view tests/test-absorb.t @ 40326:fed697fa1734

sqlitestore: file storage backend using SQLite

This commit provides an extension which uses SQLite to store file
data (as opposed to revlogs).

As the inline documentation describes, there are still several
aspects to the extension that are incomplete. But it's a start.
The extension does support basic clone, checkout, and commit
workflows, which makes it suitable for simple use cases.

One notable missing feature is support for "bundlerepos." This is
probably responsible for the most test failures when the extension
is activated as part of the test suite.

All revision data is stored in SQLite. Data is stored as zstd
compressed chunks (default if zstd is available), zlib compressed
chunks (default if zstd is not available), or raw chunks (if
configured or if a compressed delta is not smaller than the raw
delta). This makes things very similar to revlogs.

Unlike revlogs, the extension doesn't yet enforce a limit on delta
chain length. This is an obvious limitation and should be addressed.
This is somewhat mitigated by the use of zstd, which is much faster
than zlib to decompress.

There is a dedicated table for storing deltas. Deltas are stored
by the SHA-1 hash of their uncompressed content. The "fileindex" table
has columns that reference the delta for each revision and the base
delta that delta should be applied against. A recursive SQL query
is used to resolve the delta chain along with the delta data.

By storing deltas by hash, we are able to de-duplicate delta storage!
With revlogs, the same deltas in different revlogs would result in
duplicate storage of that delta. In this scheme, inserting the
duplicate delta is a no-op and delta chains simply reference the
existing delta.

When initially implementing this extension, I did not have
content-indexed deltas and deltas could be duplicated across files
(just like revlogs). When I implemented content-indexed deltas, the
size of the SQLite database for a full clone of mozilla-unified
dropped:

before: 2,554,261,504 bytes
after:  2,488,754,176 bytes

Surprisingly, this is still larger than the bytes size of revlog
files:

revlog files: 2,104,861,230 bytes
du -b:        2,254,381,614

I would have expected storage to be smaller since we're not limiting
delta chain length and since we're using zstd instead of zlib. I
suspect the SQLite indexes and per-column overhead account for the
bulk of the differences. (Keep in mind that revlog uses a 64-byte
packed struct for revision index data and deltas are stored without
padding. Aside from the 12 unused bytes in the 32 byte node field,
revlogs are pretty efficient.) Another source of overhead is file
name storage. With revlogs, file names are stored in the filesystem.
But with SQLite, we need to store file names in the database. This is
roughly equivalent to the size of the fncache file, which for the
mozilla-unified repository is ~34MB.

Since the SQLite database isn't append-only and since delta chains
can reference any delta, this opens some interesting possibilities.
For example, we could store deltas in reverse, such that fulltexts
are stored for newer revisions and deltas are applied to reconstruct
older revisions. This is likely a more optimal storage strategy for
version control, as new data tends to be more frequently accessed
than old data. We would obviously need wire protocol support for
transferring revision data from newest to oldest. And we would
probably need some kind of mechanism for "re-encoding" stores. But
it should be doable.

This extension is very much experimental quality. There are a handful
of features that don't work. It probably isn't suitable for day-to-day
use. But it could be used in limited cases (e.g. read-only checkouts
like in CI). And it is also a good proving ground for alternate
storage backends. As we continue to define interfaces for all things
storage, it will be useful to have a viable alternate storage backend
to see how things shake out in practice.

test-storage.py passes on Python 2 and introduces no new test failures on
Python 3. Having the storage-level unit tests has proved to be insanely
useful when developing this extension. Those tests caught numerous bugs
during development and I'm convinced this style of testing is the way
forward for ensuring alternate storage backends work as intended. Of
course, test coverage isn't close to what it needs to be. But it is
a start. And what coverage we have gives me confidence that basic store
functionality is implemented properly.

