# Copyright 2007 Bryan O'Sullivan <bos@serpentine.com>

# Copyright 2007 Alexis S. L. Carvalho <alexis@cecm.usp.br>

#

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

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

from __future__ import absolute_import, print_function

import posixpath

from mercurial.i18n import _

from mercurial import (

    error,

    pycompat,

)

from . import common

SKIPREV = common.SKIPREV

def rpairs(path):

    '''Yield tuples with path split at '/', starting with the full path.

    No leading, trailing or double '/', please.

    >>> for x in rpairs(b'foo/bar/baz'): print(x)

    ('foo/bar/baz', '')

    ('foo/bar', 'baz')

    ('foo', 'bar/baz')

    ('.', 'foo/bar/baz')

    '''

    i = len(path)

    while i != -1:

        yield path[:i], path[i + 1 :]

def normalize(path):

    ''' We use posixpath.normpath to support cross-platform path format.

    However, it doesn't handle None input. So we wrap it up. '''

    if path is None:

        return None

    return posixpath.normpath(path)

class filemapper(object):

    '''Map and filter filenames when importing.

    A name can be mapped to itself, a new name, or None (omit from new

    repository).'''

    def __init__(self, ui, path=None):

        self.ui = ui

        self.include = {}

        self.exclude = {}

        self.rename = {}

        self.targetprefixes = None

        if path:

            if self.parse(path):

    def parse(self, path):

        errs = 0

        def check(name, mapping, listname):

            if not name:

                self.ui.warn(

                    % (lex.infile, lex.lineno, listname)

                )

                return 1

            if name in mapping:

                self.ui.warn(

                    % (lex.infile, lex.lineno, name, listname)

                )

                return 1

                self.ui.warn(

                    % (lex.infile, lex.lineno, listname, pycompat.bytestr(name))

                )

                return 1

            return 0

        lex = common.shlexer(

        )

        cmd = lex.get_token()

        while cmd:

                name = normalize(lex.get_token())

                self.include[name] = name

                name = normalize(lex.get_token())

                self.exclude[name] = name

                src = normalize(lex.get_token())

                dest = normalize(lex.get_token())

                self.rename[src] = dest

                errs += self.parse(normalize(lex.get_token()))

            else:

                self.ui.warn(

                    % (lex.infile, lex.lineno, pycompat.bytestr(cmd))

                )

                errs += 1

            cmd = lex.get_token()

        return errs

    def lookup(self, name, mapping):

        name = normalize(name)

        for pre, suf in rpairs(name):

            try:

                return mapping[pre], pre, suf

            except KeyError:

                pass

    def istargetfile(self, filename):

        """Return true if the given target filename is covered as a destination

        of the filemap. This is useful for identifying what parts of the target

        repo belong to the source repo and what parts don't."""

        if self.targetprefixes is None:

            self.targetprefixes = set()

            for before, after in self.rename.iteritems():

                self.targetprefixes.add(after)

        # If "." is a target, then all target files are considered from the

        # source.

            return True

        filename = normalize(filename)

        for pre, suf in rpairs(filename):

            # This check is imperfect since it doesn't account for the

            # include/exclude list, but it should work in filemaps that don't

            # apply include/exclude to the same source directories they are

            # renaming.

            if pre in self.targetprefixes:

                return True

        return False

    def __call__(self, name):

        if self.include:

            inc = self.lookup(name, self.include)[0]

        else:

            inc = name

        if self.exclude:

            exc = self.lookup(name, self.exclude)[0]

        else:

        if (not self.include and exc) or (len(inc) <= len(exc)):

            return None

        newpre, pre, suf = self.lookup(name, self.rename)

        if newpre:

                return suf

            if suf:

                    return newpre + suf

            return newpre

        return name

    def active(self):

        return bool(self.include or self.exclude or self.rename)

# This class does two additional things compared to a regular source:

#

# - Filter and rename files.  This is mostly wrapped by the filemapper

#   class above. We hide the original filename in the revision that is

#   returned by getchanges to be able to find things later in getfile.

#

# - Return only revisions that matter for the files we're interested in.

#   This involves rewriting the parents of the original revision to

#   create a graph that is restricted to those revisions.

#

#   This set of revisions includes not only revisions that directly

#   touch files we're interested in, but also merges that merge two

#   or more interesting revisions.

class filemap_source(common.converter_source):

    def __init__(self, ui, baseconverter, filemap):

        super(filemap_source, self).__init__(ui, baseconverter.repotype)

        self.base = baseconverter

        self.filemapper = filemapper(ui, filemap)

        self.commits = {}

        # if a revision rev has parent p in the original revision graph, then

        # rev will have parent self.parentmap[p] in the restricted graph.

        self.parentmap = {}

        # self.wantedancestors[rev] is the set of all ancestors of rev that

        # are in the restricted graph.

        self.wantedancestors = {}

        self.convertedorder = None

        self._rebuilt = False

        self.origparents = {}

        self.children = {}

        self.seenchildren = {}

        # experimental config: convert.ignoreancestorcheck

        self.ignoreancestorcheck = self.ui.configbool(

        )

    def before(self):

        self.base.before()

    def after(self):

        self.base.after()

    def setrevmap(self, revmap):

