match: sort patterns before compiling them into a regex
While investigating cripping performance for `hg cat` in some context, I
discovered that, for large inputs, building a regex from out of order patterns
result may result in a *much* slower regex and a much slower associated
matcher's performance.
So we are now sorting the patterns to help the regex engine.
There is more to the story as we rely on regexp more than we should. See the
next changeset for details.
Benchmarks
==========
In the following benchmark we are comparing the `hg cat` and `hg files` run
time when matching against the full list of files in the repository. They are
run:
- without the rust extensions
- with the standard python enfine (so without re2)
sort vs non-sorted - Before this changeset (3f5137543773)
---------------------------------------------------------
###### hg files ###############################################################
### mercurial-2018-08-01-zstd-sparse-revlog
sorted: 0.230092 seconds
shuffled: 0.234235 seconds (+1.80%)
### pypy-2018-08-01-zstd-sparse-revlog
sorted: 0.613567 seconds
shuffled: 0.801880 seconds (+30.69%)
### mozilla-central-2018-08-01-zstd-sparse-revlog
sorted: 62.474221 seconds
shuffled: 1364.180218 seconds (+2083.59%)
### netbeans-2018-08-01-zstd-sparse-revlog
sorted: 21.541828 seconds
shuffled: 172.759857 seconds (+701.97%)
###### hg cat #################################################################
### mercurial-2018-08-01-zstd-sparse-revlog
sorted: 0.764407 seconds
shuffled: 0.768924 seconds
### pypy-2018-08-01-zstd-sparse-revlog
sorted: 2.065220 seconds
shuffled: 2.276388 seconds (+10.22%)
### netbeans-2018-08-01-zstd-sparse-revlog
sorted: 40.967983 seconds
shuffled: 216.388709 seconds (+428.19%)
### mozilla-central-2018-08-01-zstd-sparse-revlog
sorted: 105.228510 seconds
shuffled: 1448.722784 seconds (+1276.74%)
sort vs non-sorted - With this changeset
----------------------------------------
###### hg files ###############################################################
### mercurial-2018-08-01-zstd-sparse-revlog
all-list-pattern-sorted: 0.230069
all-list-pattern-shuffled: 0.231165
### pypy-2018-08-01-zstd-sparse-revlog
all-list-pattern-sorted: 0.616799
all-list-pattern-shuffled: 0.616393
### netbeans-2018-08-01-zstd-sparse-revlog
all-list-pattern-sorted: 21.586773
all-list-pattern-shuffled: 21.908197
### mozilla-central-2018-08-01-zstd-sparse-revlog
all-list-pattern-sorted: 61.279490
all-list-pattern-shuffled: 62.473549
###### hg cat #################################################################
### mercurial-2018-08-01-zstd-sparse-revlog
sorted: 0.763883 seconds
shuffled: 0.765848 seconds
### pypy-2018-08-01-zstd-sparse-revlog
sorted: 2.070498 seconds
shuffled: 2.069197 seconds
### netbeans-2018-08-01-zstd-sparse-revlog
sorted: 41.392423 seconds
shuffled: 41.648689 seconds
### mozilla-central-2018-08-01-zstd-sparse-revlog
sorted: 103.315670 seconds
shuffled: 104.369358 seconds
$ fileset() {
> hg debugfileset --all-files "$@"
> }
$ hg init repo
$ cd repo
$ echo a > a1
$ echo a > a2
$ echo b > b1
$ echo b > b2
$ hg ci -Am addfiles
adding a1
adding a2
adding b1
adding b2
Test operators and basic patterns
$ fileset -v a1
(symbol 'a1')
* matcher:
<patternmatcher patterns='a1$'>
a1
$ fileset -v 'a*'
(symbol 'a*')
* matcher:
<patternmatcher patterns='a[^/]*$'>
a1
a2
$ fileset -v '"re:a\d"'
(string 're:a\\d')
* matcher:
<patternmatcher patterns='a\\d'>
a1
a2
$ fileset -v '!re:"a\d"'
(not
(kindpat
(symbol 're')
(string 'a\\d')))
* matcher:
<predicatenmatcher
pred=<not
<patternmatcher patterns='a\\d'>>>
b1
b2
$ fileset -v 'path:a1 or glob:b?'
