# Copyright 2009-2010 Gregory P. Ward

# Copyright 2009-2010 Intelerad Medical Systems Incorporated

# Copyright 2010-2011 Fog Creek Software

# Copyright 2010-2011 Unity Technologies

#

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

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

'''largefiles utility code: must not import other modules in this package.'''

from __future__ import absolute_import

import copy

import hashlib

import os

import stat

from mercurial.i18n import _

from mercurial import (

    dirstate,

    encoding,

    error,

    httpconnection,

    match as matchmod,

    node,

    pycompat,

    scmutil,

    sparse,

    util,

    vfs as vfsmod,

)

shortname = '.hglf'

shortnameslash = shortname + '/'

longname = 'largefiles'

# -- Private worker functions ------------------------------------------

def getminsize(ui, assumelfiles, opt, default=10):

    lfsize = opt

    if not lfsize and assumelfiles:

        lfsize = ui.config(longname, 'minsize', default=default)

    if lfsize:

        try:

            lfsize = float(lfsize)

        except ValueError:

            raise error.Abort(_('largefiles: size must be number (not %s)\n')

                             % lfsize)

    if lfsize is None:

        raise error.Abort(_('minimum size for largefiles must be specified'))

    return lfsize

def link(src, dest):

    """Try to create hardlink - if that fails, efficiently make a copy."""

    util.makedirs(os.path.dirname(dest))

    try:

        util.oslink(src, dest)

    except OSError:

        # if hardlinks fail, fallback on atomic copy

        with open(src, 'rb') as srcf, util.atomictempfile(dest) as dstf:

            for chunk in util.filechunkiter(srcf):

                dstf.write(chunk)

        os.chmod(dest, os.stat(src).st_mode)

def usercachepath(ui, hash):

    '''Return the correct location in the "global" largefiles cache for a file

    with the given hash.

    This cache is used for sharing of largefiles across repositories - both

    to preserve download bandwidth and storage space.'''

    return os.path.join(_usercachedir(ui), hash)

def _usercachedir(ui):

    '''Return the location of the "global" largefiles cache.'''

    path = ui.configpath(longname, 'usercache', None)

    if path:

        return path

    if pycompat.iswindows:

        appdata = encoding.environ.get('LOCALAPPDATA',\

                        encoding.environ.get('APPDATA'))

        if appdata:

            return os.path.join(appdata, longname)

        home = encoding.environ.get('HOME')

        if home:

            return os.path.join(home, 'Library', 'Caches', longname)

    elif pycompat.isposix:

        path = encoding.environ.get('XDG_CACHE_HOME')

        if path:

            return os.path.join(path, longname)

        home = encoding.environ.get('HOME')

        if home:

            return os.path.join(home, '.cache', longname)

    else:

        raise error.Abort(_('unknown operating system: %s\n')

                          % pycompat.osname)

    raise error.Abort(_('unknown %s usercache location') % longname)

def inusercache(ui, hash):

    path = usercachepath(ui, hash)

    return os.path.exists(path)

def findfile(repo, hash):

    '''Return store path of the largefile with the specified hash.

    As a side effect, the file might be linked from user cache.

    Return None if the file can't be found locally.'''

    path, exists = findstorepath(repo, hash)

    if exists:

        repo.ui.note(_('found %s in store\n') % hash)

        return path

    elif inusercache(repo.ui, hash):

        repo.ui.note(_('found %s in system cache\n') % hash)

        path = storepath(repo, hash)

        link(usercachepath(repo.ui, hash), path)

        return path

    return None

class largefilesdirstate(dirstate.dirstate):

    def __getitem__(self, key):

        return super(largefilesdirstate, self).__getitem__(unixpath(key))

    def normal(self, f):

        return super(largefilesdirstate, self).normal(unixpath(f))

    def remove(self, f):

        return super(largefilesdirstate, self).remove(unixpath(f))

    def add(self, f):

        return super(largefilesdirstate, self).add(unixpath(f))

    def drop(self, f):

        return super(largefilesdirstate, self).drop(unixpath(f))

    def forget(self, f):

        return super(largefilesdirstate, self).forget(unixpath(f))

    def normallookup(self, f):

        return super(largefilesdirstate, self).normallookup(unixpath(f))

    def _ignore(self, f):

        return False

    def write(self, tr=False):

        # (1) disable PENDING mode always

        #     (lfdirstate isn't yet managed as a part of the transaction)

        # (2) avoid develwarn 'use dirstate.write with ....'

        super(largefilesdirstate, self).write(None)

def openlfdirstate(ui, repo, create=True):

    '''

    Return a dirstate object that tracks largefiles: i.e. its root is

    the repo root, but it is saved in .hg/largefiles/dirstate.

