convert: fix --datesort ordering

Two branches a and b starting at root, with commits interleaved like:
root a0 a1 b0 a2 a3 b1
were converted in the following order:
root a0 b0 a1 b1 a2 a3

Replace depth based toposort with a more classic traversal method.

tests/test-convert-datesort

@@ -0,0 +1,40
+#!/bin/sh
+
+cat >> $HGRCPATH <<EOF
+[extensions]
+convert=
+graphlog=
+EOF
+
+hg init t
+cd t
+echo a >> a
+hg ci -Am a0 -d '1 0'
+hg branch brancha
+echo a >> a
+hg ci -m a1 -d '2 0'
+echo a >> a
+hg ci -m a2 -d '3 0'
+echo a >> a
+hg ci -m a3 -d '4 0'
+hg up -C 0
+hg branch branchb
+echo b >> b
+hg ci -Am b0 -d '5 0'
+hg up -C brancha
+echo a >> a
+hg ci -m a4 -d '6 0'
+echo a >> a
+hg ci -m a5 -d '7 0'
+echo a >> a
+hg ci -m a6 -d '8 0'
+hg up -C branchb
+echo b >> b
+hg ci -m b1 -d '9 0'
+cd ..
+
+echo % convert with datesort
+hg convert --datesort t t2
+echo % graph converted repo
+hg -R t2 glog --template '#rev# "#desc#"\n'
+

tests/test-convert-datesort.out

@@ -0,0 +1,40
+adding a
+marked working directory as branch brancha
+1 files updated, 0 files merged, 0 files removed, 0 files unresolved
+marked working directory as branch branchb
+adding b
+1 files updated, 0 files merged, 1 files removed, 0 files unresolved
+2 files updated, 0 files merged, 0 files removed, 0 files unresolved
+% convert with datesort
+initializing destination t2 repository
+scanning source...
+sorting...
+converting...
+8 a0
+7 a1
+6 a2
+5 a3
+4 b0
+3 a4
+2 a5
+1 a6
+0 b1
+% graph converted repo
+o  8 "b1"
+|
+| o  7 "a6"
+| |
+| o  6 "a5"
+| |
+| o  5 "a4"
+| |
+o |  4 "b0"
+| |
+| o  3 "a3"
+| |
+| o  2 "a2"
+| |
+| o  1 "a1"
+|/
+o  0 "a0"
+

hgext/convert/convcmd.py

@@ -1,331 +1,342
 # convcmd - convert extension commands definition
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms
 # of the GNU General Public License, incorporated herein by reference.
 
 from common import NoRepo, SKIPREV, converter_source, converter_sink
 from cvs import convert_cvs
 from darcs import darcs_source
 from git import convert_git
 from hg import mercurial_source, mercurial_sink
 from subversion import svn_source, debugsvnlog
 import filemap
 
 import os, shutil
 from mercurial import hg, util
 from mercurial.i18n import _
 
 source_converters = [
     ('cvs', convert_cvs),
     ('git', convert_git),
     ('svn', svn_source),
     ('hg', mercurial_source),
     ('darcs', darcs_source),
     ]
 
 sink_converters = [
     ('hg', mercurial_sink),
     ]
 
 def convertsource(ui, path, type, rev):
     exceptions = []
     for name, source in source_converters:
         try:
             if not type or name == type:
                 return source(ui, path, rev)
         except NoRepo, inst:
             exceptions.append(inst)
     if not ui.quiet:
         for inst in exceptions:
             ui.write(_("%s\n") % inst)
     raise util.Abort('%s: unknown repository type' % path)
 
 def convertsink(ui, path, type):
     for name, sink in sink_converters:
         try:
             if not type or name == type:
                 return sink(ui, path)
         except NoRepo, inst:
             ui.note(_("convert: %s\n") % inst)
     raise util.Abort('%s: unknown repository type' % path)
 
 class converter(object):
     def __init__(self, ui, source, dest, revmapfile, opts):
 
         self.source = source
         self.dest = dest
         self.ui = ui
         self.opts = opts
         self.commitcache = {}
         self.revmapfile = revmapfile
         self.revmapfilefd = None
         self.authors = {}
         self.authorfile = None
 
         self.maporder = []
         self.map = {}
         try:
             origrevmapfile = open(self.revmapfile, 'r')
             for l in origrevmapfile:
                 sv, dv = l[:-1].split()
                 if sv not in self.map:
                     self.maporder.append(sv)
                 self.map[sv] = dv
             origrevmapfile.close()
         except IOError:
             pass
 
