2005-08-28 01:21:25 +04:00
|
|
|
# manifest.py - manifest revision class for mercurial
|
|
|
|
#
|
2007-06-19 10:51:34 +04:00
|
|
|
# Copyright 2005-2007 Matt Mackall <mpm@selenic.com>
|
2005-08-28 01:21:25 +04:00
|
|
|
#
|
2009-04-26 03:08:54 +04:00
|
|
|
# This software may be used and distributed according to the terms of the
|
2010-01-20 07:20:08 +03:00
|
|
|
# GNU General Public License version 2 or any later version.
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2015-12-22 08:35:46 +03:00
|
|
|
from __future__ import absolute_import
|
|
|
|
|
2015-09-04 12:57:58 +03:00
|
|
|
import heapq
|
2015-12-22 08:35:46 +03:00
|
|
|
import os
|
|
|
|
import struct
|
|
|
|
|
|
|
|
from .i18n import _
|
2017-03-19 08:11:37 +03:00
|
|
|
from .node import (
|
|
|
|
bin,
|
|
|
|
hex,
|
|
|
|
)
|
2015-12-22 08:35:46 +03:00
|
|
|
from . import (
|
|
|
|
error,
|
|
|
|
mdiff,
|
|
|
|
parsers,
|
|
|
|
revlog,
|
|
|
|
util,
|
|
|
|
)
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2015-03-14 01:19:54 +03:00
|
|
|
propertycache = util.propertycache
|
2015-03-07 19:42:05 +03:00
|
|
|
|
2015-03-28 08:26:41 +03:00
|
|
|
def _parsev1(data):
|
2015-03-28 01:02:43 +03:00
|
|
|
# This method does a little bit of excessive-looking
|
|
|
|
# precondition checking. This is so that the behavior of this
|
|
|
|
# class exactly matches its C counterpart to try and help
|
|
|
|
# prevent surprise breakage for anyone that develops against
|
|
|
|
# the pure version.
|
|
|
|
if data and data[-1] != '\n':
|
|
|
|
raise ValueError('Manifest did not end in a newline.')
|
|
|
|
prev = None
|
|
|
|
for l in data.splitlines():
|
|
|
|
if prev is not None and prev > l:
|
|
|
|
raise ValueError('Manifest lines not in sorted order.')
|
|
|
|
prev = l
|
|
|
|
f, n = l.split('\0')
|
|
|
|
if len(n) > 40:
|
2017-03-12 10:30:15 +03:00
|
|
|
yield f, bin(n[:40]), n[40:]
|
2015-03-28 01:02:43 +03:00
|
|
|
else:
|
2017-03-12 10:30:15 +03:00
|
|
|
yield f, bin(n), ''
|
2015-03-28 01:02:43 +03:00
|
|
|
|
2015-03-28 08:26:41 +03:00
|
|
|
def _parsev2(data):
|
|
|
|
metadataend = data.find('\n')
|
|
|
|
# Just ignore metadata for now
|
|
|
|
pos = metadataend + 1
|
|
|
|
prevf = ''
|
|
|
|
while pos < len(data):
|
|
|
|
end = data.find('\n', pos + 1) # +1 to skip stem length byte
|
|
|
|
if end == -1:
|
|
|
|
raise ValueError('Manifest ended with incomplete file entry.')
|
|
|
|
stemlen = ord(data[pos])
|
|
|
|
items = data[pos + 1:end].split('\0')
|
|
|
|
f = prevf[:stemlen] + items[0]
|
|
|
|
if prevf > f:
|
|
|
|
raise ValueError('Manifest entries not in sorted order.')
|
|
|
|
fl = items[1]
|
|
|
|
# Just ignore metadata (items[2:] for now)
|
|
|
|
n = data[end + 1:end + 21]
|
|
|
|
yield f, n, fl
|
|
|
|
pos = end + 22
|
|
|
|
prevf = f
|
|
|
|
|
|
|
|
def _parse(data):
|
|
|
|
"""Generates (path, node, flags) tuples from a manifest text"""
|
|
|
|
if data.startswith('\0'):
|
|
|
|
return iter(_parsev2(data))
|
|
|
|
else:
|
|
|
|
return iter(_parsev1(data))
|
|
|
|
|
2015-04-01 00:01:33 +03:00
|
|
|
def _text(it, usemanifestv2):
|
2015-03-28 01:37:46 +03:00
|
|
|
"""Given an iterator over (path, node, flags) tuples, returns a manifest
|
|
|
|
text"""
|
2015-04-01 00:01:33 +03:00
|
|
|
if usemanifestv2:
|
|
|
|
return _textv2(it)
|
|
|
|
else:
|
|
|
|
return _textv1(it)
|
|
|
|
|
|
|
|
def _textv1(it):
|
2015-03-28 01:37:46 +03:00
|
|
|
files = []
|
|
|
|
lines = []
|
|
|
|
_hex = revlog.hex
|
|
|
|
for f, n, fl in it:
|
|
|
|
files.append(f)
|
|
|
|
# if this is changed to support newlines in filenames,
|
|
|
|
# be sure to check the templates/ dir again (especially *-raw.tmpl)
|
|
|
|
lines.append("%s\0%s%s\n" % (f, _hex(n), fl))
|
|
|
|
|
|
|
|
_checkforbidden(files)
|
|
|
|
return ''.join(lines)
|
|
|
|
|
2015-04-01 00:01:33 +03:00
|
|
|
def _textv2(it):
|
|
|
|
files = []
|
|
|
|
lines = ['\0\n']
|
|
|
|
prevf = ''
|
|
|
|
for f, n, fl in it:
|
|
|
|
files.append(f)
|
|
|
|
stem = os.path.commonprefix([prevf, f])
|
|
|
|
stemlen = min(len(stem), 255)
|
|
|
|
lines.append("%c%s\0%s\n%s\n" % (stemlen, f[stemlen:], fl, n))
|
|
|
|
prevf = f
|
|
|
|
_checkforbidden(files)
|
|
|
|
return ''.join(lines)
|
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
class lazymanifestiter(object):
|
|
|
|
def __init__(self, lm):
|
|
|
|
self.pos = 0
|
|
|
|
self.lm = lm
|
2015-03-07 20:04:39 +03:00
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def __iter__(self):
|
|
|
|
return self
|
|
|
|
|
|
|
|
def next(self):
|
|
|
|
try:
|
|
|
|
data, pos = self.lm._get(self.pos)
|
|
|
|
except IndexError:
|
|
|
|
raise StopIteration
|
|
|
|
if pos == -1:
|
|
|
|
self.pos += 1
|
|
|
|
return data[0]
|
|
|
|
self.pos += 1
|
|
|
|
zeropos = data.find('\x00', pos)
|
|
|
|
return data[pos:zeropos]
|
|
|
|
|
2017-03-12 08:43:20 +03:00
|
|
|
__next__ = next
|
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
class lazymanifestiterentries(object):
|
|
|
|
def __init__(self, lm):
|
|
|
|
self.lm = lm
|
|
|
|
self.pos = 0
|
2015-03-07 20:04:39 +03:00
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def __iter__(self):
|
|
|
|
return self
|
|
|
|
|
|
|
|
def next(self):
|
|
|
|
try:
|
|
|
|
data, pos = self.lm._get(self.pos)
|
|
|
|
except IndexError:
|
|
|
|
raise StopIteration
|
|
|
|
if pos == -1:
|
|
|
|
self.pos += 1
|
|
|
|
return data
|
|
|
|
zeropos = data.find('\x00', pos)
|
|
|
|
hashval = unhexlify(data, self.lm.extrainfo[self.pos],
|
|
|
|
zeropos + 1, 40)
|
|
|
|
flags = self.lm._getflags(data, self.pos, zeropos)
|
|
|
|
self.pos += 1
|
|
|
|
return (data[pos:zeropos], hashval, flags)
|
|
|
|
|
2017-03-12 08:43:20 +03:00
|
|
|
__next__ = next
|
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def unhexlify(data, extra, pos, length):
|
2017-03-12 10:29:48 +03:00
|
|
|
s = bin(data[pos:pos + length])
|
2016-09-12 14:37:14 +03:00
|
|
|
if extra:
|
|
|
|
s += chr(extra & 0xff)
|
|
|
|
return s
|
|
|
|
|
|
|
|
def _cmp(a, b):
|
|
|
|
return (a > b) - (a < b)
|
|
|
|
|
|
|
|
class _lazymanifest(object):
|
|
|
|
def __init__(self, data, positions=None, extrainfo=None, extradata=None):
|
|
|
|
if positions is None:
|
|
|
|
self.positions = self.findlines(data)
|
|
|
|
self.extrainfo = [0] * len(self.positions)
|
|
|
|
self.data = data
|
|
|
|
self.extradata = []
|
|
|
|
else:
|
|
|
|
self.positions = positions[:]
|
|
|
|
self.extrainfo = extrainfo[:]
|
|
|
|
self.extradata = extradata[:]
|
|
|
|
self.data = data
|
|
|
|
|
|
|
|
def findlines(self, data):
|
|
|
|
if not data:
|
|
|
|
return []
|
|
|
|
pos = data.find("\n")
|
2017-03-12 08:44:21 +03:00
|
|
|
if pos == -1 or data[-1:] != '\n':
|
2016-09-12 14:37:14 +03:00
|
|
|
raise ValueError("Manifest did not end in a newline.")
|
|
|
|
positions = [0]
|
|
|
|
prev = data[:data.find('\x00')]
|
|
|
|
while pos < len(data) - 1 and pos != -1:
|
|
|
|
positions.append(pos + 1)
|
|
|
|
nexts = data[pos + 1:data.find('\x00', pos + 1)]
|
|
|
|
if nexts < prev:
|
|
|
|
raise ValueError("Manifest lines not in sorted order.")
|
|
|
|
prev = nexts
|
|
|
|
pos = data.find("\n", pos + 1)
|
|
|
|
return positions
|
|
|
|
|
|
|
|
def _get(self, index):
|
|
|
|
# get the position encoded in pos:
|
|
|
|
# positive number is an index in 'data'
|
|
|
|
# negative number is in extrapieces
|
|
|
|
pos = self.positions[index]
|
|
|
|
if pos >= 0:
|
|
|
|
return self.data, pos
|
|
|
|
return self.extradata[-pos - 1], -1
|
|
|
|
|
|
|
|
def _getkey(self, pos):
|
|
|
|
if pos >= 0:
|
|
|
|
return self.data[pos:self.data.find('\x00', pos + 1)]
|
|
|
|
return self.extradata[-pos - 1][0]
|
|
|
|
|
|
|
|
def bsearch(self, key):
|
|
|
|
first = 0
|
|
|
|
last = len(self.positions) - 1
|
|
|
|
|
|
|
|
while first <= last:
|
|
|
|
midpoint = (first + last)//2
|
|
|
|
nextpos = self.positions[midpoint]
|
|
|
|
candidate = self._getkey(nextpos)
|
|
|
|
r = _cmp(key, candidate)
|
|
|
|
if r == 0:
|
|
|
|
return midpoint
|
|
|
|
else:
|
|
|
|
if r < 0:
|
|
|
|
last = midpoint - 1
|
|
|
|
else:
|
|
|
|
first = midpoint + 1
|
|
|
|
return -1
|
|
|
|
|
|
|
|
def bsearch2(self, key):
|
|
|
|
# same as the above, but will always return the position
|
|
|
|
# done for performance reasons
|
|
|
|
first = 0
|
|
|
|
last = len(self.positions) - 1
|
|
|
|
|
|
|
|
while first <= last:
|
|
|
|
midpoint = (first + last)//2
|
|
|
|
nextpos = self.positions[midpoint]
|
|
|
|
candidate = self._getkey(nextpos)
|
|
|
|
r = _cmp(key, candidate)
|
|
|
|
if r == 0:
|
|
|
|
return (midpoint, True)
|
|
|
|
else:
|
|
|
|
if r < 0:
|
|
|
|
last = midpoint - 1
|
|
|
|
else:
|
|
|
|
first = midpoint + 1
|
|
|
|
return (first, False)
|
2015-03-07 19:43:12 +03:00
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def __contains__(self, key):
|
|
|
|
return self.bsearch(key) != -1
|
2015-03-07 20:04:39 +03:00
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def _getflags(self, data, needle, pos):
|
|
|
|
start = pos + 41
|
|
|
|
end = data.find("\n", start)
|
|
|
|
if end == -1:
|
|
|
|
end = len(data) - 1
|
|
|
|
if start == end:
|
|
|
|
return ''
|
|
|
|
return self.data[start:end]
|
2015-03-07 20:04:39 +03:00
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def __getitem__(self, key):
|
2017-03-12 10:31:54 +03:00
|
|
|
if not isinstance(key, bytes):
|
|
|
|
raise TypeError("getitem: manifest keys must be a bytes.")
