# Code dedicated to an cache around obsolescence property

#

# This module content aims at being upstreamed.

#

# Copyright 2017 Pierre-Yves David <pierre-yves.david@ens-lyon.org>

#

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

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

import hashlib

import struct

import weakref

import errno

from mercurial import (

    error,

    localrepo,

    obsolete,

    phases,

    node,

    util,

)

from mercurial.i18n import _

from . import (

    exthelper,

)

eh = exthelper.exthelper()

try:

    obsstorefilecache = localrepo.localrepository.obsstore

except AttributeError:

    # XXX hg-3.8 compat

    #

    # mercurial 3.8 has issue with accessing file cache property from their

    # cache. This is fix by 36fbd72c2f39fef8ad52d7c559906c2bc388760c in core

    # and shipped in 3.9

    obsstorefilecache = localrepo.localrepository.__dict__['obsstore']

# obsstore is a filecache so we have do to some spacial dancing

@eh.wrapfunction(obsstorefilecache, 'func')

def obsstorewithcache(orig, repo):

    obsstore = orig(repo)

    obsstore.obscache = obscache(repo.unfiltered())

    class cachekeyobsstore(obsstore.__class__):

        _obskeysize = 200

        def cachekey(self, index=None):

            """return (current-length, cachekey)

            'current-length': is the current length of the obsstore storage file,

            'cachekey' is the hash of the last 200 bytes ending at 'index'.

            if 'index' is unspecified, current obsstore length is used.

            Cacheckey will be set to null id if the obstore is empty.

            If the index specified is higher than the current obsstore file

            length, cachekey will be set to None."""

            # default value

            obsstoresize = 0

            keydata = ''

            # try to get actual data from the obsstore

            try:

                with self.svfs('obsstore') as obsfile:

                    obsfile.seek(0, 2)

                    obsstoresize = obsfile.tell()

                    if index is None:

                        index = obsstoresize

                    elif obsstoresize < index:

                        return obsstoresize, None

                    actualsize = min(index, self._obskeysize)

                    if actualsize:

                        obsfile.seek(index - actualsize, 0)

                        keydata = obsfile.read(actualsize)

            except (OSError, IOError) as e:

                if e.errno != errno.ENOENT:

                    raise

            key = hashlib.sha1(keydata).digest()

            return obsstoresize, key

    obsstore.__class__ = cachekeyobsstore

    return obsstore

emptykey = (node.nullrev, node.nullid, 0, 0, node.nullid)

def upgradeneeded(repo, key):

    """return (valid, start-rev, start-obs-idx)

    'valid': is "False" if older cache value needs invalidation,

    'start-rev': first revision not in the cache. None if cache is up to date,

    'start-obs-idx': index of the first obs-markers not in the cache. None is

                     up to date.

    """

    reset = False

    status = _checkkey(repo, key)

    if status is None:

        reset = True

        key = emptykey

    else:

        tiprev, obssize, obskey = status

    keytiprev, keytipnode, keyobslength, keyobssize, keyobskey = key

    if not reset and keytiprev == tiprev and keyobssize == obssize:

        return None # nothing to upgrade

    ### cache is valid, is there anything to update

    # any new changesets ?

    revs = ()

    if keytiprev < tiprev:

    # any new markers

    markers = ()

    if keyobssize < obssize:

        # XXX Three are a small race change here. Since the obsstore might have

        # move forward between the time we computed the cache key and we access

        # the data. To fix this we need so "up to" argument when fetching the

        # markers here. Otherwise we might return more markers than covered by

        # the cache key.

        #

        # In pratice the cache is only updated after each transaction within a

        # lock. So we should be fine. We could enforce this with a new repository

        # requirement (or fix the race, that is not too hard).

    return reset, revs, markers, (obssize, obskey)

def _checkkey(repo, key):