(or
(kindpat
(symbol 'path')
(symbol 'a1'))
(kindpat
(symbol 'glob')
(symbol 'b?')))
* matcher:
<patternmatcher patterns='b.$|a1(?:/|$)'>
a1
b1
b2
$ fileset -v --no-show-matcher 'a1 or a2'
(or
(symbol 'a1')
(symbol 'a2'))
a1
a2
$ fileset 'a1 | a2'
a1
a2
$ fileset 'a* and "*1"'
a1
$ fileset 'a* & "*1"'
a1
$ fileset 'not (r"a*")'
b1
b2
$ fileset '! ("a*")'
b1
b2
$ fileset 'a* - a1'
a2
$ fileset 'a_b'
$ fileset '"\xy"'
hg: parse error: invalid \x escape* (glob)
[10]
Test invalid syntax
$ fileset -v '"added"()'
(func
(string 'added')
None)
hg: parse error: not a symbol
[10]
$ fileset -v '()()'
(func
(group
None)
None)
hg: parse error: not a symbol
[10]
$ fileset -v -- '-x'
(negate
(symbol 'x'))
hg: parse error: can't use negate operator in this context
[10]
$ fileset -v -- '-()'
(negate
(group
None))
hg: parse error: can't use negate operator in this context
[10]
$ fileset -p parsed 'a, b, c'
* parsed:
(list
(symbol 'a')
(symbol 'b')
(symbol 'c'))
hg: parse error: can't use a list in this context
(see 'hg help "filesets.x or y"')
[10]
$ fileset '"path":.'
hg: parse error: not a symbol
[10]
$ fileset 'path:foo bar'
hg: parse error at 9: invalid token
[10]
$ fileset 'foo:bar:baz'
hg: parse error: not a symbol
[10]
$ fileset 'foo:bar()'
hg: parse error: pattern must be a string
[10]
$ fileset 'foo:bar'
hg: parse error: invalid pattern kind: foo
[10]
Show parsed tree at stages:
$ fileset -p unknown a
abort: invalid stage name: unknown
[255]
$ fileset -p parsed 'path:a1 or glob:b?'
* parsed:
(or
(kindpat
(symbol 'path')
(symbol 'a1'))
(kindpat
(symbol 'glob')
(symbol 'b?')))
a1
b1
b2
$ fileset -p all -s 'a1 or a2 or (grep("b") & clean())'
* parsed:
(or
(symbol 'a1')
(symbol 'a2')
(group
(and
(func
(symbol 'grep')
(string 'b'))
(func
(symbol 'clean')
None))))
* analyzed:
(or
(symbol 'a1')
(symbol 'a2')
(and
(func
(symbol 'grep')
(string 'b'))
(withstatus
(func
(symbol 'clean')
None)
(string 'clean'))))
* optimized:
(or
(patterns
(symbol 'a1')
(symbol 'a2'))
(and
(withstatus
(func
(symbol 'clean')
None)
(string 'clean'))
(func
(symbol 'grep')
(string 'b'))))
* matcher:
<unionmatcher matchers=[
<patternmatcher patterns='a1$|a2$'>,
<intersectionmatcher
m1=<predicatenmatcher pred=clean>,
m2=<predicatenmatcher pred=grep('b')>>]>
a1
a2
b1
b2
Union of basic patterns:
$ fileset -p optimized -s -r. 'a1 or a2 or path:b1'
* optimized:
(patterns
(symbol 'a1')
(symbol 'a2')
(kindpat
(symbol 'path')
(symbol 'b1')))
* matcher:
<patternmatcher patterns='a1$|a2$|b1(?:/|$)'>
a1
a2
b1
OR expression should be reordered by weight:
$ fileset -p optimized -s -r. 'grep("a") or a1 or grep("b") or b2'
* optimized:
(or
(patterns
(symbol 'a1')
(symbol 'b2'))
(func
(symbol 'grep')
(string 'a'))
(func
(symbol 'grep')
(string 'b')))
* matcher:
<unionmatcher matchers=[
<patternmatcher patterns='a1$|b2$'>,
<predicatenmatcher pred=grep('a')>,
<predicatenmatcher pred=grep('b')>]>
a1
a2
b1
b2
Use differencematcher for 'x and not y':
$ fileset -p optimized -s 'a* and not a1'
* optimized:
(minus
(symbol 'a*')
(symbol 'a1'))
* matcher:
<differencematcher
m1=<patternmatcher patterns='a[^/]*$'>,
m2=<patternmatcher patterns='a1$'>>
a2
