sparse-revlog: implement algorithm to write sparse delta chains (
issue5480)
The classic behavior of revlog._isgooddeltainfo is to consider the span size
of the whole delta chain, and limit it to 4 * textlen.
Once sparse-revlog writing is allowed (and enforced with a requirement),
revlog._isgooddeltainfo considers the span of the largest chunk as the
distance used in the verification, instead of using the span of the whole
delta chain.
In order to compute the span of the largest chunk, we need to slice into
chunks a chain with the new revision at the top of the revlog, and take the
maximal span of these chunks. The sparse read density is a parameter to the
slicing, as it will stop when the global read density reaches this threshold.
For instance, a density of 50% means that 2 of 4 read bytes are actually used
for the reconstruction of the revision (the others are part of other chains).
This allows a new revision to be potentially stored with a diff against
another revision anywhere in the history, instead of forcing it in the last 4
* textlen. The result is a much better compression on repositories that have
many concurrent branches. Here are a comparison between using deltas from
current upstream (aggressive-merge-deltas on by default) and deltas from a
sparse-revlog
Comparison of `.hg/store/` size:
mercurial (6.74% merges):
before: 46,831,873 bytes
after: 46,795,992 bytes (no relevant change)
pypy (8.30% merges):
before: 333,524,651 bytes
after: 308,417,511 bytes -8%
netbeans (34.21% merges):
before: 1,141,847,554 bytes
after: 1,131,093,161 bytes -1%
mozilla-central (4.84% merges):
before: 2,344,248,850 bytes
after: 2,328,459,258 bytes -1%
large-private-repo-A (merge 19.73%)
before: 41,510,550,163 bytes
after: 8,121,763,428 bytes -80%
large-private-repo-B (23.77%)
before: 58,702,221,709 bytes
after: 8,351,588,828 bytes -76%
Comparison of `00manifest.d` size:
mercurial (6.74% merges):
before: 6,143,044 bytes
after: 6,107,163 bytes
pypy (8.30% merges):
before: 52,941,780 bytes
after: 27,834,082 bytes -48%
netbeans (34.21% merges):
before: 130,088,982 bytes
after: 119,337,636 bytes -10%
mozilla-central (4.84% merges):
before: 215,096,339 bytes
after: 199,496,863 bytes -8%
large-private-repo-A (merge 19.73%)
before: 33,725,285,081 bytes
after: 390,302,545 bytes -99%
large-private-repo-B (23.77%)
before: 49,457,701,645 bytes
after: 1,366,752,187 bytes -97%
The better delta chains provide a performance boost in relevant repositories:
pypy, bundling 1000 revisions:
before: 1.670s
after: 1.149s -31%
Unbundling got a bit slower. probably because the sparse algorithm is still
pure
python.
pypy, unbundling 1000 revisions:
before: 4.062s
after: 4.507s +10%
Performance of bundle/unbundle in repository with few concurrent branches (eg:
mercurial) are unaffected.
No significant differences have been noticed then timing `hg push` and `hg
pull` locally. More state timings are being gathered.
Same as for aggressive-merge-delta, better delta comes with longer delta
chains. Longer chains have a performance impact. For example. The length of
the chain needed to get the manifest of pypy's tip moves from 82 item to 1929
items. This moves the restore time from 3.88ms to 11.3ms.
Delta chain length is an independent issue that affects repository without
this changes. It will be dealt with independently.
No significant differences have been observed on repositories where
`sparse-revlog` have not much effect (mercurial, unity, netbeans). On pypy,
small differences have been observed on some operation affected by delta chain
building and retrieval.
pypy, perfmanifest
before: 0.006162s
after: 0.017899s +190%
pypy, commit:
before: 0.382
after: 0.376 -1%
pypy, status:
before: 0.157
after: 0.168 +7%
More comprehensive and stable timing comparisons are in progress.
#testcases obsstore-on obsstore-off
$ cat > $TESTTMP/editor.py <<EOF
> #!$PYTHON
> import os
> import sys
> path = os.path.join(os.environ['TESTTMP'], 'messages')
> messages = open(path).read().split('--\n')
> prompt = open(sys.argv[1]).read()
> sys.stdout.write(''.join('EDITOR: %s' % l for l in prompt.splitlines(True)))
> sys.stdout.flush()
> with open(sys.argv[1], 'w') as f:
> f.write(messages[0])
> with open(path, 'w') as f:
> f.write('--\n'.join(messages[1:]))
> EOF
$ cat >> $HGRCPATH <<EOF
> [extensions]
> drawdag=$TESTDIR/drawdag.py
> split=
> [ui]
> interactive=1
> color=no
> paginate=never
> [diff]
> git=1
> unified=0
> [alias]
> glog=log -G -T '{rev}:{node|short} {desc} {bookmarks}\n'
> EOF
#if obsstore-on
$ cat >> $HGRCPATH <<EOF
> [experimental]
> evolution=all
> EOF
#endif
$ hg init a
$ cd a
Nothing to split
$ hg split
nothing to split
[1]
$ hg commit -m empty --config ui.allowemptycommit=1
$ hg split
abort: cannot split an empty revision
[255]
$ rm -rf .hg
$ hg init
Cannot split working directory
$ hg split -r 'wdir()'
abort: cannot split working directory
[255]
Generate some content. The sed filter drop CR on Windows, which is dropped in
the a > b line.