Differential Revision: https://phab.mercurial-scm.org/D4928

  $ cat >> $HGRCPATH << EOF
  > [extensions]
  > absorb=
  > EOF

  $ sedi() { # workaround check-code
  > pattern="$1"
  > shift
  > for i in "$@"; do
  >     sed "$pattern" "$i" > "$i".tmp
  >     mv "$i".tmp "$i"
  > done
  > }

  $ hg init repo1
  $ cd repo1

Do not crash with empty repo:

  $ hg absorb
  abort: no mutable changeset to change
  [255]

Make some commits:

  $ for i in 1 2 3 4 5; do
  >   echo $i >> a
  >   hg commit -A a -m "commit $i" -q
  > done

  $ hg annotate a
  0: 1
  1: 2
  2: 3
  3: 4
  4: 5

Change a few lines:

  $ cat > a <<EOF
  > 1a
  > 2b
  > 3
  > 4d
  > 5e
  > EOF

Preview absorb changes:

  $ hg absorb --print-changes --dry-run
  showing changes for a
          @@ -0,2 +0,2 @@
  4ec16f8 -1
  5c5f952 -2
  4ec16f8 +1a
  5c5f952 +2b
          @@ -3,2 +3,2 @@
  ad8b8b7 -4
  4f55fa6 -5
  ad8b8b7 +4d
  4f55fa6 +5e
  
  4 changesets affected
  4f55fa6 commit 5
  ad8b8b7 commit 4
  5c5f952 commit 2
  4ec16f8 commit 1

Run absorb:

  $ hg absorb --apply-changes
  saved backup bundle to * (glob)
  2 of 2 chunk(s) applied
  $ hg annotate a
  0: 1a
  1: 2b
  2: 3
  3: 4d
  4: 5e

Delete a few lines and related commits will be removed if they will be empty:

  $ cat > a <<EOF
  > 2b
  > 4d
  > EOF
  $ echo y | hg absorb --config ui.interactive=1
  showing changes for a
          @@ -0,1 +0,0 @@
  f548282 -1a
          @@ -2,1 +1,0 @@
  ff5d556 -3
          @@ -4,1 +2,0 @@
  84e5416 -5e
  
  3 changesets affected
  84e5416 commit 5
  ff5d556 commit 3
  f548282 commit 1
  apply changes (yn)?  y
  saved backup bundle to * (glob)
  3 of 3 chunk(s) applied
  $ hg annotate a
  1: 2b
  2: 4d
  $ hg log -T '{rev} {desc}\n' -Gp
  @  2 commit 4
  |  diff -r 1cae118c7ed8 -r 58a62bade1c6 a
  |  --- a/a	Thu Jan 01 00:00:00 1970 +0000
  |  +++ b/a	Thu Jan 01 00:00:00 1970 +0000
  |  @@ -1,1 +1,2 @@
  |   2b
  |  +4d
  |
  o  1 commit 2
  |  diff -r 84add69aeac0 -r 1cae118c7ed8 a
  |  --- a/a	Thu Jan 01 00:00:00 1970 +0000
  |  +++ b/a	Thu Jan 01 00:00:00 1970 +0000
  |  @@ -0,0 +1,1 @@
  |  +2b
  |
  o  0 commit 1
  

Non 1:1 map changes will be ignored:

  $ echo 1 > a
  $ hg absorb --apply-changes
  nothing applied
  [1]

Insertaions:

  $ cat > a << EOF
  > insert before 2b
  > 2b
  > 4d
  > insert aftert 4d
  > EOF
  $ hg absorb -q --apply-changes
  $ hg status
  $ hg annotate a
  1: insert before 2b
  1: 2b
  2: 4d
  2: insert aftert 4d

Bookmarks are moved:

  $ hg bookmark -r 1 b1
  $ hg bookmark -r 2 b2
  $ hg bookmark ba
  $ hg bookmarks
     b1                        1:b35060a57a50
     b2                        2:946e4bc87915
   * ba                        2:946e4bc87915
  $ sedi 's/insert/INSERT/' a
  $ hg absorb -q --apply-changes
  $ hg status
  $ hg bookmarks
     b1                        1:a4183e9b3d31
     b2                        2:c9b20c925790
   * ba                        2:c9b20c925790

Non-mofified files are ignored:

  $ touch b
  $ hg commit -A b -m b
  $ touch c
  $ hg add c
  $ hg rm b
  $ hg absorb --apply-changes
  nothing applied
  [1]
  $ sedi 's/INSERT/Insert/' a
  $ hg absorb --apply-changes
  saved backup bundle to * (glob)
  2 of 2 chunk(s) applied
  $ hg status
  A c
  R b

Public commits will not be changed:

  $ hg phase -p 1
  $ sedi 's/Insert/insert/' a
  $ hg absorb -pn
  showing changes for a
          @@ -0,1 +0,1 @@
          -Insert before 2b
          +insert before 2b
          @@ -3,1 +3,1 @@
  85b4e0e -Insert aftert 4d
  85b4e0e +insert aftert 4d
  