$ fileset -p optimized -s '!binary() and a*'
* optimized:
(minus
(symbol 'a*')
(func
(symbol 'binary')
None))
* matcher:
<differencematcher
m1=<patternmatcher patterns='a[^/]*$'>,
m2=<predicatenmatcher pred=binary>>
a1
a2
'x - y' is rewritten to 'x and not y' first so the operands can be reordered:
$ fileset -p analyzed -p optimized -s 'a* - a1'
* analyzed:
(and
(symbol 'a*')
(not
(symbol 'a1')))
* optimized:
(minus
(symbol 'a*')
(symbol 'a1'))
* matcher:
<differencematcher
m1=<patternmatcher patterns='a[^/]*$'>,
m2=<patternmatcher patterns='a1$'>>
a2
$ fileset -p analyzed -p optimized -s 'binary() - a*'
* analyzed:
(and
(func
(symbol 'binary')
None)
(not
(symbol 'a*')))
* optimized:
(and
(not
(symbol 'a*'))
(func
(symbol 'binary')
None))
* matcher:
<intersectionmatcher
m1=<predicatenmatcher
pred=<not
<patternmatcher patterns='a[^/]*$'>>>,
m2=<predicatenmatcher pred=binary>>
Test files status
$ rm a1
$ hg rm a2
$ echo b >> b2
$ hg cp b1 c1
$ echo c > c2
$ echo c > c3
$ cat > .hgignore <<EOF
> \.hgignore
> 2$
> EOF
$ fileset 'modified()'
b2
$ fileset 'added()'
c1
$ fileset 'removed()'
a2
$ fileset 'deleted()'
a1
$ fileset 'missing()'
a1
$ fileset 'unknown()'
c3
$ fileset 'ignored()'
.hgignore
c2
$ fileset 'hgignore()'
.hgignore
a2
b2
c2
$ fileset 'clean()'
b1
$ fileset 'copied()'
c1
Test files status in different revisions
$ hg status -m
M b2
$ fileset -r0 'revs("wdir()", modified())' --traceback
b2
$ hg status -a
A c1
$ fileset -r0 'revs("wdir()", added())'
c1
$ hg status --change 0 -a
A a1
A a2
A b1
A b2
$ hg status -mru
M b2
R a2
? c3
$ fileset -r0 'added() and revs("wdir()", modified() or removed() or unknown())'
a2
b2
$ fileset -r0 'added() or revs("wdir()", added())'
a1
a2
b1
b2
c1
Test insertion of status hints
$ fileset -p optimized 'added()'
* optimized:
(withstatus
(func
(symbol 'added')
None)
(string 'added'))
c1
$ fileset -p optimized 'a* & removed()'
* optimized:
(and
(symbol 'a*')
(withstatus
(func
(symbol 'removed')
None)
(string 'removed')))
a2
$ fileset -p optimized 'a* - removed()'
* optimized:
(minus
(symbol 'a*')
(withstatus
(func
(symbol 'removed')
None)
(string 'removed')))
a1
$ fileset -p analyzed -p optimized '(added() + removed()) - a*'
* analyzed:
(and
(withstatus
(or
(func
(symbol 'added')
None)
(func
(symbol 'removed')
None))
(string 'added removed'))
(not
(symbol 'a*')))
* optimized:
(and
(not
(symbol 'a*'))
(withstatus
(or
(func
(symbol 'added')
None)
(func
(symbol 'removed')
None))
(string 'added removed')))
c1
$ fileset -p optimized 'a* + b* + added() + unknown()'
* optimized:
(withstatus
(or
(patterns
(symbol 'a*')
(symbol 'b*'))
(func
(symbol 'added')
None)
(func
(symbol 'unknown')
None))
(string 'added unknown'))
a1
a2
b1
b2
c1
c3
$ fileset -p analyzed -p optimized 'removed() & missing() & a*'
* analyzed:
(and
(withstatus
(and
(func
(symbol 'removed')
None)
(func
(symbol 'missing')
None))
(string 'removed missing'))
(symbol 'a*'))
* optimized:
(and
(symbol 'a*')
(withstatus
(and
(func
(symbol 'removed')
None)
(func
(symbol 'missing')
None))
(string 'removed missing')))
$ fileset -p optimized 'clean() & revs(0, added())'
* optimized:
(and
(withstatus
(func
(symbol 'clean')
