Pekka Enberg
a72b07e949
Merge 'Fix VDBE program abort' from Nikita Sivukhin
...
This PR add proper program abort in case of unfinished statement reset
and interruption.
Also, this PR makes rollback methods non-failing because otherwise of
their callers usually unclear (if rollback failed - what is the state of
statement/connection/transaction?)
Reviewed-by: Preston Thorpe <preston@turso.tech >
Closes #3591
2025-10-07 09:07:07 +03:00
bit-aloo
66c69461fb
Add getter/setter for checkpoint threshold in LogicalLog
...
Wire threshold access through Storage
Add checkpoint threshold accessors to MvStore
2025-10-07 10:17:04 +05:30
Pekka Enberg
f8bdc02986
Merge 'Rename Completion methods' from Pedro Muniz
...
Reviewed-by: bit-aloo (@Shourya742)
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com >
Closes #3579
2025-10-06 20:12:21 +03:00
Pere Diaz Bou
59c08c1062
Merge 'core/mvcc: implement PartialOrd for RowId' from Pere Diaz Bou
...
Closes #3597
2025-10-06 16:42:04 +02:00
pedrocarlo
5a7390735d
rename Completion functions
2025-10-06 11:07:06 -03:00
Nikita Sivukhin
e2f7310617
add explicit tracker for Txn cleanup necessary for statement
2025-10-06 17:51:43 +04:00
Pekka Enberg
41d909e1fb
Merge 'MVCC: do checkpoint writes in ascending order of rowid' from Jussi Saurio
...
For insert-heavy checkpoints this gives a much higher chance of using
the balance-quick subalgorithm instead of the more complex and slower
balance-nonroot.
Closes #3589
2025-10-06 16:30:19 +03:00
Pere Diaz Bou
aed255d2e6
core/mvcc: implement PartialOrd for RowId
2025-10-06 13:43:42 +02:00
Pere Diaz Bou
fc7e1639a0
core/mvcc: filter out seek results where is not same table_id
2025-10-06 13:29:10 +02:00
Pere Diaz Bou
b9b9831d17
core/mvcc: test seek with empty table
2025-10-06 13:28:51 +02:00
Nikita Sivukhin
8dae601fac
make rollback non-failing method
2025-10-06 13:21:45 +04:00
Nikita Sivukhin
38d2630969
remove unnecessary SchemaLocked error
...
- lock() return error in case when another thread panicked while holding the same lock
- we better to just panic too in any such case
2025-10-06 12:15:15 +04:00
Jussi Saurio
fd9008f123
MVCC: do checkpoint writes in ascending order of rowid
...
For insert-heavy checkpoints this gives a much higher chance of using
the balance-quick subalgorithm instead of the more complex and slower
balance-nonroot.
2025-10-06 10:11:19 +03:00
Pere Diaz Bou
e3f2bbf90b
core/mvcc: remove unnecessary fields in CommitState::Checkpoint
2025-10-03 22:27:03 +02:00
Pere Diaz Bou
c612a51df7
fmt
2025-10-03 22:24:38 +02:00
Pere Diaz Bou
8e80af7221
core/mvcc: checkpoint if needed after commit
2025-10-03 22:24:38 +02:00
Pere Diaz Bou
a30c95d988
core/mvcc/logical-log: add threshold for checkpoints in logical log
2025-10-03 22:24:33 +02:00
Pere Diaz Bou
55f5cd2ad7
core/mvcc/logical-log: test automatic checkpoint
2025-10-03 17:23:59 +02:00
Pere Diaz Bou
7db5a09005
core/mvcc/logical-log: checkpoint support not updating transaction_state if needed
2025-10-03 17:23:35 +02:00
Pekka Enberg
297aaf4887
core/mvcc: Rename "-lg" to "-log"
...
The "lg" name is just weird.
2025-10-03 10:08:02 +03:00
pedrocarlo
131a5b8048
adjust logical log IO functions to return Completions and not IOResult
2025-10-03 01:44:41 -03:00
Jussi Saurio
9e4ea6ea34
Merge 'core/mvcc/logical-log: fail in read_more_data if couldn't read enough' from Pere Diaz Bou
...
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com >
Closes #3500
2025-10-02 06:58:28 +03:00
Jussi Saurio
bb4e54ca73
Merge 'fix/mvcc: deserialize table_id as i64' from Jussi Saurio
...
