| Submitter | Philippe Pepiot |
|---|---|
| Date | April 5, 2017, 4:14 p.m. |
| Message ID | <c0231296470330eeae91.1491408887@sh71.tls.logilab.fr> |
| Download | mbox | patch |
| Permalink | /patch/19972/ |
| State | Superseded |
| Headers | show |
Comments
On Wed, 05 Apr 2017 18:14:47 +0200, Philippe Pepiot wrote: > # HG changeset patch > # User Philippe Pepiot <philippe.pepiot@logilab.fr> > # Date 1491391351 -7200 > # Wed Apr 05 13:22:31 2017 +0200 > # Node ID c0231296470330eeae91058b9a3d7a6785e16cbe > # Parent 04ec317b81280c189fcea33a05c8cbbac3c186b1 > # Available At https://hg.logilab.org/users/ppepiot/hg > # hg pull https://hg.logilab.org/users/ppepiot/hg -r c02312964703 > perf: add historical portability for util.timer > > util.timer has been introduced in ae5d60bb and used in perf.py since 22fbca1d. > For historical portability, use locally defined timer. > > diff --git a/contrib/perf.py b/contrib/perf.py > --- a/contrib/perf.py > +++ b/contrib/perf.py > @@ -67,6 +67,14 @@ def safehasattr(thing, attr): > setattr(util, 'safehasattr', safehasattr) > > # for "historical portability": > +# use locally defined timer instead of util.timer, because util.timer has been > +# available since ae5d60bb70c9 > +if safehasattr(time, 'perf_counter'): > + timer = time.perf_counter > +else: > + timer = time.time Any reason to drop time.clock() support? And I found "timer" is widely used in perf.py for different meaning. Perhaps we'll need another name. Sorry I didn't notice it yesterday.
On 04/06/2017 02:31 PM, Yuya Nishihara wrote: > On Wed, 05 Apr 2017 18:14:47 +0200, Philippe Pepiot wrote: >> # HG changeset patch >> # User Philippe Pepiot <philippe.pepiot@logilab.fr> >> # Date 1491391351 -7200 >> # Wed Apr 05 13:22:31 2017 +0200 >> # Node ID c0231296470330eeae91058b9a3d7a6785e16cbe >> # Parent 04ec317b81280c189fcea33a05c8cbbac3c186b1 >> # Available At https://hg.logilab.org/users/ppepiot/hg >> # hg pull https://hg.logilab.org/users/ppepiot/hg -r c02312964703 >> perf: add historical portability for util.timer >> >> util.timer has been introduced in ae5d60bb and used in perf.py since 22fbca1d. >> For historical portability, use locally defined timer. >> >> diff --git a/contrib/perf.py b/contrib/perf.py >> --- a/contrib/perf.py >> +++ b/contrib/perf.py >> @@ -67,6 +67,14 @@ def safehasattr(thing, attr): >> setattr(util, 'safehasattr', safehasattr) >> >> # for "historical portability": >> +# use locally defined timer instead of util.timer, because util.timer has been >> +# available since ae5d60bb70c9 >> +if safehasattr(time, 'perf_counter'): >> + timer = time.perf_counter >> +else: >> + timer = time.time > > Any reason to drop time.clock() support? I dropped windows support by mistake (previous version of perf.py was only using time.time). > > And I found "timer" is widely used in perf.py for different meaning. Perhaps > we'll need another name. Sorry I didn't notice it yesterday. > Good catch, sorry to not have seen it before submission. I propose to forcibly define util.timer instead (like util.safehasattr). Will send a V3
Patch
diff --git a/contrib/perf.py b/contrib/perf.py --- a/contrib/perf.py +++ b/contrib/perf.py @@ -67,6 +67,14 @@ def safehasattr(thing, attr): setattr(util, 'safehasattr', safehasattr) # for "historical portability": +# use locally defined timer instead of util.timer, because util.timer has been +# available since ae5d60bb70c9 +if safehasattr(time, 'perf_counter'): + timer = time.perf_counter +else: + timer = time.time + +# for "historical portability": # use locally defined empty option list, if formatteropts isn't # available, because commands.formatteropts has been available since # 3.2 (or 7a7eed5176a4), even though formatting itself has been @@ -193,13 +201,13 @@ def stub_timer(fm, func, title=None): def _timer(fm, func, title=None): gc.collect() results = [] - begin = util.timer() + begin = timer() count = 0 while True: ostart = os.times() - cstart = util.timer() + cstart = timer() r = func() - cstop = util.timer() + cstop = timer() ostop = os.times() count += 1 a, b = ostart, ostop diff --git a/tests/test-check-code.t b/tests/test-check-code.t --- a/tests/test-check-code.t +++ b/tests/test-check-code.t @@ -9,7 +9,7 @@ New errors are not allowed. Warnings are $ hg locate -X contrib/python-zstandard -X hgext/fsmonitor/pywatchman | > sed 's-\\-/-g' | xargs "$check_code" --warnings --per-file=0 || false - contrib/perf.py:859: + contrib/perf.py:867: > r.revision(r.node(x)) don't covert rev to node before passing to revision(nodeorrev) Skipping i18n/polib.py it has no-che?k-code (glob) diff --git a/tests/test-contrib-perf.t b/tests/test-contrib-perf.t --- a/tests/test-contrib-perf.t +++ b/tests/test-contrib-perf.t @@ -165,7 +165,7 @@ Check perf.py for historical portability $ (hg files -r 1.2 glob:mercurial/*.c glob:mercurial/*.py; > hg files -r tip glob:mercurial/*.c glob:mercurial/*.py) | > "$TESTDIR"/check-perf-code.py contrib/perf.py - contrib/perf.py:859: + contrib/perf.py:867: > r.revision(r.node(x)) don't covert rev to node before passing to revision(nodeorrev) [1]