|
2016-09-12 14:37:14 +03:00
|
|
|
needle = self.bsearch(key)
|
|
|
|
if needle == -1:
|
|
|
|
raise KeyError
|
|
|
|
data, pos = self._get(needle)
|
|
|
|
if pos == -1:
|
|
|
|
return (data[1], data[2])
|
|
|
|
zeropos = data.find('\x00', pos)
|
|
|
|
assert 0 <= needle <= len(self.positions)
|
|
|
|
assert len(self.extrainfo) == len(self.positions)
|
|
|
|
hashval = unhexlify(data, self.extrainfo[needle], zeropos + 1, 40)
|
|
|
|
flags = self._getflags(data, needle, zeropos)
|
|
|
|
return (hashval, flags)
|
2015-03-13 04:53:44 +03:00
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def __delitem__(self, key):
|
|
|
|
needle, found = self.bsearch2(key)
|
|
|
|
if not found:
|
|
|
|
raise KeyError
|
|
|
|
cur = self.positions[needle]
|
|
|
|
self.positions = self.positions[:needle] + self.positions[needle + 1:]
|
|
|
|
self.extrainfo = self.extrainfo[:needle] + self.extrainfo[needle + 1:]
|
|
|
|
if cur >= 0:
|
|
|
|
self.data = self.data[:cur] + '\x00' + self.data[cur + 1:]
|
|
|
|
|
|
|
|
def __setitem__(self, key, value):
|
2017-03-19 08:12:03 +03:00
|
|
|
if not isinstance(key, bytes):
|
|
|
|
raise TypeError("setitem: manifest keys must be a byte string.")
|
2016-09-12 14:37:14 +03:00
|
|
|
if not isinstance(value, tuple) or len(value) != 2:
|
|
|
|
raise TypeError("Manifest values must be a tuple of (node, flags).")
|
|
|
|
hashval = value[0]
|
2017-03-19 08:12:03 +03:00
|
|
|
if not isinstance(hashval, bytes) or not 20 <= len(hashval) <= 22:
|
|
|
|
raise TypeError("node must be a 20-byte byte string")
|
2016-09-12 14:37:14 +03:00
|
|
|
flags = value[1]
|
|
|
|
if len(hashval) == 22:
|
|
|
|
hashval = hashval[:-1]
|
2017-03-19 08:12:03 +03:00
|
|
|
if not isinstance(flags, bytes) or len(flags) > 1:
|
2016-09-12 14:37:14 +03:00
|
|
|
raise TypeError("flags must a 0 or 1 byte string, got %r", flags)
|
|
|
|
needle, found = self.bsearch2(key)
|
|
|
|
if found:
|
|
|
|
# put the item
|
|
|
|
pos = self.positions[needle]
|
|
|
|
if pos < 0:
|
|
|
|
self.extradata[-pos - 1] = (key, hashval, value[1])
|
|
|
|
else:
|
|
|
|
# just don't bother
|
|
|
|
self.extradata.append((key, hashval, value[1]))
|
|
|
|
self.positions[needle] = -len(self.extradata)
|
|
|
|
else:
|
|
|
|
# not found, put it in with extra positions
|
|
|
|
self.extradata.append((key, hashval, value[1]))
|
|
|
|
self.positions = (self.positions[:needle] + [-len(self.extradata)]
|
|
|
|
+ self.positions[needle:])
|
|
|
|
self.extrainfo = (self.extrainfo[:needle] + [0] +
|
|
|
|
self.extrainfo[needle:])
|
2015-03-13 04:18:29 +03:00
|
|
|
|
2006-07-16 11:59:20 +04:00
|
|
|
def copy(self):
|
2016-09-12 14:37:14 +03:00
|
|
|
# XXX call _compact like in C?
|
|
|
|
return _lazymanifest(self.data, self.positions, self.extrainfo,
|
|
|
|
self.extradata)
|
|
|
|
|
|
|
|
def _compact(self):
|
|
|
|
# hopefully not called TOO often
|
|
|
|
if len(self.extradata) == 0:
|
|
|
|
return
|
|
|
|
l = []
|
|
|
|
last_cut = 0
|
|
|
|
i = 0
|
|
|
|
offset = 0
|
|
|
|
self.extrainfo = [0] * len(self.positions)
|
|
|
|
while i < len(self.positions):
|
|
|
|
if self.positions[i] >= 0:
|
|
|
|
cur = self.positions[i]
|
|
|
|
last_cut = cur
|
|
|
|
while True:
|
|
|
|
self.positions[i] = offset
|
|
|
|
i += 1
|
|
|
|
if i == len(self.positions) or self.positions[i] < 0:
|
|
|
|
break
|
|
|
|
offset += self.positions[i] - cur
|
|
|
|
cur = self.positions[i]
|
|
|
|
end_cut = self.data.find('\n', cur)
|
|
|
|
if end_cut != -1:
|
|
|
|
end_cut += 1
|
|
|
|
offset += end_cut - cur
|
|
|
|
l.append(self.data[last_cut:end_cut])
|
|
|
|
else:
|
|
|
|
while i < len(self.positions) and self.positions[i] < 0:
|
|
|
|
cur = self.positions[i]
|
|
|
|
t = self.extradata[-cur - 1]
|
|
|
|
l.append(self._pack(t))
|
|
|
|
self.positions[i] = offset
|
|
|
|
if len(t[1]) > 20:
|
|
|
|
self.extrainfo[i] = ord(t[1][21])
|
|
|
|
offset += len(l[-1])
|
|
|
|
i += 1
|
|
|
|
self.data = ''.join(l)
|
|
|
|
self.extradata = []
|
|
|
|
|
|
|
|
def _pack(self, d):
|
2017-03-19 08:11:37 +03:00
|
|
|
return d[0] + '\x00' + hex(d[1][:20]) + d[2] + '\n'
|
2016-09-12 14:37:14 +03:00
|
|
|
|
|
|
|
def text(self):
|
|
|
|
self._compact()
|
|
|
|
return self.data
|
2015-03-07 20:04:39 +03:00
|
|
|
|
|
|
|
def diff(self, m2, clean=False):
|
|
|
|
'''Finds changes between the current manifest and m2.'''
|
2016-09-12 14:37:14 +03:00
|
|
|
# XXX think whether efficiency matters here
|
2015-03-07 20:04:39 +03:00
|
|
|
diff = {}
|
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
for fn, e1, flags in self.iterentries():
|
2015-03-07 20:04:39 +03:00
|
|
|
if fn not in m2:
|
2016-09-12 14:37:14 +03:00
|
|
|
diff[fn] = (e1, flags), (None, '')
|
2015-03-07 20:04:39 +03:00
|
|
|
else:
|
|
|
|
e2 = m2[fn]
|
2016-09-12 14:37:14 +03:00
|
|
|
if (e1, flags) != e2:
|
|
|
|
diff[fn] = (e1, flags), e2
|
2015-03-07 20:04:39 +03:00
|
|
|
elif clean:
|
|
|
|
diff[fn] = None
|
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
for fn, e2, flags in m2.iterentries():
|
2015-03-07 20:04:39 +03:00
|
|
|
if fn not in self:
|
2016-09-12 14:37:14 +03:00
|
|
|
diff[fn] = (None, ''), (e2, flags)
|
2015-03-07 20:04:39 +03:00
|
|
|
|
|
|
|
return diff
|
|
|
|
|
2016-09-12 14:37:14 +03:00
|
|
|
def iterentries(self):
|
|
|
|
return lazymanifestiterentries(self)
|
|
|
|
|
|
|
|
def iterkeys(self):
|
|
|
|
return lazymanifestiter(self)
|
|
|
|
|
|
|
|
def __iter__(self):
|
|
|
|
return lazymanifestiter(self)
|
|
|
|
|
|
|
|
def __len__(self):
|
|
|
|
return len(self.positions)
|
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
def filtercopy(self, filterfn):
|
2016-09-12 14:37:14 +03:00
|
|
|
# XXX should be optimized
|
2015-03-07 20:04:39 +03:00
|
|
|
c = _lazymanifest('')
|
2015-03-13 04:18:29 +03:00
|
|
|
for f, n, fl in self.iterentries():
|
2015-03-07 20:04:39 +03:00
|
|
|
if filterfn(f):
|
|
|
|
c[f] = n, fl
|
|
|
|
return c
|
|
|
|
|
2015-03-07 05:29:47 +03:00
|
|
|
try:
|
|
|
|
_lazymanifest = parsers.lazymanifest
|
|
|
|
except AttributeError:
|
|
|
|
pass
|
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
class manifestdict(object):
|
|
|
|
def __init__(self, data=''):
|
2015-03-28 08:26:41 +03:00
|
|
|
if data.startswith('\0'):
|
|
|
|
#_lazymanifest can not parse v2
|
|
|
|
self._lm = _lazymanifest('')
|
|
|
|
for f, n, fl in _parsev2(data):
|
|
|
|
self._lm[f] = n, fl
|
|
|
|
else:
|
|
|
|
self._lm = _lazymanifest(data)
|
2015-03-07 20:04:39 +03:00
|
|
|
|
|
|
|
def __getitem__(self, key):
|
|
|
|
return self._lm[key][0]
|
|
|
|
|
2015-03-11 02:26:13 +03:00
|
|
|
def find(self, key):
|
|
|
|
return self._lm[key]
|
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
def __len__(self):
|
|
|
|
return len(self._lm)
|
|
|
|
|
2016-11-04 03:31:14 +03:00
|
|
|
def __nonzero__(self):
|
|
|
|
# nonzero is covered by the __len__ function, but implementing it here
|
|
|
|
# makes it easier for extensions to override.
|
|
|
|
return len(self._lm) != 0
|
|
|
|
|
2017-03-13 22:40:14 +03:00
|
|
|
__bool__ = __nonzero__
|
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
def __setitem__(self, key, node):
|
|
|
|
self._lm[key] = node, self.flags(key, '')
|
|
|
|
|
|
|
|
def __contains__(self, key):
|
|
|
|
return key in self._lm
|
|
|
|
|
|
|
|
def __delitem__(self, key):
|
|
|
|
del self._lm[key]
|
|
|
|
|
2015-03-11 23:46:15 +03:00
|
|
|
def __iter__(self):
|
2015-03-13 04:18:29 +03:00
|
|
|
return self._lm.__iter__()
|
2015-03-07 20:04:39 +03:00
|
|
|
|
|
|
|
def iterkeys(self):
|
2015-03-11 23:46:15 +03:00
|
|
|
return self._lm.iterkeys()
|
|
|
|
|
|
|
|
def keys(self):
|
|
|
|
return list(self.iterkeys())
|
2015-03-07 20:04:39 +03:00
|
|
|
|
2017-03-07 20:56:11 +03:00
|
|
|
def filesnotin(self, m2, match=None):
|
2015-02-28 00:57:37 +03:00
|
|
|
'''Set of files in this manifest that are not in the other'''
|
2017-03-07 20:56:11 +03:00
|
|
|
if match:
|
|
|
|
m1 = self.matches(match)
|
|
|
|
m2 = m2.matches(match)
|
|
|
|
return m1.filesnotin(m2)
|
2016-05-03 01:22:16 +03:00
|
|
|
diff = self.diff(m2)
|
|
|
|
files = set(filepath
|
|
|
|
for filepath, hashflags in diff.iteritems()
|
|
|
|
if hashflags[1][0] is None)
|
2015-02-28 00:57:37 +03:00
|
|
|
return files
|
|
|
|
|
2015-03-14 01:19:54 +03:00
|
|
|
@propertycache
|
|
|
|
def _dirs(self):
|
2015-04-07 00:36:08 +03:00
|
|
|
return util.dirs(self)
|
2015-03-14 01:19:54 +03:00
|
|
|
|
|
|
|
def dirs(self):
|
|
|
|
return self._dirs
|
|
|
|
|
2015-03-14 01:25:01 +03:00
|
|
|
def hasdir(self, dir):
|
|
|
|
return dir in self._dirs
|
|
|
|
|
2015-04-08 19:38:09 +03:00
|
|
|
def _filesfastpath(self, match):
|
|
|
|
'''Checks whether we can correctly and quickly iterate over matcher
|
|
|
|
files instead of over manifest files.'''