  1 changesets affected
  85b4e0e commit 4
  $ hg absorb --apply-changes
  saved backup bundle to * (glob)
  1 of 2 chunk(s) applied
  $ hg diff -U 0
  diff -r 1c8eadede62a a
  --- a/a	Thu Jan 01 00:00:00 1970 +0000
  +++ b/a	* (glob)
  @@ -1,1 +1,1 @@
  -Insert before 2b
  +insert before 2b
  $ hg annotate a
  1: Insert before 2b
  1: 2b
  2: 4d
  2: insert aftert 4d

Make working copy clean:

  $ hg revert -q -C a b
  $ hg forget c
  $ rm c
  $ hg status

Merge commit will not be changed:

  $ echo 1 > m1
  $ hg commit -A m1 -m m1
  $ hg bookmark -q -i m1
  $ hg update -q '.^'
  $ echo 2 > m2
  $ hg commit -q -A m2 -m m2
  $ hg merge -q m1
  $ hg commit -m merge
  $ hg bookmark -d m1
  $ hg log -G -T '{rev} {desc} {phase}\n'
  @    6 merge draft
  |\
  | o  5 m2 draft
  | |
  o |  4 m1 draft
  |/
  o  3 b draft
  |
  o  2 commit 4 draft
  |
  o  1 commit 2 public
  |
  o  0 commit 1 public
  
  $ echo 2 >> m1
  $ echo 2 >> m2
  $ hg absorb --apply-changes
  abort: no mutable changeset to change
  [255]
  $ hg revert -q -C m1 m2

Use a new repo:

  $ cd ..
  $ hg init repo2
  $ cd repo2

Make some commits to multiple files:

  $ for f in a b; do
  >   for i in 1 2; do
  >     echo $f line $i >> $f
  >     hg commit -A $f -m "commit $f $i" -q
  >   done
  > done

Use pattern to select files to be fixed up:

  $ sedi 's/line/Line/' a b
  $ hg status
  M a
  M b
  $ hg absorb --apply-changes a
  saved backup bundle to * (glob)
  1 of 1 chunk(s) applied
  $ hg status
  M b
  $ hg absorb --apply-changes --exclude b
  nothing applied
  [1]
  $ hg absorb --apply-changes b
  saved backup bundle to * (glob)
  1 of 1 chunk(s) applied
  $ hg status
  $ cat a b
  a Line 1
  a Line 2
  b Line 1
  b Line 2

Test config option absorb.max-stack-size:

  $ sedi 's/Line/line/' a b
  $ hg log -T '{rev}:{node} {desc}\n'
  3:712d16a8f445834e36145408eabc1d29df05ec09 commit b 2
  2:74cfa6294160149d60adbf7582b99ce37a4597ec commit b 1
  1:28f10dcf96158f84985358a2e5d5b3505ca69c22 commit a 2
  0:f9a81da8dc53380ed91902e5b82c1b36255a4bd0 commit a 1
  $ hg --config absorb.max-stack-size=1 absorb -pn
  absorb: only the recent 1 changesets will be analysed
  showing changes for a
          @@ -0,2 +0,2 @@
          -a Line 1
          -a Line 2
          +a line 1
          +a line 2
  showing changes for b
          @@ -0,2 +0,2 @@
          -b Line 1
  712d16a -b Line 2
          +b line 1
  712d16a +b line 2
  
  1 changesets affected
  712d16a commit b 2

Test obsolete markers creation:

  $ cat >> $HGRCPATH << EOF
  > [experimental]
  > evolution=createmarkers
  > [absorb]
  > add-noise=1
  > EOF

  $ hg --config absorb.max-stack-size=3 absorb -a
  absorb: only the recent 3 changesets will be analysed
  2 of 2 chunk(s) applied
  $ hg log -T '{rev}:{node|short} {desc} {get(extras, "absorb_source")}\n'
  6:3dfde4199b46 commit b 2 712d16a8f445834e36145408eabc1d29df05ec09
  5:99cfab7da5ff commit b 1 74cfa6294160149d60adbf7582b99ce37a4597ec
  4:fec2b3bd9e08 commit a 2 28f10dcf96158f84985358a2e5d5b3505ca69c22
  0:f9a81da8dc53 commit a 1 
  $ hg absorb --apply-changes
  1 of 1 chunk(s) applied
  $ hg log -T '{rev}:{node|short} {desc} {get(extras, "absorb_source")}\n'
  10:e1c8c1e030a4 commit b 2 3dfde4199b4610ea6e3c6fa9f5bdad8939d69524
  9:816c30955758 commit b 1 99cfab7da5ffdaf3b9fc6643b14333e194d87f46
  8:5867d584106b commit a 2 fec2b3bd9e0834b7cb6a564348a0058171aed811
  7:8c76602baf10 commit a 1 f9a81da8dc53380ed91902e5b82c1b36255a4bd0

