view tests/test-convert-hg-sink.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]
  > convert=
  > [convert]
  > hg.saverev=False
  > EOF
  $ hg init orig
  $ cd orig
  $ echo foo > foo
  $ echo bar > bar
  $ hg ci -qAm 'add foo and bar'
  $ hg rm foo
  $ hg ci -m 'remove foo'
  $ mkdir foo
  $ echo file > foo/file
  $ hg ci -qAm 'add foo/file'
  $ hg tag some-tag
  $ hg tag -l local-tag
  $ echo '1234567890123456789012345678901234567890 missing_tag' >> .hgtags
  $ hg ci -m 'add a missing tag'
  $ hg log
  changeset:   4:3fb95ee23a66
  tag:         tip
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  summary:     add a missing tag
  
  changeset:   3:593cbf6fb2b4
  tag:         local-tag
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  summary:     Added tag some-tag for changeset ad681a868e44
  
  changeset:   2:ad681a868e44
  tag:         some-tag
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  summary:     add foo/file
  
  changeset:   1:cbba8ecc03b7
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  summary:     remove foo
  
  changeset:   0:327daa9251fa
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  summary:     add foo and bar
  
  $ hg phase --public -r tip
  $ cd ..
  $ hg convert orig new 2>&1 | grep -v 'subversion python bindings could not be loaded'
  initializing destination new repository
  scanning source...
  sorting...
  converting...
  4 add foo and bar
  3 remove foo
  2 add foo/file
  1 Added tag some-tag for changeset ad681a868e44
  0 add a missing tag
  missing tag entry: "1234567890123456789012345678901234567890 missing_tag"
  $ cd new
  $ hg log -G --template '{rev} {node|short} ({phase}) "{desc}"\n'
  o  4 3fb95ee23a66 (public) "add a missing tag"
  |
  o  3 593cbf6fb2b4 (public) "Added tag some-tag for changeset ad681a868e44"
  |
  o  2 ad681a868e44 (public) "add foo/file"
  |
  o  1 cbba8ecc03b7 (public) "remove foo"
  |
  o  0 327daa9251fa (public) "add foo and bar"
  

  $ hg out ../orig
  comparing with ../orig
  searching for changes
  no changes found
  [1]

dirstate should be empty:

  $ hg debugstate
  $ hg parents -q
  $ hg up -C
  3 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ hg copy bar baz

put something in the dirstate:

  $ hg debugstate > debugstate
  $ grep baz debugstate
  a   0         -1 unset               baz
  copy: bar -> baz

add a new revision in the original repo

  $ cd ../orig
  $ echo baz > baz
  $ hg ci -qAm 'add baz'
  $ cd ..
  $ hg convert orig new 2>&1 | grep -v 'subversion python bindings could not be loaded'
  scanning source...
  sorting...
  converting...
  0 add baz
  $ cd new
  $ hg out ../orig
  comparing with ../orig
  searching for changes
  no changes found
  [1]

dirstate should be the same (no output below):

  $ hg debugstate > new-debugstate
  $ diff debugstate new-debugstate

no copies

  $ hg up -C
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ hg debugrename baz
  baz not renamed
  $ cd ..

test tag rewriting

  $ cat > filemap <<EOF
  > exclude foo
  > EOF
  $ hg convert --filemap filemap orig new-filemap 2>&1 | grep -v 'subversion python bindings could not be loaded'
  initializing destination new-filemap repository
  scanning source...
  sorting...
  converting...
  5 add foo and bar
  4 remove foo
  3 add foo/file
  2 Added tag some-tag for changeset ad681a868e44
  1 add a missing tag
  missing tag entry: "1234567890123456789012345678901234567890 missing_tag"
  0 add baz
  $ cd new-filemap
  $ hg tags
  tip                                3:7bb553f2c68a
  some-tag                           0:ba8636729451
  $ cd ..