    """internal function"""

    if key is None:

        return None

    ### Is the cache valid ?

    keytiprev, keytipnode, keyobslength, keyobssize, keyobskey = key

    # check for changelog strip

    cl = repo.changelog

    tiprev = len(cl) - 1

    if (tiprev < keytiprev

            or cl.node(keytiprev) != keytipnode):

        return None

    # check for obsstore strip

    obssize, obskey = repo.obsstore.cachekey(index=keyobssize)

    if obskey != keyobskey:

        return None

    return tiprev, obssize, obskey

# XXX copied as is from Mercurial 4.2 and added the "offset" parameters

@util.nogc

def _readmarkers(data, offset=None):

    """Read and enumerate markers from raw data"""

    off = 0

    diskversion = struct.unpack('>B', data[off:off + 1])[0]

    if offset is None:

        off += 1

    else:

        assert 1 <= offset

        off = offset

    if diskversion not in obsolete.formats:

        raise error.Abort(_('parsing obsolete marker: unknown version %r')

                          % diskversion)

    return diskversion, obsolete.formats[diskversion][0](data, off)

def markersfrom(obsstore, byteoffset, firstmarker):

    if not firstmarker:

        return list(obsstore)

    elif '_all' in vars(obsstore):

        # if the data are in memory, just use that

        return obsstore._all[firstmarker:]

    else:

        obsdata = obsstore.svfs.tryread('obsstore')

        return _readmarkers(obsdata, byteoffset)[1]

class obscache(object):

    """cache the "does a rev" is the precursors of some obsmarkers data

    This is not directly holding the "is this revision obsolete" information,

    because phases data gets into play here. However, it allow to compute the

    "obsolescence" set without reading the obsstore content.

    Implementation note #1:

      The obsstore is implementing only half of the transaction logic it

      should. It properly record the starting point of the obsstore to allow

      clean rollback. However it still write to the obsstore file directly

      during the transaction. Instead it should be keeping data in memory and

      write to a '.pending' file to make the data vailable for hooks.

      This cache is not going futher than what the obstore is doing, so it does

      not has any '.pending' logic. When the obsstore gains proper '.pending'

      support, adding it to this cache should not be too hard. As the flag

      always move from 0 to 1, we could have a second '.pending' cache file to

      be read. If flag is set in any of them, the value is 1. For the same

      reason, updating the file in place should be possible.

    Implementation note #2:

      Instead of having a large final update run, we could update this cache at

      the level adding a new changeset or a new obsmarkers. More on this in the

      'update code'.

    Implementation note #3:

        Storage-wise, we could have a "start rev" to avoid storing useless

        zero. That would be especially useful for the '.pending' overlay.

    """

    _filepath = 'cache/evoext-obscache-00'

    _headerformat = '>q20sQQ20s'

    def __init__(self, repo):

        self._vfs = repo.vfs

        # cache key covering the changesets and obsmarkers content

        #

        # It contains the following data. Combined with 'upgradeneeded' it allows to

        # do iterative upgrade for cache depending of theses two pieces of data.

        #

        # The cache key parts are"

        # - tip-rev,

        # - tip-node,

        # - obsstore-length (nb markers),

        # - obsstore-file-size (in bytes),

        # - obsstore "cache key"

        self._cachekey = None

        self._ondiskkey = None

        self._data = bytearray()

    def get(self, rev):

        """return True if "rev" is used as "precursors for any obsmarkers

        Make sure the cache has been updated to match the repository content before using it"""

        return self._data[rev]

    def clear(self, reset=False):

        """invalidate the cache content"""

        self._cachekey = emptykey if reset else None

        self._data = bytearray()

    def uptodate(self, repo):

        if self._cachekey is None:

            self.load(repo)

        status = _checkkey(repo, self._cachekey)

        return (status is not None

                and status[0] == self._cachekey[0] # tiprev

                and status[1] == self._cachekey[3]) # obssize

    def update(self, repo):

        """Iteratively update the cache with new repository data"""

        # If we do not have any data, try loading from disk

        if self._cachekey is None:

            self.load(repo)

        assert repo.filtername is None

        cl = repo.changelog

        upgrade = upgradeneeded(repo, self._cachekey)

        if upgrade is None:

            return

        reset, revs, obsmarkers, obskeypair = upgrade

        if reset or self._cachekey is None:

            self.clear(reset=True)

        if revs:

            self._updaterevs(repo, revs)

            assert len(self._data) == len(cl), (len(self._data), len(cl))

        if obsmarkers:

            self._updatemarkers(repo, obsmarkers)