|
|
|
|
files = match.files()
|
|
|
|
return (len(files) < 100 and (match.isexact() or
|
2015-05-19 21:16:20 +03:00
|
|
|
(match.prefix() and all(fn in self for fn in files))))
|
2015-04-08 19:38:09 +03:00
|
|
|
|
2015-04-08 01:18:52 +03:00
|
|
|
def walk(self, match):
|
|
|
|
'''Generates matching file names.
|
|
|
|
|
|
|
|
Equivalent to manifest.matches(match).iterkeys(), but without creating
|
|
|
|
an entirely new manifest.
|
|
|
|
|
|
|
|
It also reports nonexistent files by marking them bad with match.bad().
|
|
|
|
'''
|
2015-04-08 07:08:23 +03:00
|
|
|
if match.always():
|
|
|
|
for f in iter(self):
|
|
|
|
yield f
|
|
|
|
return
|
|
|
|
|
2015-04-08 01:18:52 +03:00
|
|
|
fset = set(match.files())
|
|
|
|
|
|
|
|
# avoid the entire walk if we're only looking for specific files
|
2015-04-08 19:38:09 +03:00
|
|
|
if self._filesfastpath(match):
|
2015-04-08 08:35:44 +03:00
|
|
|
for fn in sorted(fset):
|
|
|
|
yield fn
|
2015-04-08 08:36:17 +03:00
|
|
|
return
|
2015-04-08 01:18:52 +03:00
|
|
|
|
|
|
|
for fn in self:
|
|
|
|
if fn in fset:
|
|
|
|
# specified pattern is the exact name
|
|
|
|
fset.remove(fn)
|
|
|
|
if match(fn):
|
|
|
|
yield fn
|
|
|
|
|
|
|
|
# for dirstate.walk, files=['.'] means "walk the whole tree".
|
|
|
|
# follow that here, too
|
|
|
|
fset.discard('.')
|
|
|
|
|
|
|
|
for fn in sorted(fset):
|
|
|
|
if not self.hasdir(fn):
|
|
|
|
match.bad(fn, None)
|
|
|
|
|
2014-10-23 08:38:30 +04:00
|
|
|
def matches(self, match):
|
|
|
|
'''generate a new manifest filtered by the match argument'''
|
|
|
|
if match.always():
|
|
|
|
return self.copy()
|
|
|
|
|
2015-04-08 19:38:09 +03:00
|
|
|
if self._filesfastpath(match):
|
2015-04-08 20:01:31 +03:00
|
|
|
m = manifestdict()
|
|
|
|
lm = self._lm
|
|
|
|
for fn in match.files():
|
|
|
|
if fn in lm:
|
|
|
|
m._lm[fn] = lm[fn]
|
|
|
|
return m
|
2014-10-23 08:38:30 +04:00
|
|
|
|
2015-04-11 04:13:01 +03:00
|
|
|
m = manifestdict()
|
2015-04-08 20:06:05 +03:00
|
|
|
m._lm = self._lm.filtercopy(match)
|
|
|
|
return m
|
2014-10-23 08:38:30 +04:00
|
|
|
|
2017-03-07 20:56:11 +03:00
|
|
|
def diff(self, m2, match=None, clean=False):
|
2014-12-16 00:04:28 +03:00
|
|
|
'''Finds changes between the current manifest and m2.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
m2: the manifest to which this manifest should be compared.
|
|
|
|
clean: if true, include files unchanged between these manifests
|
|
|
|
with a None value in the returned dictionary.
|
|
|
|
|
|
|
|
The result is returned as a dict with filename as key and
|
|
|
|
values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
|
|
|
|
nodeid in the current/other manifest and fl1/fl2 is the flag
|
|
|
|
in the current/other manifest. Where the file does not exist,
|
|
|
|
the nodeid will be None and the flags will be the empty
|
|
|
|
string.
|
|
|
|
'''
|
2017-03-07 20:56:11 +03:00
|
|
|
if match:
|
|
|
|
m1 = self.matches(match)
|
|
|
|
m2 = m2.matches(match)
|
|
|
|
return m1.diff(m2, clean=clean)
|
2015-03-07 20:04:39 +03:00
|
|
|
return self._lm.diff(m2._lm, clean)
|
manifest: for diff(), only iterate over files, not flags
From manifest.diff(), we return a dict from filename to pairs of pairs
of file nodeids and flags (values of the form ((n1,n2),(fl1,fl2))). To
create this dict, we currently generate one dict for files (with
(n1,n2) values) and one for flags (with (fl1,fl2) values) and then
join these dicts. Missing files are represented by None and missing
flags by '', but due to the dict joining, the inner pairs themselves
can also be None. The only caller, merge.manifestmerge(), then unpacks
these values while checking for None values.
By inlining the calls to dicthelpers and simplifying it to only
iterate over files (ignoring flags-only differences), we can simplify
life for our caller.
2014-10-15 09:48:44 +04:00
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
def setflag(self, key, flag):
|
|
|
|
self._lm[key] = self[key], flag
|
manifest: for diff(), only iterate over files, not flags
From manifest.diff(), we return a dict from filename to pairs of pairs
of file nodeids and flags (values of the form ((n1,n2),(fl1,fl2))). To
create this dict, we currently generate one dict for files (with
(n1,n2) values) and one for flags (with (fl1,fl2) values) and then
join these dicts. Missing files are represented by None and missing
flags by '', but due to the dict joining, the inner pairs themselves
can also be None. The only caller, merge.manifestmerge(), then unpacks
these values while checking for None values.
By inlining the calls to dicthelpers and simplifying it to only
iterate over files (ignoring flags-only differences), we can simplify
life for our caller.
2014-10-15 09:48:44 +04:00
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
def get(self, key, default=None):
|
|
|
|
try:
|
|
|
|
return self._lm[key][0]
|
|
|
|
except KeyError:
|
|
|
|
return default
|
2006-07-16 11:59:20 +04:00
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
def flags(self, key, default=''):
|
|
|
|
try:
|
|
|
|
return self._lm[key][1]
|
|
|
|
except KeyError:
|
|
|
|
return default
|
2014-10-08 22:47:30 +04:00
|
|
|
|
2015-03-07 20:04:39 +03:00
|
|
|
def copy(self):
|
2015-04-11 04:13:01 +03:00
|
|
|
c = manifestdict()
|
2015-03-07 20:04:39 +03:00
|
|
|
c._lm = self._lm.copy()
|
|
|
|
return c
|
|
|
|
|
|
|
|
def iteritems(self):
|
2015-03-13 04:18:29 +03:00
|
|
|
return (x[:2] for x in self._lm.iterentries())
|
2015-03-07 20:04:39 +03:00
|
|
|
|
2016-02-08 08:13:24 +03:00
|
|
|
def iterentries(self):
|
|
|
|
return self._lm.iterentries()
|
|
|
|
|
2015-04-01 00:01:33 +03:00
|
|
|
def text(self, usemanifestv2=False):
|
|
|
|
if usemanifestv2:
|
|
|
|
return _textv2(self._lm.iterentries())
|
|
|
|
else:
|
|
|
|
# use (probably) native version for v1
|
|
|
|
return self._lm.text()
|
2014-08-07 17:46:05 +04:00
|
|
|
|
2014-10-08 23:20:14 +04:00
|
|
|
def fastdelta(self, base, changes):
|
2017-04-03 18:45:24 +03:00
|
|
|
"""Given a base manifest text as a bytearray and a list of changes
|
2014-10-08 23:20:14 +04:00
|
|
|
relative to that text, compute a delta that can be used by revlog.
|
|
|
|
"""
|
|
|
|
delta = []
|
|
|
|
dstart = None
|
|
|
|
dend = None
|
|
|
|
dline = [""]
|
|
|
|
start = 0
|
|
|
|
# zero copy representation of base as a buffer
|
|
|
|
addbuf = util.buffer(base)
|
|
|
|
|
2015-11-06 05:56:40 +03:00
|
|
|
changes = list(changes)
|
|
|
|
if len(changes) < 1000:
|
|
|
|
# start with a readonly loop that finds the offset of
|
|
|
|
# each line and creates the deltas
|
|
|
|
for f, todelete in changes:
|
|
|
|
# bs will either be the index of the item or the insert point
|
|
|
|
start, end = _msearch(addbuf, f, start)
|
|
|
|
if not todelete:
|
|
|
|
h, fl = self._lm[f]
|
|
|
|
l = "%s\0%s%s\n" % (f, revlog.hex(h), fl)
|
|
|
|
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
|
2014-10-08 23:20:14 +04:00
|
|
|
dend = end
|
2015-11-06 05:56:40 +03:00
|
|
|
dline = [l]
|
|
|
|
|
|
|
|
if dstart is not None:
|
|
|
|
delta.append([dstart, dend, "".join(dline)])
|
|
|
|
# apply the delta to the base, and get a delta for addrevision
|
|
|
|
deltatext, arraytext = _addlistdelta(base, delta)
|
|
|
|
else:
|
|
|
|
# For large changes, it's much cheaper to just build the text and
|
|
|
|
# diff it.
|
2017-03-12 10:32:21 +03:00
|
|
|
arraytext = bytearray(self.text())
|
|
|
|
deltatext = mdiff.textdiff(
|
|
|
|
util.buffer(base), util.buffer(arraytext))
|
2015-11-06 05:56:40 +03:00
|
|
|
|
2014-10-08 23:20:14 +04:00
|
|
|
return arraytext, deltatext
|
|
|
|
|
2014-10-08 23:21:59 +04:00
|
|
|
def _msearch(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.
|
|
|
|
|
2017-03-26 13:06:48 +03:00
|
|
|
m should be a buffer, a memoryview or a byte string.
|
|
|
|
s is a byte string'''
|
2014-10-08 23:21:59 +04:00
|
|
|
def advance(i, c):
|
2017-03-26 13:06:48 +03:00
|
|
|
while i < lenm and m[i:i + 1] != c:
|
2014-10-08 23:21:59 +04:00
|
|
|
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
|
2017-03-26 13:06:48 +03:00
|
|
|
while start > 0 and m[start - 1:start] != '\n':
|
2014-10-08 23:21:59 +04:00
|
|
|
start -= 1
|
|
|
|
end = advance(start, '\0')
|
2017-03-26 13:06:48 +03:00
|
|
|
if bytes(m[start:end]) < s:
|
2014-10-08 23:21:59 +04:00
|
|
|
# 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)
|
|
|
|
|
2014-09-11 18:14:34 +04:00
|
|
|
def _checkforbidden(l):
|
2014-08-07 17:46:05 +04:00
|
|
|
"""Check filenames for illegal characters."""