Test cases for hg-hg roundtrip

Helper

  $ glog()
  > {
  >     hg log -G --template '{rev} {node|short} ({phase}) "{desc}" files: {files}\n' $*
  > }

Create a tricky source repo

  $ hg init source
  $ cd source

  $ echo 0 > 0
  $ hg ci -Aqm '0: add 0'
  $ echo a > a
  $ mkdir dir
  $ echo b > dir/b
  $ hg ci -qAm '1: add a and dir/b'
  $ echo c > dir/c
  $ hg ci -qAm '2: add dir/c'
  $ hg copy a e
  $ echo b >> b
  $ hg ci -qAm '3: copy a to e, change b'
  $ hg up -qr -3
  $ echo a >> a
  $ hg ci -qAm '4: change a'
  $ hg merge
  merging a and e to e
  2 files updated, 1 files merged, 0 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ hg copy b dir/d
  $ hg ci -qAm '5: merge 2 and 3, copy b to dir/d'
  $ echo a >> a
  $ hg ci -qAm '6: change a'

  $ hg mani
  0
  a
  b
  dir/b
  dir/c
  dir/d
  e
  $ hg phase --public -r tip
  $ glog
  @  6 0613c8e59a3d (public) "6: change a" files: a
  |
  o    5 717e9b37cdb7 (public) "5: merge 2 and 3, copy b to dir/d" files: dir/d e
  |\
  | o  4 86a55cb968d5 (public) "4: change a" files: a
  | |
  o |  3 0e6e235919dd (public) "3: copy a to e, change b" files: b e
  | |
  o |  2 0394b0d5e4f7 (public) "2: add dir/c" files: dir/c
  |/
  o  1 333546584845 (public) "1: add a and dir/b" files: a dir/b
  |
  o  0 d1a24e2ebd23 (public) "0: add 0" files: 0
  
  $ cd ..

Convert excluding rev 0 and dir/ (and thus rev2):

  $ cat << EOF > filemap
  > exclude dir
  > EOF

  $ hg convert --filemap filemap source dest --config convert.hg.revs=1::
  initializing destination dest repository
  scanning source...
  sorting...
  converting...
  5 1: add a and dir/b
  4 2: add dir/c
  3 3: copy a to e, change b
  2 4: change a
  1 5: merge 2 and 3, copy b to dir/d
  0 6: change a

Verify that conversion skipped rev 2:

  $ glog -R dest
  o  4 78814e84a217 (draft) "6: change a" files: a
  |
  o    3 f7cff662c5e5 (draft) "5: merge 2 and 3, copy b to dir/d" files: e
  |\
  | o  2 ab40a95b0072 (draft) "4: change a" files: a
  | |
  o |  1 bd51f17597bf (draft) "3: copy a to e, change b" files: b e
  |/
  o  0 a4a1dae0fe35 (draft) "1: add a and dir/b" files: 0 a
  

Verify mapping correct in both directions:

  $ cat source/.hg/shamap
  a4a1dae0fe3514cefd9b8541b7abbc8f44f946d5 333546584845f70c4cfecb992341aaef0e708166
  bd51f17597bf32268e68a560b206898c3960cda2 0e6e235919dd8e9285ba8eb5adf703af9ad99378
  ab40a95b00725307e79c2fd271000aa8af9759f4 86a55cb968d51770cba2a1630d6cc637b574580a
  f7cff662c5e581e6f3f1a85ffdd2bcb35825f6ba 717e9b37cdb7eb9917ca8e30aa3f986e6d5b177d
  78814e84a217894517c2de392b903ed05e6871a4 0613c8e59a3ddb9789072ef52f1ed13496489bb4
  $ cat dest/.hg/shamap
  333546584845f70c4cfecb992341aaef0e708166 a4a1dae0fe3514cefd9b8541b7abbc8f44f946d5
  0394b0d5e4f761ced559fd0bbdc6afc16cb3f7d1 a4a1dae0fe3514cefd9b8541b7abbc8f44f946d5
  0e6e235919dd8e9285ba8eb5adf703af9ad99378 bd51f17597bf32268e68a560b206898c3960cda2
  86a55cb968d51770cba2a1630d6cc637b574580a ab40a95b00725307e79c2fd271000aa8af9759f4
  717e9b37cdb7eb9917ca8e30aa3f986e6d5b177d f7cff662c5e581e6f3f1a85ffdd2bcb35825f6ba
  0613c8e59a3ddb9789072ef52f1ed13496489bb4 78814e84a217894517c2de392b903ed05e6871a4

Verify meta data converted correctly:

