of a chain

        while (endidx > 1

                and endidx > startidx

                and length(revs[endidx - 1]) == 0):

            endidx -= 1

    return revs[startidx:endidx]

def segmentspan(revlog, revs):

    """Get the byte span of a segment of revisions

    revs is a sorted array of revision numbers

    >>> revlog = _testrevlog([

    ...  5,  #0

    ...  10, #1

    ...  12, #2

    ...  12, #3 (empty)

    ...  17, #4

    ... ])

    >>> segmentspan(revlog, [0, 1, 2, 3, 4])

    17

    >>> segmentspan(revlog, [0, 4])

    17

    >>> segmentspan(revlog, [3, 4])

    5

    >>> segmentspan(revlog, [1, 2, 3,])

    7

    >>> segmentspan(revlog, [1, 3])

    7

    """

    if not revs:

        return 0

    end = revlog.end(revs[-1])

    return end - revlog.start(revs[0])

def _textfromdelta(fh, revlog, baserev, delta, p1, p2, flags, expectednode):

    """build full text from a (base, delta) pair and other metadata"""

    # special case deltas which replace entire base; no need to decode

    # base revision. this neatly avoids censored bases, which throw when

    # they're decoded.

    hlen = struct.calcsize(">lll")

    if delta[:hlen] == mdiff.replacediffheader(revlog.rawsize(baserev),

                                               len(delta) - hlen):

        fulltext = delta[hlen:]

    else:

        # deltabase is rawtext before changed by flag processors, which is

        # equivalent to non-raw text

        basetext = revlog.revision(baserev, _df=fh, raw=False)

        fulltext = mdiff.patch(basetext, delta)

    try:

        res = revlog._processflags(fulltext, flags, 'read', raw=True)

        fulltext, validatehash = res

        if validatehash:

            revlog.checkhash(fulltext, expectednode, p1=p1, p2=p2)

        if flags & REVIDX_ISCENSORED:

            raise error.StorageError(_('node %s is not censored') %

                                     expectednode)

    except error.CensoredNodeError:

        # must pass the censored index flag to add censored revisions

        if not flags & REVIDX_ISCENSORED:

            raise

    return fulltext

@attr.s(slots=True, frozen=True)

class _deltainfo(object):

    distance = attr.ib()

    deltalen = attr.ib()

    data = attr.ib()

    base = attr.ib()

    chainbase = attr.ib()

    chainlen = attr.ib()

    compresseddeltalen = attr.ib()

    snapshotdepth = attr.ib()

def isgooddeltainfo(revlog, deltainfo, revinfo):

    """Returns True if the given delta is good. Good means that it is within

    the disk span, disk size, and chain length bounds that we know to be

    performant."""

    if deltainfo is None:

        return False

    # - 'deltainfo.distance' is the distance from the base revision --

    #   bounding it limits the amount of I/O we need to do.

    # - 'deltainfo.compresseddeltalen' is the sum of the total size of

    #   deltas we need to apply -- bounding it limits the amount of CPU

    #   we consume.

    textlen = revinfo.textlen

    defaultmax = textlen * 4

    maxdist = revlog._maxdeltachainspan

    if not maxdist:

        maxdist = deltainfo.distance # ensure the conditional pass

    maxdist = max(maxdist, defaultmax)

    # Bad delta from read span:

    #

    #   If the span of data read is larger than the maximum allowed.

    #

    #   In the sparse-revlog case, we rely on the associated "sparse reading"

    #   to avoid issue related to the span of data. In theory, it would be

    #   possible to build pathological revlog where delta pattern would lead

    #   to too many reads. However, they do not happen in practice at all. So

    #   we skip the span check entirely.

    if not revlog._sparserevlog and maxdist < deltainfo.distance:

        return False

    # Bad delta from new delta size:

    #

    #   If the delta size is larger than the target text, storing the

    #   delta will be inefficient.