|
|
|
|
for f in l:
|
|
|
|
if '\n' in f or '\r' in f:
|
|
|
|
raise error.RevlogError(
|
|
|
|
_("'\\n' and '\\r' disallowed in filenames: %r") % f)
|
|
|
|
|
|
|
|
|
2014-08-07 20:47:20 +04:00
|
|
|
# apply the changes collected during the bisect loop to our addlist
|
|
|
|
# return a delta suitable for addrevision
|
2014-09-11 18:14:34 +04:00
|
|
|
def _addlistdelta(addlist, x):
|
2014-08-07 20:47:20 +04:00
|
|
|
# for large addlist arrays, building a new array is cheaper
|
|
|
|
# than repeatedly modifying the existing one
|
|
|
|
currentposition = 0
|
2017-03-12 10:32:21 +03:00
|
|
|
newaddlist = bytearray()
|
2014-08-07 20:47:20 +04:00
|
|
|
|
|
|
|
for start, end, content in x:
|
|
|
|
newaddlist += addlist[currentposition:start]
|
|
|
|
if content:
|
2017-03-12 10:32:21 +03:00
|
|
|
newaddlist += bytearray(content)
|
2014-08-07 20:47:20 +04:00
|
|
|
|
|
|
|
currentposition = end
|
|
|
|
|
|
|
|
newaddlist += addlist[currentposition:]
|
|
|
|
|
|
|
|
deltatext = "".join(struct.pack(">lll", start, end, len(content))
|
|
|
|
+ content for start, end, content in x)
|
|
|
|
return deltatext, newaddlist
|
|
|
|
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
def _splittopdir(f):
|
|
|
|
if '/' in f:
|
|
|
|
dir, subpath = f.split('/', 1)
|
|
|
|
return dir + '/', subpath
|
|
|
|
else:
|
|
|
|
return '', f
|
|
|
|
|
2015-09-26 05:54:46 +03:00
|
|
|
_noop = lambda s: None
|
2015-04-10 03:14:35 +03:00
|
|
|
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
class treemanifest(object):
|
2015-02-23 21:57:57 +03:00
|
|
|
def __init__(self, dir='', text=''):
|
|
|
|
self._dir = dir
|
2015-04-14 09:21:02 +03:00
|
|
|
self._node = revlog.nullid
|
2015-09-26 05:54:46 +03:00
|
|
|
self._loadfunc = _noop
|
|
|
|
self._copyfunc = _noop
|
2015-02-26 19:16:13 +03:00
|
|
|
self._dirty = False
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
self._dirs = {}
|
|
|
|
# Using _lazymanifest here is a little slower than plain old dicts
|
|
|
|
self._files = {}
|
|
|
|
self._flags = {}
|
2015-02-26 19:16:13 +03:00
|
|
|
if text:
|
|
|
|
def readsubtree(subdir, subm):
|
|
|
|
raise AssertionError('treemanifest constructor only accepts '
|
|
|
|
'flat manifests')
|
|
|
|
self.parse(text, readsubtree)
|
|
|
|
self._dirty = True # Mark flat manifest dirty after parsing
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2015-02-23 21:57:57 +03:00
|
|
|
def _subpath(self, path):
|
|
|
|
return self._dir + path
|
|
|
|
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
def __len__(self):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
size = len(self._files)
|
|
|
|
for m in self._dirs.values():
|
|
|
|
size += m.__len__()
|
|
|
|
return size
|
|
|
|
|
2015-03-31 03:21:49 +03:00
|
|
|
def _isempty(self):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load() # for consistency; already loaded by all callers
|
2015-03-31 03:21:49 +03:00
|
|
|
return (not self._files and (not self._dirs or
|
2015-05-16 21:34:19 +03:00
|
|
|
all(m._isempty() for m in self._dirs.values())))
|
2015-03-31 03:21:49 +03:00
|
|
|
|
2015-09-26 00:17:36 +03:00
|
|
|
def __repr__(self):
|
|
|
|
return ('<treemanifest dir=%s, node=%s, loaded=%s, dirty=%s at 0x%x>' %
|
2015-04-10 03:14:35 +03:00
|
|
|
(self._dir, revlog.hex(self._node),
|
2015-09-26 05:54:46 +03:00
|
|
|
bool(self._loadfunc is _noop),
|
2015-09-26 00:17:36 +03:00
|
|
|
self._dirty, id(self)))
|
2015-04-14 09:21:02 +03:00
|
|
|
|
|
|
|
def dir(self):
|
|
|
|
'''The directory that this tree manifest represents, including a
|
|
|
|
trailing '/'. Empty string for the repo root directory.'''
|
|
|
|
return self._dir
|
|
|
|
|
|
|
|
def node(self):
|
|
|
|
'''This node of this instance. nullid for unsaved instances. Should
|
|
|
|
be updated when the instance is read or written from a revlog.
|
|
|
|
'''
|
2015-02-26 19:16:13 +03:00
|
|
|
assert not self._dirty
|
2015-04-14 09:21:02 +03:00
|
|
|
return self._node
|
|
|
|
|
|
|
|
def setnode(self, node):
|
|
|
|
self._node = node
|
2015-02-26 19:16:13 +03:00
|
|
|
self._dirty = False
|
2015-02-23 21:57:57 +03:00
|
|
|
|
2016-02-08 08:14:01 +03:00
|
|
|
def iterentries(self):
|
|
|
|
self._load()
|
|
|
|
for p, n in sorted(self._dirs.items() + self._files.items()):
|
|
|
|
if p in self._files:
|
|
|
|
yield self._subpath(p), n, self._flags.get(p, '')
|
|
|
|
else:
|
|
|
|
for x in n.iterentries():
|
|
|
|
yield x
|
|
|
|
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
def iteritems(self):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
for p, n in sorted(self._dirs.items() + self._files.items()):
|
|
|
|
if p in self._files:
|
2015-02-23 21:57:57 +03:00
|
|
|
yield self._subpath(p), n
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
else:
|
2015-02-23 21:57:57 +03:00
|
|
|
for f, sn in n.iteritems():
|
|
|
|
yield f, sn
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
|
|
|
def iterkeys(self):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
for p in sorted(self._dirs.keys() + self._files.keys()):
|
|
|
|
if p in self._files:
|
2015-02-23 21:57:57 +03:00
|
|
|
yield self._subpath(p)
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
else:
|
|
|
|
for f in self._dirs[p].iterkeys():
|
2015-02-23 21:57:57 +03:00
|
|
|
yield f
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
|
|
|
def keys(self):
|
|
|
|
return list(self.iterkeys())
|
|
|
|
|
|
|
|
def __iter__(self):
|
|
|
|
return self.iterkeys()
|
|
|
|
|
|
|
|
def __contains__(self, f):
|
|
|
|
if f is None:
|
|
|
|
return False
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
if dir not in self._dirs:
|
|
|
|
return False
|
|
|
|
return self._dirs[dir].__contains__(subpath)
|
|
|
|
else:
|
|
|
|
return f in self._files
|
|
|
|
|
|
|
|
def get(self, f, default=None):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
if dir not in self._dirs:
|
|
|
|
return default
|
|
|
|
return self._dirs[dir].get(subpath, default)
|
|
|
|
else:
|
|
|
|
return self._files.get(f, default)
|
|
|
|
|
|
|
|
def __getitem__(self, f):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
return self._dirs[dir].__getitem__(subpath)
|
|
|
|
else:
|
|
|
|
return self._files[f]
|
|
|
|
|
|
|
|
def flags(self, f):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
if dir not in self._dirs:
|
|
|
|
return ''
|
|
|
|
return self._dirs[dir].flags(subpath)
|
|
|
|
else:
|
|
|
|
if f in self._dirs:
|
|
|
|
return ''
|
|
|
|
return self._flags.get(f, '')
|
|
|
|
|
|
|
|
def find(self, f):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
return self._dirs[dir].find(subpath)
|
|
|
|
else:
|
|
|
|
return self._files[f], self._flags.get(f, '')
|
|
|
|
|
|
|
|
def __delitem__(self, f):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
self._dirs[dir].__delitem__(subpath)
|
|
|
|
# If the directory is now empty, remove it
|
2015-03-31 03:21:49 +03:00
|
|
|
if self._dirs[dir]._isempty():
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
del self._dirs[dir]
|
|
|
|
else:
|
|
|
|
del self._files[f]
|
|
|
|
if f in self._flags:
|
|
|
|
del self._flags[f]
|
2015-02-26 19:16:13 +03:00
|
|
|
self._dirty = True
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
|
|
|
def __setitem__(self, f, n):
|
|
|
|
assert n is not None
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
if dir not in self._dirs:
|
2015-02-23 21:57:57 +03:00
|
|
|
self._dirs[dir] = treemanifest(self._subpath(dir))
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
self._dirs[dir].__setitem__(subpath, n)
|
|
|
|
else:
|
2015-03-26 19:42:21 +03:00
|
|
|
self._files[f] = n[:21] # to match manifestdict's behavior
|
2015-02-26 19:16:13 +03:00
|
|
|
self._dirty = True
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2015-09-26 05:54:46 +03:00
|
|
|
def _load(self):
|
|
|
|
if self._loadfunc is not _noop:
|
|
|
|
lf, self._loadfunc = self._loadfunc, _noop
|
|
|
|
lf(self)
|
|
|
|
elif self._copyfunc is not _noop:
|
|
|
|
cf, self._copyfunc = self._copyfunc, _noop
|
|
|
|
cf(self)
|
|
|
|
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
def setflag(self, f, flags):
|
|
|
|
"""Set the flags (symlink, executable) for path f."""
