# hg.py - repository classes 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 sys, struct, os

import util

from revlog import *

from demandload import *

demandload(globals(), "re lock urllib urllib2 transaction time socket")

demandload(globals(), "tempfile httprangereader bdiff urlparse stat")

demandload(globals(), "bisect select")

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 t

        s = t.find('\1\n', 2)

        mt = t[2:s]

        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 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))

class manifest(revlog):

    def __init__(self, opener):

        self.mapcache = None

        self.listcache = None

        self.addlist = None

        revlog.__init__(self, opener, "00manifest.i", "00manifest.d")

    def read(self, node):

        if node == nullid: return {} # don't upset local cache

        if self.mapcache and self.mapcache[0] == node:

            return self.mapcache[1]

        text = self.revision(node)

        map = {}

        flag = {}

        self.listcache = (text, text.splitlines(1))

        for l in self.listcache[1]:

            (f, n) = l.split('\0')

            map[f] = bin(n[:40])

            flag[f] = (n[40:-1] == "x")

        self.mapcache = (node, map, flag)

        return map

    def readflags(self, node):

        if node == nullid: return {} # don't upset local cache

        if not self.mapcache or self.mapcache[0] != node:

            self.read(node)

        return self.mapcache[2]

    def diff(self, a, b):

        # this is sneaky, as we're not actually using a and b

        if self.listcache and self.addlist and self.listcache[0] == a:

            d = mdiff.diff(self.listcache[1], self.addlist, 1)

            if mdiff.patch(a, d) != b:

                sys.stderr.write("*** sortdiff failed, falling back ***\n")

                return mdiff.textdiff(a, b)

            return d

        else:

            return mdiff.textdiff(a, b)

    def add(self, map, flags, transaction, link, p1=None, p2=None,

            changed=None):

        # directly generate the mdiff delta from the data collected during

        # the bisect loop below

        def gendelta(delta):

            i = 0

            result = []

            while i < len(delta):

                start = delta[i][2]

                end = delta[i][3]

                l = delta[i][4]

                if l == None:

                    l = ""

                while i < len(delta) - 1 and start <= delta[i+1][2] \

                          and end >= delta[i+1][2]:

                    if delta[i+1][3] > end:

                        end = delta[i+1][3]

                    if delta[i+1][4]:

                        l += delta[i+1][4]

                    i += 1

                result.append(struct.pack(">lll", start, end, len(l)) +  l)

                i += 1

            return result

        # apply the changes collected during the bisect loop to our addlist

        def addlistdelta(addlist, delta):

            # apply the deltas to the addlist.  start from the bottom up

            # so changes to the offsets don't mess things up.

            i = len(delta)

            while i > 0:

                i -= 1

                start = delta[i][0]

                end = delta[i][1]

                if delta[i][4]:

                    addlist[start:end] = [delta[i][4]]

                else:

                    del addlist[start:end]

            return addlist

        # calculate the byte offset of the start of each line in the

        # manifest

        def calcoffsets(addlist):

            offsets = [0] * (len(addlist) + 1)

            offset = 0

            i = 0

            while i < len(addlist):

                offsets[i] = offset

                offset += len(addlist[i])

                i += 1

            offsets[i] = offset

            return offsets

        # if we're using the listcache, make sure it is valid and

        # parented by the same node we're diffing against

        if not changed or not self.listcache or not p1 or \

               self.mapcache[0] != p1:

            files = map.keys()

            files.sort()

            self.addlist = ["%s\000%s%s\n" %

                            (f, hex(map[f]), flags[f] and "x" or '')

                            for f in files]

            cachedelta = None

        else:

            addlist = self.listcache[1]

            # find the starting offset for each line in the add list

            offsets = calcoffsets(addlist)

            # combine the changed lists into one list for sorting

            work = [[x, 0] for x in changed[0]]

            work[len(work):] = [[x, 1] for x in changed[1]]

            work.sort()

            delta = []

            bs = 0

            for w in work:

                f = w[0]

                # bs will either be the index of the item or the insert point

                bs = bisect.bisect(addlist, f, bs)

                if bs < len(addlist):

                    fn = addlist[bs][:addlist[bs].index('\0')]

                else:

                    fn = None

                if w[1] == 0:

                    l = "%s\000%s%s\n" % (f, hex(map[f]),

                                          flags[f] and "x" or '')

                else:

                    l = None

                start = bs

                if fn != f:

                    # item not found, insert a new one

                    end = bs

                    if w[1] == 1:

                        sys.stderr.write("failed to remove %s from manifest\n"

                                         % f)

                        sys.exit(1)

                else:

                    # item is found, replace/delete the existing line

                    end = bs + 1

                delta.append([start, end, offsets[start], offsets[end], l])

            self.addlist = addlistdelta(addlist, delta)

            if self.mapcache[0] == self.tip():

                cachedelta = "".join(gendelta(delta))

            else:

                cachedelta = None

        text = "".join(self.addlist)

        if cachedelta and mdiff.patch(self.listcache[0], cachedelta) != text:

            sys.stderr.write("manifest delta failure\n")

            sys.exit(1)

