convert: optimize convert of files that are unmodified from p2 in merges

Conversion of a merge starts with p1 and re-adds the files that were changed in
the merge or came unmodified from p2. Files that are unmodified from p1 will
thus not be touched and take no time. Files that are unmodified from p2 would be
retrieved and rehashed. They would end up getting the same hash as in p2 and end
up reusing the filelog entry and look like the p1 case ... but it was slow.

Instead, make getchanges also return 'files that are unmodified from p2' so the
sink can reuse the existing p2 entry instead of calling getfile.

Reuse of filelog entries can make a big difference when files are big and with
long revlong chains so they take time to retrieve and hash, or when using an
expensive custom getfile function (think
http://mercurial.selenic.com/wiki/ConvertExtension#Customization with a code
reformatter).

This in combination with changes to reuse filectx entries in
localrepo._filecommit make 'unchanged from p2' almost as fast as 'unchanged
from p1'.

This is so far only implemented for the combination of hg source and hg sink.

This is a refactoring/optimization. It is covered by existing tests and show no
changes - which is a good thing.

hgext/convert/bzr.py

@@ -143,7 +143,8 b' class bzr_source(converter_source):'
         parentids = self._parentids.pop(version)
         # only diff against first parent id
         prevtree = self.sourcerepo.revision_tree(parentids[0])
-        return self._gettreechanges(self._revtree, prevtree)
+        files, changes = self._gettreechanges(self._revtree, prevtree)
+        return files, changes, set()
 
     def getcommit(self, version):
         rev = self.sourcerepo.get_revision(version)

hgext/convert/common.py

@@ -97,7 +97,7 b' class converter_source(object):'
         raise NotImplementedError
 
     def getchanges(self, version, full):
-        """Returns a tuple of (files, copies).
+        """Returns a tuple of (files, copies, cleanp2).
 
         files is a sorted list of (filename, id) tuples for all files
         changed between version and its first parent returned by
@@ -105,6 +105,10 b' class converter_source(object):'
         id is the source revision id of the file.
 
         copies is a dictionary of dest: source
+
+        cleanp2 is the set of files filenames that are clean against p2.
+        (Files that are clean against p1 are already not in files (unless
+        full). This makes it possible to handle p2 clean files similarly.)
         """
         raise NotImplementedError
 
@@ -215,7 +219,8 b' class converter_sink(object):'
         mapping equivalent authors identifiers for each system."""
         return None
 
-    def putcommit(self, files, copies, parents, commit, source, revmap, full):
+    def putcommit(self, files, copies, parents, commit, source, revmap, full,
+                  cleanp2):
         """Create a revision with all changed files listed in 'files'
         and having listed parents. 'commit' is a commit object
         containing at a minimum the author, date, and message for this
@@ -225,6 +230,8 b' class converter_sink(object):'
         of source revisions to converted revisions. Only getfile() and
         lookuprev() should be called on 'source'. 'full' means that 'files'
         is complete and all other files should be removed.
+        'cleanp2' is a set of the filenames that are unchanged from p2
+        (only in the common merge case where there two parents).
 
         Note that the sink repository is not told to update itself to
         a particular revision (or even what that revision would be)

hgext/convert/convcmd.py

@@ -397,7 +397,7 b' class converter(object):'
                 dest = self.map[changes]
             self.map[rev] = dest
             return
-        files, copies = changes
+        files, copies, cleanp2 = changes
         pbranches = []
         if commit.parents:
             for prev in commit.parents:
@@ -413,6 +413,8 b' class converter(object):'
             parents = [self.map.get(p, p) for p in parents]
         except KeyError:
             parents = [b[0] for b in pbranches]
+        if len(pbranches) != 2:
+            cleanp2 = set()
         if len(parents) < 3:
             source = progresssource(self.ui, self.source, len(files))
         else:
@@ -423,7 +425,7 b' class converter(object):'
             source = progresssource(self.ui, self.source,
                                     len(files) * (len(parents) - 1))
         newnode = self.dest.putcommit(files, copies, parents, commit,
-                                      source, self.map, full)
+                                      source, self.map, full, cleanp2)
         source.close()
         self.source.converted(rev, newnode)
         self.map[rev] = newnode

hgext/convert/cvs.py

@@ -262,7 +262,7 b' class convert_cvs(converter_source):'
         if full:
             raise util.Abort(_("convert from cvs do not support --full"))
         self._parse()
-        return sorted(self.files[rev].iteritems()), {}
+        return sorted(self.files[rev].iteritems()), {}, set()
 
     def getcommit(self, rev):
         self._parse()

hgext/convert/darcs.py

@@ -188,7 +188,7 b' class darcs_source(converter_source, com'
                 changes.append((elt.text.strip(), rev))
         self.pull(rev)
         self.lastrev = rev
-        return sorted(changes), copies
+        return sorted(changes), copies, set()
 