$ $TESTDIR/seq.py 1 5 | sed 's/\r$//' >> a
$ hg ci -m a1 -A a -q
$ hg bookmark -i r1
$ sed 's/1/11/;s/3/33/;s/5/55/' a > b
$ mv b a
$ hg ci -m a2 -q
$ hg bookmark -i r2
Cannot split a public changeset
$ hg phase --public -r 'all()'
$ hg split .
abort: cannot split public changeset
(see 'hg help phases' for details)
[255]
$ hg phase --draft -f -r 'all()'
Cannot split while working directory is dirty
$ touch dirty
$ hg add dirty
$ hg split .
abort: uncommitted changes
[255]
$ hg forget dirty
$ rm dirty
Make a clean directory for future tests to build off of
$ cp -R . ../clean
Split a head
$ hg bookmark r3
$ hg split 'all()'
abort: cannot split multiple revisions
[255]
$ runsplit() {
> cat > $TESTTMP/messages <<EOF
> split 1
> --
> split 2
> --
> split 3
> EOF
> cat <<EOF | hg split "$@"
> y
> y
> y
> y
> y
> y
> EOF
> }
$ HGEDITOR=false runsplit
diff --git a/a b/a
1 hunks, 1 lines changed
examine changes to 'a'? [Ynesfdaq?] y
@@ -5,1 +5,1 @@ 4
-5
+55
record this change to 'a'? [Ynesfdaq?] y
transaction abort!
rollback completed
abort: edit failed: false exited with status 1
[255]
$ hg status
$ HGEDITOR="\"$PYTHON\" $TESTTMP/editor.py"
$ runsplit
diff --git a/a b/a
1 hunks, 1 lines changed
examine changes to 'a'? [Ynesfdaq?] y
@@ -5,1 +5,1 @@ 4
-5
+55
record this change to 'a'? [Ynesfdaq?] y
EDITOR: HG: Splitting 1df0d5c5a3ab. Write commit message for the first split changeset.
EDITOR: a2
EDITOR:
EDITOR:
EDITOR: HG: Enter commit message. Lines beginning with 'HG:' are removed.
EDITOR: HG: Leave message empty to abort commit.
EDITOR: HG: --
EDITOR: HG: user: test
EDITOR: HG: branch 'default'
EDITOR: HG: changed a
created new head
diff --git a/a b/a
1 hunks, 1 lines changed
examine changes to 'a'? [Ynesfdaq?] y
@@ -3,1 +3,1 @@ 2
-3
+33
record this change to 'a'? [Ynesfdaq?] y
EDITOR: HG: Splitting 1df0d5c5a3ab. So far it has been split into:
EDITOR: HG: - e704349bd21b: split 1
EDITOR: HG: Write commit message for the next split changeset.
EDITOR: a2
EDITOR:
EDITOR:
EDITOR: HG: Enter commit message. Lines beginning with 'HG:' are removed.
EDITOR: HG: Leave message empty to abort commit.
EDITOR: HG: --
EDITOR: HG: user: test
EDITOR: HG: branch 'default'
EDITOR: HG: changed a
diff --git a/a b/a
1 hunks, 1 lines changed
examine changes to 'a'? [Ynesfdaq?] y
@@ -1,1 +1,1 @@
-1
+11
record this change to 'a'? [Ynesfdaq?] y
EDITOR: HG: Splitting 1df0d5c5a3ab. So far it has been split into:
EDITOR: HG: - e704349bd21b: split 1
EDITOR: HG: - a09ad58faae3: split 2
EDITOR: HG: Write commit message for the next split changeset.
EDITOR: a2
EDITOR:
EDITOR:
EDITOR: HG: Enter commit message. Lines beginning with 'HG:' are removed.
EDITOR: HG: Leave message empty to abort commit.
EDITOR: HG: --
EDITOR: HG: user: test
EDITOR: HG: branch 'default'
EDITOR: HG: changed a
saved backup bundle to $TESTTMP/a/.hg/strip-backup/1df0d5c5a3ab-8341b760-split.hg (obsstore-off !)
