# 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 errno

import hashlib

import os

import struct

import time

import weakref

from mercurial import (

    error,

    localrepo,

    obsolete,

    phases,

    pycompat,

    node,

    util,

)

from mercurial.i18n import _

from . import (

    exthelper,

)

eh = exthelper.exthelper()

# prior to hg-4.2 there are not util.timer

if util.safehasattr(util, 'timer'):

    timer = util.timer

elif util.safehasattr(time, "perf_counter"):

    timer = time.perf_counter

elif getattr(pycompat, 'osname', os.name) == 'nt':

    timer = time.clock

else:

    timer = time.time

# hg < 4.2 compat

try:

    from mercurial import vfs as vfsmod

    vfsmod.vfs

except ImportError:

    from mercurial import scmutil as vfsmod

obsstorefilecache = localrepo.localrepository.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

            if keydata:

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

            else:

                # reusing an existing "empty" value make it easier to define a

                # default cachekey for 'no data'.

                key = node.nullid

            return obsstoresize, key

    obsstore.__class__ = cachekeyobsstore

    return obsstore

if obsolete._fm0readmarkers.__code__.co_argcount > 1:

    # hg-4.3+ has the "offset" parameter, and _fm?readmarkers also have an

    # extra "stop" parameter

    # Note that _readmarkers is wrapped by @util.nogc, so its co_argcount is

    # misleadingly 0. So we check _fm0readmarkers instead, which increased its

    # argument count in the same changeset (5d3ba439).

    _readmarkers = obsolete._readmarkers

else:

    # 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 dualsourcecache(object):

    """An abstract class for cache that needs both changelog and obsstore

    This class handle the tracking of changelog and obsstore update. It provide

    data to performs incremental update (see the 'updatefrom' function for

    details).  This class can also detect stripping of the changelog or the

    obsstore and can reset the cache in this cache (see the 'clear' function

    for details).

    """

    # default key used for an empty cache

    #

    # The cache key covering the changesets and obsmarkers content

    #

    # The cache key parts are:

    # - tip-rev,

    # - tip-node,

    # - obsstore-length (nb markers),

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

    # - obsstore "cache key"

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

    _cachename = None # used for error message

    def __init__(self):

        super(dualsourcecache, self).__init__()

        self._cachekey = None

    def _updatefrom(self, repo, revs, obsmarkers):

        """override this method to update your cache data incrementally

        revs:      list of new revision in the changelog

        obsmarker: list of new obsmarkers in the obsstore

        """

        raise NotImplementedError

    def clear(self, reset=False):

        """invalidate the cache content

        if 'reset' is passed, we detected a strip and the cache will have to be

        recomputed.

        """

        # /!\ IMPORTANT /!\

        # You must overide this method to actually

        if reset:

            self._cachekey = self.emptykey if reset else None

        else:

            self._cachekey = None

    def load(self, repo):

        """Load data from disk

        Do not forget to restore the "cachekey" attribute while doing so.

        """

        raise NotImplementedError

    # Useful public function (no need to override them)

    def uptodate(self, repo):

        """return True if the cache content is up to date False otherwise

        This method can be used to detect of the cache is lagging behind new

        data in either changelog or obsstore.

        """

        if self._cachekey is None:

            self.load(repo)

        status = self._checkkey(repo.changelog, repo.obsstore)

        return (status is not None

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

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

    def update(self, repo):

        """update the cache with new repository data

        The update will be incremental when possible"""

        repo = repo.unfiltered()

        # 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 = self._upgradeneeded(repo)

        if upgrade is None:

            return

        reset, revs, obsmarkers, obskeypair = upgrade

        if reset or self._cachekey is None:

            repo.ui.log('evoext-cache', 'strip detected, %s cache reset\n' % self._cachename)

            self.clear(reset=True)

        starttime = timer()

        self._updatefrom(repo, revs, obsmarkers)

        duration = timer() - starttime

        repo.ui.log('evoext-cache', 'updated %s in %.4f seconds (%sr, %so)\n',

                    self._cachename, duration, len(revs), len(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)

    # from here, there are internal function only

    def _checkkey(self, changelog, obsstore):

        """internal function"""

        key = self._cachekey

        if key is None:

            return None

        ### Is the cache valid ?

        keytiprev, keytipnode, keyobslength, keyobssize, keyobskey = key

        # check for changelog strip

        tiprev = len(changelog) - 1

        if (tiprev < keytiprev

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

            return None

        # check for obsstore strip

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

        if obskey != keyobskey:

            return None

        if obssize != keyobssize:

            # we want to return the obskey for the new size

            __, obskey = obsstore.cachekey(index=obssize)

        return tiprev, obssize, obskey

    def _upgradeneeded(self, repo):

        """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.

        """

        # We need to ensure we use the same changelog and obsstore through the

        # processing. Otherwise some invalidation could update the object and their

        # content after we computed the cache key.

        cl = repo.changelog

        obsstore = repo.obsstore

        key = self._cachekey

        reset = False

        status = self._checkkey(cl, obsstore)

        if status is None:

            reset = True

            key = self.emptykey

            obssize, obskey = obsstore.cachekey()

            tiprev = len(cl) - 1

        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:

            revs = list(cl.revs(start=keytiprev + 1, stop=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).

            markers = markersfrom(obsstore, keyobssize, keyobslength)

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

def getcachevfs(repo):

    cachevfs = getattr(repo, 'cachevfs', None)

    if cachevfs is None:

        cachevfs = vfsmod.vfs(repo.vfs.join('cache'))

        cachevfs.createmode = repo.store.createmode

    return cachevfs

class obscache(dualsourcecache):

    """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 = 'evoext-obscache-00'

    _headerformat = '>q20sQQ20s'

    _cachename = 'evo-ext-obscache' # used for error message

    def __init__(self, repo):

        super(obscache, self).__init__()

        self._ondiskkey = None

        self._vfs = getcachevfs(repo)

    @util.propertycache

    def get(self):

        """final signature: obscache.get(rev)

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

        IMPORTANT: make sure the cache has been updated to match the repository

        content before using it

        We use a property cache to skip the attribute resolution overhead in

        hot loops."""

        return self._data.__getitem__

    def _setdata(self, data):

        """set a new bytearray data, invalidating the 'get' shortcut if needed"""

        self._data = data

        if 'get' in vars(self):

            del self.get

    def clear(self, reset=False):

        """invalidate the cache content"""

        super(obscache, self).clear(reset=reset)

        self._setdata(bytearray())

    def _updatefrom(self, repo, revs, obsmarkers):