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

        self.mapcache = (n, map, flags)

        self.listcache = (text, self.addlist)

        self.addlist = None

        return n

class changelog(revlog):

    def __init__(self, opener):

        revlog.__init__(self, opener, "00changelog.i", "00changelog.d")

    def extract(self, text):

        if not text:

            return (nullid, "", "0", [], "")

        last = text.index("\n\n")

        desc = text[last + 2:]

        l = text[:last].splitlines()

        manifest = bin(l[0])

        user = l[1]

        date = l[2]

        files = l[3:]

        return (manifest, user, date, files, desc)

    def read(self, node):

        return self.extract(self.revision(node))

    def add(self, manifest, list, desc, transaction, p1=None, p2=None,

                  user=None, date=None):

        date = date or "%d %d" % (time.time(), time.timezone)

        list.sort()

        l = [hex(manifest), user, date] + list + ["", desc]

        text = "\n".join(l)

        return self.addrevision(text, transaction, self.count(), p1, p2)

class dirstate:

    def __init__(self, opener, ui, root):

        self.opener = opener

        self.root = root

        self.dirty = 0

        self.ui = ui

        self.map = None

        self.pl = None

        self.copies = {}

        self.ignorefunc = None

    def wjoin(self, f):

        return os.path.join(self.root, f)

    def ignore(self, f):

        if not self.ignorefunc:

            bigpat = []

            try:

                l = file(self.wjoin(".hgignore"))

                for pat in l:

                    if pat != "\n":

                        p = util.pconvert(pat[:-1])

                        try:

                            r = re.compile(p)

                        except:

                            self.ui.warn("ignoring invalid ignore"

                                         + " regular expression '%s'\n" % p)

                        else:

                            bigpat.append(util.pconvert(pat[:-1]))

            except IOError: pass

            if bigpat:

                s = "(?:%s)" % (")|(?:".join(bigpat))

                r = re.compile(s)

                self.ignorefunc = r.search

            else:

                self.ignorefunc = util.never

        return self.ignorefunc(f)

    def __del__(self):

        if self.dirty:

            self.write()

    def __getitem__(self, key):

        try:

            return self.map[key]

        except TypeError:

            self.read()

            return self[key]

    def __contains__(self, key):

        if not self.map: self.read()

        return key in self.map

    def parents(self):

        if not self.pl:

            self.read()

        return self.pl

    def markdirty(self):

        if not self.dirty:

            self.dirty = 1

    def setparents(self, p1, p2 = nullid):

        self.markdirty()

        self.pl = p1, p2

    def state(self, key):

        try:

            return self[key][0]

        except KeyError:

            return "?"

    def read(self):

        if self.map is not None: return self.map

        self.map = {}

        self.pl = [nullid, nullid]

        try:

            st = self.opener("dirstate").read()

            if not st: return

        except: return

        self.pl = [st[:20], st[20: 40]]

        pos = 40

        while pos < len(st):

            e = struct.unpack(">cllll", st[pos:pos+17])

            l = e[4]

            pos += 17

            f = st[pos:pos + l]

            if '\0' in f:

                f, c = f.split('\0')

                self.copies[f] = c

            self.map[f] = e[:4]

            pos += l

    def copy(self, source, dest):

        self.read()

        self.markdirty()

        self.copies[dest] = source

    def copied(self, file):

        return self.copies.get(file, None)

        ''' current states:

        n  normal

        m  needs merging

        r  marked for removal

        a  marked for addition'''

        if not files: return

        self.read()

        self.markdirty()

        for f in files:

            if state == "r":

                self.map[f] = ('r', 0, 0, 0)

            else:

                s = os.stat(os.path.join(self.root, f))

    def forget(self, files):

        if not files: return

        self.read()

        self.markdirty()

        for f in files:

            try:

                del self.map[f]

            except KeyError:

                self.ui.warn("not in dirstate: %s!\n" % f)

                pass

    def clear(self):

        self.map = {}

        self.markdirty()

    def write(self):

        st = self.opener("dirstate", "w")

        st.write("".join(self.pl))

        for f, e in self.map.items():

            c = self.copied(f)

            if c:

                f = f + "\0" + c

            e = struct.pack(">cllll", e[0], e[1], e[2], e[3], len(f))

            st.write(e + f)

        self.dirty = 0

    def walk(self, files = None, match = util.always):

        self.read()

        dc = self.map.copy()

        # walk all files by default

        if not files: files = [self.root]

        known = {'.hg': 1}

        def seen(fn):

            if fn in known: return True

            known[fn] = 1

        def traverse():

            for f in util.unique(files):

                f = os.path.join(self.root, f)

                if os.path.isdir(f):

                    for dir, subdirs, fl in os.walk(f):

                        d = dir[len(self.root) + 1:]

                        nd = os.path.normpath(d)

                        if seen(nd):

                            subdirs[:] = []

                            continue

                        for sd in subdirs:

                            ds = os.path.join(nd, sd +'/')

                            if self.ignore(ds) or not match(ds):

                                subdirs.remove(sd)

                        subdirs.sort()

                        fl.sort()

                        for fn in fl:

                            fn = util.pconvert(os.path.join(d, fn))

                            yield 'f', fn

                else:

                    yield 'f', f[len(self.root) + 1:]

            ks = dc.keys()

            ks.sort()

            for k in ks:

                yield 'm', k

        # yield only files that match: all in dirstate, others only if

        # not in .hgignore

        for src, fn in util.unique(traverse()):

            fn = os.path.normpath(fn)

            if seen(fn): continue

            if fn in dc:

                del dc[fn]

            elif self.ignore(fn):

                continue

            if match(fn):

                yield src, fn