    '''

    vfs = repo.vfs

    lfstoredir = longname

    opener = vfsmod.vfs(vfs.join(lfstoredir))

    lfdirstate = largefilesdirstate(opener, ui, repo.root,

                                    repo.dirstate._validate,

                                    lambda: sparse.matcher(repo))

    # If the largefiles dirstate does not exist, populate and create

    # it. This ensures that we create it on the first meaningful

    # largefiles operation in a new clone.

    if create and not vfs.exists(vfs.join(lfstoredir, 'dirstate')):

        matcher = getstandinmatcher(repo)

        standins = repo.dirstate.walk(matcher, subrepos=[], unknown=False,

                                      ignored=False)

        if len(standins) > 0:

            vfs.makedirs(lfstoredir)

        for standin in standins:

            lfile = splitstandin(standin)

            lfdirstate.normallookup(lfile)

    return lfdirstate

def lfdirstatestatus(lfdirstate, repo):

    pctx = repo['.']

    match = matchmod.always(repo.root, repo.getcwd())

    unsure, s = lfdirstate.status(match, subrepos=[], ignored=False,

                                  clean=False, unknown=False)

    modified, clean = s.modified, s.clean

    for lfile in unsure:

        try:

            fctx = pctx[standin(lfile)]

        except LookupError:

            fctx = None

        if not fctx or readasstandin(fctx) != hashfile(repo.wjoin(lfile)):

            modified.append(lfile)

        else:

            clean.append(lfile)

            lfdirstate.normal(lfile)

    return s

def listlfiles(repo, rev=None, matcher=None):

    '''return a list of largefiles in the working copy or the

    specified changeset'''

    if matcher is None:

        matcher = getstandinmatcher(repo)

    # ignore unknown files in working directory

    return [splitstandin(f)

            for f in repo[rev].walk(matcher)

            if rev is not None or repo.dirstate[f] != '?']

def instore(repo, hash, forcelocal=False):

    '''Return true if a largefile with the given hash exists in the store'''

    return os.path.exists(storepath(repo, hash, forcelocal))

def storepath(repo, hash, forcelocal=False):

    '''Return the correct location in the repository largefiles store for a

    file with the given hash.'''

    if not forcelocal and repo.shared():

        return repo.vfs.reljoin(repo.sharedpath, longname, hash)

    return repo.vfs.join(longname, hash)

def findstorepath(repo, hash):

    '''Search through the local store path(s) to find the file for the given

    hash.  If the file is not found, its path in the primary store is returned.

    The return value is a tuple of (path, exists(path)).

    '''

    # For shared repos, the primary store is in the share source.  But for

    # backward compatibility, force a lookup in the local store if it wasn't

    # found in the share source.

    path = storepath(repo, hash, False)

    if instore(repo, hash):

        return (path, True)

    elif repo.shared() and instore(repo, hash, True):

        return storepath(repo, hash, True), True

    return (path, False)

def copyfromcache(repo, hash, filename):

    '''Copy the specified largefile from the repo or system cache to

    filename in the repository. Return true on success or false if the

    file was not found in either cache (which should not happened:

    this is meant to be called only after ensuring that the needed

    largefile exists in the cache).'''

    wvfs = repo.wvfs

    path = findfile(repo, hash)

    if path is None:

        return False

    wvfs.makedirs(wvfs.dirname(wvfs.join(filename)))

    # The write may fail before the file is fully written, but we

    # don't use atomic writes in the working copy.

    with open(path, 'rb') as srcfd, wvfs(filename, 'wb') as destfd:

        gothash = copyandhash(

            util.filechunkiter(srcfd), destfd)

    if gothash != hash:

        repo.ui.warn(_('%s: data corruption in %s with hash %s\n')