#if obsstore-off
$ hg bookmark
r1 0:a61bcde8c529
r2 3:00eebaf8d2e2
* r3 3:00eebaf8d2e2
$ hg glog -p
@ 3:00eebaf8d2e2 split 3 r2 r3
| diff --git a/a b/a
| --- a/a
| +++ b/a
| @@ -1,1 +1,1 @@
| -1
| +11
|
o 2:a09ad58faae3 split 2
| diff --git a/a b/a
| --- a/a
| +++ b/a
| @@ -3,1 +3,1 @@
| -3
| +33
|
o 1:e704349bd21b split 1
| diff --git a/a b/a
| --- a/a
| +++ b/a
| @@ -5,1 +5,1 @@
| -5
| +55
|
o 0:a61bcde8c529 a1 r1
diff --git a/a b/a
new file mode 100644
--- /dev/null
+++ b/a
@@ -0,0 +1,5 @@
+1
+2
+3
+4
+5
#else
$ hg bookmark
r1 0:a61bcde8c529
r2 4:00eebaf8d2e2
* r3 4:00eebaf8d2e2
$ hg glog
@ 4:00eebaf8d2e2 split 3 r2 r3
|
o 3:a09ad58faae3 split 2
|
o 2:e704349bd21b split 1
|
o 0:a61bcde8c529 a1 r1
#endif
Split a head while working parent is not that head
$ cp -R $TESTTMP/clean $TESTTMP/b
$ cd $TESTTMP/b
$ hg up 0 -q
$ hg bookmark r3
$ runsplit tip >/dev/null
#if obsstore-off
$ hg bookmark
r1 0:a61bcde8c529
r2 3:00eebaf8d2e2
* r3 0:a61bcde8c529
$ hg glog
o 3:00eebaf8d2e2 split 3 r2
|
o 2:a09ad58faae3 split 2
|
o 1:e704349bd21b split 1
|
@ 0:a61bcde8c529 a1 r1 r3
#else
$ hg bookmark
r1 0:a61bcde8c529
r2 4:00eebaf8d2e2
* r3 0:a61bcde8c529
$ hg glog
o 4:00eebaf8d2e2 split 3 r2
|
o 3:a09ad58faae3 split 2
|
o 2:e704349bd21b split 1
|
@ 0:a61bcde8c529 a1 r1 r3
#endif
Split a non-head
$ cp -R $TESTTMP/clean $TESTTMP/c
$ cd $TESTTMP/c
$ echo d > d
$ hg ci -m d1 -A d
$ hg bookmark -i d1
$ echo 2 >> d
$ hg ci -m d2
$ echo 3 >> d
$ hg ci -m d3
$ hg bookmark -i d3
$ hg up '.^' -q
$ hg bookmark d2
$ cp -R . ../d
$ runsplit -r 1 | grep rebasing
rebasing 2:b5c5ea414030 "d1" (d1)
rebasing 3:f4a0a8d004cc "d2" (d2)
rebasing 4:777940761eba "d3" (d3)
#if obsstore-off
$ hg bookmark
d1 4:c4b449ef030e
* d2 5:c9dd00ab36a3
d3 6:19f476bc865c
r1 0:a61bcde8c529
r2 3:00eebaf8d2e2
$ hg glog -p
o 6:19f476bc865c d3 d3
| diff --git a/d b/d
| --- a/d
| +++ b/d
| @@ -2,0 +3,1 @@
| +3
|
@ 5:c9dd00ab36a3 d2 d2
| diff --git a/d b/d
| --- a/d
| +++ b/d
| @@ -1,0 +2,1 @@
| +2
|
o 4:c4b449ef030e d1 d1
| diff --git a/d b/d
| new file mode 100644
| --- /dev/null
| +++ b/d
| @@ -0,0 +1,1 @@
| +d
|
o 3:00eebaf8d2e2 split 3 r2
| diff --git a/a b/a
| --- a/a
| +++ b/a
| @@ -1,1 +1,1 @@
| -1
| +11
|
o 2:a09ad58faae3 split 2
| diff --git a/a b/a
| --- a/a
| +++ b/a
| @@ -3,1 +3,1 @@
| -3
| +33
|
o 1:e704349bd21b split 1
| diff --git a/a b/a
| --- a/a
| +++ b/a
| @@ -5,1 +5,1 @@
| -5
| +55
|
o 0:a61bcde8c529 a1 r1
diff --git a/a b/a
new file mode 100644
--- /dev/null
+++ b/a
@@ -0,0 +1,5 @@
+1
+2
+3
+4
+5
#else
$ hg bookmark
d1 8:c4b449ef030e
* d2 9:c9dd00ab36a3
d3 10:19f476bc865c
r1 0:a61bcde8c529
r2 7:00eebaf8d2e2
$ hg glog
o 10:19f476bc865c d3 d3
|
@ 9:c9dd00ab36a3 d2 d2
|
o 8:c4b449ef030e d1 d1
|
o 7:00eebaf8d2e2 split 3 r2
|
o 6:a09ad58faae3 split 2
|
o 5:e704349bd21b split 1
|
o 0:a61bcde8c529 a1 r1
#endif
Split a non-head without rebase
