view hgext/remotefilelog/contentstore.py @ 49777:e1953a34c110

bundle: emit full snapshot as is, without doing a redelta

With the new `forced` delta-reused policy, it become important to be able to
send full snapshot where full snapshot are needed. Otherwise, the fallback delta
will simply be used on the client side… creating monstrous delta chain, since
revision that are meant as a reset of delta-chain chain becoming too complex are
simply adding a new full delta-tree on the leaf of another one.

In the `non-forced` cases, client process full snapshot from the bundle
differently from deltas, so client will still try to convert the full snapshot
into a delta if possible. So this will no lead to pathological storage
explosion.

I have considered making this configurable, but the impact seems limited enough
that it does not seems to be worth it. Especially with the current
sparse-revlog format that use "delta-tree" with multiple level snapshots, full
snapshot are much less frequent and not that different from other intermediate
snapshot that we are already sending over the wire anyway.

CPU wise, this will help the bundling side a little as it will not need to
reconstruct revisions and compute deltas. The unbundling side might save a tiny
amount of CPU as it won't need to reconstruct the delta-base to reconstruct the
revision full text. This only slightly visible in some of the benchmarks. And
have no real impact on most of them.

### data-env-vars.name      = pypy-2018-08-01-zstd-sparse-revlog
  # benchmark.name          = perf-bundle
  # benchmark.variants.revs = last-40000
before:          11.467186 seconds
just-emit-full:  11.190576 seconds   (-2.41%)
with-pull-force: 11.041091 seconds   (-3.72%)
  # benchmark.name          = perf-unbundle
  # benchmark.variants.revs = last-40000
before:     16.744862
just-emit-full:: 16.561036 seconds   (-1.10%)
with-pull-force: 16.389344 seconds   (-2.12%)
  # benchmark.name          = pull
  # benchmark.variants.revs = last-40000
before:     26.870569
just-emit-full:  26.391188 seconds   (-1.78%)
with-pull-force: 25.633184 seconds   (-4.60%)


Space wise (so network-wise) the impact is fairly small. When taking compression into
account.

Below are tests the size of `hg bundle --all` for a handful of benchmark repositories
(with bzip, zstd compression and without it)

This show a small increase in the bundle size, but nothing really significant
except maybe for mozilla-try (+12%) that nobody really pulls large chunk of anyway.
Mozilla-try is also the repository that benefit the most for not having to
recompute deltas client size.

### mercurial:
bzip-before:     26 406 342 bytes
bzip-after:      26 691 543 bytes  +1.08%

zstd-before:     27 918 645 bytes
zstd-after:      28 075 896 bytes  +0.56%

none-before:     98 675 601 bytes
none-after:     100 411 237 bytes  +1.76%


### pypy
bzip-before:    201 295 752 bytes
bzip-after:     209 780 282 bytes  +4.21%

zstd-before:    202 974 795 bytes
zstd-after:     205 165 780 bytes  +1.08%

none-before:    871 070 261 bytes
none-after:     993 595 057 bytes +14.07%


### netbeans
bzip-before:    601 314 330 bytes
bzip-after:     614 246 241 bytes  +2.15%

zstd-before:    604 745 136 bytes
zstd-after:     615 497 705 bytes  +1.78%

none-before:  3 338 238 571 bytes
none-after:   3 439 422 535 bytes  +3.03%


### mozilla-central
bzip-before:  1 493 006 921 bytes
bzip-after:   1 549 650 570 bytes  +3.79%

zstd-before:  1 481 910 102 bytes
zstd-after:   1 513 052 415 bytes  +2.10%

none-before:  6 535 929 910 bytes
none-after:   7 010 191 342 bytes  +7.26%


### mozilla-try
bzip-before:  6 583 425 999 bytes
bzip-after:   7 423 536 928 bytes +12.76%

zstd-before:  6 021 009 212 bytes
zstd-after:   6 674 922 420 bytes +10.86%

none-before: 22 954 739 558 bytes
none-after:  26 013 854 771 bytes +13.32%

import threading

from mercurial.node import (
    hex,
    sha1nodeconstants,
)
from mercurial.pycompat import getattr
from mercurial import (
    mdiff,
    revlog,
)
from . import (
    basestore,
    constants,
    shallowutil,
)


class ChainIndicies:
    """A static class for easy reference to the delta chain indicies."""