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
dir, subpath = _splittopdir(f)
|
|
|
|
if dir:
|
|
|
|
if dir not in self._dirs:
|
2015-02-23 21:57:57 +03:00
|
|
|
self._dirs[dir] = treemanifest(self._subpath(dir))
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
self._dirs[dir].setflag(subpath, flags)
|
|
|
|
else:
|
|
|
|
self._flags[f] = flags
|
2015-02-26 19:16:13 +03:00
|
|
|
self._dirty = True
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
|
|
|
def copy(self):
|
2015-02-23 21:57:57 +03:00
|
|
|
copy = treemanifest(self._dir)
|
2015-04-14 09:21:02 +03:00
|
|
|
copy._node = self._node
|
2015-02-26 19:16:13 +03:00
|
|
|
copy._dirty = self._dirty
|
2015-09-26 05:54:46 +03:00
|
|
|
if self._copyfunc is _noop:
|
|
|
|
def _copyfunc(s):
|
|
|
|
self._load()
|
|
|
|
for d in self._dirs:
|
|
|
|
s._dirs[d] = self._dirs[d].copy()
|
|
|
|
s._files = dict.copy(self._files)
|
|
|
|
s._flags = dict.copy(self._flags)
|
|
|
|
if self._loadfunc is _noop:
|
|
|
|
_copyfunc(copy)
|
|
|
|
else:
|
|
|
|
copy._copyfunc = _copyfunc
|
|
|
|
else:
|
|
|
|
copy._copyfunc = self._copyfunc
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
return copy
|
|
|
|
|
2017-03-07 20:56:11 +03:00
|
|
|
def filesnotin(self, m2, match=None):
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
'''Set of files in this manifest that are not in the other'''
|
2017-03-07 20:56:11 +03:00
|
|
|
if match:
|
|
|
|
m1 = self.matches(match)
|
|
|
|
m2 = m2.matches(match)
|
|
|
|
return m1.filesnotin(m2)
|
|
|
|
|
2015-03-04 00:50:06 +03:00
|
|
|
files = set()
|
|
|
|
def _filesnotin(t1, t2):
|
2015-02-26 19:16:13 +03:00
|
|
|
if t1._node == t2._node and not t1._dirty and not t2._dirty:
|
|
|
|
return
|
2015-04-10 03:14:35 +03:00
|
|
|
t1._load()
|
|
|
|
t2._load()
|
2015-03-04 00:50:06 +03:00
|
|
|
for d, m1 in t1._dirs.iteritems():
|
|
|
|
if d in t2._dirs:
|
|
|
|
m2 = t2._dirs[d]
|
|
|
|
_filesnotin(m1, m2)
|
|
|
|
else:
|
|
|
|
files.update(m1.iterkeys())
|
|
|
|
|
|
|
|
for fn in t1._files.iterkeys():
|
|
|
|
if fn not in t2._files:
|
|
|
|
files.add(t1._subpath(fn))
|
|
|
|
|
|
|
|
_filesnotin(self, m2)
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
return files
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _alldirs(self):
|
2015-04-07 00:36:08 +03:00
|
|
|
return util.dirs(self)
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
|
|
|
def dirs(self):
|
|
|
|
return self._alldirs
|
|
|
|
|
|
|
|
def hasdir(self, dir):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
2015-03-17 02:01:16 +03:00
|
|
|
topdir, subdir = _splittopdir(dir)
|
|
|
|
if topdir:
|
|
|
|
if topdir in self._dirs:
|
|
|
|
return self._dirs[topdir].hasdir(subdir)
|
|
|
|
return False
|
|
|
|
return (dir + '/') in self._dirs
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2015-04-08 01:18:52 +03:00
|
|
|
def walk(self, match):
|
|
|
|
'''Generates matching file names.
|
|
|
|
|
|
|
|
Equivalent to manifest.matches(match).iterkeys(), but without creating
|
|
|
|
an entirely new manifest.
|
|
|
|
|
|
|
|
It also reports nonexistent files by marking them bad with match.bad().
|
|
|
|
'''
|
2015-04-08 07:08:23 +03:00
|
|
|
if match.always():
|
|
|
|
for f in iter(self):
|
|
|
|
yield f
|
|
|
|
return
|
|
|
|
|
2015-04-08 01:18:52 +03:00
|
|
|
fset = set(match.files())
|
|
|
|
|
2015-04-08 01:18:52 +03:00
|
|
|
for fn in self._walk(match):
|
2015-04-08 01:18:52 +03:00
|
|
|
if fn in fset:
|
|
|
|
# specified pattern is the exact name
|
|
|
|
fset.remove(fn)
|
2015-04-08 01:18:52 +03:00
|
|
|
yield fn
|
2015-04-08 01:18:52 +03:00
|
|
|
|
|
|
|
# for dirstate.walk, files=['.'] means "walk the whole tree".
|
|
|
|
# follow that here, too
|
|
|
|
fset.discard('.')
|
|
|
|
|
|
|
|
for fn in sorted(fset):
|
|
|
|
if not self.hasdir(fn):
|
|
|
|
match.bad(fn, None)
|
|
|
|
|
2015-05-07 01:59:35 +03:00
|
|
|
def _walk(self, match):
|
|
|
|
'''Recursively generates matching file names for walk().'''
|
|
|
|
if not match.visitdir(self._dir[:-1] or '.'):
|
|
|
|
return
|
2015-04-08 01:18:52 +03:00
|
|
|
|
|
|
|
# yield this dir's files and walk its submanifests
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
2015-04-08 01:18:52 +03:00
|
|
|
for p in sorted(self._dirs.keys() + self._files.keys()):
|
|
|
|
if p in self._files:
|
|
|
|
fullp = self._subpath(p)
|
|
|
|
if match(fullp):
|
|
|
|
yield fullp
|
|
|
|
else:
|
2015-05-07 01:59:35 +03:00
|
|
|
for f in self._dirs[p]._walk(match):
|
2015-04-08 01:18:52 +03:00
|
|
|
yield f
|
|
|
|
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
def matches(self, match):
|
|
|
|
'''generate a new manifest filtered by the match argument'''
|
|
|
|
if match.always():
|
|
|
|
return self.copy()
|
|
|
|
|
2015-03-31 04:10:59 +03:00
|
|
|
return self._matches(match)
|
|
|
|
|
2015-05-07 01:59:35 +03:00
|
|
|
def _matches(self, match):
|
2015-03-31 04:10:59 +03:00
|
|
|
'''recursively generate a new manifest filtered by the match argument.
|
2015-05-07 01:59:35 +03:00
|
|
|
'''
|
|
|
|
|
treemanifest: don't iterate entire matching submanifests on match()
Before a4236180df5e (match: remove unnecessary optimization where
visitdir() returns 'all', 2015-05-06), match.visitdir() used to return
the special value 'all' to indicate that it was known that all
subdirectories would also be included in the match. The purpose for
that value was to avoid calling the matcher on all the paths. It
turned out that calling the matcher was not a problem, so the special
return value was removed and the code was simplified. However, if we
use the same special value for not just avoiding calling the matcher
on each file, but to avoid iterating over each file, it's a much
bigger win. On commands like
hg st --rev .^ --rev . dom/
we run the matcher (dom/) on the two manifests, then diff the narrowed
manifest. If the size of the match is much larger than the size of the
diff, this is wasteful. In the above case, we would end up iterating
over the 15k-or-so files in dom/ for each of the manifests, only to
later discover that they are mostly the same. This means that runningt
the command above is usually slower than getting the status for the
entire repo, because that code avoids calling treemanifest.match() and
only calls treemanifest.diff(), which loads only what's needed for the
diff.
Let's fix this by reintroducing the 'all' value in match.visitdir()
and making treemanifest.match() return a lazy copy of the manifest
from dom/ and down (in the above case). This speeds up the above
command on the Firefox repo from 0.357s to 0.137s (best of 5). The
wider the match, the bigger the speedup.
2015-12-12 20:57:05 +03:00
|
|
|
visit = match.visitdir(self._dir[:-1] or '.')
|
|
|
|
if visit == 'all':
|
|
|
|
return self.copy()
|
|
|
|
ret = treemanifest(self._dir)
|
|
|
|
if not visit:
|
2015-05-07 01:59:35 +03:00
|
|
|
return ret
|
2015-03-31 04:10:59 +03:00
|
|
|
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
2015-03-31 04:10:59 +03:00
|
|
|
for fn in self._files:
|
|
|
|
fullp = self._subpath(fn)
|
|
|
|
if not match(fullp):
|
|
|
|
continue
|
|
|
|
ret._files[fn] = self._files[fn]
|
|
|
|
if fn in self._flags:
|
|
|
|
ret._flags[fn] = self._flags[fn]
|
|
|
|
|
|
|
|
for dir, subm in self._dirs.iteritems():
|
2015-05-07 01:59:35 +03:00
|
|
|
m = subm._matches(match)
|
2015-03-31 04:10:59 +03:00
|
|
|
if not m._isempty():
|
|
|
|
ret._dirs[dir] = m
|
|
|
|
|
2015-02-26 19:16:13 +03:00
|
|
|
if not ret._isempty():
|
|
|
|
ret._dirty = True
|
2015-03-31 04:10:59 +03:00
|
|
|
return ret
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2017-03-07 20:56:11 +03:00
|
|
|
def diff(self, m2, match=None, clean=False):
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
'''Finds changes between the current manifest and m2.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
m2: the manifest to which this manifest should be compared.
|
|
|
|
clean: if true, include files unchanged between these manifests
|
|
|
|
with a None value in the returned dictionary.
|
|
|
|
|
|
|
|
The result is returned as a dict with filename as key and
|
|
|
|
values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
|
|
|
|
nodeid in the current/other manifest and fl1/fl2 is the flag
|
|
|
|
in the current/other manifest. Where the file does not exist,
|
|
|
|
the nodeid will be None and the flags will be the empty
|
|
|
|
string.
|
|
|
|
'''
|
2017-03-07 20:56:11 +03:00
|
|
|
if match:
|
|
|
|
m1 = self.matches(match)
|
|
|
|
m2 = m2.matches(match)
|
|
|
|
return m1.diff(m2, clean=clean)
|
2015-02-20 04:13:35 +03:00
|
|
|
result = {}
|
|
|
|
emptytree = treemanifest()
|
|
|
|
def _diff(t1, t2):
|
2015-02-26 19:16:13 +03:00
|
|
|
if t1._node == t2._node and not t1._dirty and not t2._dirty:
|
|
|
|
return
|
2015-04-10 03:14:35 +03:00
|
|
|
t1._load()
|
|
|
|
t2._load()
|
2015-02-20 04:13:35 +03:00
|
|
|
for d, m1 in t1._dirs.iteritems():
|
|
|
|
m2 = t2._dirs.get(d, emptytree)
|
|
|
|
_diff(m1, m2)
|
|
|
|
|
|
|
|
for d, m2 in t2._dirs.iteritems():
|
|
|
|
if d not in t1._dirs:
|
|
|
|
_diff(emptytree, m2)
|
|
|
|
|
|
|
|
for fn, n1 in t1._files.iteritems():
|
|
|
|
fl1 = t1._flags.get(fn, '')
|
|
|
|
n2 = t2._files.get(fn, None)
|
|
|
|
fl2 = t2._flags.get(fn, '')
|
|
|
|
if n1 != n2 or fl1 != fl2:
|
|
|
|
result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2))
|
|
|
|
elif clean:
|
|
|
|
result[t1._subpath(fn)] = None
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2015-02-20 04:13:35 +03:00
|
|
|
for fn, n2 in t2._files.iteritems():
|
|
|
|
if fn not in t1._files:
|
|
|
|
fl2 = t2._flags.get(fn, '')
|
|
|
|
result[t2._subpath(fn)] = ((None, ''), (n2, fl2))
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2015-02-20 04:13:35 +03:00
|
|
|
_diff(self, m2)
|
|
|
|
return result
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2015-05-19 07:31:40 +03:00
|
|
|
def unmodifiedsince(self, m2):
|
|
|
|
return not self._dirty and not m2._dirty and self._node == m2._node
|
|
|
|
|
2015-04-14 09:21:02 +03:00
|
|
|
def parse(self, text, readsubtree):
|
2015-04-13 09:01:18 +03:00
|
|
|
for f, n, fl in _parse(text):
|
2015-12-05 01:24:45 +03:00
|
|
|
if fl == 't':
|
2015-04-14 09:21:02 +03:00
|
|
|
f = f + '/'
|
|
|
|
self._dirs[f] = readsubtree(self._subpath(f), n)
|
2015-02-26 19:16:13 +03:00
|
|
|
elif '/' in f:
|
|
|
|
# This is a flat manifest, so use __setitem__ and setflag rather
|
|
|
|
# than assigning directly to _files and _flags, so we can
|
|
|
|
# assign a path in a subdirectory, and to mark dirty (compared
|
|
|
|
# to nullid).
|
2015-04-14 09:21:02 +03:00
|
|
|
self[f] = n
|
|
|
|
if fl:
|
|
|
|
self.setflag(f, fl)
|
2015-02-26 19:16:13 +03:00
|
|
|
else:
|
|
|
|
# Assigning to _files and _flags avoids marking as dirty,
|
|
|
|
# and should be a little faster.
|
|
|
|
self._files[f] = n
|
|
|
|
if fl:
|
|
|
|
self._flags[f] = fl
|
2015-04-13 09:01:18 +03:00
|
|
|
|
2015-04-01 00:01:33 +03:00
|
|
|
def text(self, usemanifestv2=False):
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
"""Get the full data of this manifest as a bytestring."""
