view mercurial/manifest.py @ 14365:a8e3931e3fb5

revlog: linearize created changegroups in generaldelta revlogs

This greatly improves the speed of the bundling process, and often reduces the
bundle size considerably. (Although if the repository is already ordered, this
has little effect on both time and bundle size.)

For non-generaldelta clients, the reduced bundle size translates to a reduced
repository size, similar to shrinking the revlogs (which uses the exact same
algorithm). For generaldelta clients the difference is minor.

When the new bundle format comes, reordering will not be necessary since we
can then store the deltaparent relationsships directly. The eventual default
behavior for clients and servers is presented in the table below, where "new"
implies support for GD as well as the new bundle format:

                    old client                    new client
old server          old bundle, no reorder        old bundle, no reorder
new server, non-GD  old bundle, no reorder[1]     old bundle, no reorder[2]
new server, GD      old bundle, reorder[3]        new bundle, no reorder[4]

[1] reordering is expensive on the server in this case, skip it
[2] client can choose to do its own redelta here
[3] reordering is needed because otherwise the pull does a lot of extra
    work on the server
[4] reordering isn't needed because client can get deltabase in bundle
    format

Currently, the default is to reorder on GD-servers, and not otherwise. A new
setting, bundle.reorder, has been added to override the default reordering
behavior. It can be set to either 'auto' (the default), or any true or false
value as a standard boolean setting, to either force the reordering on or off
regardless of generaldelta.


Some timing data from a relatively branch test repository follows. All
bundling is done with --all --type none options.

Non-generaldelta, non-shrunk repo:
-----------------------------------
Size: 276M

Without reorder (default):
Bundle time: 14.4 seconds
Bundle size: 939M

With reorder:
Bundle time: 1 minute, 29.3 seconds
Bundle size: 381M

Generaldelta, non-shrunk repo:
-----------------------------------
Size: 87M

Without reorder:
Bundle time: 2 minutes, 1.4 seconds
Bundle size: 939M

With reorder (default):
Bundle time: 25.5 seconds
Bundle size: 381M

# manifest.py - manifest revision class for mercurial
#
# Copyright 2005-2007 Matt Mackall <[e-mail]>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

from i18n import _
import mdiff, parsers, error, revlog
import array, struct

class manifestdict(dict):
    def __init__(self, mapping=None, flags=None):
        if mapping is None:
            mapping = {}
        if flags is None:
            flags = {}
        dict.__init__(self, mapping)
        self._flags = flags
    def flags(self, f):
        return self._flags.get(f, "")
    def set(self, f, flags):
        self._flags[f] = flags
    def copy(self):
        return manifestdict(self, dict.copy(self._flags))

class manifest(revlog.revlog):
    def __init__(self, opener):
        self._mancache = None
        revlog.revlog.__init__(self, opener, "00manifest.i")

    def parse(self, lines):
        mfdict = manifestdict()
        parsers.parse_manifest(mfdict, mfdict._flags, lines)
        return mfdict

    def readdelta(self, node):
        r = self.rev(node)
        return self.parse(mdiff.patchtext(self.revdiff(self.deltaparent(r), r)))

    def readfast(self, node):
        '''use the faster of readdelta or read'''
        r = self.rev(node)
        deltaparent = self.deltaparent(r)
        if deltaparent != revlog.nullrev and deltaparent in self.parentrevs(r):
            return self.readdelta(node)
        return self.read(node)

    def read(self, node):
        if node == revlog.nullid:
            return manifestdict() # don't upset local cache
        if self._mancache and self._mancache[0] == node:
            return self._mancache[1]
        text = self.revision(node)
        arraytext = array.array('c', text)
        mapping = self.parse(text)
        self._mancache = (node, mapping, arraytext)
        return mapping

    def _search(self, m, s, lo=0, hi=None):
        '''return a tuple (start, end) that says where to find s within m.