     def getfile(self, name, rev):
         if rev != self.lastrev:

hgext/convert/filemap.py

@@ -384,12 +384,15 b' class filemap_source(converter_source):'
         # Get the real changes and do the filtering/mapping. To be
         # able to get the files later on in getfile, we hide the
         # original filename in the rev part of the return value.
-        changes, copies = self.base.getchanges(rev, full)
+        changes, copies, cleanp2 = self.base.getchanges(rev, full)
         files = {}
+        ncleanp2 = set(cleanp2)
         for f, r in changes:
             newf = self.filemapper(f)
             if newf and (newf != f or newf not in files):
                 files[newf] = (f, r)
+                if newf != f:
+                    ncleanp2.discard(f)
         files = sorted(files.items())
 
         ncopies = {}
@@ -400,7 +403,7 b' class filemap_source(converter_source):'
                 if newsource:
                     ncopies[newc] = newsource
 
-        return files, ncopies
+        return files, ncopies, ncleanp2
 
     def getfile(self, name, rev):
         realname, realrev = rev

hgext/convert/git.py

@@ -264,7 +264,7 b' class convert_git(converter_source):'
             else:
                 self.retrievegitmodules(version)
                 changes.append(('.hgsubstate', ''))
-        return (changes, copies)
+        return (changes, copies, set())
 
     def getcommit(self, version):
         c = self.catfile(version, "commit") # read the commit hash

hgext/convert/gnuarch.py

@@ -171,7 +171,7 b' class gnuarch_source(converter_source, c'
             copies.update(cps)
 
         self.lastrev = rev
-        return sorted(set(changes)), copies
+        return sorted(set(changes)), copies, set()
 
     def getcommit(self, rev):
         changes = self.changes[rev]

hgext/convert/hg.py

@@ -132,9 +132,14 b' class mercurial_sink(converter_sink):'
             fp.write('%s %s\n' % (revid, s[1]))
         return fp.getvalue()
 
-    def putcommit(self, files, copies, parents, commit, source, revmap, full):
+    def putcommit(self, files, copies, parents, commit, source, revmap, full,
+                  cleanp2):
         files = dict(files)
+
         def getfilectx(repo, memctx, f):
+            if p2ctx and f in cleanp2 and f not in copies:
+                self.ui.debug('reusing %s from p2\n' % f)
+                return p2ctx[f]
             try:
                 v = files[f]
             except KeyError:
@@ -199,6 +204,9 b' class mercurial_sink(converter_sink):'
         while parents:
             p1 = p2
             p2 = parents.pop(0)
+            p2ctx = None
+            if p2 != nullid:
+                p2ctx = self.repo[p2]
             fileset = set(files)
             if full:
                 fileset.update(self.repo[p1])
@@ -382,9 +390,13 b' class mercurial_source(converter_source)'
         # getcopies() is also run for roots and before filtering so missing
         # revlogs are detected early
         copies = self.getcopies(ctx, parents, copyfiles)
+        cleanp2 = set()
+        if len(parents) == 2:
+            cleanp2.update(self.repo.status(parents[1].node(), ctx.node(),
+                                            clean=True).clean)
         changes = [(f, rev) for f in files if f not in self.ignored]
         changes.sort()
-        return changes, copies
+        return changes, copies, cleanp2
 
     def getcopies(self, ctx, parents, files):
         copies = {}

hgext/convert/monotone.py

@@ -280,7 +280,7 b' class monotone_source(converter_source, '
             for fromfile in renamed.values():
                 files[fromfile] = rev
 
-        return (files.items(), copies)
+        return (files.items(), copies, set())
 
     def getfile(self, name, rev):
         if not self.mtnisfile(name, rev):

hgext/convert/p4.py

@@ -195,7 +195,7 b' class p4_source(converter_source):'
     def getchanges(self, rev, full):
         if full:
             raise util.Abort(_("convert from p4 do not support --full"))
-        return self.files[rev], {}
+        return self.files[rev], {}, set()
 
     def getcommit(self, rev):
         return self.changeset[rev]

hgext/convert/subversion.py

@@ -474,7 +474,7 b' class svn_source(converter_source):'
             (files, copies) = self._getchanges(rev, full)
             # caller caches the result, so free it here to release memory
             del self.paths[rev]
-        return (files, copies)
+        return (files, copies, set())
 
     def getchangedfiles(self, rev, i):
         # called from filemap - cache computed values for reuse in getchanges
@@ -1240,7 +1240,8 b' class svn_sink(converter_sink, commandli'
     def revid(self, rev):
         return u"svn:%s@%s" % (self.uuid, rev)
 
-    def putcommit(self, files, copies, parents, commit, source, revmap, full):
+    def putcommit(self, files, copies, parents, commit, source, revmap, full,
+                  cleanp2):
         for parent in parents:
             try:
                 return self.revid(self.childmap[parent])