Add support for cloning with hardlinks on windows. In order to use hardlinks, the win32file module is needed, and this is present in ActivePython. If it isn't present, or hardlinks are not supported on the underlying filesystem, a regular copy is used. When using hardlinks the biggest benefit is probably the saving in space, but cloning can be much quicker. For example cloning the Xen tree (non trivial) without an update goes from about 95s to 15s. Unix-like platforms should be unaffected, although should be more tolerant on filesystems that don't support hard links. (tweaked by mpm to deal with new copyfiles function) --- hg.orig/mercurial/commands.py 2005-09-13 19:32:53.000000000 -0500 +++ hg/mercurial/commands.py 2005-09-14 12:11:34.000000000 -0500 @@ -620,10 +620,6 @@ def clone(ui, source, dest=None, **opts) if other.dev() != -1: abspath = os.path.abspath(source) - copyfile = (os.stat(dest).st_dev == other.dev() - and getattr(os, 'link', None) or shutil.copy2) - if copyfile is not shutil.copy2: - ui.note("cloning by hardlink\n") # we use a lock here because if we race with commit, we can # end up with extra data in the cloned revlogs that's not @@ -638,7 +634,7 @@ def clone(ui, source, dest=None, **opts) for f in files.split(): src = os.path.join(source, ".hg", f) dst = os.path.join(dest, ".hg", f) - util.copyfiles(src, dst, copyfile) + util.copyfiles(src, dst) repo = hg.repository(ui, dest) Index: hg/mercurial/util.py =================================================================== --- hg.orig/mercurial/util.py 2005-09-08 00:15:25.000000000 -0500 +++ hg/mercurial/util.py 2005-09-14 12:16:49.000000000 -0500 @@ -12,7 +12,7 @@ platform-specific details from the core. import os, errno from demandload import * -demandload(globals(), "re cStringIO") +demandload(globals(), "re cStringIO shutil") def binary(s): """return true if a string is binary data using diff's heuristic""" @@ -217,17 +217,28 @@ def rename(src, dst): os.unlink(dst) os.rename(src, dst) -def copyfiles(src, dst, copyfile): - """Copy a directory tree, files are copied using 'copyfile'.""" +def copyfiles(src, dst, hardlink=None): + """Copy a directory tree using hardlinks if possible""" + + if hardlink is None: + hardlink = (os.stat(src).st_dev == + os.stat(os.path.dirname(dst)).st_dev) if os.path.isdir(src): os.mkdir(dst) for name in os.listdir(src): srcname = os.path.join(src, name) dstname = os.path.join(dst, name) - copyfiles(srcname, dstname, copyfile) + copyfiles(srcname, dstname, hardlink) else: - copyfile(src, dst) + if hardlink: + try: + os_link(src, dst) + except: + hardlink = False + shutil.copy2(src, dst) + else: + shutil.copy2(src, dst) def opener(base): """ @@ -244,13 +255,13 @@ def opener(base): if mode[0] != "r": try: - s = os.stat(f) + nlink = nlinks(f) except OSError: d = os.path.dirname(f) if not os.path.isdir(d): os.makedirs(d) else: - if s.st_nlink > 1: + if nlink > 1: file(f + ".tmp", "wb").write(file(f, "rb").read()) rename(f+".tmp", f) @@ -266,10 +277,41 @@ def _makelock_file(info, pathname): def _readlock_file(pathname): return file(pathname).read() +def nlinks(pathname): + """Return number of hardlinks for the given file.""" + return os.stat(pathname).st_nlink + +if hasattr(os, 'link'): + os_link = os.link +else: + def os_link(src, dst): + raise OSError(0, "Hardlinks not supported") + # Platform specific variants if os.name == 'nt': nulldev = 'NUL:' + try: # ActivePython can create hard links using win32file module + import win32file + + def os_link(src, dst): # NB will only succeed on NTFS + win32file.CreateHardLink(dst, src) + + def nlinks(pathname): + """Return number of hardlinks for the given file.""" + try: + fh = win32file.CreateFile(pathname, + win32file.GENERIC_READ, win32file.FILE_SHARE_READ, + None, win32file.OPEN_EXISTING, 0, None) + res = win32file.GetFileInformationByHandle(fh) + fh.Close() + return res[7] + except: + return os.stat(pathname).st_nlink + + except ImportError: + pass + def is_exec(f, last): return last

# filelog.py - file history class for mercurial
#
# Copyright 2005 Matt Mackall <mpm@selenic.com>
#
# This software may be used and distributed according to the terms
# of the GNU General Public License, incorporated herein by reference.

import os
from revlog import *
from demandload import *
demandload(globals(), "bdiff")

class filelog(revlog):
    def __init__(self, opener, path):
        revlog.__init__(self, opener,
                        os.path.join("data", self.encodedir(path + ".i")),
                        os.path.join("data", self.encodedir(path + ".d")))

    # This avoids a collision between a file named foo and a dir named
    # foo.i or foo.d
    def encodedir(self, path):
        return (path
                .replace(".hg/", ".hg.hg/")
                .replace(".i/", ".i.hg/")
                .replace(".d/", ".d.hg/"))

    def decodedir(self, path):
        return (path
                .replace(".d.hg/", ".d/")
                .replace(".i.hg/", ".i/")
                .replace(".hg.hg/", ".hg/"))

    def read(self, node):
        t = self.revision(node)
        if not t.startswith('\1\n'):
            return t
        s = t.find('\1\n', 2)
        return t[s+2:]

    def readmeta(self, node):
        t = self.revision(node)
        if not t.startswith('\1\n'):
            return {}
        s = t.find('\1\n', 2)
        mt = t[2:s]
        m = {}
        for l in mt.splitlines():
            k, v = l.split(": ", 1)
            m[k] = v
        return m

    def add(self, text, meta, transaction, link, p1=None, p2=None):
        if meta or text.startswith('\1\n'):
            mt = ""
            if meta:
                mt = [ "%s: %s\n" % (k, v) for k,v in meta.items() ]
            text = "\1\n" + "".join(mt) + "\1\n" + text
        return self.addrevision(text, transaction, link, p1, p2)

    def renamed(self, node):
        if 0 and self.parents(node)[0] != nullid:
            return False
        m = self.readmeta(node)
        if m and m.has_key("copy"):
            return (m["copy"], bin(m["copyrev"]))
        return False

    def annotate(self, node):

        def decorate(text, rev):
            return ([rev] * len(text.splitlines()), text)

        def pair(parent, child):
            for a1, a2, b1, b2 in bdiff.blocks(parent[1], child[1]):
                child[0][b1:b2] = parent[0][a1:a2]
            return child

        # find all ancestors
        needed = {node:1}
        visit = [node]
        while visit:
            n = visit.pop(0)
            for p in self.parents(n):
                if p not in needed:
                    needed[p] = 1
                    visit.append(p)
                else:
                    # count how many times we'll use this
                    needed[p] += 1

        # sort by revision which is a topological order
        visit = [ (self.rev(n), n) for n in needed.keys() ]
        visit.sort()
        hist = {}

        for r,n in visit:
            curr = decorate(self.read(n), self.linkrev(n))
            for p in self.parents(n):
                if p != nullid:
                    curr = pair(hist[p], curr)
                    # trim the history of unneeded revs
                    needed[p] -= 1
                    if not needed[p]:
                        del hist[p]
            hist[n] = curr

        return zip(hist[n][0], hist[n][1].splitlines(1))