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
2016-02-21 10:57:21 +03:00
|
|
|
return _text(self.iterentries(), usemanifestv2)
|
treemanifest: create treemanifest class
There are a number of problems with large and flat manifests. Copying
from http://mercurial.selenic.com/wiki/ManifestShardingPlan:
* manifest too large for RAM
* manifest resolution too much CPU (long delta chains)
* committing is slow because entire manifest has to be hashed
* impossible for narrow clone to leave out part of manifest as all is
needed to calculate new hash
* diffing two revisions involves traversing entire subdirectories
even if identical
This is a first step in a series introducing a manifest revlog per
directory.
This change adds a new manifest class: treemanifest, which is a tree
where each node has a dict of files (nodeids), a dict of flags, and a
dict of subdirectories (treemanifests). So far, it behaves just like
manifestdict, but it will later help us write one manifest revlog per
directory. The new class is still unused; it will be used after the
next change.
The code is not yet optimized. Running with it (see below) makes most
or all operations slower. Once we start storing manifest revlogs for
every directory, it should be possible to make many of these
operations much faster. The fastdelta() optimization has been
intentionally not implemented for the treemanifests. We can implement
it later if necessary.
All tests pass when run with the following patch (and without, of
couse):
--- a/mercurial/manifest.py Thu Mar 19 11:08:42 2015 -0700
+++ b/mercurial/manifest.py Thu Mar 19 11:15:50 2015 -0700
@@ -596,7 +596,7 @@ class manifest(revlog.revlog):
return None, None
def add(self, m, transaction, link, p1, p2, added, removed):
- if p1 in self._mancache:
+ if False and p1 in self._mancache:
# If our first parent is in the manifest cache, we can
# compute a delta here using properties we know about the
# manifest up-front, which may save time later for the
@@ -626,3 +626,5 @@ class manifest(revlog.revlog):
self._mancache[n] = (m, arraytext)
return n
+
+manifestdict = treemanifest
2015-03-19 21:08:42 +03:00
|
|
|
|
2015-04-14 09:21:02 +03:00
|
|
|
def dirtext(self, usemanifestv2=False):
|
|
|
|
"""Get the full data of this directory as a bytestring. Make sure that
|
|
|
|
any submanifests have been written first, so their nodeids are correct.
|
|
|
|
"""
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load()
|
2015-04-14 09:21:02 +03:00
|
|
|
flags = self.flags
|
2015-12-05 01:24:45 +03:00
|
|
|
dirs = [(d[:-1], self._dirs[d]._node, 't') for d in self._dirs]
|
2015-04-14 09:21:02 +03:00
|
|
|
files = [(f, self._files[f], flags(f)) for f in self._files]
|
|
|
|
return _text(sorted(dirs + files), usemanifestv2)
|
|
|
|
|
2015-04-10 03:14:35 +03:00
|
|
|
def read(self, gettext, readsubtree):
|
2015-09-26 05:54:46 +03:00
|
|
|
def _load_for_read(s):
|
|
|
|
s.parse(gettext(), readsubtree)
|
|
|
|
s._dirty = False
|
|
|
|
self._loadfunc = _load_for_read
|
2015-04-10 03:14:35 +03:00
|
|
|
|
2015-04-14 09:21:02 +03:00
|
|
|
def writesubtrees(self, m1, m2, writesubtree):
|
2015-04-10 03:14:35 +03:00
|
|
|
self._load() # for consistency; should never have any effect here
|
2016-08-30 03:48:14 +03:00
|
|
|
m1._load()
|
|
|
|
m2._load()
|
2015-04-14 09:21:02 +03:00
|
|
|
emptytree = treemanifest()
|
|
|
|
for d, subm in self._dirs.iteritems():
|
|
|
|
subp1 = m1._dirs.get(d, emptytree)._node
|
|
|
|
subp2 = m2._dirs.get(d, emptytree)._node
|
|
|
|
if subp1 == revlog.nullid:
|
|
|
|
subp1, subp2 = subp2, subp1
|
|
|
|
writesubtree(subm, subp1, subp2)
|
|
|
|
|
2017-04-10 23:07:47 +03:00
|
|
|
def walksubtrees(self, matcher=None):
|
|
|
|
"""Returns an iterator of the subtrees of this manifest, including this
|
|
|
|
manifest itself.
|
|
|
|
|
|
|
|
If `matcher` is provided, it only returns subtrees that match.
|
|
|
|
"""
|
|
|
|
if matcher and not matcher.visitdir(self._dir[:-1] or '.'):
|
|
|
|
return
|
|
|
|
if not matcher or matcher(self._dir[:-1]):
|
|
|
|
yield self
|
|
|
|
|
|
|
|
self._load()
|
|
|
|
for d, subm in self._dirs.iteritems():
|
|
|
|
for subtree in subm.walksubtrees(matcher=matcher):
|
|
|
|
yield subtree
|
|
|
|
|
2016-08-17 23:25:13 +03:00
|
|
|
class manifestrevlog(revlog.revlog):
|
|
|
|
'''A revlog that stores manifest texts. This is responsible for caching the
|
|
|
|
full-text manifest contents.
|
|
|
|
'''
|
2017-03-02 03:35:57 +03:00
|
|
|
def __init__(self, opener, dir='', dirlogcache=None, indexfile=None):
|
|
|
|
"""Constructs a new manifest revlog
|
|
|
|
|
|
|
|
`indexfile` - used by extensions to have two manifests at once, like
|
|
|
|
when transitioning between flatmanifeset and treemanifests.
|
|
|
|
"""
|
2016-08-17 23:25:13 +03:00
|
|
|
# During normal operations, we expect to deal with not more than four
|
|
|
|
# revs at a time (such as during commit --amend). When rebasing large
|
|
|
|
# stacks of commits, the number can go up, hence the config knob below.
|
|
|
|
cachesize = 4
|
2016-09-14 02:00:41 +03:00
|
|
|
usetreemanifest = False
|
|
|
|
usemanifestv2 = False
|
2016-08-17 23:25:13 +03:00
|
|
|
opts = getattr(opener, 'options', None)
|
|
|
|
if opts is not None:
|
|
|
|
cachesize = opts.get('manifestcachesize', cachesize)
|
2016-09-14 02:00:41 +03:00
|
|
|
usetreemanifest = opts.get('treemanifest', usetreemanifest)
|
|
|
|
usemanifestv2 = opts.get('manifestv2', usemanifestv2)
|
|
|
|
|
|
|
|
self._treeondisk = usetreemanifest
|
|
|
|
self._usemanifestv2 = usemanifestv2
|
|
|
|
|
2016-08-17 23:25:13 +03:00
|
|
|
self._fulltextcache = util.lrucachedict(cachesize)
|
|
|
|
|
2016-09-14 02:00:41 +03:00
|
|
|
if dir:
|
|
|
|
assert self._treeondisk, 'opts is %r' % opts
|
|
|
|
if not dir.endswith('/'):
|
|
|
|
dir = dir + '/'
|
2017-03-02 03:35:57 +03:00
|
|
|
|
|
|
|
if indexfile is None:
|
|
|
|
indexfile = '00manifest.i'
|
|
|
|
if dir:
|
|
|
|
indexfile = "meta/" + dir + indexfile
|
|
|
|
|
2016-09-14 02:00:41 +03:00
|
|
|
self._dir = dir
|
2016-09-14 02:00:41 +03:00
|
|
|
# The dirlogcache is kept on the root manifest log
|
|
|
|
if dir:
|
|
|
|
self._dirlogcache = dirlogcache
|
|
|
|
else:
|
|
|
|
self._dirlogcache = {'': self}
|
2016-09-14 02:00:41 +03:00
|
|
|
|
2016-09-22 15:51:58 +03:00
|
|
|
super(manifestrevlog, self).__init__(opener, indexfile,
|
|
|
|
checkambig=bool(dir))
|
2016-09-14 02:00:41 +03:00
|
|
|
|
2016-08-17 23:25:13 +03:00
|
|
|
@property
|
|
|
|
def fulltextcache(self):
|
|
|
|
return self._fulltextcache
|
|
|
|
|
|
|
|
def clearcaches(self):
|
|
|
|
super(manifestrevlog, self).clearcaches()
|
|
|
|
self._fulltextcache.clear()
|
2016-09-14 02:00:41 +03:00
|
|
|
self._dirlogcache = {'': self}
|
|
|
|
|
|
|
|
def dirlog(self, dir):
|
|
|
|
if dir:
|
|
|
|
assert self._treeondisk
|
|
|
|
if dir not in self._dirlogcache:
|
|
|
|
self._dirlogcache[dir] = manifestrevlog(self.opener, dir,
|
|
|
|
self._dirlogcache)
|
|
|
|
return self._dirlogcache[dir]
|
2016-08-17 23:25:13 +03:00
|
|
|
|
2016-11-10 13:13:19 +03:00
|
|
|
def add(self, m, transaction, link, p1, p2, added, removed, readtree=None):
|
2016-09-20 22:24:01 +03:00
|
|
|
if (p1 in self.fulltextcache and util.safehasattr(m, 'fastdelta')
|
|
|
|
and not self._usemanifestv2):
|
|
|
|
# If our first parent is in the manifest cache, we can
|
|
|
|
# compute a delta here using properties we know about the
|
|
|
|
# manifest up-front, which may save time later for the
|
|
|
|
# revlog layer.
|
|
|
|
|
|
|
|
_checkforbidden(added)
|
|
|
|
# combine the changed lists into one sorted iterator
|
|
|
|
work = heapq.merge([(x, False) for x in added],
|
|
|
|
[(x, True) for x in removed])
|
|
|
|
|
|
|
|
arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work)
|
|
|
|
cachedelta = self.rev(p1), deltatext
|
|
|
|
text = util.buffer(arraytext)
|
|
|
|
n = self.addrevision(text, transaction, link, p1, p2, cachedelta)
|
|
|
|
else:
|
|
|
|
# The first parent manifest isn't already loaded, so we'll
|
|
|
|
# just encode a fulltext of the manifest and pass that
|
|
|
|
# through to the revlog layer, and let it handle the delta
|
|
|
|
# process.
|
|
|
|
if self._treeondisk:
|
2016-11-10 13:13:19 +03:00
|
|
|
assert readtree, "readtree must be set for treemanifest writes"
|
|
|
|
m1 = readtree(self._dir, p1)
|
|
|
|
m2 = readtree(self._dir, p2)
|
|
|
|
n = self._addtree(m, transaction, link, m1, m2, readtree)
|
2016-09-20 22:24:01 +03:00
|
|
|
arraytext = None
|
|
|
|
else:
|
|
|
|
text = m.text(self._usemanifestv2)
|
|
|
|
n = self.addrevision(text, transaction, link, p1, p2)
|
2017-03-12 10:32:21 +03:00
|
|
|
arraytext = bytearray(text)
|
2016-09-20 22:24:01 +03:00
|
|
|
|
2016-10-18 08:51:22 +03:00
|
|
|
if arraytext is not None:
|
|
|
|
self.fulltextcache[n] = arraytext
|
2016-09-20 22:24:01 +03:00
|
|
|
|
|
|
|
return n
|
|
|
|
|
2016-11-10 13:13:19 +03:00
|
|
|
def _addtree(self, m, transaction, link, m1, m2, readtree):
|
2016-09-20 22:24:01 +03:00
|
|
|
# If the manifest is unchanged compared to one parent,
|
|
|
|
# don't write a new revision
|
treemanifest: make node reuse match flat manifest behavior
In a flat manifest, a node with the same content but different parents is still
considered a new node. In the current tree manifests however, if the content is
the same, we ignore the parents entirely and just reuse the existing node.