    # The filename of this revision delta
    NAME = 0
    # The mercurial file node for this revision delta
    NODE = 1
    # The filename of the delta base's revision. This is useful when delta
    # between different files (like in the case of a move or copy, we can delta
    # against the original file content).
    BASENAME = 2
    # The mercurial file node for the delta base revision. This is the nullid if
    # this delta is a full text.
    BASENODE = 3
    # The actual delta or full text data.
    DATA = 4


class unioncontentstore(basestore.baseunionstore):
    def __init__(self, *args, **kwargs):
        super(unioncontentstore, self).__init__(*args, **kwargs)

        self.stores = args
        self.writestore = kwargs.get('writestore')

        # If allowincomplete==True then the union store can return partial
        # delta chains, otherwise it will throw a KeyError if a full
        # deltachain can't be found.
        self.allowincomplete = kwargs.get('allowincomplete', False)

    def get(self, name, node):
        """Fetches the full text revision contents of the given name+node pair.
        If the full text doesn't exist, throws a KeyError.

        Under the hood, this uses getdeltachain() across all the stores to build
        up a full chain to produce the full text.
        """
        chain = self.getdeltachain(name, node)

        if chain[-1][ChainIndicies.BASENODE] != sha1nodeconstants.nullid:
            # If we didn't receive a full chain, throw
            raise KeyError((name, hex(node)))

        # The last entry in the chain is a full text, so we start our delta
        # applies with that.
        fulltext = chain.pop()[ChainIndicies.DATA]

        text = fulltext
        while chain:
            delta = chain.pop()[ChainIndicies.DATA]
            text = mdiff.patches(text, [delta])

        return text

    @basestore.baseunionstore.retriable
    def getdelta(self, name, node):
        """Return the single delta entry for the given name/node pair."""
        for store in self.stores:
            try:
                return store.getdelta(name, node)
            except KeyError:
                pass

        raise KeyError((name, hex(node)))

    def getdeltachain(self, name, node):
        """Returns the deltachain for the given name/node pair.

        Returns an ordered list of:

          [(name, node, deltabasename, deltabasenode, deltacontent),...]

        where the chain is terminated by a full text entry with a nullid
        deltabasenode.
        """
        chain = self._getpartialchain(name, node)
        while chain[-1][ChainIndicies.BASENODE] != sha1nodeconstants.nullid:
            x, x, deltabasename, deltabasenode, x = chain[-1]
            try:
                morechain = self._getpartialchain(deltabasename, deltabasenode)
                chain.extend(morechain)
            except KeyError:
                # If we allow incomplete chains, don't throw.
                if not self.allowincomplete:
                    raise
                break

        return chain

    @basestore.baseunionstore.retriable
    def getmeta(self, name, node):
        """Returns the metadata dict for given node."""
        for store in self.stores:
            try:
                return store.getmeta(name, node)
            except KeyError:
                pass
        raise KeyError((name, hex(node)))

    def getmetrics(self):
        metrics = [s.getmetrics() for s in self.stores]
        return shallowutil.sumdicts(*metrics)

    @basestore.baseunionstore.retriable
    def _getpartialchain(self, name, node):
        """Returns a partial delta chain for the given name/node pair.