        # update the key from the new data

        key = list(self._cachekey)

        if revs:

            key[0] = len(cl) - 1

            key[1] = cl.node(key[0])

        if obsmarkers:

            key[2] += len(obsmarkers)

            key[3], key[4] = obskeypair

        self._cachekey = tuple(key)

    def _updaterevs(self, repo, revs):

        """update the cache with new revisions

        Newly added changeset might be affected by obsolescence markers

        we already have locally. So we needs to have some global

        knowledge about the markers to handle that question.

        Right now this requires parsing all markers in the obsstore. We could

        imagine using various optimisation (eg: another cache, network

        exchange, etc).

        A possible approach to this is to build a set of all node used as

        precursors in `obsstore._obscandidate`. If markers are not loaded yet,

        we could initialize it by doing a quick scan through the obsstore data

        and filling a (pre-sized) set. Doing so would be much faster than

        parsing all the obsmarkers since we would access less data, not create

        any object beside the nodes and not have to decode any complex data.

        For now we stick to the simpler approach of paying the

        performance cost on new changesets.

        """

        node = repo.changelog.node

        succs = repo.obsstore.successors

        for r in revs:

            if node(r) in succs:

                val = 1

            else:

                val = 0

            self._data.append(val)

    def _updatemarkers(self, repo, obsmarkers):

        """update the cache with new markers"""

        rev = repo.changelog.nodemap.get

        for m in obsmarkers:

            r = rev(m[0])

            if r is not None:

                self._data[r] = 1

    def save(self, repo):

        """save the data to disk"""

        # XXX it happens that the obsstore is (buggilly) always up to date on disk

        if self._cachekey is None or self._cachekey == self._ondiskkey:

            return

        cachefile = repo.vfs(self._filepath, 'w', atomictemp=True)

        headerdata = struct.pack(self._headerformat, *self._cachekey)

        cachefile.write(headerdata)

        cachefile.write(self._data)

        cachefile.close()

    def load(self, repo):

        """load data from disk"""

        assert repo.filtername is None

        data = repo.vfs.tryread(self._filepath)

        if not data:

            self._cachekey = emptykey

            self._data = bytearray()

        else:

            headersize = struct.calcsize(self._headerformat)

            self._cachekey = struct.unpack(self._headerformat, data[:headersize])

            self._data = bytearray(data[headersize:])

        self._ondiskkey = self._cachekey

def _computeobsoleteset(orig, repo):

    """the set of obsolete revisions"""

    obs = set()

    repo = repo.unfiltered()

    if util.safehasattr(repo._phasecache, 'getrevset'):

        notpublic = repo._phasecache.getrevset(repo, (phases.draft, phases.secret))

    else:

        # < hg-4.2 compat

        notpublic = repo.revs("not public()")

    if notpublic:

        obscache = repo.obsstore.obscache

        # Since we warm the cache at the end of every transaction, the cache

        # should be up to date. However a non-enabled client might have touched

        # the repository.

        #

        # Updating the cache without a lock is sloppy, so we fallback to the

        # old method and rely on the fact the next transaction will write the

        # cache down anyway.

        #

        # With the current implementation updating the cache will requires to

        # load the obsstore anyway. Once loaded, hitting the obsstore directly

        # will be about as fast...

        if not obscache.uptodate(repo):

            if repo.currenttransaction() is None:

                repo.ui.log('evoext-obscache',

                            'obscache is out of date, '

                            'falling back to slower obsstore version\n')

                repo.ui.debug('obscache is out of date')

                return orig(repo)

            else:

                # If a transaction is open, it is worthwhile to update and use

                # the cache, the lock prevent race and it will be written on

                # disk when the transaction close.

                obscache.update(repo)

        isobs = obscache.get

    for r in notpublic:

        if isobs(r):

            obs.add(r)