view tests/test-lfs-largefiles.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

#require no-reposimplestore no-chg

This tests the interaction between the largefiles and lfs extensions, and
conversion from largefiles -> lfs.

  $ cat >> $HGRCPATH << EOF
  > [extensions]
  > largefiles =
  > 
  > [lfs]
  > # standin files are 41 bytes.  Stay bigger for clarity.
  > threshold = 42
  > EOF

Setup a repo with a normal file and a largefile, above and below the lfs
threshold to test lfconvert.  *.txt start life as a normal file; *.bin start as
an lfs/largefile.

  $ hg init largefiles
  $ cd largefiles
  $ echo 'normal' > normal.txt
  $ echo 'normal above lfs threshold 0000000000000000000000000' > lfs.txt
  $ hg ci -Am 'normal.txt'
  adding lfs.txt
  adding normal.txt
  $ echo 'largefile' > large.bin
  $ echo 'largefile above lfs threshold 0000000000000000000000' > lfs.bin
  $ hg add --large large.bin lfs.bin
  $ hg ci -m 'add largefiles'

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

Add an lfs file and normal file that collide with files on the other branch.
large.bin is added as a normal file, and is named as such only to clash with the
largefile on the other branch.

  $ hg up -q '.^'
  $ echo 'below lfs threshold' > large.bin
  $ echo 'lfs above the lfs threshold for length 0000000000000' > lfs.bin
  $ hg ci -Am 'add with lfs extension'
  adding large.bin
  adding lfs.bin
  created new head

  $ hg log -G
  @  changeset:   2:e989d0fa3764
  |  tag:         tip
  |  parent:      0:29361292f54d
  |  user:        test
  |  date:        Thu Jan 01 00:00:00 1970 +0000
  |  summary:     add with lfs extension
  |
  | o  changeset:   1:6513aaab9ca0
  |/   user:        test
  |    date:        Thu Jan 01 00:00:00 1970 +0000
  |    summary:     add largefiles
  |
  o  changeset:   0:29361292f54d
     user:        test
     date:        Thu Jan 01 00:00:00 1970 +0000
     summary:     normal.txt
  
--------------------------------------------------------------------------------
Merge largefiles into lfs branch

The largefiles extension will prompt to use the normal or largefile when merged
into the lfs files.  `hg manifest` will show standins if present.  They aren't,
because largefiles merge doesn't merge content.  If it did, selecting (n)ormal
would convert to lfs on commit, if appropriate.

BUG: Largefiles isn't running the merge tool, like when two lfs files are
merged.  This is probably by design, but it should probably at least prompt if
content should be taken from (l)ocal or (o)ther as well.

  $ hg --config ui.interactive=True merge 6513aaab9ca0 <<EOF
  > n
  > n
  > EOF
  remote turned local normal file large.bin into a largefile
  use (l)argefile or keep (n)ormal file? n
  remote turned local normal file lfs.bin into a largefile
  use (l)argefile or keep (n)ormal file? n
  0 files updated, 0 files merged, 0 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ hg ci -m 'merge lfs with largefiles -> normal'
  $ hg manifest
  large.bin
  lfs.bin
  lfs.txt
  normal.txt

The merged lfs.bin resolved to lfs because the (n)ormal option was picked.  The
lfs.txt file is unchanged by the merge, because it was added before lfs was
enabled, and the content didn't change.
  $ hg debugdata lfs.bin 0
  version https://git-lfs.github.com/spec/v1
  oid sha256:81c7492b2c05e130431f65a87651b54a30c5da72c99ce35a1e9b9872a807312b
  size 53
  x-is-binary 0
  $ hg debugdata lfs.txt 0
  normal above lfs threshold 0000000000000000000000000

Another filelog entry is NOT made by the merge, so nothing is committed as lfs.
  $ hg log -r . -T '{join(lfs_files, ", ")}\n'
  

Replay the last merge, but pick (l)arge this time.  The manifest will show any
standins.