None)
(string 'clean'))
(func
(symbol 'revs')
(list
(symbol '0')
(withstatus
(func
(symbol 'added')
None)
(string 'added')))))
b1
$ fileset -p optimized 'clean() & status(null, 0, b* & added())'
* optimized:
(and
(withstatus
(func
(symbol 'clean')
None)
(string 'clean'))
(func
(symbol 'status')
(list
(symbol 'null')
(symbol '0')
(and
(symbol 'b*')
(withstatus
(func
(symbol 'added')
None)
(string 'added'))))))
b1
Test files properties
>>> open('bin', 'wb').write(b'\0a') and None
$ fileset 'binary()'
bin
$ fileset 'binary() and unknown()'
bin
$ echo '^bin$' >> .hgignore
$ fileset 'binary() and ignored()'
bin
$ hg add bin
$ fileset 'binary()'
bin
$ fileset -p optimized -s 'binary() and b*'
* optimized:
(and
(symbol 'b*')
(func
(symbol 'binary')
None))
* matcher:
<intersectionmatcher
m1=<patternmatcher patterns='b[^/]*$'>,
m2=<predicatenmatcher pred=binary>>
bin
$ fileset 'grep("b{1}")'
.hgignore
b1
b2
c1
$ fileset 'grep("missingparens(")'
hg: parse error: invalid match pattern: (unbalanced parenthesis|missing \)).* (re)
[10]
#if execbit
$ chmod +x b2
$ fileset 'exec()'
b2
#endif
#if symlink
$ ln -s b2 b2link
$ fileset 'symlink() and unknown()'
b2link
$ hg add b2link
#endif
#if no-windows
$ echo foo > con.xml
$ fileset 'not portable()'
con.xml
$ hg --config ui.portablefilenames=ignore add con.xml
#endif
>>> open('1k', 'wb').write(b' '*1024) and None
>>> open('2k', 'wb').write(b' '*2048) and None
$ hg add 1k 2k
$ fileset 'size("bar")'
hg: parse error: couldn't parse size: bar
[10]
$ fileset '(1k, 2k)'
hg: parse error: can't use a list in this context
(see 'hg help "filesets.x or y"')
[10]
$ fileset 'size(1k)'
1k
$ fileset '(1k or 2k) and size("< 2k")'
1k
$ fileset '(1k or 2k) and size("<=2k")'
1k
2k
$ fileset '(1k or 2k) and size("> 1k")'
2k
$ fileset '(1k or 2k) and size(">=1K")'
1k
2k
$ fileset '(1k or 2k) and size(".5KB - 1.5kB")'
1k
$ fileset 'size("1M")'
$ fileset 'size("1 GB")'
Test merge states
$ hg ci -m manychanges
$ hg file -r . 'set:copied() & modified()'
[1]
$ hg up -C 0
* files updated, 0 files merged, * files removed, 0 files unresolved (glob)
$ echo c >> b2
$ hg ci -m diverging b2
created new head
$ fileset 'resolved()'
$ fileset 'unresolved()'
$ hg merge
merging b2
warning: conflicts while merging b2! (edit, then use 'hg resolve --mark')
* files updated, 0 files merged, 1 files removed, 1 files unresolved (glob)
use 'hg resolve' to retry unresolved file merges or 'hg merge --abort' to abandon
[1]
$ fileset 'resolved()'
$ fileset 'unresolved()'
b2
$ echo e > b2
$ hg resolve -m b2
(no more unresolved files)
$ fileset 'resolved()'
b2
$ fileset 'unresolved()'
$ hg ci -m merge
Test subrepo predicate
$ hg init sub
$ echo a > sub/suba
$ hg -R sub add sub/suba
$ hg -R sub ci -m sub
$ echo 'sub = sub' > .hgsub
$ hg init sub2
$ echo b > sub2/b
$ hg -R sub2 ci -Am sub2
adding b
$ echo 'sub2 = sub2' >> .hgsub
$ fileset 'subrepo()'
$ hg add .hgsub
$ fileset 'subrepo()'
sub
sub2
$ fileset 'subrepo("sub")'
sub
$ fileset 'subrepo("glob:*")'
sub
sub2
$ hg ci -m subrepo
Test that .hgsubstate is updated as appropriate during a conversion. The
saverev property is enough to alter the hashes of the subrepo.