         # Read first the dst author map if any
         authorfile = self.dest.authorfile()
         if authorfile and os.path.exists(authorfile):
             self.readauthormap(authorfile)
         # Extend/Override with new author map if necessary
         if opts.get('authors'):
             self.readauthormap(opts.get('authors'))
             self.authorfile = self.dest.authorfile()
 
     def walktree(self, heads):
         '''Return a mapping that identifies the uncommitted parents of every
         uncommitted changeset.'''
         visit = heads
         known = {}
         parents = {}
         while visit:
             n = visit.pop(0)
             if n in known or n in self.map: continue
             known[n] = 1
             commit = self.cachecommit(n)
             parents[n] = []
             for p in commit.parents:
                 parents[n].append(p)
                 visit.append(p)
 
         return parents
 
     def toposort(self, parents):
         '''Return an ordering such that every uncommitted changeset is
         preceeded by all its uncommitted ancestors.'''
         visit = parents.keys()
         seen = {}
         children = {}
+        actives = []
 
         while visit:
             n = visit.pop(0)
             if n in seen: continue
             seen[n] = 1
             # Ensure that nodes without parents are present in the 'children'
             # mapping.
             children.setdefault(n, [])
+            hasparent = False
             for p in parents[n]:
                 if not p in self.map:
                     visit.append(p)
+                    hasparent = True
                 children.setdefault(p, []).append(n)
+            if not hasparent:
+                actives.append(n)
+
+        del seen
+        del visit
+
+        if self.opts.get('datesort'):
+            dates = {}
+            def getdate(n):
+                if n not in dates:
+                    dates[n] = util.parsedate(self.commitcache[n].date)
+                return dates[n]
+
+            def picknext(nodes):
+                return min([(getdate(n), n) for n in nodes])[1]
+        else:
+            prev = [None]
+            def picknext(nodes):
+                # Return the first eligible child of the previously converted
+                # revision, or any of them.
+                next = nodes[0]
+                for n in nodes:
+                    if prev[0] in parents[n]:
+                        next = n
+                        break
+                prev[0] = next
+                return next
 
         s = []
-        removed = {}
-        visit = children.keys()
-        while visit:
-            n = visit.pop(0)
-            if n in removed: continue
-            dep = 0
-            if n in parents:
-                for p in parents[n]:
-                    if p in self.map: continue
-                    if p not in removed:
-                        # we're still dependent
-                        visit.append(n)
-                        dep = 1
-                        break
+        pendings = {}
+        while actives:
+            n = picknext(actives)
+            actives.remove(n)
+            s.append(n)
 
-            if not dep:
-                # all n's parents are in the list
-                removed[n] = 1
-                if n not in self.map:
-                    s.append(n)
-                if n in children:
-                    for c in children[n]:
-                        visit.insert(0, c)
+            # Update dependents list
+            for c in children.get(n, []):
+                if c not in pendings:
+                    pendings[c] = [p for p in parents[c] if p not in self.map]
+                pendings[c].remove(n)
+                if not pendings[c]:
+                    # Parents are converted, node is eligible
+                    actives.insert(0, c)
+                    pendings[c] = None
 
-        if self.opts.get('datesort'):
-            depth = {}
-            for n in s:
-                depth[n] = 0
-                pl = [p for p in self.commitcache[n].parents
-                      if p not in self.map]
-                if pl:
-                    depth[n] = max([depth[p] for p in pl]) + 1
-
-            s = [(depth[n], util.parsedate(self.commitcache[n].date), n)
-                 for n in s]
-            s.sort()
-            s = [e[2] for e in s]
+        if len(s) != len(parents):
+            raise util.Abort(_("not all revisions were sorted"))
 
         return s
 
     def mapentry(self, src, dst):
         if self.revmapfilefd is None:
             try:
                 self.revmapfilefd = open(self.revmapfile, "a")
             except IOError, (errno, strerror):
                 raise util.Abort("Could not open map file %s: %s, %s\n" % (self.revmapfile, errno, strerror))
         self.map[src] = dst
         self.revmapfilefd.write("%s %s\n" % (src, dst))
         self.revmapfilefd.flush()
 
     def writeauthormap(self):
         authorfile = self.authorfile
         if authorfile:
            self.ui.status('Writing author map file %s\n' % authorfile)
            ofile = open(authorfile, 'w+')
            for author in self.authors:
                ofile.write("%s=%s\n" % (author, self.authors[author]))
            ofile.close()
 
     def readauthormap(self, authorfile):
         afile = open(authorfile, 'r')
         for line in afile:
             try:
                 srcauthor = line.split('=')[0].strip()
                 dstauthor = line.split('=')[1].strip()
                 if srcauthor in self.authors and dstauthor != self.authors[srcauthor]:
                     self.ui.status(
                         'Overriding mapping for author %s, was %s, will be %s\n'
                         % (srcauthor, self.authors[srcauthor], dstauthor))
                 else:
                     self.ui.debug('Mapping author %s to %s\n'
                                   % (srcauthor, dstauthor))
                     self.authors[srcauthor] = dstauthor
             except IndexError:
                 self.ui.warn(
                     'Ignoring bad line in author file map %s: %s\n'
                     % (authorfile, line))
         afile.close()
 
