view tests/test-fix-topology.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

A script that implements uppercasing all letters in a file.

  $ UPPERCASEPY="$TESTTMP/uppercase.py"
  $ cat > $UPPERCASEPY <<EOF
  > import sys
  > from mercurial.utils.procutil import setbinary
  > setbinary(sys.stdin)
  > setbinary(sys.stdout)
  > sys.stdout.write(sys.stdin.read().upper())
  > EOF
  $ TESTLINES="foo\nbar\nbaz\n"
  $ printf $TESTLINES | "$PYTHON" $UPPERCASEPY
  FOO
  BAR
  BAZ

Tests for the fix extension's behavior around non-trivial history topologies.
Looks for correct incremental fixing and reproduction of parent/child
relationships. We indicate fixed file content by uppercasing it.

  $ cat >> $HGRCPATH <<EOF
  > [extensions]
  > fix =
  > [fix]
  > uppercase-whole-file:command="$PYTHON" $UPPERCASEPY
  > uppercase-whole-file:fileset=set:**
  > EOF

This tests the only behavior that should really be affected by obsolescence, so
we'll test it with evolution off and on. This only changes the revision
numbers, if all is well.

#testcases obsstore-off obsstore-on
#if obsstore-on
  $ cat >> $HGRCPATH <<EOF
  > [experimental]
  > evolution.createmarkers=True
  > evolution.allowunstable=True
  > EOF
#endif

Setting up the test topology. Scroll down to see the graph produced. We make it
clear which files were modified in each revision. It's enough to test at the
file granularity, because that demonstrates which baserevs were diffed against.
The computation of changed lines is orthogonal and tested separately.

  $ hg init repo
  $ cd repo

  $ printf "aaaa\n" > a
  $ hg commit -Am "change A"
  adding a
  $ printf "bbbb\n" > b
  $ hg commit -Am "change B"
  adding b
  $ printf "cccc\n" > c
  $ hg commit -Am "change C"
  adding c
  $ hg checkout 0
  0 files updated, 0 files merged, 2 files removed, 0 files unresolved
  $ printf "dddd\n" > d
  $ hg commit -Am "change D"
  adding d
  created new head
  $ hg merge -r 2
  2 files updated, 0 files merged, 0 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ printf "eeee\n" > e
  $ hg commit -Am "change E"
  adding e
  $ hg checkout 0
  0 files updated, 0 files merged, 4 files removed, 0 files unresolved
  $ printf "ffff\n" > f
  $ hg commit -Am "change F"
  adding f
  created new head
  $ hg checkout 0
  0 files updated, 0 files merged, 1 files removed, 0 files unresolved
  $ printf "gggg\n" > g
  $ hg commit -Am "change G"
  adding g
  created new head
  $ hg merge -r 5
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ printf "hhhh\n" > h
  $ hg commit -Am "change H"
  adding h
  $ hg merge -r 4
  4 files updated, 0 files merged, 0 files removed, 0 files unresolved
  (branch merge, don't forget to commit)
  $ printf "iiii\n" > i
  $ hg commit -Am "change I"
  adding i
  $ hg checkout 2
  0 files updated, 0 files merged, 6 files removed, 0 files unresolved
  $ printf "jjjj\n" > j
  $ hg commit -Am "change J"
  adding j
  created new head
  $ hg checkout 7
  3 files updated, 0 files merged, 3 files removed, 0 files unresolved
  $ printf "kkkk\n" > k
  $ hg add
  adding k

  $ hg log --graph --template '{rev} {desc}\n'
  o  9 change J
  |
  | o    8 change I
  | |\
  | | @    7 change H
  | | |\
  | | | o  6 change G
  | | | |
  | | o |  5 change F
  | | |/
  | o |  4 change E
  |/| |
  | o |  3 change D
  | |/
  o |  2 change C
  | |
  o |  1 change B
  |/
  o  0 change A
  

Fix all but the root revision and its four children.

#if obsstore-on
  $ hg fix -r '2|4|7|8|9' --working-dir
#else
  $ hg fix -r '2|4|7|8|9' --working-dir
  saved backup bundle to * (glob)
#endif

The five revisions remain, but the other revisions were fixed and replaced. All
parent pointers have been accurately set to reproduce the previous topology
(though it is rendered in a slightly different order now).

#if obsstore-on
  $ hg log --graph --template '{rev} {desc}\n'
  o  14 change J
  |
  | o    13 change I
  | |\
  | | @    12 change H
  | | |\
  | o | |  11 change E
  |/| | |
  o | | |  10 change C
  | | | |
  | | | o  6 change G
  | | | |
  | | o |  5 change F
  | | |/
  | o /  3 change D
  | |/
  o /  1 change B
  |/
  o  0 change A
  
  $ C=10
  $ E=11
  $ H=12
  $ I=13
  $ J=14
#else
  $ hg log --graph --template '{rev} {desc}\n'
  o  9 change J
  |
  | o    8 change I
  | |\
  | | @    7 change H
  | | |\
  | o | |  6 change E
  |/| | |
  o | | |  5 change C
  | | | |
  | | | o  4 change G
  | | | |
  | | o |  3 change F
  | | |/
  | o /  2 change D
  | |/
  o /  1 change B
  |/
  o  0 change A
  
  $ C=5
  $ E=6
  $ H=7
  $ I=8
  $ J=9
#endif

Change C is a root of the set being fixed, so all we fix is what has changed
since its parent. That parent, change B, is its baserev.