  $ hg -R dest log -r 1 --debug -p --git
  changeset:   1:bd51f17597bf32268e68a560b206898c3960cda2
  phase:       draft
  parent:      0:a4a1dae0fe3514cefd9b8541b7abbc8f44f946d5
  parent:      -1:0000000000000000000000000000000000000000
  manifest:    1:040c72ed9b101773c24ac314776bfc846943781f
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  files+:      b e
  extra:       branch=default
  description:
  3: copy a to e, change b
  
  
  diff --git a/b b/b
  new file mode 100644
  --- /dev/null
  +++ b/b
  @@ -0,0 +1,1 @@
  +b
  diff --git a/a b/e
  copy from a
  copy to e
  
Verify files included and excluded correctly:

  $ hg -R dest manifest -r tip
  0
  a
  b
  e


Make changes in dest and convert back:

  $ hg -R dest up -q
  $ echo dest > dest/dest
  $ hg -R dest ci -Aqm 'change in dest'
  $ hg -R dest tip
  changeset:   5:a2e0e3cc6d1d
  tag:         tip
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  summary:     change in dest
  

(converting merges back after using a filemap will probably cause chaos so we
exclude merges.)

  $ hg convert dest source --config convert.hg.revs='!merge()'
  scanning source...
  sorting...
  converting...
  0 change in dest

Verify the conversion back:

  $ hg -R source log --debug -r tip
  changeset:   7:e6d364a69ff1248b2099e603b0c145504cade6f0
  tag:         tip
  phase:       draft
  parent:      6:0613c8e59a3ddb9789072ef52f1ed13496489bb4
  parent:      -1:0000000000000000000000000000000000000000
  manifest:    7:aa3e9542f3b76d4f1f1b2e9c7ce9dbb48b6a95ec
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  files+:      dest
  extra:       branch=default
  description:
  change in dest
  
  
Files that had been excluded are still present:

  $ hg -R source manifest -r tip
  0
  a
  b
  dest
  dir/b
  dir/c
  dir/d
  e

More source changes

  $ cd source
  $ echo 1 >> a
  $ hg ci -m '8: source first branch'
  created new head
  $ hg up -qr -2
  $ echo 2 >> a
  $ hg ci -m '9: source second branch'
  $ hg merge -q --tool internal:local
  $ hg ci -m '10: source merge'
  $ echo >> a
  $ hg ci -m '11: source change'

  $ hg mani
  0
  a
  b
  dest
  dir/b
  dir/c
  dir/d
  e

  $ glog -r 6:
  @  11 0c8927d1f7f4 (draft) "11: source change" files: a
  |
  o    10 9ccb7ee8d261 (draft) "10: source merge" files: a
  |\
  | o  9 f131b1518dba (draft) "9: source second branch" files: a
  | |
  o |  8 669cf0e74b50 (draft) "8: source first branch" files: a
  | |
  | o  7 e6d364a69ff1 (draft) "change in dest" files: dest
  |/
  o  6 0613c8e59a3d (public) "6: change a" files: a
  |
  ~
  $ cd ..

  $ hg convert --filemap filemap source dest --config convert.hg.revs=3:
  scanning source...
  sorting...
  converting...
  3 8: source first branch
  2 9: source second branch
  1 10: source merge
  0 11: source change

  $ glog -R dest
  o  9 8432d597b263 (draft) "11: source change" files: a
  |
  o    8 632ffacdcd6f (draft) "10: source merge" files: a
  |\
  | o  7 049cfee90ee6 (draft) "9: source second branch" files: a
  | |
  o |  6 9b6845e036e5 (draft) "8: source first branch" files: a
  | |
  | @  5 a2e0e3cc6d1d (draft) "change in dest" files: dest
  |/
  o  4 78814e84a217 (draft) "6: change a" files: a
  |
  o    3 f7cff662c5e5 (draft) "5: merge 2 and 3, copy b to dir/d" files: e
  |\
  | o  2 ab40a95b0072 (draft) "4: change a" files: a
  | |
  o |  1 bd51f17597bf (draft) "3: copy a to e, change b" files: b e
  |/
  o  0 a4a1dae0fe35 (draft) "1: add a and dir/b" files: 0 a
  
  $ cd ..

