[Bug 5538] New: followlines() spends a lot of time in linkrev adjustment

[mercurial-bugs at mercurial-scm.org]

https://bz.mercurial-scm.org/show_bug.cgi?id=5538

            Bug ID: 5538
           Summary: followlines() spends a lot of time in linkrev
                    adjustment
           Product: Mercurial
           Version: default branch
          Hardware: All
                OS: All
            Status: UNCONFIRMED
          Severity: bug
          Priority: normal
         Component: Mercurial
          Assignee: bugzilla at mercurial-scm.org
          Reporter: gregory.szorc at gmail.com
                CC: mercurial-devel at mercurial-scm.org,
                    pierre-yves.david at ens-lyon.org

On the Firefox repo, `hg log -r 'followlines("layout/generic/nsBlockFrame.cpp",
2750:2850)'` spends ~85% of execution time in linkrev adjustment:

$ hg log -r 'followlines("layout/generic/nsBlockFrame.cpp", 2750:2850)'
--pager=false -T '{rev}\n' --profile
| 100.0%  dispatch.py:    callcatch                line 45: 
mercurial.dispatch.run()
| 100.0%  scmutil.py:     callcatch                line 234:  return
scmutil.callcatch(ui...
| 100.0%  dispatch.py:    _runcatchfunc            line 145:  return func()
| 100.0%  blackbox.py:    log                      line 215:  return
_dispatch(req)
| 99.8%  commands.py:    log                       line 3546:  return orig(ui,
repo, *args...
| 99.8%  cmdutil.py:     getlogrevs                line 3418:  revs, expr,
filematcher = c...
| 98.2%  cmdutil.py:     _makelogrevset            line 2138:  expr,
filematcher = _makelo...
| 98.2%  smartset.py:    __iter__                  line 1961:  it = iter(revs)
| 98.2%  smartset.py:    _consumegen               line 815:  for x in
self._consumegen():
| 98.2%  revset.py:      <genexpr>                 line 846:  for item in
self._gen:
| 98.2%  context.py:     blockancestors            line 954:  (c.rev() for c,
_linerange
 \ 85.9%  util.py:        __get__                  line 763:  return
self._changeid
   | 85.9%  context.py:     _changeid              line 815:  result =
self.func(obj)
   | 85.9%  context.py:     _adjustlinkrev         line 700:  return
self._adjustlinkrev(...
   | 85.9%  ancestor.py:    __contains__           line 860:  if lkr not in
memberanc:
   | 29.0%  changelog.py:   parentrevs             line 353:  for parent in
parentrevs(-h...
   |  9.7%  revlog.py:      parentrevs             line 359:  return
super(changelog, sel...
 \ 12.2%  context.py:     _changesrange            line 1200:  inrangep,
linerange1 = _cha...
   | 11.6%  context.py:     data                   line 1178:  blocks =
mdiff.allblocks(fc...
   | 11.6%  filelog.py:     read                   line 1132:  return
self._filelog.read(s...
   | 11.6%  revlog.py:      revision               line 48:  t =
self.revision(node)
   |  8.0%  revlog.py:      _chunks                line 1313:  bins =
self._chunks(chain, ...
   |  8.0%  revlog.py:      decompress             line 1232: 
ladd(decomp(buffer(data, ch...
---
Sample count: 11546
Total time: 38.470000 seconds

This makes followlines() extremely slow on large repos, especially with files
with many revisions.

-- 
You are receiving this mail because:
You are on the CC list for the bug.