diff: cache textconv output

Running a textconv filter can take a long time. It's
particularly bad for a large file which needs to be spooled
to disk, but even for small files, the fork+exec overhead
can add up for something like "git log -p".

This patch uses the notes-cache mechanism to keep a fast
cache of textconv output. Caches are stored in
refs/notes/textconv/$x, where $x is the userdiff driver
defined in gitattributes.

Caching is enabled only if diff.$x.cachetextconv is true.

In my test repo, on a commit with 45 jpg and avi files
changed and a textconv to show their exif tags:

  [before]
  $ time git show >/dev/null
  real    0m13.724s
  user    0m12.057s
  sys     0m1.624s

  [after, first run]
  $ git config diff.mfo.cachetextconv true
  $ time git show >/dev/null
  real    0m14.252s
  user    0m12.197s
  sys     0m1.800s

  [after, subsequent runs]
  $ time git show >/dev/null
  real    0m0.352s
  user    0m0.148s
  sys     0m0.200s

So for a slight (3.8%) cost on the first run, we achieve an
almost 40x speed up on subsequent runs.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

1 files changed, 109 insertions, 0 deletions

diff --git a/t/t4042-diff-textconv-caching.sh b/t/t4042-diff-textconv-caching.sh
new file mode 100755
index 0000000..91f8198
--- /dev/null
+++ b/t/t4042-diff-textconv-caching.sh
@@ -0,0 +1,109 @@
+#!/bin/sh
+
+test_description='test textconv caching'
+. ./test-lib.sh
+
+cat >helper <<'EOF'
+#!/bin/sh
+sed 's/^/converted: /' "$@" >helper.out
+cat helper.out
+EOF
+chmod +x helper
+
+test_expect_success 'setup' '
+	echo foo content 1 >foo.bin &&
+	echo bar content 1 >bar.bin &&
+	git add . &&
+	git commit -m one &&
+	echo foo content 2 >foo.bin &&
+	echo bar content 2 >bar.bin &&
+	git commit -a -m two &&
+	echo "*.bin diff=magic" >.gitattributes &&
+	git config diff.magic.textconv ./helper &&
+	git config diff.magic.cachetextconv true
+'
+
+cat >expect <<EOF
+diff --git a/bar.bin b/bar.bin
+index fcf9166..28283d5 100644
+--- a/bar.bin
++++ b/bar.bin
+@@ -1 +1 @@
+-converted: bar content 1
++converted: bar content 2
+diff --git a/foo.bin b/foo.bin
+index d5b9fe3..1345db2 100644
+--- a/foo.bin
++++ b/foo.bin
+@@ -1 +1 @@
+-converted: foo content 1
++converted: foo content 2
+EOF
+
+test_expect_success 'first textconv works' '
+	git diff HEAD^ HEAD >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success 'cached textconv produces same output' '
+	git diff HEAD^ HEAD >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success 'cached textconv does not run helper' '
+	rm -f helper.out &&
+	git diff HEAD^ HEAD >actual &&
+	test_cmp expect actual &&
+	! test -r helper.out
+'
+
+cat >expect <<EOF
+diff --git a/bar.bin b/bar.bin
+index fcf9166..28283d5 100644
+--- a/bar.bin
++++ b/bar.bin
+@@ -1,2 +1,2 @@
+ converted: other
+-converted: bar content 1
++converted: bar content 2
+diff --git a/foo.bin b/foo.bin
+index d5b9fe3..1345db2 100644
+--- a/foo.bin
++++ b/foo.bin
+@@ -1,2 +1,2 @@
+ converted: other
+-converted: foo content 1
++converted: foo content 2
+EOF
+test_expect_success 'changing textconv invalidates cache' '
+	echo other >other &&
+	git config diff.magic.textconv "./helper other" &&
+	git diff HEAD^ HEAD >actual &&
+	test_cmp expect actual
+'
+
+cat >expect <<EOF
+diff --git a/bar.bin b/bar.bin
+index fcf9166..28283d5 100644
+--- a/bar.bin
++++ b/bar.bin
+@@ -1,2 +1,2 @@
+ converted: other
+-converted: bar content 1
++converted: bar content 2
+diff --git a/foo.bin b/foo.bin
+index d5b9fe3..1345db2 100644
+--- a/foo.bin
++++ b/foo.bin
+@@ -1 +1 @@
+-converted: foo content 1
++converted: foo content 2
+EOF
+test_expect_success 'switching diff driver produces correct results' '
+	git config diff.moremagic.textconv ./helper &&
+	echo foo.bin diff=moremagic >>.gitattributes &&
+	git diff HEAD^ HEAD >actual &&
+	test_cmp expect actual
+'
+
+test_done