Reviewed-by: Nikita Sivukhin (@sivukhin)
Closes #3492
2025-10-02 06:58:01 +03:00
Pere Diaz Bou
b89df44339
fmt
2025-10-01 14:31:30 +02:00
Pere Diaz Bou
dc0245d758
core/mvcc/logical-log: fail in read_more_data if couldn't read enough
2025-10-01 14:25:22 +02:00
Jussi Saurio
e9f0c59bcc
fix/mvcc: set log offset to end of file after recovery finishes
...
otherwise we start overwriting existing log entries
2025-10-01 12:46:24 +03:00
Jussi Saurio
bcb941f33b
fix/mvcc: deserialize table_id as i64
2025-10-01 10:26:23 +03:00
Jussi Saurio
d4d50b564a
fix even more tests
2025-09-30 23:22:07 +03:00
Jussi Saurio
adc5b7b27f
remove monkey print
2025-09-30 22:57:21 +03:00
Jussi Saurio
fe871188bf
fix tests again
2025-09-30 22:54:48 +03:00
Jussi Saurio
fb2878973f
fix sort order of write set
2025-09-30 22:54:36 +03:00
Jussi Saurio
fd84fd0683
fix test compilation errors
2025-09-30 22:27:28 +03:00
Jussi Saurio
e68c652f8f
Add some table ID integrity checks to logical log recovery
2025-09-30 22:27:28 +03:00
Jussi Saurio
64ce33bd5c
Move resolution of tableid/rootpage inside MvCursor constructor
2025-09-30 17:04:37 +03:00
Jussi Saurio
7c897d382f
Implement MvTableId newtype for better type safety of table ids
2025-09-30 16:54:22 +03:00
Jussi Saurio
0ba4c6c00e
use negative table id in mvcc tests
2025-09-30 16:53:12 +03:00
Jussi Saurio
a52dbb7842
Handle table ID / rootpages properly for both checkpointed and non-checkpointed tables
...
Table ID is an opaque identifier that is only meaningful to the MV store.
Each checkpointed MVCC table corresponds to a single B-tree on the pager,
which naturally has a root page.
We cannot use root page as the MVCC table ID directly because:
- We assign table IDs during MVCC commit, but
- we commit pages to the pager only during checkpoint
which means the root page is not easily knowable ahead of time.
Hence, we:
- store the mapping between table id and btree rootpage
- sqlite_schema rows will have a negative rootpage column if the
table has not been checkpointed yet.
2025-09-30 16:53:12 +03:00
Pekka Enberg
9788f6d005
Merge 'core/mvcc: Optimize exclusive transaction check' from Pekka Enberg
...
The check is in fastpath so switch to an atomic instead.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com >
Closes #3458
2025-09-30 16:02:42 +03:00
Pekka Enberg
3d327ba63c
core/mvcc: Optimize exclusive transaction check
...
The check is in fastpath so switch to an atomic instead.
2025-09-30 15:00:24 +03:00
Pere Diaz Bou
0f631101df
core: change page idx type from usize to i64
...
MVCC is like the annoying younger cousin (I know because I was him) that
needs to be treated differently. MVCC requires us to use root_pages that
might not be allocated yet, and the plan is to use negative root_pages
for that case. Therefore, we need i64 in order to fit this change.
2025-09-29 18:38:43 +02:00
Pekka Enberg
65b382b9e1
Merge 'Make MVCC code Send and Sync' from Pekka Enberg
...
Closes #3361
2025-09-27 08:21:01 +03:00
Pekka Enberg
a932ac5450
Merge 'core: recover logical log on Database::connect ' from Pere Diaz Bou
...
If we open database and logical log is not empty we need to recover from
it. We also make sure a single recover executes concurrently and other
connections just wait for it to finish.
I also changed the fuzz tester to use `restart` instead of calling
`load_logical_log` manually to test this behaviour.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com >
Closes #3359
2025-09-27 08:20:41 +03:00
Pere Diaz Bou
99adf73168
core/mvcc/logical-log: rename to needs_recovery
2025-09-26 16:59:57 +02:00
Pere Diaz Bou
9e47cc3700
clippy
2025-09-26 14:16:11 +02:00
Pekka Enberg
1402e9841e
core/mvcc: Wrap StreamingLogicalLogReader::buffer with RwLock
2025-09-26 14:19:16 +03:00
Pekka Enberg
96accef06c
core/mvcc: Wrap header with RwLock
2025-09-26 14:10:18 +03:00
Pere Diaz Bou
9c1d94a355
core/mvcc/logical-log: assert we don't call begin_load_tx twice
2025-09-26 12:59:13 +02:00
Pere Diaz Bou
4cdf293a2b
core/mvcc/logical-log: fuzz test recover use db.restart
2025-09-26 12:56:58 +02:00
Pere Diaz Bou
ae994146af
core/mvcc/logical-log: on mvcc restart clear DATABASE_MANAGER
2025-09-26 12:56:43 +02:00
Pere Diaz Bou
83d8a7c775
core/mvcc/logical-log: lock recover logical log process
2025-09-26 12:47:31 +02:00