$ cd $TESTTMP/d
#if obsstore-off
$ runsplit -r 1 --no-rebase
abort: cannot split changeset with children without rebase
[255]
#else
$ runsplit -r 1 --no-rebase >/dev/null
3 new orphan changesets
$ hg bookmark
d1 2:b5c5ea414030
* d2 3:f4a0a8d004cc
d3 4:777940761eba
r1 0:a61bcde8c529
r2 7:00eebaf8d2e2
$ hg glog
o 7:00eebaf8d2e2 split 3 r2
|
o 6:a09ad58faae3 split 2
|
o 5:e704349bd21b split 1
|
| * 4:777940761eba d3 d3
| |
| @ 3:f4a0a8d004cc d2 d2
| |
| * 2:b5c5ea414030 d1 d1
| |
| x 1:1df0d5c5a3ab a2
|/
o 0:a61bcde8c529 a1 r1
#endif
Split a non-head with obsoleted descendants
#if obsstore-on
$ hg init $TESTTMP/e
$ cd $TESTTMP/e
$ hg debugdrawdag <<'EOS'
> H I J
> | | |
> F G1 G2 # amend: G1 -> G2
> | | / # prune: F
> C D E
> \|/
> B
> |
> A
> EOS
2 new orphan changesets
$ eval `hg tags -T '{tag}={node}\n'`
$ rm .hg/localtags
$ hg split $B --config experimental.evolution=createmarkers
abort: split would leave orphaned changesets behind
[255]
$ cat > $TESTTMP/messages <<EOF
> Split B
> EOF
$ cat <<EOF | hg split $B
> y
> y
> EOF
diff --git a/B b/B
new file mode 100644
examine changes to 'B'? [Ynesfdaq?] y
@@ -0,0 +1,1 @@
+B
\ No newline at end of file
record this change to 'B'? [Ynesfdaq?] y
EDITOR: HG: Splitting 112478962961. Write commit message for the first split changeset.
EDITOR: B
EDITOR:
EDITOR:
EDITOR: HG: Enter commit message. Lines beginning with 'HG:' are removed.
EDITOR: HG: Leave message empty to abort commit.
EDITOR: HG: --
EDITOR: HG: user: test
EDITOR: HG: branch 'default'
EDITOR: HG: added B
created new head
rebasing 2:26805aba1e60 "C"
rebasing 3:be0ef73c17ad "D"
rebasing 4:49cb92066bfd "E"
rebasing 7:97a6268cc7ef "G2"
rebasing 10:e2f1e425c0db "J"
$ hg glog -r 'sort(all(), topo)'
o 16:556c085f8b52 J
|
o 15:8761f6c9123f G2
|
o 14:a7aeffe59b65 E
|
| o 13:e1e914ede9ab D
|/
| o 12:01947e9b98aa C
|/
o 11:0947baa74d47 Split B
|
| * 9:88ede1d5ee13 I
| |
| x 6:af8cbf225b7b G1
| |
| x 3:be0ef73c17ad D
| |
| | * 8:74863e5b5074 H
| | |
| | x 5:ee481a2a1e69 F
| | |
| | x 2:26805aba1e60 C
| |/
| x 1:112478962961 B
|/
o 0:426bada5c675 A
#endif
Preserve secret phase in split
$ cp -R $TESTTMP/clean $TESTTMP/phases1
$ cd $TESTTMP/phases1
$ hg phase --secret -fr tip
$ hg log -T '{short(node)} {phase}\n'
1df0d5c5a3ab secret
a61bcde8c529 draft
$ runsplit tip >/dev/null
$ hg log -T '{short(node)} {phase}\n'
00eebaf8d2e2 secret
a09ad58faae3 secret
e704349bd21b secret
a61bcde8c529 draft
Do not move things to secret even if phases.new-commit=secret
$ cp -R $TESTTMP/clean $TESTTMP/phases2
$ cd $TESTTMP/phases2
$ cat >> .hg/hgrc <<EOF
> [phases]
> new-commit=secret
> EOF
$ hg log -T '{short(node)} {phase}\n'
1df0d5c5a3ab draft
a61bcde8c529 draft
$ runsplit tip >/dev/null
$ hg log -T '{short(node)} {phase}\n'
00eebaf8d2e2 draft
a09ad58faae3 draft
e704349bd21b draft
a61bcde8c529 draft