$ hg init ../converted
$ hg --config extensions.convert= convert --config convert.hg.saverev=True \
> sub ../converted/sub
initializing destination ../converted/sub repository
scanning source...
sorting...
converting...
0 sub
$ hg clone -U sub2 ../converted/sub2
$ hg --config extensions.convert= convert --config convert.hg.saverev=True \
> . ../converted
scanning source...
sorting...
converting...
4 addfiles
3 manychanges
2 diverging
1 merge
0 subrepo
no ".hgsubstate" updates will be made for "sub2"
$ hg up -q -R ../converted -r tip
$ hg --cwd ../converted cat sub/suba sub2/b -r tip
a
b
$ oldnode=`hg log -r tip -T "{node}\n"`
$ newnode=`hg log -R ../converted -r tip -T "{node}\n"`
$ [ "$oldnode" != "$newnode" ] || echo "nothing changed"
Test with a revision
$ hg log -G --template '{rev} {desc}\n'
@ 4 subrepo
|
o 3 merge
|\
| o 2 diverging
| |
o | 1 manychanges
|/
o 0 addfiles
$ echo unknown > unknown
$ fileset -r1 'modified()'
b2
$ fileset -r1 'added() and c1'
c1
$ fileset -r1 'removed()'
a2
$ fileset -r1 'deleted()'
$ fileset -r1 'unknown()'
$ fileset -r1 'ignored()'
$ fileset -r1 'hgignore()'
.hgignore
a2
b2
bin
c2
sub2
$ fileset -r1 'binary()'
bin
$ fileset -r1 'size(1k)'
1k
$ fileset -r3 'resolved()'
$ fileset -r3 'unresolved()'
#if execbit
$ fileset -r1 'exec()'
b2
#endif
#if symlink
$ fileset -r1 'symlink()'
b2link
#endif
#if no-windows
$ fileset -r1 'not portable()'
con.xml
$ hg forget 'con.xml'
#endif
$ fileset -r4 'subrepo("re:su.*")'
sub
sub2
$ fileset -r4 'subrepo(re:su.*)'
sub
sub2
$ fileset -r4 'subrepo("sub")'
sub
$ fileset -r4 'b2 or c1'
b2
c1
>>> open('dos', 'wb').write(b"dos\r\n") and None
>>> open('mixed', 'wb').write(b"dos\r\nunix\n") and None
>>> open('mac', 'wb').write(b"mac\r") and None
$ hg add dos mixed mac
(remove a1, to examine safety of 'eol' on removed files)
$ rm a1
$ fileset 'eol(dos)'
dos
mixed
$ fileset 'eol(unix)'
.hgignore
.hgsub
.hgsubstate
b1
b2
b2.orig
c1
c2
c3
con.xml (no-windows !)
mixed
unknown
$ fileset 'eol(mac)'
mac
Test safety of 'encoding' on removed files
$ fileset 'encoding("ascii")'
.hgignore
.hgsub
.hgsubstate
1k
2k
b1
b2
b2.orig
b2link (symlink !)
bin
c1
c2
c3
con.xml (no-windows !)
dos
mac
mixed
unknown
Test 'revs(...)'