     def cachecommit(self, rev):
         commit = self.source.getcommit(rev)
         commit.author = self.authors.get(commit.author, commit.author)
         self.commitcache[rev] = commit
         return commit
 
     def copy(self, rev):
         commit = self.commitcache[rev]
         do_copies = hasattr(self.dest, 'copyfile')
         filenames = []
 
         changes = self.source.getchanges(rev)
         if isinstance(changes, basestring):
             if changes == SKIPREV:
                 dest = SKIPREV
             else:
                 dest = self.map[changes]
             self.mapentry(rev, dest)
             return
         files, copies = changes
         pbranches = []
         if commit.parents:
             for prev in commit.parents:
                 if prev not in self.commitcache:
                     self.cachecommit(prev)
                 pbranches.append((self.map[prev], 
                                   self.commitcache[prev].branch))
         self.dest.setbranch(commit.branch, pbranches)
         for f, v in files:
             filenames.append(f)
             try:
                 data = self.source.getfile(f, v)
             except IOError, inst:
                 self.dest.delfile(f)
             else:
                 e = self.source.getmode(f, v)
                 self.dest.putfile(f, e, data)
                 if do_copies:
                     if f in copies:
                         copyf = copies[f]
                         # Merely marks that a copy happened.
                         self.dest.copyfile(copyf, f)
 
         parents = [b[0] for b in pbranches]
         newnode = self.dest.putcommit(filenames, parents, commit)
         self.mapentry(rev, newnode)
 
     def convert(self):
         try:
             self.source.before()
             self.dest.before()
             self.source.setrevmap(self.map, self.maporder)
             self.ui.status("scanning source...\n")
             heads = self.source.getheads()
             parents = self.walktree(heads)
             self.ui.status("sorting...\n")
             t = self.toposort(parents)
             num = len(t)
             c = None
 
             self.ui.status("converting...\n")
             for c in t:
                 num -= 1
                 desc = self.commitcache[c].desc
                 if "\n" in desc:
                     desc = desc.splitlines()[0]
                 self.ui.status("%d %s\n" % (num, desc))
                 self.copy(c)
 
             tags = self.source.gettags()
             ctags = {}
             for k in tags:
                 v = tags[k]
                 if self.map.get(v, SKIPREV) != SKIPREV:
                     ctags[k] = self.map[v]
 
             if c and ctags:
                 nrev = self.dest.puttags(ctags)
                 # write another hash correspondence to override the previous
                 # one so we don't end up with extra tag heads
                 if nrev:
                     self.mapentry(c, nrev)
 
             self.writeauthormap()
         finally:
             self.cleanup()
 
     def cleanup(self):
         try:
             self.dest.after()
         finally:
             self.source.after()
         if self.revmapfilefd:
             self.revmapfilefd.close()
 
 def convert(ui, src, dest=None, revmapfile=None, **opts):
     util._encoding = 'UTF-8'
 
     if not dest:
         dest = hg.defaultdest(src) + "-hg"
         ui.status("assuming destination %s\n" % dest)
 
     destc = convertsink(ui, dest, opts.get('dest_type'))
 
     try:
         srcc = convertsource(ui, src, opts.get('source_type'),
                              opts.get('rev'))
     except Exception:
         for path in destc.created:
             shutil.rmtree(path, True)
         raise
 
     fmap = opts.get('filemap')
     if fmap:
         srcc = filemap.filemap_source(ui, srcc, fmap)
         destc.setfilemapmode(True)
 
     if not revmapfile:
         try:
             revmapfile = destc.revmapfile()
         except:
             revmapfile = os.path.join(destc, "map")
 
     c = converter(ui, srcc, destc, revmapfile, opts)
     c.convert()
 

tests/test-convert-clonebranches.out

@@ -1,29 +1,29
 marked working directory as branch branch0
 1 files updated, 0 files merged, 0 files removed, 0 files unresolved
 (branch merge, don't forget to commit)
 % convert
 3 adda
 2 addb
 pulling from default into branch0
 1 changesets found
 1 changea
 0 mergeab
 pulling from default into branch0
 1 changesets found
 marked working directory as branch branch1
 marked working directory as branch branch2
 1 files updated, 0 files merged, 0 files removed, 0 files unresolved
 (branch merge, don't forget to commit)
 marked working directory as branch branch3
 % incremental conversion
-2 c1
-pulling from branch0 into branch1
+2 c2
+pulling from branch0 into branch2
 2 changesets found
-1 c2
-pulling from branch0 into branch2
+1 c1
+pulling from branch0 into branch1
 2 changesets found
 0 c3
 pulling from branch2 into branch3
 3 changesets found
 pulling from branch1 into branch3
 1 changesets found