  $ hg up -Cq e989d0fa3764

  $ hg --config ui.interactive=True merge 6513aaab9ca0 <<EOF
  > l
  > l
  > EOF
  remote turned local normal file large.bin into a largefile
  use (l)argefile or keep (n)ormal file? l
  remote turned local normal file lfs.bin into a largefile
  use (l)argefile or keep (n)ormal file? l
  getting changed largefiles
  2 largefiles updated, 0 removed
  2 files updated, 0 files merged, 2 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ hg ci -m 'merge lfs with largefiles -> large'
  created new head
  $ hg manifest
  .hglf/large.bin
  .hglf/lfs.bin
  lfs.txt
  normal.txt

--------------------------------------------------------------------------------
Merge lfs into largefiles branch

  $ hg up -Cq 6513aaab9ca0
  $ hg --config ui.interactive=True merge e989d0fa3764 <<EOF
  > n
  > n
  > EOF
  remote turned local largefile large.bin into a normal file
  keep (l)argefile or use (n)ormal file? n
  remote turned local largefile lfs.bin into a normal file
  keep (l)argefile or use (n)ormal file? n
  getting changed largefiles
  0 largefiles updated, 0 removed
  2 files updated, 0 files merged, 2 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ hg ci -m 'merge largefiles with lfs -> normal'
  created new head
  $ hg manifest
  large.bin
  lfs.bin
  lfs.txt
  normal.txt

The merged lfs.bin got converted to lfs because the (n)ormal option was picked.
The lfs.txt file is unchanged by the merge, because it was added before lfs was
enabled.
  $ hg debugdata lfs.bin 0
  version https://git-lfs.github.com/spec/v1
  oid sha256:81c7492b2c05e130431f65a87651b54a30c5da72c99ce35a1e9b9872a807312b
  size 53
  x-is-binary 0
  $ hg debugdata lfs.txt 0
  normal above lfs threshold 0000000000000000000000000

Another filelog entry is NOT made by the merge, so nothing is committed as lfs.
  $ hg log -r . -T '{join(lfs_files, ", ")}\n'
  

Replay the last merge, but pick (l)arge this time.  The manifest will show the
standins.

  $ hg up -Cq 6513aaab9ca0

  $ hg --config ui.interactive=True merge e989d0fa3764 <<EOF
  > l
  > l
  > EOF
  remote turned local largefile large.bin into a normal file
  keep (l)argefile or use (n)ormal file? l
  remote turned local largefile lfs.bin into a normal file
  keep (l)argefile or use (n)ormal file? l
  0 files updated, 0 files merged, 0 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ hg ci -m 'merge largefiles with lfs -> large'
  created new head
  $ hg manifest
  .hglf/large.bin
  .hglf/lfs.bin
  lfs.txt
  normal.txt

--------------------------------------------------------------------------------

When both largefiles and lfs are configured to add by size, the tie goes to
largefiles since it hooks cmdutil.add() and lfs hooks the filelog write in the
commit.  By the time the commit occurs, the tracked file is smaller than the
threshold (assuming it is > 41, so the standins don't become lfs objects).

  $ "$PYTHON" -c 'import sys ; sys.stdout.write("y\n" * 1048576)' > large_by_size.bin
  $ hg --config largefiles.minsize=1 ci -Am 'large by size'
  adding large_by_size.bin as a largefile
  $ hg manifest
  .hglf/large.bin
  .hglf/large_by_size.bin
  .hglf/lfs.bin
  lfs.txt
  normal.txt

  $ hg rm large_by_size.bin
  $ hg ci -m 'remove large_by_size.bin'

Largefiles doesn't do anything special with diff, so it falls back to diffing
the standins.  Extdiff also is standin based comparison.  Diff and extdiff both
work on the original file for lfs objects.

Largefile -> lfs transition
  $ hg diff -r 1 -r 3
  diff -r 6513aaab9ca0 -r dcc5ce63e252 .hglf/large.bin
  --- a/.hglf/large.bin	Thu Jan 01 00:00:00 1970 +0000
  +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
  @@ -1,1 +0,0 @@
  -cef9a458373df9b0743a0d3c14d0c66fb19b8629
  diff -r 6513aaab9ca0 -r dcc5ce63e252 .hglf/lfs.bin
  --- a/.hglf/lfs.bin	Thu Jan 01 00:00:00 1970 +0000
  +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
  @@ -1,1 +0,0 @@
  -557fb6309cef935e1ac2c8296508379e4b15a6e6
  diff -r 6513aaab9ca0 -r dcc5ce63e252 large.bin
  --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
  +++ b/large.bin	Thu Jan 01 00:00:00 1970 +0000
  @@ -0,0 +1,1 @@
  +below lfs threshold
  diff -r 6513aaab9ca0 -r dcc5ce63e252 lfs.bin
  --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
  +++ b/lfs.bin	Thu Jan 01 00:00:00 1970 +0000
  @@ -0,0 +1,1 @@
  +lfs above the lfs threshold for length 0000000000000

