bdiff: balance recursion to avoid quadratic behavior (issue4704)
For highly structured files like JSON or XML dumps with large numbers
of duplicate lines (eg braces) and isolated matching lines, bdiff
could find large numbers of equally good spans. Because it prefers
earlier matches, this would result in pathologically unbalance
recursion that resulted in quadratic performance.
This patch makes it prefer matches closer to the middle that tend to
balance recursion. This change improves the speed of a pathological
test case from 1100s to 9s.
Included is a smaller test that has a roughly 50x safety margin on the
performance it accepts. It's likely to fail on pure builds because
difflib also has a recursion-balancing problem.
Testing diff --change
$ hg init a
$ cd a
$ echo "first" > file.txt
$ hg add file.txt
$ hg commit -m 'first commit' # 0
$ echo "second" > file.txt
$ hg commit -m 'second commit' # 1
$ echo "third" > file.txt
$ hg commit -m 'third commit' # 2
$ hg diff --nodates --change 1
diff -r 4bb65dda5db4 -r e9b286083166 file.txt
--- a/file.txt
+++ b/file.txt
@@ -1,1 +1,1 @@
-first
+second
$ hg diff --change e9b286083166
diff -r 4bb65dda5db4 -r e9b286083166 file.txt
--- a/file.txt Thu Jan 01 00:00:00 1970 +0000
+++ b/file.txt Thu Jan 01 00:00:00 1970 +0000
@@ -1,1 +1,1 @@
-first
+second
$ cd ..
Test dumb revspecs: top-level "x:y", "x:", ":y" and ":" ranges should be handled
as pairs even if x == y, but not for "f(x:y)" nor "x::y" (issue3474, issue4774)
$ hg clone -q a dumbspec
$ cd dumbspec
$ echo "wdir" > file.txt
$ hg diff -r 2:2
$ hg diff -r 2:.
$ hg diff -r 2:
$ hg diff -r :0
$ hg diff -r '2:first(2:2)'
$ hg diff -r 'first(2:2)' --nodates
diff -r bf5ff72eb7e0 file.txt
--- a/file.txt
+++ b/file.txt
@@ -1,1 +1,1 @@
-third
+wdir
$ hg diff -r 2::2 --nodates
diff -r bf5ff72eb7e0 file.txt
--- a/file.txt
+++ b/file.txt
@@ -1,1 +1,1 @@
-third
+wdir
$ hg diff -r "2 and 1"
abort: empty revision range
[255]
$ cd ..
$ hg clone -qr0 a dumbspec-rev0
$ cd dumbspec-rev0
$ echo "wdir" > file.txt
$ hg diff -r :
$ hg diff -r 'first(:)' --nodates
diff -r 4bb65dda5db4 file.txt
--- a/file.txt
+++ b/file.txt
@@ -1,1 +1,1 @@
-first
+wdir
$ cd ..
Testing diff --change when merge:
$ cd a
$ for i in 1 2 3 4 5 6 7 8 9 10; do
> echo $i >> file.txt
> done
$ hg commit -m "lots of text" # 3
$ sed -e 's,^2$,x,' file.txt > file.txt.tmp
$ mv file.txt.tmp file.txt
$ hg commit -m "change 2 to x" # 4
$ hg up -r 3
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ sed -e 's,^8$,y,' file.txt > file.txt.tmp
$ mv file.txt.tmp file.txt
$ hg commit -m "change 8 to y"
created new head
$ hg up -C -r 4
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ hg merge -r 5
merging file.txt
0 files updated, 1 files merged, 0 files removed, 0 files unresolved
(branch merge, don't forget to commit)
$ hg commit -m "merge 8 to y" # 6
$ hg diff --change 5
diff -r ae119d680c82 -r 9085c5c02e52 file.txt
--- a/file.txt Thu Jan 01 00:00:00 1970 +0000
+++ b/file.txt Thu Jan 01 00:00:00 1970 +0000
@@ -6,6 +6,6 @@
5
6
7
-8
+y
9
10
must be similar to 'hg diff --change 5':
$ hg diff -c 6
diff -r 273b50f17c6d -r 979ca961fd2e file.txt
--- a/file.txt Thu Jan 01 00:00:00 1970 +0000
+++ b/file.txt Thu Jan 01 00:00:00 1970 +0000
@@ -6,6 +6,6 @@
5
6
7
-8
+y
9
10
$ cd ..