[PATCH 6 of 6 V3] revset: optimize out sort() according to ordering policy
Yuya Nishihara
yuya at tcha.org
Fri Jun 17 10:45:32 EDT 2016
# HG changeset patch
# User Yuya Nishihara <yuya at tcha.org>
# Date 1462250172 -32400
# Tue May 03 13:36:12 2016 +0900
# Node ID e5243019c82eb1c22bdcad3bab6d775d7946a311
# Parent 249e5cd2f9a97e17872e78b6f1e16e709e01bca3
revset: optimize out sort() according to ordering policy
See the previous patch of reverse() for why. Unlike reverse(), we validate
the arguments by _getsortargs().
diff --git a/mercurial/revset.py b/mercurial/revset.py
--- a/mercurial/revset.py
+++ b/mercurial/revset.py
@@ -2517,6 +2517,8 @@ def _optimize(x, small, order):
wa, ta = _optimize(x[2], small, d)
if f == 'reverse' and order != _defineorder:
return wa, ta
+ if f == 'sort' and order != _defineorder:
+ return wa, _getsortargs(ta)[0]
if f in ("author branch closed date desc file grep keyword "
"outgoing user"):
w = 10 # slow
diff --git a/tests/test-revset.t b/tests/test-revset.t
--- a/tests/test-revset.t
+++ b/tests/test-revset.t
@@ -1191,8 +1191,7 @@ Test order of revisions in compound expr
(see hg help "revsets.x or y")
[255]
- BROKEN: sort() will need different workaround since it modifies the flag
- of input set:
+ sort() should be optimized out if order doesn't matter:
$ try --optimize '0:2 & sort(all(), -rev)'
(and
@@ -1214,23 +1213,24 @@ Test order of revisions in compound expr
('symbol', '2')
('string', 'define'))
(func
- ('symbol', '_reorder')
- (func
- ('symbol', 'sort')
- (list
- (func
- ('symbol', 'all')
- None)
- ('string', '-rev')))))
+ ('symbol', 'all')
+ None))
* set:
<filteredset
- <spanset- 0:2>,
- <filteredset
- <spanset- 0:2>,
- <spanset+ 0:9>>>
+ <spanset+ 0:2>,
+ <spanset+ 0:9>>
+ 0
+ 1
2
- 1
- 0
+
+ invalid argument for sort() to be optimized out:
+
+ $ log '0:2 & sort()'
+ hg: parse error: sort requires one or two arguments
+ [255]
+ $ log '0:2 & sort(all(), -invalid)'
+ hg: parse error: unknown sort key '-invalid'
+ [255]
for 'A & f(B)', 'B' should not be affected by the order of 'A':
More information about the Mercurial-devel
mailing list