================
small reminder of the repository state
$ hg log -G
@ changeset: 4:* (glob)
| tag: tip
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: subrepo
|
o changeset: 3:* (glob)
|\ parent: 2:55b05bdebf36
| | parent: 1:* (glob)
| | user: test
| | date: Thu Jan 01 00:00:00 1970 +0000
| | summary: merge
| |
| o changeset: 2:55b05bdebf36
| | parent: 0:8a9576c51c1f
| | user: test
| | date: Thu Jan 01 00:00:00 1970 +0000
| | summary: diverging
| |
o | changeset: 1:* (glob)
|/ user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: manychanges
|
o changeset: 0:8a9576c51c1f
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: addfiles
$ hg status --change 0
A a1
A a2
A b1
A b2
$ hg status --change 1
M b2
A 1k
A 2k
A b2link (symlink !)
A bin
A c1
A con.xml (no-windows !)
R a2
$ hg status --change 2
M b2
$ hg status --change 3
M b2
A 1k
A 2k
A b2link (symlink !)
A bin
A c1
A con.xml (no-windows !)
R a2
$ hg status --change 4
A .hgsub
A .hgsubstate
$ hg status
A dos
A mac
A mixed
R con.xml (no-windows !)
! a1
? b2.orig
? c3
? unknown
Test files at -r0 should be filtered by files at wdir
-----------------------------------------------------
$ fileset -r0 'tracked() and revs("wdir()", tracked())'
a1
b1
b2
Test that "revs()" work at all
------------------------------
$ fileset "revs('2', modified())"
b2
Test that "revs()" work for file missing in the working copy/current context
----------------------------------------------------------------------------
(a2 not in working copy)
$ fileset "revs('0', added())"
a1
a2
b1
b2
(none of the file exist in "0")
$ fileset -r 0 "revs('4', added())"
.hgsub
.hgsubstate
Call with empty revset
--------------------------
$ fileset "revs('2-2', modified())"
Call with revset matching multiple revs
---------------------------------------
$ fileset "revs('0+4', added())"
.hgsub
.hgsubstate
a1
a2
b1
b2
overlapping set
$ fileset "revs('1+2', modified())"
b2
test 'status(...)'
=================
Simple case
-----------
$ fileset "status(3, 4, added())"
.hgsub
.hgsubstate
use rev to restrict matched file
-----------------------------------------
$ hg status --removed --rev 0 --rev 1
R a2
$ fileset "status(0, 1, removed())"
a2
$ fileset "tracked() and status(0, 1, removed())"
$ fileset -r 4 "status(0, 1, removed())"
a2
$ fileset -r 4 "tracked() and status(0, 1, removed())"
$ fileset "revs('4', tracked() and status(0, 1, removed()))"
$ fileset "revs('0', tracked() and status(0, 1, removed()))"
a2
check wdir()
------------
$ hg status --removed --rev 4
R con.xml (no-windows !)
$ fileset "status(4, 'wdir()', removed())"
con.xml (no-windows !)
$ hg status --removed --rev 2
R a2
$ fileset "status('2', 'wdir()', removed())"
a2
test backward status
--------------------
$ hg status --removed --rev 0 --rev 4
R a2
$ hg status --added --rev 4 --rev 0
A a2
$ fileset "status(4, 0, added())"
a2
test cross branch status
------------------------
$ hg status --added --rev 1 --rev 2
A a2
$ fileset "status(1, 2, added())"
a2
test with multi revs revset
---------------------------
$ hg status --added --rev 0:1 --rev 3:4
A .hgsub
A .hgsubstate
A 1k
A 2k
A b2link (symlink !)
A bin
A c1
A con.xml (no-windows !)
$ fileset "status('0:1', '3:4', added())"
.hgsub
.hgsubstate
1k
2k
b2link (symlink !)
bin
c1
con.xml (no-windows !)
tests with empty value
----------------------
Fully empty revset
$ fileset "status('', '4', added())"
hg: parse error: first argument to status must be a revision
[10]
$ fileset "status('2', '', added())"
hg: parse error: second argument to status must be a revision
[10]
Empty revset will error at the revset layer
$ fileset "status(' ', '4', added())"
hg: parse error at 1: not a prefix: end
(
^ here)
[10]
$ fileset "status('2', ' ', added())"
hg: parse error at 1: not a prefix: end
(
^ here)
[10]