In our external treemanifest extension, we want to allow having one treemanifest
for every flat manifests, as a way of easeing the migration to treemanifests. To
make this possible, let's change the root node treemanifest behavior to match
the behavior for flat manifests, so we can have a 1:1 relationship.
While this sounds like a BC breakage, it's not actually a state users can
normally get in because: A) you can't make empty commits, and B) even if you try
to make an empty commit (by making a commit then amending it's changes away),
the higher level commit logic in localrepo.commitctx() forces the commit to use
the original p1 manifest node if no files were changed. So this would only
affect extensions and automation that reached passed the normal
localrepo.commit() logic straight into the manifest logic.
2017-03-02 03:19:41 +03:00
|
|
|
if self._dir != '' and (m.unmodifiedsince(m1) or m.unmodifiedsince(m2)):
|
2016-09-20 22:24:01 +03:00
|
|
|
return m.node()
|
|
|
|
def writesubtree(subm, subp1, subp2):
|
|
|
|
sublog = self.dirlog(subm.dir())
|
2016-11-10 13:13:19 +03:00
|
|
|
sublog.add(subm, transaction, link, subp1, subp2, None, None,
|
|
|
|
readtree=readtree)
|
2016-09-20 22:24:01 +03:00
|
|
|
m.writesubtrees(m1, m2, writesubtree)
|
|
|
|
text = m.dirtext(self._usemanifestv2)
|
treemanifest: make node reuse match flat manifest behavior
In a flat manifest, a node with the same content but different parents is still
considered a new node. In the current tree manifests however, if the content is
the same, we ignore the parents entirely and just reuse the existing node.
In our external treemanifest extension, we want to allow having one treemanifest
for every flat manifests, as a way of easeing the migration to treemanifests. To
make this possible, let's change the root node treemanifest behavior to match
the behavior for flat manifests, so we can have a 1:1 relationship.
While this sounds like a BC breakage, it's not actually a state users can
normally get in because: A) you can't make empty commits, and B) even if you try
to make an empty commit (by making a commit then amending it's changes away),
the higher level commit logic in localrepo.commitctx() forces the commit to use
the original p1 manifest node if no files were changed. So this would only
affect extensions and automation that reached passed the normal
localrepo.commit() logic straight into the manifest logic.
2017-03-02 03:19:41 +03:00
|
|
|
n = None
|
|
|
|
if self._dir != '':
|
|
|
|
# Double-check whether contents are unchanged to one parent
|
|
|
|
if text == m1.dirtext(self._usemanifestv2):
|
|
|
|
n = m1.node()
|
|
|
|
elif text == m2.dirtext(self._usemanifestv2):
|
|
|
|
n = m2.node()
|
|
|
|
|
|
|
|
if not n:
|
2016-09-20 22:24:01 +03:00
|
|
|
n = self.addrevision(text, transaction, link, m1.node(), m2.node())
|
treemanifest: make node reuse match flat manifest behavior
In a flat manifest, a node with the same content but different parents is still
considered a new node. In the current tree manifests however, if the content is
the same, we ignore the parents entirely and just reuse the existing node.
In our external treemanifest extension, we want to allow having one treemanifest
for every flat manifests, as a way of easeing the migration to treemanifests. To
make this possible, let's change the root node treemanifest behavior to match
the behavior for flat manifests, so we can have a 1:1 relationship.
While this sounds like a BC breakage, it's not actually a state users can
normally get in because: A) you can't make empty commits, and B) even if you try
to make an empty commit (by making a commit then amending it's changes away),
the higher level commit logic in localrepo.commitctx() forces the commit to use
the original p1 manifest node if no files were changed. So this would only
affect extensions and automation that reached passed the normal
localrepo.commit() logic straight into the manifest logic.
2017-03-02 03:19:41 +03:00
|
|
|
|
2016-09-20 22:24:01 +03:00
|
|
|
# Save nodeid so parent manifest can calculate its nodeid
|
|
|
|
m.setnode(n)
|
|
|
|
return n
|
|
|
|
|
2016-08-17 23:25:13 +03:00
|
|
|
class manifestlog(object):
|
|
|
|
"""A collection class representing the collection of manifest snapshots
|
|
|
|
referenced by commits in the repository.
|
|
|
|
|
|
|
|
In this situation, 'manifest' refers to the abstract concept of a snapshot
|
|
|
|
of the list of files in the given commit. Consumers of the output of this
|
|
|
|
class do not care about the implementation details of the actual manifests
|
|
|
|
they receive (i.e. tree or flat or lazily loaded, etc)."""
|
2016-08-17 23:25:13 +03:00
|
|
|
def __init__(self, opener, repo):
|
2016-09-20 22:24:01 +03:00
|
|
|
usetreemanifest = False
|
2016-11-10 13:13:19 +03:00
|
|
|
cachesize = 4
|
2016-09-20 22:24:01 +03:00
|
|
|
|
|
|
|
opts = getattr(opener, 'options', None)
|
|
|
|
if opts is not None:
|
|
|
|
usetreemanifest = opts.get('treemanifest', usetreemanifest)
|
2016-11-10 13:13:19 +03:00
|
|
|
cachesize = opts.get('manifestcachesize', cachesize)
|
2016-09-20 22:24:01 +03:00
|
|
|
self._treeinmem = usetreemanifest
|
|
|
|
|
2016-10-19 03:33:39 +03:00
|
|
|
self._oldmanifest = repo._constructmanifest()
|
|
|
|
self._revlog = self._oldmanifest
|
|
|
|
|
2016-11-03 03:10:47 +03:00
|
|
|
# A cache of the manifestctx or treemanifestctx for each directory
|
|
|
|
self._dirmancache = {}
|
2016-11-10 13:13:19 +03:00
|
|
|
self._dirmancache[''] = util.lrucachedict(cachesize)
|
2016-11-03 03:10:47 +03:00
|
|
|
|
2016-11-10 13:13:19 +03:00
|
|
|
self.cachesize = cachesize
|
2016-08-17 23:25:13 +03:00
|
|
|
|
2016-08-17 23:25:13 +03:00
|
|
|
def __getitem__(self, node):
|
2016-11-03 03:33:31 +03:00
|
|
|
"""Retrieves the manifest instance for the given node. Throws a
|
|
|
|
LookupError if not found.
|
2016-08-17 23:25:13 +03:00
|
|
|
"""
|
2016-11-03 03:24:06 +03:00
|
|
|
return self.get('', node)
|
|
|
|
|
2016-11-15 02:17:27 +03:00
|
|
|
def get(self, dir, node, verify=True):
|
2016-11-03 03:24:06 +03:00
|
|
|
"""Retrieves the manifest instance for the given node. Throws a
|
|
|
|
LookupError if not found.
|
2016-11-15 02:17:27 +03:00
|
|
|
|
|
|
|
`verify` - if True an exception will be thrown if the node is not in
|
|
|
|
the revlog
|
2016-11-03 03:24:06 +03:00
|
|
|
"""
|
2016-11-03 03:10:47 +03:00
|
|
|
if node in self._dirmancache.get(dir, ()):
|
|
|
|
cachemf = self._dirmancache[dir][node]
|
|
|
|
# The old manifest may put non-ctx manifests in the cache, so
|
|
|
|
# skip those since they don't implement the full api.
|
|
|
|
if (isinstance(cachemf, manifestctx) or
|
|
|
|
isinstance(cachemf, treemanifestctx)):
|
|
|
|
return cachemf
|
|
|
|
|
2016-11-03 03:24:06 +03:00
|
|
|
if dir:
|
|
|
|
if self._revlog._treeondisk:
|
2016-11-15 02:17:27 +03:00
|
|
|
if verify:
|
|
|
|
dirlog = self._revlog.dirlog(dir)
|
|
|
|
if node not in dirlog.nodemap:
|
|
|
|
raise LookupError(node, dirlog.indexfile,
|
|
|
|
_('no node'))
|
2017-03-02 03:39:48 +03:00
|
|
|
m = treemanifestctx(self, dir, node)
|
2016-11-03 03:24:06 +03:00
|
|
|
else:
|
|
|
|
raise error.Abort(
|
|
|
|
_("cannot ask for manifest directory '%s' in a flat "
|
|
|
|
"manifest") % dir)
|
2016-08-31 23:29:49 +03:00
|
|
|
else:
|
2016-11-15 02:17:27 +03:00
|
|
|
if verify:
|
|
|
|
if node not in self._revlog.nodemap:
|
|
|
|
raise LookupError(node, self._revlog.indexfile,
|
|
|
|
_('no node'))
|
2016-11-03 03:24:06 +03:00
|
|
|
if self._treeinmem:
|
2017-03-02 03:39:48 +03:00
|
|
|
m = treemanifestctx(self, '', node)
|
2016-11-03 03:24:06 +03:00
|
|
|
else:
|
2017-03-02 03:39:48 +03:00
|
|
|
m = manifestctx(self, node)
|
2016-11-03 03:10:47 +03:00
|
|
|
|
|
|
|
if node != revlog.nullid:
|
|
|
|
mancache = self._dirmancache.get(dir)
|
|
|
|
if not mancache:
|
|
|
|
mancache = util.lrucachedict(self.cachesize)
|
|
|
|
self._dirmancache[dir] = mancache
|
|
|
|
mancache[node] = m
|
2016-08-17 23:25:13 +03:00
|
|
|
return m
|
|
|
|
|
2016-11-10 13:13:19 +03:00
|
|
|
def clearcaches(self):
|
|
|
|
self._dirmancache.clear()
|
|
|
|
self._revlog.clearcaches()
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
class memmanifestctx(object):
|
2017-03-02 03:39:48 +03:00
|
|
|
def __init__(self, manifestlog):
|
|
|
|
self._manifestlog = manifestlog
|
2016-11-08 19:03:43 +03:00
|
|
|
self._manifestdict = manifestdict()
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def _revlog(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
return self._manifestlog._revlog
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def new(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
return memmanifestctx(self._manifestlog)
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def copy(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
memmf = memmanifestctx(self._manifestlog)
|
2016-11-08 19:03:43 +03:00
|
|
|
memmf._manifestdict = self.read().copy()
|
|
|
|
return memmf
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def read(self):
|
|
|
|
return self._manifestdict
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def write(self, transaction, link, p1, p2, added, removed):
|
|
|
|
return self._revlog().add(self._manifestdict, transaction, link, p1, p2,
|
|
|
|
added, removed)
|
|
|
|
|
2016-09-12 20:55:43 +03:00
|
|
|
class manifestctx(object):
|
2016-08-17 23:25:13 +03:00
|
|
|
"""A class representing a single revision of a manifest, including its
|
|
|
|
contents, its parent revs, and its linkrev.