  $ hg cat -r $C 'set:**'
  aaaa
  bbbb
  CCCC

Change E is a merge with only one parent being fixed. Its baserevs are the
unfixed parent plus the baserevs of the other parent. This evaluates to changes
B and D. We now have to decide what it means to incrementally fix a merge
commit. We choose to fix anything that has changed versus any baserev. Only the
undisturbed content of the common ancestor, change A, is unfixed.

  $ hg cat -r $E 'set:**'
  aaaa
  BBBB
  CCCC
  DDDD
  EEEE

Change H is a merge with neither parent being fixed. This is essentially
equivalent to the previous case because there is still only one baserev for
each parent of the merge.

  $ hg cat -r $H 'set:**'
  aaaa
  FFFF
  GGGG
  HHHH

Change I is a merge that has four baserevs; two from each parent. We handle
multiple baserevs in the same way regardless of how many came from each parent.
So, fixing change H will fix any files that were not exactly the same in each
baserev.

  $ hg cat -r $I 'set:**'
  aaaa
  BBBB
  CCCC
  DDDD
  EEEE
  FFFF
  GGGG
  HHHH
  IIII

Change J is a simple case with one baserev, but its baserev is not its parent,
change C. Its baserev is its grandparent, change B.

  $ hg cat -r $J 'set:**'
  aaaa
  bbbb
  CCCC
  JJJJ

The working copy was dirty, so it is treated much like a revision. The baserevs
for the working copy are inherited from its parent, change H, because it is
also being fixed.

  $ cat *
  aaaa
  FFFF
  GGGG
  HHHH
  KKKK

Change A was never a baserev because none of its children were to be fixed.

  $ cd ..

The --all flag should fix anything that wouldn't cause a problem if you fixed
it, including the working copy. Obsolete revisions are not fixed because that
could cause divergence. Public revisions would cause an abort because they are
immutable. We can fix orphans because their successors are still just orphans
of the original obsolete parent. When obsolesence is off, we're just fixing and
replacing anything that isn't public.

  $ hg init fixall
  $ cd fixall

#if obsstore-on
  $ printf "one\n" > foo.whole
  $ hg commit -Aqm "first"
  $ hg phase --public
  $ hg tag --local root
  $ printf "two\n" > foo.whole
  $ hg commit -m "second"
  $ printf "three\n" > foo.whole
  $ hg commit -m "third" --secret
  $ hg tag --local secret
  $ hg checkout root
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ printf "four\n" > foo.whole
  $ hg commit -m "fourth"
  created new head
  $ printf "five\n" > foo.whole
  $ hg commit -m "fifth"
  $ hg tag --local replaced
  $ printf "six\n" > foo.whole
  $ hg commit -m "sixth"
  $ hg checkout replaced
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ printf "seven\n" > foo.whole
  $ hg commit --amend
  1 new orphan changesets
  $ hg checkout secret
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ printf "uncommitted\n" > foo.whole

  $ hg log --graph --template '{rev} {desc} {phase}\n'
  o  6 fifth draft
  |
  | *  5 sixth draft
  | |
  | x  4 fifth draft
  |/
  o  3 fourth draft
  |
  | @  2 third secret
  | |
  | o  1 second draft
  |/
  o  0 first public
  

  $ hg fix --all

  $ hg log --graph --template '{rev} {desc}\n' -r 'sort(all(), topo)' --hidden
  o  11 fifth
  |
  o  9 fourth
  |
  | @  8 third
  | |
  | o  7 second
  |/
  | *  10 sixth
  | |
  | | x  5 sixth
  | |/
  | x  4 fifth
  | |
  | | x  6 fifth
  | |/
  | x  3 fourth
  |/
  | x  2 third
  | |
  | x  1 second
  |/
  o  0 first
  

  $ hg cat -r 7 foo.whole
  TWO
  $ hg cat -r 8 foo.whole
  THREE
  $ hg cat -r 9 foo.whole
  FOUR
  $ hg cat -r 10 foo.whole
  SIX
  $ hg cat -r 11 foo.whole
  SEVEN
  $ cat foo.whole
  UNCOMMITTED
#else
  $ printf "one\n" > foo.whole
  $ hg commit -Aqm "first"
  $ hg phase --public
  $ hg tag --local root
  $ printf "two\n" > foo.whole
  $ hg commit -m "second"
  $ printf "three\n" > foo.whole
  $ hg commit -m "third" --secret
  $ hg tag --local secret
  $ hg checkout root
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ printf "four\n" > foo.whole
  $ hg commit -m "fourth"
  created new head
  $ printf "uncommitted\n" > foo.whole

  $ hg log --graph --template '{rev} {desc} {phase}\n'
  @  3 fourth draft
  |
  | o  2 third secret
  | |
  | o  1 second draft
  |/
  o  0 first public
  

  $ hg fix --all
  saved backup bundle to * (glob)

  $ hg log --graph --template '{rev} {desc} {phase}\n'
  @  3 fourth draft
  |
  | o  2 third secret
  | |
  | o  1 second draft
  |/
  o  0 first public
  
  $ hg cat -r 0 foo.whole
  one
  $ hg cat -r 1 foo.whole
  TWO
  $ hg cat -r 2 foo.whole
  THREE
  $ hg cat -r 3 foo.whole
  FOUR
  $ cat foo.whole
  UNCOMMITTED
#endif

  $ cd ..