lfs -> largefiles transition
  $ hg diff -r 2 -r 6
  diff -r e989d0fa3764 -r 95e1e80325c8 .hglf/large.bin
  --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
  +++ b/.hglf/large.bin	Thu Jan 01 00:00:00 1970 +0000
  @@ -0,0 +1,1 @@
  +cef9a458373df9b0743a0d3c14d0c66fb19b8629
  diff -r e989d0fa3764 -r 95e1e80325c8 .hglf/lfs.bin
  --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
  +++ b/.hglf/lfs.bin	Thu Jan 01 00:00:00 1970 +0000
  @@ -0,0 +1,1 @@
  +557fb6309cef935e1ac2c8296508379e4b15a6e6
  diff -r e989d0fa3764 -r 95e1e80325c8 large.bin
  --- a/large.bin	Thu Jan 01 00:00:00 1970 +0000
  +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
  @@ -1,1 +0,0 @@
  -below lfs threshold
  diff -r e989d0fa3764 -r 95e1e80325c8 lfs.bin
  --- a/lfs.bin	Thu Jan 01 00:00:00 1970 +0000
  +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
  @@ -1,1 +0,0 @@
  -lfs above the lfs threshold for length 0000000000000

A largefiles repo can be converted to lfs.  The lfconvert command uses the
convert extension under the hood with --to-normal.  So the --config based
parameters are available, but not --authormap, --branchmap, etc.

  $ cd ..
  $ hg lfconvert --to-normal largefiles nolargefiles 2>&1
  initializing destination nolargefiles
  0 additional largefiles cached
  scanning source...
  sorting...
  converting...
  8 normal.txt
  7 add largefiles
  6 add with lfs extension
  5 merge lfs with largefiles -> normal
  4 merge lfs with largefiles -> large
  3 merge largefiles with lfs -> normal
  2 merge largefiles with lfs -> large
  1 large by size
  0 remove large_by_size.bin
  $ cd nolargefiles

The requirement is added to the destination repo, and the extension is enabled
locally.

  $ cat .hg/requires
  dotencode
  fncache
  generaldelta
  lfs
  revlogv1
  store
  $ hg config --debug extensions | grep lfs
  $TESTTMP/nolargefiles/.hg/hgrc:*: extensions.lfs= (glob)

  $ hg log -r 'all()' -G -T '{rev} {join(lfs_files, ", ")} ({desc})\n'
  o  8 large_by_size.bin (remove large_by_size.bin)
  |
  o  7 large_by_size.bin (large by size)
  |
  o    6  (merge largefiles with lfs -> large)
  |\
  +---o  5  (merge largefiles with lfs -> normal)
  | |/
  +---o  4 lfs.bin (merge lfs with largefiles -> large)
  | |/
  +---o  3  (merge lfs with largefiles -> normal)
  | |/
  | o  2 lfs.bin (add with lfs extension)
  | |
  o |  1 lfs.bin (add largefiles)
  |/
  o  0 lfs.txt (normal.txt)
  
  $ hg debugdata lfs.bin 0
  version https://git-lfs.github.com/spec/v1
  oid sha256:2172a5bd492dd41ec533b9bb695f7691b6351719407ac797f0ccad5348c81e62
  size 53
  x-is-binary 0
  $ hg debugdata lfs.bin 1
  version https://git-lfs.github.com/spec/v1
  oid sha256:81c7492b2c05e130431f65a87651b54a30c5da72c99ce35a1e9b9872a807312b
  size 53
  x-is-binary 0
  $ hg debugdata lfs.bin 2
  version https://git-lfs.github.com/spec/v1
  oid sha256:2172a5bd492dd41ec533b9bb695f7691b6351719407ac797f0ccad5348c81e62
  size 53
  x-is-binary 0
  $ hg debugdata lfs.bin 3
  abort: invalid revision identifier 3
  [255]

No diffs when comparing merge and p1 that kept p1's changes.  Diff of lfs to
largefiles no longer operates in standin files.

This `head -n 20` looks dumb (since we expect no output), but if something
breaks you can get 1048576 lines of +y in the output, which takes a looooooong
time to print.
  $ hg diff -r 2:3 | head -n 20
  $ hg diff -r 2:6
  diff -r e989d0fa3764 -r 752e3a0d8488 large.bin
  --- a/large.bin	Thu Jan 01 00:00:00 1970 +0000
  +++ b/large.bin	Thu Jan 01 00:00:00 1970 +0000
  @@ -1,1 +1,1 @@
  -below lfs threshold
  +largefile
  diff -r e989d0fa3764 -r 752e3a0d8488 lfs.bin
  --- a/lfs.bin	Thu Jan 01 00:00:00 1970 +0000
  +++ b/lfs.bin	Thu Jan 01 00:00:00 1970 +0000
  @@ -1,1 +1,1 @@
  -lfs above the lfs threshold for length 0000000000000
  +largefile above lfs threshold 0000000000000000000000