|
|
|
|
"""
|
2017-03-02 03:39:48 +03:00
|
|
|
def __init__(self, manifestlog, node):
|
|
|
|
self._manifestlog = manifestlog
|
2016-09-12 20:55:43 +03:00
|
|
|
self._data = None
|
2016-08-17 23:25:13 +03:00
|
|
|
|
|
|
|
self._node = node
|
2016-08-31 23:29:49 +03:00
|
|
|
|
|
|
|
# TODO: We eventually want p1, p2, and linkrev exposed on this class,
|
|
|
|
# but let's add it later when something needs it and we can load it
|
|
|
|
# lazily.
|
|
|
|
#self.p1, self.p2 = revlog.parents(node)
|
|
|
|
#rev = revlog.rev(node)
|
|
|
|
#self.linkrev = revlog.linkrev(rev)
|
2016-08-17 23:25:13 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def _revlog(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
return self._manifestlog._revlog
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-08-17 23:25:13 +03:00
|
|
|
def node(self):
|
|
|
|
return self._node
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def new(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
return memmanifestctx(self._manifestlog)
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def copy(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
memmf = memmanifestctx(self._manifestlog)
|
2016-11-08 19:03:43 +03:00
|
|
|
memmf._manifestdict = self.read().copy()
|
|
|
|
return memmf
|
|
|
|
|
2016-11-17 21:59:15 +03:00
|
|
|
@propertycache
|
|
|
|
def parents(self):
|
|
|
|
return self._revlog().parents(self._node)
|
|
|
|
|
2016-09-12 20:55:43 +03:00
|
|
|
def read(self):
|
2017-02-26 21:16:47 +03:00
|
|
|
if self._data is None:
|
2016-09-12 20:55:43 +03:00
|
|
|
if self._node == revlog.nullid:
|
|
|
|
self._data = manifestdict()
|
|
|
|
else:
|
2016-11-08 19:03:43 +03:00
|
|
|
rl = self._revlog()
|
2016-10-19 03:44:26 +03:00
|
|
|
text = rl.revision(self._node)
|
2017-03-12 10:32:21 +03:00
|
|
|
arraytext = bytearray(text)
|
2016-10-19 03:44:26 +03:00
|
|
|
rl._fulltextcache[self._node] = arraytext
|
2016-09-12 20:55:43 +03:00
|
|
|
self._data = manifestdict(text)
|
|
|
|
return self._data
|
|
|
|
|
2016-11-03 03:10:47 +03:00
|
|
|
def readfast(self, shallow=False):
|
2016-11-03 03:10:47 +03:00
|
|
|
'''Calls either readdelta or read, based on which would be less work.
|
|
|
|
readdelta is called if the delta is against the p1, and therefore can be
|
|
|
|
read quickly.
|
|
|
|
|
|
|
|
If `shallow` is True, nothing changes since this is a flat manifest.
|
|
|
|
'''
|
2016-11-08 19:03:43 +03:00
|
|
|
rl = self._revlog()
|
2016-09-14 02:26:30 +03:00
|
|
|
r = rl.rev(self._node)
|
|
|
|
deltaparent = rl.deltaparent(r)
|
|
|
|
if deltaparent != revlog.nullrev and deltaparent in rl.parentrevs(r):
|
|
|
|
return self.readdelta()
|
|
|
|
return self.read()
|
|
|
|
|
2016-11-03 03:10:47 +03:00
|
|
|
def readdelta(self, shallow=False):
|
2016-11-03 03:10:47 +03:00
|
|
|
'''Returns a manifest containing just the entries that are present
|
|
|
|
in this manifest, but not in its p1 manifest. This is efficient to read
|
|
|
|
if the revlog delta is already p1.
|
|
|
|
|
|
|
|
Changing the value of `shallow` has no effect on flat manifests.
|
|
|
|
'''
|
2016-11-08 19:03:43 +03:00
|
|
|
revlog = self._revlog()
|
2016-09-14 02:25:21 +03:00
|
|
|
if revlog._usemanifestv2:
|
|
|
|
# Need to perform a slow delta
|
|
|
|
r0 = revlog.deltaparent(revlog.rev(self._node))
|
2017-03-02 03:39:48 +03:00
|
|
|
m0 = self._manifestlog[revlog.node(r0)].read()
|
2016-09-14 02:25:21 +03:00
|
|
|
m1 = self.read()
|
|
|
|
md = manifestdict()
|
|
|
|
for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
|
|
|
|
if n1:
|
|
|
|
md[f] = n1
|
|
|
|
if fl1:
|
|
|
|
md.setflag(f, fl1)
|
|
|
|
return md
|
|
|
|
|
|
|
|
r = revlog.rev(self._node)
|
|
|
|
d = mdiff.patchtext(revlog.revdiff(revlog.deltaparent(r), r))
|
|
|
|
return manifestdict(d)
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def find(self, key):
|
|
|
|
return self.read().find(key)
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
class memtreemanifestctx(object):
|
2017-03-02 03:39:48 +03:00
|
|
|
def __init__(self, manifestlog, dir=''):
|
|
|
|
self._manifestlog = manifestlog
|
2016-11-08 19:03:43 +03:00
|
|
|
self._dir = dir
|
|
|
|
self._treemanifest = treemanifest()
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def _revlog(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
return self._manifestlog._revlog
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def new(self, dir=''):
|
2017-03-02 03:39:48 +03:00
|
|
|
return memtreemanifestctx(self._manifestlog, dir=dir)
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def copy(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
|
2016-11-08 19:03:43 +03:00
|
|
|
memmf._treemanifest = self._treemanifest.copy()
|
|
|
|
return memmf
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def read(self):
|
|
|
|
return self._treemanifest
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def write(self, transaction, link, p1, p2, added, removed):
|
2016-11-10 13:13:19 +03:00
|
|
|
def readtree(dir, node):
|
2017-03-02 03:39:48 +03:00
|
|
|
return self._manifestlog.get(dir, node).read()
|
2016-11-08 19:03:43 +03:00
|
|
|
return self._revlog().add(self._treemanifest, transaction, link, p1, p2,
|
2016-11-10 13:13:19 +03:00
|
|
|
added, removed, readtree=readtree)
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-09-12 20:55:43 +03:00
|
|
|
class treemanifestctx(object):
|
2017-03-02 03:39:48 +03:00
|
|
|
def __init__(self, manifestlog, dir, node):
|
|
|
|
self._manifestlog = manifestlog
|
2016-08-31 23:29:49 +03:00
|
|
|
self._dir = dir
|
2016-09-12 20:55:43 +03:00
|
|
|
self._data = None
|
2016-08-31 23:29:49 +03:00
|
|
|
|
|
|
|
self._node = node
|
|
|
|
|
|
|
|
# TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that
|
|
|
|
# we can instantiate treemanifestctx objects for directories we don't
|
|
|
|
# have on disk.
|
|
|
|
#self.p1, self.p2 = revlog.parents(node)
|
|
|
|
#rev = revlog.rev(node)
|
|
|
|
#self.linkrev = revlog.linkrev(rev)
|
|
|
|
|
2016-10-19 03:44:42 +03:00
|
|
|
def _revlog(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
return self._manifestlog._revlog.dirlog(self._dir)
|
2016-10-19 03:44:42 +03:00
|
|
|
|
2016-09-12 20:55:43 +03:00
|
|
|
def read(self):
|
2017-02-26 21:16:47 +03:00
|
|
|
if self._data is None:
|
2016-10-19 03:44:42 +03:00
|
|
|
rl = self._revlog()
|
2016-09-12 20:55:43 +03:00
|
|
|
if self._node == revlog.nullid:
|
|
|
|
self._data = treemanifest()
|
2016-10-19 03:44:42 +03:00
|
|
|
elif rl._treeondisk:
|
2016-09-12 20:55:43 +03:00
|
|
|
m = treemanifest(dir=self._dir)
|
|
|
|
def gettext():
|
2016-10-19 03:44:42 +03:00
|
|
|
return rl.revision(self._node)
|
2016-09-12 20:55:43 +03:00
|
|
|
def readsubtree(dir, subm):
|
2016-11-15 02:24:07 +03:00
|
|
|
# Set verify to False since we need to be able to create
|
|
|
|
# subtrees for trees that don't exist on disk.
|
2017-03-02 03:39:48 +03:00
|
|
|
return self._manifestlog.get(dir, subm, verify=False).read()
|
2016-09-12 20:55:43 +03:00
|
|
|
m.read(gettext, readsubtree)
|
|
|
|
m.setnode(self._node)
|
|
|
|
self._data = m
|
|
|
|
else:
|
2016-11-04 23:49:15 +03:00
|
|
|
text = rl.revision(self._node)
|
2017-03-12 10:32:21 +03:00
|
|
|
arraytext = bytearray(text)
|
2016-10-19 03:44:42 +03:00
|
|
|
rl.fulltextcache[self._node] = arraytext
|
2016-09-12 20:55:43 +03:00
|
|
|
self._data = treemanifest(dir=self._dir, text=text)
|
|
|
|
|
|
|
|
return self._data
|
2016-08-31 23:29:49 +03:00
|
|
|
|
|
|
|
def node(self):
|
|
|
|
return self._node
|
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def new(self, dir=''):
|
2017-03-02 03:39:48 +03:00
|
|
|
return memtreemanifestctx(self._manifestlog, dir=dir)
|
2016-11-08 19:03:43 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def copy(self):
|
2017-03-02 03:39:48 +03:00
|
|
|
memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
|
2016-11-08 19:03:43 +03:00
|
|
|
memmf._treemanifest = self.read().copy()
|
|
|
|
return memmf
|
|
|
|
|
2016-11-17 21:59:15 +03:00
|
|
|
@propertycache
|
|
|
|
def parents(self):
|
|
|
|
return self._revlog().parents(self._node)
|
|
|
|
|
2016-11-03 03:10:47 +03:00
|
|
|
def readdelta(self, shallow=False):
|
2016-11-03 03:10:47 +03:00
|
|
|
'''Returns a manifest containing just the entries that are present
|
|
|
|
in this manifest, but not in its p1 manifest. This is efficient to read
|
|
|
|
if the revlog delta is already p1.
|
|
|
|
|
|
|
|
If `shallow` is True, this will read the delta for this directory,
|
|
|
|
without recursively reading subdirectory manifests. Instead, any
|
|
|
|
subdirectory entry will be reported as it appears in the manifest, i.e.
|
|
|
|
the subdirectory will be reported among files and distinguished only by
|
|
|
|
its 't' flag.
|
|
|
|
'''
|
2016-10-19 03:44:42 +03:00
|
|
|
revlog = self._revlog()
|
2016-11-03 03:10:47 +03:00
|
|
|
if shallow and not revlog._usemanifestv2:
|
|
|
|
r = revlog.rev(self._node)
|
|
|
|
d = mdiff.patchtext(revlog.revdiff(revlog.deltaparent(r), r))
|
|
|
|
return manifestdict(d)
|
|
|
|
else:
|
|
|
|
# Need to perform a slow delta
|
|
|
|
r0 = revlog.deltaparent(revlog.rev(self._node))
|
2017-03-02 03:39:48 +03:00
|
|
|
m0 = self._manifestlog.get(self._dir, revlog.node(r0)).read()
|
2016-11-03 03:10:47 +03:00
|
|
|
m1 = self.read()
|
|
|
|
md = treemanifest(dir=self._dir)
|
|
|
|
for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
|
|
|
|
if n1:
|
|
|
|
md[f] = n1
|
|
|
|
if fl1:
|
|
|
|
md.setflag(f, fl1)
|
|
|
|
return md
|
|
|
|
|
|
|
|
def readfast(self, shallow=False):
|
2016-11-03 03:10:47 +03:00
|
|
|
'''Calls either readdelta or read, based on which would be less work.
|
|
|
|
readdelta is called if the delta is against the p1, and therefore can be
|
|
|
|
read quickly.
|
|
|
|
|
|
|
|
If `shallow` is True, it only returns the entries from this manifest,
|
|
|
|
and not any submanifests.
|
|
|
|
'''
|
2016-10-19 03:44:42 +03:00
|
|
|
rl = self._revlog()
|
2016-09-14 02:26:30 +03:00
|
|
|
r = rl.rev(self._node)
|
|
|
|
deltaparent = rl.deltaparent(r)
|
2016-11-03 03:10:47 +03:00
|
|
|
if (deltaparent != revlog.nullrev and
|
|
|
|
deltaparent in rl.parentrevs(r)):
|
|
|
|
return self.readdelta(shallow=shallow)
|
|
|
|
|
|
|
|
if shallow:
|
|
|
|
return manifestdict(rl.revision(self._node))
|
|
|
|
else:
|
|
|
|
return self.read()
|
2016-09-14 02:26:30 +03:00
|
|
|
|
2016-11-08 19:03:43 +03:00
|
|
|
def find(self, key):
|
|
|
|
return self.read().find(key)
|