        A partial chain is a chain that may not be terminated in a full-text.
        """
        for store in self.stores:
            try:
                return store.getdeltachain(name, node)
            except KeyError:
                pass

        raise KeyError((name, hex(node)))

    def add(self, name, node, data):
        raise RuntimeError(
            b"cannot add content only to remotefilelog contentstore"
        )

    def getmissing(self, keys):
        missing = keys
        for store in self.stores:
            if missing:
                missing = store.getmissing(missing)
        return missing

    def addremotefilelognode(self, name, node, data):
        if self.writestore:
            self.writestore.addremotefilelognode(name, node, data)
        else:
            raise RuntimeError(b"no writable store configured")

    def markledger(self, ledger, options=None):
        for store in self.stores:
            store.markledger(ledger, options)


class remotefilelogcontentstore(basestore.basestore):
    def __init__(self, *args, **kwargs):
        super(remotefilelogcontentstore, self).__init__(*args, **kwargs)
        self._threaddata = threading.local()

    def get(self, name, node):
        # return raw revision text
        data = self._getdata(name, node)

        offset, size, flags = shallowutil.parsesizeflags(data)
        content = data[offset : offset + size]

        ancestormap = shallowutil.ancestormap(data)
        p1, p2, linknode, copyfrom = ancestormap[node]
        copyrev = None
        if copyfrom:
            copyrev = hex(p1)

        self._updatemetacache(node, size, flags)

        # lfs tracks renames in its own metadata, remove hg copy metadata,
        # because copy metadata will be re-added by lfs flag processor.
        if flags & revlog.REVIDX_EXTSTORED:
            copyrev = copyfrom = None
        revision = shallowutil.createrevlogtext(content, copyfrom, copyrev)
        return revision

    def getdelta(self, name, node):
        # Since remotefilelog content stores only contain full texts, just
        # return that.
        revision = self.get(name, node)
        return (
            revision,
            name,
            sha1nodeconstants.nullid,
            self.getmeta(name, node),
        )

    def getdeltachain(self, name, node):
        # Since remotefilelog content stores just contain full texts, we return
        # a fake delta chain that just consists of a single full text revision.
        # The nullid in the deltabasenode slot indicates that the revision is a
        # fulltext.
        revision = self.get(name, node)
        return [(name, node, None, sha1nodeconstants.nullid, revision)]

    def getmeta(self, name, node):
        self._sanitizemetacache()
        if node != self._threaddata.metacache[0]:
            data = self._getdata(name, node)
            offset, size, flags = shallowutil.parsesizeflags(data)
            self._updatemetacache(node, size, flags)
        return self._threaddata.metacache[1]

    def add(self, name, node, data):
        raise RuntimeError(
            b"cannot add content only to remotefilelog contentstore"
        )

    def _sanitizemetacache(self):
        metacache = getattr(self._threaddata, 'metacache', None)
        if metacache is None:
            self._threaddata.metacache = (None, None)  # (node, meta)

    def _updatemetacache(self, node, size, flags):
        self._sanitizemetacache()
        if node == self._threaddata.metacache[0]:
            return
        meta = {constants.METAKEYFLAG: flags, constants.METAKEYSIZE: size}
        self._threaddata.metacache = (node, meta)


class remotecontentstore:
    def __init__(self, ui, fileservice, shared):
        self._fileservice = fileservice
        # type(shared) is usually remotefilelogcontentstore
        self._shared = shared

    def get(self, name, node):
        self._fileservice.prefetch(
            [(name, hex(node))], force=True, fetchdata=True
        )
        return self._shared.get(name, node)

    def getdelta(self, name, node):
        revision = self.get(name, node)
        return (
            revision,
            name,
            sha1nodeconstants.nullid,
            self._shared.getmeta(name, node),
        )

    def getdeltachain(self, name, node):
        # Since our remote content stores just contain full texts, we return a
        # fake delta chain that just consists of a single full text revision.
        # The nullid in the deltabasenode slot indicates that the revision is a
        # fulltext.
        revision = self.get(name, node)
        return [(name, node, None, sha1nodeconstants.nullid, revision)]

    def getmeta(self, name, node):
        self._fileservice.prefetch(
            [(name, hex(node))], force=True, fetchdata=True
        )
        return self._shared.getmeta(name, node)

    def add(self, name, node, data):
        raise RuntimeError(b"cannot add to a remote store")