Two way tests

  $ hg init 0
  $ echo f > 0/f
  $ echo a > 0/a-only
  $ echo b > 0/b-only
  $ hg -R 0 ci -Aqm0

  $ cat << EOF > filemap-a
  > exclude b-only
  > EOF
  $ cat << EOF > filemap-b
  > exclude a-only
  > EOF
  $ hg convert --filemap filemap-a 0 a
  initializing destination a repository
  scanning source...
  sorting...
  converting...
  0 0
  $ hg -R a up -q
  $ echo a > a/f
  $ hg -R a ci -ma

  $ hg convert --filemap filemap-b 0 b
  initializing destination b repository
  scanning source...
  sorting...
  converting...
  0 0
  $ hg -R b up -q
  $ echo b > b/f
  $ hg -R b ci -mb

  $ tail 0/.hg/shamap
  86f3f774ffb682bffb5dc3c1d3b3da637cb9a0d6 8a028c7c77f6c7bd6d63bc3f02ca9f779eabf16a
  dd9f218eb91fb857f2a62fe023e1d64a4e7812fe 8a028c7c77f6c7bd6d63bc3f02ca9f779eabf16a
  $ tail a/.hg/shamap
  8a028c7c77f6c7bd6d63bc3f02ca9f779eabf16a 86f3f774ffb682bffb5dc3c1d3b3da637cb9a0d6
  $ tail b/.hg/shamap
  8a028c7c77f6c7bd6d63bc3f02ca9f779eabf16a dd9f218eb91fb857f2a62fe023e1d64a4e7812fe

  $ hg convert a 0
  scanning source...
  sorting...
  converting...
  0 a

  $ hg convert b 0
  scanning source...
  sorting...
  converting...
  0 b

  $ hg -R 0 log -G
  o  changeset:   2:637fbbbe96b6
  |  tag:         tip
  |  parent:      0:8a028c7c77f6
  |  user:        test
  |  date:        Thu Jan 01 00:00:00 1970 +0000
  |  summary:     b
  |
  | o  changeset:   1:ec7b9c96e692
  |/   user:        test
  |    date:        Thu Jan 01 00:00:00 1970 +0000
  |    summary:     a
  |
  @  changeset:   0:8a028c7c77f6
     user:        test
     date:        Thu Jan 01 00:00:00 1970 +0000
     summary:     0
  
  $ hg convert --filemap filemap-b 0 a --config convert.hg.revs=1::
  scanning source...
  sorting...
  converting...

  $ hg -R 0 up -r1
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ echo f >> 0/f
  $ hg -R 0 ci -mx

  $ hg convert --filemap filemap-b 0 a --config convert.hg.revs=1::
  scanning source...
  sorting...
  converting...
  0 x

  $ hg -R a log -G -T '{rev} {desc|firstline} ({files})\n'
  o  2 x (f)
  |
  @  1 a (f)
  |
  o  0 0 (a-only f)
  
  $ hg -R a mani -r tip
  a-only
  f

An additional round, demonstrating that unchanged files don't get converted

  $ echo f >> 0/f
  $ echo f >> 0/a-only
  $ hg -R 0 ci -m "extra f+a-only change"

  $ hg convert --filemap filemap-b 0 a --config convert.hg.revs=1::
  scanning source...
  sorting...
  converting...
  0 extra f+a-only change

  $ hg -R a log -G -T '{rev} {desc|firstline} ({files})\n'
  o  3 extra f+a-only change (f)
  |
  o  2 x (f)
  |
  @  1 a (f)
  |
  o  0 0 (a-only f)
  

Conversion after rollback

  $ hg -R a rollback -f
  repository tip rolled back to revision 2 (undo convert)

  $ hg convert --filemap filemap-b 0 a --config convert.hg.revs=1::
  scanning source...
  sorting...
  converting...
  0 extra f+a-only change

  $ hg -R a log -G -T '{rev} {desc|firstline} ({files})\n'
  o  3 extra f+a-only change (f)
  |
  o  2 x (f)
  |
  @  1 a (f)
  |
  o  0 0 (a-only f)
  
Convert with --full adds and removes files that didn't change

  $ echo f >> 0/f
  $ hg -R 0 ci -m "f"
  $ hg convert --filemap filemap-b --full 0 a --config convert.hg.revs=1::
  scanning source...
  sorting...
  converting...
  0 f
  $ hg -R a status --change tip
  M f
  A b-only
  R a-only