Mercurial > hg
view tests/test-progress.t @ 36820:f33a87cf60cc
xdiff: add a preprocessing step that trims files
xdiff has a `xdl_trim_ends` step that removes common lines, unmatchable
lines. That is in theory good, but happens too late - after splitting,
hashing, and adjusting the hash values so they are unique. Those splitting,
hashing and adjusting hash values steps could have noticeable overhead.
Diffing two large files with minor (one-line-ish) changes are not uncommon.
In that case, the raw performance of those preparation steps seriously
matter. Even allocating an O(N) array and storing line offsets to it is
expensive. Therefore my previous attempts [1] [2] cannot be good enough
since they do not remove the O(N) array assignment.
This patch adds a preprocessing step - `xdl_trim_files` that runs before
other preprocessing steps. It counts common prefix and suffix and lines in
them (needed for displaying line number), without doing anything else.
Testing with a crafted large (169MB) file, with minor change:
```
open('a','w').write(''.join('%s\n' % (i % 100000) for i in xrange(30000000) if i != 6000000))
open('b','w').write(''.join('%s\n' % (i % 100000) for i in xrange(30000000) if i != 6003000))
```
Running xdiff by a simple binary [3], this patch improves the xdiff perf by
more than 10x for the above case:
```
# xdiff before this patch
2.41s user 1.13s system 98% cpu 3.592 total
# xdiff after this patch
0.14s user 0.16s system 98% cpu 0.309 total
# gnu diffutils
0.12s user 0.15s system 98% cpu 0.272 total
# (best of 20 runs)
```
It's still slightly slower than GNU diffutils. But it's pretty close now.
Testing with real repo data:
For the whole repo, this patch makes xdiff 25% faster:
```
# hg perfbdiff --count 100 --alldata -c d334afc585e2 --blocks [--xdiff]
# xdiff, after
! wall 0.058861 comb 0.050000 user 0.050000 sys 0.000000 (best of 100)
# xdiff, before
! wall 0.077816 comb 0.080000 user 0.080000 sys 0.000000 (best of 91)
# bdiff
! wall 0.117473 comb 0.120000 user 0.120000 sys 0.000000 (best of 67)
```
For files that are long (ex. commands.py), the speedup is more than 3x, very
significant:
```
# hg perfbdiff --count 3000 --blocks commands.py.i 1 [--xdiff]
# xdiff, after
! wall 0.690583 comb 0.690000 user 0.690000 sys 0.000000 (best of 12)
# xdiff, before
! wall 2.240361 comb 2.210000 user 2.210000 sys 0.000000 (best of 4)
# bdiff
! wall 2.469852 comb 2.440000 user 2.440000 sys 0.000000 (best of 4)
```
[1]: https://phab.mercurial-scm.org/D2631
[2]: https://phab.mercurial-scm.org/D2634
[3]:
```
// Code to run xdiff from command line. No proper error handling.
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include "mercurial/thirdparty/xdiff/xdiff.h"
#define ensure(x) if (!(x)) exit(255);
mmfile_t readfile(const char *path) {
struct stat st; int fd = open(path, O_RDONLY);
fstat(fd, &st); mmfile_t file = { malloc(st.st_size), st.st_size };
ensure(read(fd, file.ptr, st.st_size) == st.st_size); close(fd);
return file;
}
int main(int argc, char const *argv[]) {
mmfile_t a = readfile(argv[1]), b = readfile(argv[2]);
xpparam_t xpp = {0}; xdemitconf_t xecfg = {0}; xdemitcb_t ecb = {0};
xdl_diff(&a, &b, &xpp, &xecfg, &ecb);
return 0;
}
```
Differential Revision: https://phab.mercurial-scm.org/D2686
| author | Jun Wu <quark@fb.com> |
|---|---|
| date | Wed, 07 Mar 2018 14:45:31 -0800 |
| parents | 12b355964de8 |
| children | 34592dd3bfa2 |
line wrap: on
line source