                     % (filename, path, gothash))

        wvfs.unlink(filename)

        return False

    return True

def copytostore(repo, ctx, file, fstandin):

    wvfs = repo.wvfs

    hash = readasstandin(ctx[fstandin])

    if instore(repo, hash):

        return

    if wvfs.exists(file):

        copytostoreabsolute(repo, wvfs.join(file), hash)

    else:

        repo.ui.warn(_("%s: largefile %s not available from local store\n") %

                     (file, hash))

def copyalltostore(repo, node):

    '''Copy all largefiles in a given revision to the store'''

    ctx = repo[node]

    for filename in ctx.files():

        realfile = splitstandin(filename)

        if realfile is not None and filename in ctx.manifest():

            copytostore(repo, ctx, realfile, filename)

def copytostoreabsolute(repo, file, hash):

    if inusercache(repo.ui, hash):

        link(usercachepath(repo.ui, hash), storepath(repo, hash))

    else:

        util.makedirs(os.path.dirname(storepath(repo, hash)))

        with open(file, 'rb') as srcf:

            with util.atomictempfile(storepath(repo, hash),

                                     createmode=repo.store.createmode) as dstf:

                for chunk in util.filechunkiter(srcf):

                    dstf.write(chunk)

        linktousercache(repo, hash)

def linktousercache(repo, hash):

    '''Link / copy the largefile with the specified hash from the store

    to the cache.'''

    path = usercachepath(repo.ui, hash)

    link(storepath(repo, hash), path)

def getstandinmatcher(repo, rmatcher=None):

    '''Return a match object that applies rmatcher to the standin directory'''

    wvfs = repo.wvfs

    standindir = shortname

    # no warnings about missing files or directories

    badfn = lambda f, msg: None

    if rmatcher and not rmatcher.always():

        pats = [wvfs.join(standindir, pat) for pat in rmatcher.files()]

        if not pats:

            pats = [wvfs.join(standindir)]

        match = scmutil.match(repo[None], pats, badfn=badfn)

    else:

        # no patterns: relative to repo root

        match = scmutil.match(repo[None], [wvfs.join(standindir)], badfn=badfn)

    return match

def composestandinmatcher(repo, rmatcher):

    '''Return a matcher that accepts standins corresponding to the

    files accepted by rmatcher. Pass the list of files in the matcher

    as the paths specified by the user.'''

    smatcher = getstandinmatcher(repo, rmatcher)

    isstandin = smatcher.matchfn

    def composedmatchfn(f):

        return isstandin(f) and rmatcher.matchfn(splitstandin(f))

    smatcher.matchfn = composedmatchfn

    return smatcher

def standin(filename):

    '''Return the repo-relative path to the standin for the specified big

    file.'''

    # Notes:

    # 1) Some callers want an absolute path, but for instance addlargefiles

    #    needs it repo-relative so it can be passed to repo[None].add().  So

    #    leave it up to the caller to use repo.wjoin() to get an absolute path.

    # 2) Join with '/' because that's what dirstate always uses, even on

    #    Windows. Change existing separator to '/' first in case we are

    #    passed filenames from an external source (like the command line).

    return shortnameslash + util.pconvert(filename)

def isstandin(filename):

    '''Return true if filename is a big file standin. filename must be

    in Mercurial's internal form (slash-separated).'''

    return filename.startswith(shortnameslash)

def splitstandin(filename):

    # Split on / because that's what dirstate always uses, even on Windows.

    # Change local separator to / first just in case we are passed filenames

    # from an external source (like the command line).

    bits = util.pconvert(filename).split('/', 1)

    if len(bits) == 2 and bits[0] == shortname:

        return bits[1]

    else:

        return None

def updatestandin(repo, lfile, standin):

    """Re-calculate hash value of lfile and write it into standin

    This assumes that "lfutil.standin(lfile) == standin", for efficiency.

    """

    file = repo.wjoin(lfile)

    if repo.wvfs.exists(lfile):

        hash = hashfile(file)

        executable = getexecutable(file)

        writestandin(repo, standin, hash, executable)

    else:

        raise error.Abort(_('%s: file not found!') % lfile)

def readasstandin(fctx):

    '''read hex hash from given filectx of standin file

    This encapsulates how "standin" data is stored into storage layer.'''

    return fctx.data().strip()

def writestandin(repo, standin, hash, executable):

    '''write hash to <repo.root>/<standin>'''

    repo.wwrite(standin, hash + '\n', executable and 'x' or '')

def copyandhash(instream, outfile):