        If the string is found m[start:end] are the line containing
        that string.  If start == end the string was not found and
        they indicate the proper sorted insertion point.  This was
        taken from bisect_left, and modified to find line start/end as
        it goes along.

        m should be a buffer or a string
        s is a string'''
        def advance(i, c):
            while i < lenm and m[i] != c:
                i += 1
            return i
        if not s:
            return (lo, lo)
        lenm = len(m)
        if not hi:
            hi = lenm
        while lo < hi:
            mid = (lo + hi) // 2
            start = mid
            while start > 0 and m[start - 1] != '\n':
                start -= 1
            end = advance(start, '\0')
            if m[start:end] < s:
                # we know that after the null there are 40 bytes of sha1
                # this translates to the bisect lo = mid + 1
                lo = advance(end + 40, '\n') + 1
            else:
                # this translates to the bisect hi = mid
                hi = start
        end = advance(lo, '\0')
        found = m[lo:end]
        if s == found:
            # we know that after the null there are 40 bytes of sha1
            end = advance(end + 40, '\n')
            return (lo, end + 1)
        else:
            return (lo, lo)

    def find(self, node, f):
        '''look up entry for a single file efficiently.
        return (node, flags) pair if found, (None, None) if not.'''
        if self._mancache and self._mancache[0] == node:
            return self._mancache[1].get(f), self._mancache[1].flags(f)
        text = self.revision(node)
        start, end = self._search(text, f)
        if start == end:
            return None, None
        l = text[start:end]
        f, n = l.split('\0')
        return revlog.bin(n[:40]), n[40:-1]

    def add(self, map, transaction, link, p1=None, p2=None,
            changed=None):
        # apply the changes collected during the bisect loop to our addlist
        # return a delta suitable for addrevision
        def addlistdelta(addlist, x):
            # start from the bottom up
            # so changes to the offsets don't mess things up.
            for start, end, content in reversed(x):
                if content:
                    addlist[start:end] = array.array('c', content)
                else:
                    del addlist[start:end]
            return "".join(struct.pack(">lll", start, end, len(content)) + content
                           for start, end, content in x)

        def checkforbidden(l):
            for f in l:
                if '\n' in f or '\r' in f:
                    raise error.RevlogError(
                        _("'\\n' and '\\r' disallowed in filenames: %r") % f)

        # if we're using the cache, make sure it is valid and
        # parented by the same node we're diffing against
        if not (changed and self._mancache and p1 and self._mancache[0] == p1):
            files = sorted(map)
            checkforbidden(files)

            # if this is changed to support newlines in filenames,
            # be sure to check the templates/ dir again (especially *-raw.tmpl)
            hex, flags = revlog.hex, map.flags
            text = ''.join("%s\000%s%s\n" % (f, hex(map[f]), flags(f))
                           for f in files)
            arraytext = array.array('c', text)
            cachedelta = None
        else:
            added, removed = changed
            addlist = self._mancache[2]

            checkforbidden(added)
            # combine the changed lists into one list for sorting
            work = [(x, False) for x in added]
            work.extend((x, True) for x in removed)
            # this could use heapq.merge() (from python2.6+) or equivalent
            # since the lists are already sorted
            work.sort()

            delta = []
            dstart = None
            dend = None
            dline = [""]
            start = 0
            # zero copy representation of addlist as a buffer
            addbuf = buffer(addlist)

            # start with a readonly loop that finds the offset of
            # each line and creates the deltas
            for f, todelete in work:
                # bs will either be the index of the item or the insert point
                start, end = self._search(addbuf, f, start)
                if not todelete:
                    l = "%s\000%s%s\n" % (f, revlog.hex(map[f]), map.flags(f))
                else:
                    if start == end:
                        # item we want to delete was not found, error out
                        raise AssertionError(
                                _("failed to remove %s from manifest") % f)
                    l = ""
                if dstart is not None and dstart <= start and dend >= start:
                    if dend < end:
                        dend = end
                    if l:
                        dline.append(l)
                else:
                    if dstart is not None:
                        delta.append([dstart, dend, "".join(dline)])
                    dstart = start
                    dend = end
                    dline = [l]

            if dstart is not None:
                delta.append([dstart, dend, "".join(dline)])
            # apply the delta to the addlist, and get a delta for addrevision
            cachedelta = (self.rev(p1), addlistdelta(addlist, delta))
            arraytext = addlist
            text = buffer(arraytext)

        n = self.addrevision(text, transaction, link, p1, p2, cachedelta)
        self._mancache = (n, map, arraytext)

        return n