    def getmissing(self, keys):
        return keys

    def markledger(self, ledger, options=None):
        pass


class manifestrevlogstore:
    def __init__(self, repo):
        self._store = repo.store
        self._svfs = repo.svfs
        self._revlogs = dict()
        self._cl = revlog.revlog(self._svfs, radix=b'00changelog.i')
        self._repackstartlinkrev = 0

    def get(self, name, node):
        return self._revlog(name).rawdata(node)

    def getdelta(self, name, node):
        revision = self.get(name, node)
        return revision, name, self._cl.nullid, self.getmeta(name, node)

    def getdeltachain(self, name, node):
        revision = self.get(name, node)
        return [(name, node, None, self._cl.nullid, revision)]

    def getmeta(self, name, node):
        rl = self._revlog(name)
        rev = rl.rev(node)
        return {
            constants.METAKEYFLAG: rl.flags(rev),
            constants.METAKEYSIZE: rl.rawsize(rev),
        }

    def getancestors(self, name, node, known=None):
        if known is None:
            known = set()
        if node in known:
            return []

        rl = self._revlog(name)
        ancestors = {}
        missing = {node}
        for ancrev in rl.ancestors([rl.rev(node)], inclusive=True):
            ancnode = rl.node(ancrev)
            missing.discard(ancnode)

            p1, p2 = rl.parents(ancnode)
            if p1 != self._cl.nullid and p1 not in known:
                missing.add(p1)
            if p2 != self._cl.nullid and p2 not in known:
                missing.add(p2)

            linknode = self._cl.node(rl.linkrev(ancrev))
            ancestors[rl.node(ancrev)] = (p1, p2, linknode, b'')
            if not missing:
                break
        return ancestors

    def getnodeinfo(self, name, node):
        cl = self._cl
        rl = self._revlog(name)
        parents = rl.parents(node)
        linkrev = rl.linkrev(rl.rev(node))
        return (parents[0], parents[1], cl.node(linkrev), None)

    def add(self, *args):
        raise RuntimeError(b"cannot add to a revlog store")

    def _revlog(self, name):
        rl = self._revlogs.get(name)
        if rl is None:
            revlogname = b'00manifesttree'
            if name != b'':
                revlogname = b'meta/%s/00manifest' % name
            rl = revlog.revlog(self._svfs, radix=revlogname)
            self._revlogs[name] = rl
        return rl

    def getmissing(self, keys):
        missing = []
        for name, node in keys:
            mfrevlog = self._revlog(name)
            if node not in mfrevlog.nodemap:
                missing.append((name, node))

        return missing

    def setrepacklinkrevrange(self, startrev, endrev):
        self._repackstartlinkrev = startrev
        self._repackendlinkrev = endrev

    def markledger(self, ledger, options=None):
        if options and options.get(constants.OPTION_PACKSONLY):
            return
        treename = b''
        rl = revlog.revlog(self._svfs, radix=b'00manifesttree')
        startlinkrev = self._repackstartlinkrev
        endlinkrev = self._repackendlinkrev
        for rev in range(len(rl) - 1, -1, -1):
            linkrev = rl.linkrev(rev)
            if linkrev < startlinkrev:
                break
            if linkrev > endlinkrev:
                continue
            node = rl.node(rev)
            ledger.markdataentry(self, treename, node)
            ledger.markhistoryentry(self, treename, node)

        for t, path, size in self._store.datafiles():
            if path[:5] != b'meta/' or path[-2:] != b'.i':
                continue

            treename = path[5 : -len(b'/00manifest')]

            rl = revlog.revlog(self._svfs, indexfile=path[:-2])
            for rev in range(len(rl) - 1, -1, -1):
                linkrev = rl.linkrev(rev)
                if linkrev < startlinkrev:
                    break
                if linkrev > endlinkrev:
                    continue
                node = rl.node(rev)
                ledger.markdataentry(self, treename, node)
                ledger.markhistoryentry(self, treename, node)

    def cleanup(self, ledger):
        pass