Executable files:

  $ cat >> $HGRCPATH << EOF
  > [diff]
  > git=True
  > EOF
  $ cd ..
  $ hg init repo3
  $ cd repo3

#if execbit
  $ echo > foo.py
  $ chmod +x foo.py
  $ hg add foo.py
  $ hg commit -mfoo
#else
  $ hg import -q --bypass - <<EOF
  > # HG changeset patch
  > foo
  > 
  > diff --git a/foo.py b/foo.py
  > new file mode 100755
  > --- /dev/null
  > +++ b/foo.py
  > @@ -0,0 +1,1 @@
  > +
  > EOF
  $ hg up -q
#endif

  $ echo bla > foo.py
  $ hg absorb --dry-run --print-changes
  showing changes for foo.py
          @@ -0,1 +0,1 @@
  99b4ae7 -
  99b4ae7 +bla
  
  1 changesets affected
  99b4ae7 foo
  $ hg absorb --apply-changes
  1 of 1 chunk(s) applied
  $ hg diff -c .
  diff --git a/foo.py b/foo.py
  new file mode 100755
  --- /dev/null
  +++ b/foo.py
  @@ -0,0 +1,1 @@
  +bla
  $ hg diff

Remove lines may delete changesets:

  $ cd ..
  $ hg init repo4
  $ cd repo4
  $ cat > a <<EOF
  > 1
  > 2
  > EOF
  $ hg commit -m a12 -A a
  $ cat > b <<EOF
  > 1
  > 2
  > EOF
  $ hg commit -m b12 -A b
  $ echo 3 >> b
  $ hg commit -m b3
  $ echo 4 >> b
  $ hg commit -m b4
  $ echo 1 > b
  $ echo 3 >> a
  $ hg absorb -pn
  showing changes for a
          @@ -2,0 +2,1 @@
  bfafb49 +3
  showing changes for b
          @@ -1,3 +1,0 @@
  1154859 -2
  30970db -3
  a393a58 -4
  
  4 changesets affected
  a393a58 b4
  30970db b3
  1154859 b12
  bfafb49 a12
  $ hg absorb -av | grep became
  0:bfafb49242db: 1 file(s) changed, became 4:1a2de97fc652
  1:115485984805: 2 file(s) changed, became 5:0c930dfab74c
  2:30970dbf7b40: became empty and was dropped
  3:a393a58b9a85: became empty and was dropped
  $ hg log -T '{rev} {desc}\n' -Gp
  @  5 b12
  |  diff --git a/b b/b
  |  new file mode 100644
  |  --- /dev/null
  |  +++ b/b
  |  @@ -0,0 +1,1 @@
  |  +1
  |
  o  4 a12
     diff --git a/a b/a
     new file mode 100644
     --- /dev/null
     +++ b/a
     @@ -0,0 +1,3 @@
     +1
     +2
     +3
  

Use revert to make the current change and its parent disappear.
This should move us to the non-obsolete ancestor.

  $ cd ..
  $ hg init repo5
  $ cd repo5
  $ cat > a <<EOF
  > 1
  > 2
  > EOF
  $ hg commit -m a12 -A a
  $ hg id
  bfafb49242db tip
  $ echo 3 >> a
  $ hg commit -m a123 a
  $ echo 4 >> a
  $ hg commit -m a1234 a
  $ hg id
  82dbe7fd19f0 tip
  $ hg revert -r 0 a
  $ hg absorb -pn
  showing changes for a
          @@ -2,2 +2,0 @@
  f1c23dd -3
  82dbe7f -4
  
  2 changesets affected
  82dbe7f a1234
  f1c23dd a123
  $ hg absorb --apply-changes --verbose
  1:f1c23dd5d08d: became empty and was dropped
  2:82dbe7fd19f0: became empty and was dropped
  a: 1 of 1 chunk(s) applied
  $ hg id
  bfafb49242db tip