        # rebuild our state to make things restartable

        #

        # To avoid calling getcommit for every revision that has already

        # been converted, we rebuild only the parentmap, delaying the

        # rebuild of wantedancestors until we need it (i.e. until a

        # merge).

        #

        # We assume the order argument lists the revisions in

        # topological order, so that we can infer which revisions were

        # wanted by previous runs.

        self._rebuilt = not revmap

        seen = {SKIPREV: SKIPREV}

        dummyset = set()

        converted = []

        for rev in revmap.order:

            mapped = revmap[rev]

            wanted = mapped not in seen

            if wanted:

                seen[mapped] = rev

                self.parentmap[rev] = rev

            else:

                self.parentmap[rev] = seen[mapped]

            self.wantedancestors[rev] = dummyset

            arg = seen[mapped]

            if arg == SKIPREV:

                arg = None

            converted.append((rev, wanted, arg))

        self.convertedorder = converted

        return self.base.setrevmap(revmap)

    def rebuild(self):

        if self._rebuilt:

            return True

        self._rebuilt = True

        self.parentmap.clear()

        self.wantedancestors.clear()

        self.seenchildren.clear()

        for rev, wanted, arg in self.convertedorder:

            if rev not in self.origparents:

                try:

                    self.origparents[rev] = self.getcommit(rev).parents

                except error.RepoLookupError:

                    continue

            if arg is not None:

                self.children[arg] = self.children.get(arg, 0) + 1

        for rev, wanted, arg in self.convertedorder:

            try:

                parents = self.origparents[rev]

            except KeyError:

                continue  # unknown revmap source

            if wanted:

                self.mark_wanted(rev, parents)

            else:

                self.mark_not_wanted(rev, arg)

            self._discard(arg, *parents)

        return True

    def getheads(self):

        return self.base.getheads()

    def getcommit(self, rev):

        # We want to save a reference to the commit objects to be able

        # to rewrite their parents later on.

        c = self.commits[rev] = self.base.getcommit(rev)

        for p in c.parents:

            self.children[p] = self.children.get(p, 0) + 1

        return c

    def numcommits(self):

        return self.base.numcommits()

    def _cachedcommit(self, rev):

        if rev in self.commits:

            return self.commits[rev]

        return self.base.getcommit(rev)

    def _discard(self, *revs):

        for r in revs:

            if r is None:

                continue

            self.seenchildren[r] = self.seenchildren.get(r, 0) + 1

            if self.seenchildren[r] == self.children[r]:

                self.wantedancestors.pop(r, None)

                self.parentmap.pop(r, None)

                del self.seenchildren[r]

                if self._rebuilt:

                    del self.children[r]

    def wanted(self, rev, i):

        # Return True if we're directly interested in rev.

        #

        # i is an index selecting one of the parents of rev (if rev

        # has no parents, i is None).  getchangedfiles will give us

        # the list of files that are different in rev and in the parent

        # indicated by i.  If we're interested in any of these files,

        # we're interested in rev.

        try:

            files = self.base.getchangedfiles(rev, i)

        except NotImplementedError:

        for f in files:

            if self.filemapper(f):

                return True

        # The include directive is documented to include nothing else (though

        # valid branch closes are included).

        if self.filemapper.include:

            return False

        # Allow empty commits in the source revision through.  The getchanges()

        # method doesn't even bother calling this if it determines that the

        # close marker is significant (i.e. all of the branch ancestors weren't

        # eliminated).  Therefore if there *is* a close marker, getchanges()

        # doesn't consider it significant, and this revision should be dropped.

    def mark_not_wanted(self, rev, p):

        # Mark rev as not interesting and update data structures.

        if p is None:

            # A root revision. Use SKIPREV to indicate that it doesn't

            # map to any revision in the restricted graph.  Put SKIPREV

            # in the set of wanted ancestors to simplify code elsewhere

            self.parentmap[rev] = SKIPREV

            self.wantedancestors[rev] = {SKIPREV}

            return

        # Reuse the data from our parent.

        self.parentmap[rev] = self.parentmap[p]

        self.wantedancestors[rev] = self.wantedancestors[p]

    def mark_wanted(self, rev, parents):

        # Mark rev ss wanted and update data structures.

        # rev will be in the restricted graph, so children of rev in

        # the original graph should still have rev as a parent in the

        # restricted graph.

        self.parentmap[rev] = rev

        # The set of wanted ancestors of rev is the union of the sets

        # of wanted ancestors of its parents. Plus rev itself.

        wrev = set()

        for p in parents:

            if p in self.wantedancestors:

                wrev.update(self.wantedancestors[p])

            else:

        wrev.add(rev)

        self.wantedancestors[rev] = wrev

    def getchanges(self, rev, full):

        parents = self.commits[rev].parents

        if len(parents) > 1 and not self.ignoreancestorcheck:

            self.rebuild()