$ cat > loop.py <<EOF > from __future__ import absolute_import > import time > from mercurial import commands, registrar > > cmdtable = {} > command = registrar.command(cmdtable) > > class incrementingtime(object): > def __init__(self): > self._time = 0.0 > def __call__(self): > self._time += 0.25 > return self._time > time.time = incrementingtime() > > @command(b'loop', > [('', 'total', '', 'override for total'), > ('', 'nested', False, 'show nested results'), > ('', 'parallel', False, 'show parallel sets of results')], > 'hg loop LOOPS', > norepo=True) > def loop(ui, loops, **opts): > loops = int(loops) > total = None > if loops >= 0: > total = loops > if opts.get('total', None): > total = int(opts.get('total')) > nested = False > if opts.get('nested', None): > nested = True > loops = abs(loops) > > for i in range(loops): > ui.progress(topiclabel, i, getloopitem(i), 'loopnum', total) > if opts.get('parallel'): > ui.progress('other', i, 'other.%d' % i, 'othernum', total) > if nested: > nested_steps = 2 > if i and i % 4 == 0: > nested_steps = 5 > for j in range(nested_steps): > ui.progress( > 'nested', j, 'nested.%d' % j, 'nestnum', nested_steps) > ui.progress( > 'nested', None, 'nested.done', 'nestnum', nested_steps) > ui.progress(topiclabel, None, 'loop.done', 'loopnum', total) > > topiclabel = 'loop' > def getloopitem(i): > return 'loop.%d' % i > > EOF $ cp $HGRCPATH $HGRCPATH.orig $ echo "[extensions]" >> $HGRCPATH $ echo "progress=" >> $HGRCPATH $ echo "loop=`pwd`/loop.py" >> $HGRCPATH $ echo "[progress]" >> $HGRCPATH $ echo "format = topic bar number" >> $HGRCPATH $ echo "assume-tty=1" >> $HGRCPATH $ echo "width=60" >> $HGRCPATH test default params, display nothing because of delay $ hg -y loop 3 $ echo "delay=0" >> $HGRCPATH $ echo "refresh=0" >> $HGRCPATH test with delay=0, refresh=0 $ hg -y loop 3 \r (no-eol) (esc) loop [ ] 0/3\r (no-eol) (esc) loop [===============> ] 1/3\r (no-eol) (esc) loop [===============================> ] 2/3\r (no-eol) (esc) \r (no-eol) (esc) no progress with --quiet $ hg -y loop 3 --quiet test plain mode exception $ HGPLAINEXCEPT=progress hg -y loop 1 \r (no-eol) (esc) loop [ ] 0/1\r (no-eol) (esc) \r (no-eol) (esc) test nested short-lived topics (which shouldn't display with nestdelay): $ hg -y loop 3 --nested \r (no-eol) (esc) loop [ ] 0/3\r (no-eol) (esc) loop [===============> ] 1/3\r (no-eol) (esc) loop [===============================> ] 2/3\r (no-eol) (esc) \r (no-eol) (esc) Test nested long-lived topic which has the same name as a short-lived peer. We shouldn't get stuck showing the short-lived inner steps, and should go back to skipping the inner steps when the slow nested step finishes. $ hg -y loop 7 --nested \r (no-eol) (esc) loop [ ] 0/7\r (no-eol) (esc) loop [=====> ] 1/7\r (no-eol) (esc) loop [============> ] 2/7\r (no-eol) (esc) loop [===================> ] 3/7\r (no-eol) (esc) loop [==========================> ] 4/7\r (no-eol) (esc) nested [==========================> ] 3/5\r (no-eol) (esc) nested [===================================> ] 4/5\r (no-eol) (esc) loop [=================================> ] 5/7\r (no-eol) (esc) loop [========================================> ] 6/7\r (no-eol) (esc) \r (no-eol) (esc) $ hg --config progress.changedelay=0 -y loop 3 --nested \r (no-eol) (esc) loop [ ] 0/3\r (no-eol) (esc) nested [ ] 0/2\r (no-eol) (esc) nested [======================> ] 1/2\r (no-eol) (esc) loop [===============> ] 1/3\r (no-eol) (esc) nested [ ] 0/2\r (no-eol) (esc) nested [======================> ] 1/2\r (no-eol) (esc) loop [===============================> ] 2/3\r (no-eol) (esc) nested [ ] 0/2\r (no-eol) (esc) nested [======================> ] 1/2\r (no-eol) (esc) \r (no-eol) (esc) test two topics being printed in parallel (as when we're doing a local --pull clone, where you get the unbundle and bundle progress at the same time): $ hg loop 3 --parallel \r (no-eol) (esc) loop [ ] 0/3\r (no-eol) (esc) loop [===============> ] 1/3\r (no-eol) (esc) loop [===============================> ] 2/3\r (no-eol) (esc) \r (no-eol) (esc) test refresh is taken in account $ hg -y --config progress.refresh=100 loop 3 test format options 1 $ hg -y --config 'progress.format=number topic item+2' loop 2 \r (no-eol) (esc) 0/2 loop lo\r (no-eol) (esc) 1/2 loop lo\r (no-eol) (esc) \r (no-eol) (esc) test format options 2 $ hg -y --config 'progress.format=number item-3 bar' loop 2 \r (no-eol) (esc) 0/2 p.0 [ ]\r (no-eol) (esc) 1/2 p.1 [=======================> ]\r (no-eol) (esc) \r (no-eol) (esc) test format options and indeterminate progress $ hg -y --config 'progress.format=number item bar' loop -- -2 \r (no-eol) (esc) 0 loop.0 [ <=> ]\r (no-eol) (esc) 1 loop.1 [ <=> ]\r (no-eol) (esc) \r (no-eol) (esc) make sure things don't fall over if count > total $ hg -y loop --total 4 6 \r (no-eol) (esc) loop [ ] 0/4\r (no-eol) (esc) loop [===========> ] 1/4\r (no-eol) (esc) loop [=======================> ] 2/4\r (no-eol) (esc) loop [===================================> ] 3/4\r (no-eol) (esc) loop [===============================================>] 4/4\r (no-eol) (esc) loop [ <=> ] 5/4\r (no-eol) (esc) \r (no-eol) (esc) test immediate progress completion $ hg -y loop 0 test delay time estimates #if no-chg $ cp $HGRCPATH.orig $HGRCPATH $ echo "[extensions]" >> $HGRCPATH $ echo "mocktime=$TESTDIR/mocktime.py" >> $HGRCPATH $ echo "progress=" >> $HGRCPATH $ echo "loop=`pwd`/loop.py" >> $HGRCPATH $ echo "[progress]" >> $HGRCPATH $ echo "assume-tty=1" >> $HGRCPATH $ echo "delay=25" >> $HGRCPATH $ echo "width=60" >> $HGRCPATH $ MOCKTIME=11 hg -y loop 8 \r (no-eol) (esc) loop [=========> ] 2/8 1m07s\r (no-eol) (esc) loop [===============> ] 3/8 56s\r (no-eol) (esc) loop [=====================> ] 4/8 45s\r (no-eol) (esc) loop [==========================> ] 5/8 34s\r (no-eol) (esc) loop [================================> ] 6/8 23s\r (no-eol) (esc) loop [=====================================> ] 7/8 12s\r (no-eol) (esc) \r (no-eol) (esc) $ MOCKTIME=10000 hg -y loop 4 \r (no-eol) (esc) loop [ ] 0/4\r (no-eol) (esc) loop [=========> ] 1/4 8h21m\r (no-eol) (esc) loop [====================> ] 2/4 5h34m\r (no-eol) (esc) loop [==============================> ] 3/4 2h47m\r (no-eol) (esc) \r (no-eol) (esc) $ MOCKTIME=1000000 hg -y loop 4 \r (no-eol) (esc) loop [ ] 0/4\r (no-eol) (esc) loop [=========> ] 1/4 5w00d\r (no-eol) (esc) loop [====================> ] 2/4 3w03d\r (no-eol) (esc) loop [=============================> ] 3/4 11d14h\r (no-eol) (esc) \r (no-eol) (esc) $ MOCKTIME=14000000 hg -y loop 4 \r (no-eol) (esc) loop [ ] 0/4\r (no-eol) (esc) loop [=========> ] 1/4 1y18w\r (no-eol) (esc) loop [===================> ] 2/4 46w03d\r (no-eol) (esc) loop [=============================> ] 3/4 23w02d\r (no-eol) (esc) \r (no-eol) (esc) Non-linear progress: $ MOCKTIME='20 20 20 20 20 20 20 20 20 20 500 500 500 500 500 20 20 20 20 20' hg -y loop 20 \r (no-eol) (esc) loop [=> ] 1/20 6m21s\r (no-eol) (esc) loop [===> ] 2/20 6m01s\r (no-eol) (esc) loop [=====> ] 3/20 5m41s\r (no-eol) (esc) loop [=======> ] 4/20 5m21s\r (no-eol) (esc) loop [=========> ] 5/20 5m01s\r (no-eol) (esc) loop [===========> ] 6/20 4m41s\r (no-eol) (esc) loop [=============> ] 7/20 4m21s\r (no-eol) (esc) loop [===============> ] 8/20 4m01s\r (no-eol) (esc) loop [================> ] 9/20 25m40s\r (no-eol) (esc) loop [===================> ] 10/20 1h06m\r (no-eol) (esc) loop [=====================> ] 11/20 1h13m\r (no-eol) (esc) loop [=======================> ] 12/20 1h07m\r (no-eol) (esc) loop [========================> ] 13/20 58m19s\r (no-eol) (esc) loop [===========================> ] 14/20 7m09s\r (no-eol) (esc) loop [=============================> ] 15/20 3m38s\r (no-eol) (esc) loop [===============================> ] 16/20 2m15s\r (no-eol) (esc) loop [=================================> ] 17/20 1m27s\r (no-eol) (esc) loop [====================================> ] 18/20 52s\r (no-eol) (esc) loop [======================================> ] 19/20 25s\r (no-eol) (esc) \r (no-eol) (esc) Time estimates should not fail when there's no end point: $ MOCKTIME=11 hg -y loop -- -4 \r (no-eol) (esc) loop [ <=> ] 2\r (no-eol) (esc) loop [ <=> ] 3\r (no-eol) (esc) \r (no-eol) (esc) #endif test line trimming by '[progress] width', when progress topic contains multi-byte characters, of which length of byte sequence and columns in display are different from each other. $ cp $HGRCPATH.orig $HGRCPATH $ cat >> $HGRCPATH <<EOF > [extensions] > progress= > loop=`pwd`/loop.py > [progress] > assume-tty = 1 > delay = 0 > refresh = 0 > EOF $ rm -f loop.pyc $ cat >> loop.py <<EOF > # use non-ascii characters as topic label of progress > # 2 x 4 = 8 columns, but 3 x 4 = 12 bytes > topiclabel = u'\u3042\u3044\u3046\u3048'.encode('utf-8') > EOF $ cat >> $HGRCPATH <<EOF > [progress] > format = topic number > width= 12 > EOF $ hg --encoding utf-8 -y loop --total 3 3 \r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88 0/3\r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88 1/3\r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88 2/3\r (no-eol) (esc) \r (no-eol) (esc) test calculation of bar width, when progress topic contains multi-byte characters, of which length of byte sequence and columns in display are different from each other. $ cat >> $HGRCPATH <<EOF > [progress] > format = topic bar > width= 21 > # progwidth should be 9 (= 21 - (8+1) - 3) > EOF $ hg --encoding utf-8 -y loop --total 3 3 \r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88 [ ]\r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88 [==> ]\r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88 [=====> ]\r (no-eol) (esc) \r (no-eol) (esc) test trimming progress items, when they contain multi-byte characters, of which length of byte sequence and columns in display are different from each other. $ rm -f loop.pyc $ rm -Rf __pycache__ $ cat >> loop.py <<EOF > # use non-ascii characters as loop items of progress > loopitems = [ > u'\u3042\u3044'.encode('utf-8'), # 2 x 2 = 4 columns > u'\u3042\u3044\u3046'.encode('utf-8'), # 2 x 3 = 6 columns > u'\u3042\u3044\u3046\u3048'.encode('utf-8'), # 2 x 4 = 8 columns > ] > def getloopitem(i): > return loopitems[i % len(loopitems)] > EOF $ cat >> $HGRCPATH <<EOF > [progress] > # trim at tail side > format = item+6 > EOF $ hg --encoding utf-8 -y loop --total 3 3 \r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84 \r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\r (no-eol) (esc) \r (no-eol) (esc) $ cat >> $HGRCPATH <<EOF > [progress] > # trim at left side > format = item-6 > EOF $ hg --encoding utf-8 -y loop --total 3 3 \r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84 \r (no-eol) (esc) \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\r (no-eol) (esc) \xe3\x81\x84\xe3\x81\x86\xe3\x81\x88\r (no-eol) (esc) \r (no-eol) (esc)