Currently we have this:
`program.alloc_cursor_id(Option<String>, CursorType)`
where the `String` is the table's name or alias ('users' or 'u' in the
query).
This is problematic because this can happen:
`SELECT * FROM t WHERE EXISTS (SELECT * FROM t)`
There are two cursors, both with identifier 't'. This causes a bug where
the program will use the same cursor for both the main query and the
subquery, since they are keyed by 't'.
Instead introduce `CursorKey`, which is a combination of:
1. `TableInternalId`, and
2. index name (`Option<Arc<Index>>` -- in case of index cursors.)
This should provide key uniqueness for cursors:
`SELECT * FROM t WHERE EXISTS (SELECT * FROM t)`
here the first 't' will have a different `TableInternalId` than the
second `t`, so there is no clash.
---
These `CursorKey`s are only required when the program needs to retrieve
the cursor ID later:
`program.resolve_cursor_id(key)`
So, there are now two methods for allocating cursors:
`program.alloc_cursor_id_keyed(key, cursor_type); // needs to be
retrieved later with same key`
`program.alloc_cursor_id(cursor_type); // does not need to be retrieved
later`
Reviewed-by: Preston Thorpe (@PThorpe92)
Closes#1604
Currently we have this:
program.alloc_cursor_id(Option<String>, CursorType)`
where the String is the table's name or alias ('users' or 'u' in
the query).
This is problematic because this can happen:
`SELECT * FROM t WHERE EXISTS (SELECT * FROM t)`
There are two cursors, both with identifier 't'. This causes a bug
where the program will use the same cursor for both the main query
and the subquery, since they are keyed by 't'.
Instead introduce `CursorKey`, which is a combination of:
1. `TableInternalId`, and
2. index name (Option<String> -- in case of index cursors.
This should provide key uniqueness for cursors:
`SELECT * FROM t WHERE EXISTS (SELECT * FROM t)`
here the first 't' will have a different `TableInternalId` than the
second `t`, so there is no clash.
This is the first step towards rollback, since we still don't spill
pages with WAL, we can simply invalidate page cache in case of failure.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#1599
Closes: https://github.com/tursodatabase/limbo/issues/1602
Fix unreachable panic when calling serialize on Value::Integer(0)
by handling `SerialTypeKind` `ConstInt0` and `ConstInt1` in
`Record::serialize()`
Changed `test_serialize_integers` to reflect this change.
Reviewed-by: Pere Diaz Bou <pere-altea@homail.com>
Closes#1603
overwrite_cell() requires that the cursor state is pointing to a valid
record, but this was not currently set properly.
Reviewed-by: Pere Diaz Bou <pere-altea@homail.com>
Closes#1598
With Blacksmith runners, it errors out as follows:
```
npm notice Publishing to https://registry.npmjs.org/ with tag next and public access
npm notice publish Signed provenance statement with source and build information from GitHub Actions
npm notice publish Provenance statement published to transparency log: https://search.sigstore.dev/?logIndex=221582002
npm error code E422
npm error 422 Unprocessable Entity - PUT https://registry.npmjs.org/@tursodatabase%2flimbo - Error verifying sigstore provenance bundle: Unsupported GitHub Actions runner environment: "self-hosted". Only "github-hosted" runners are supported when publishing with provenance.
npm error A complete log of this run can be found in: /home/runner/.npm/_logs/2025-05-28T08_42_29_830Z-debug-0.log
Error: Process completed with exit code 1.
```
Currently in the main translation logic after planning and optimization,
we don't _really_ need to pass a `&mut Vec<WhereTerm>` around anymore,
except for the fact that virtual table constraint resolution is done ad-
hoc in `init_loop()`.
Even there, the only thing we mutate is `WhereTerm::consumed` which is a
boolean indicating that the term has been "used up" by the optimizer and
shouldn't be evaluated as a normal where clause condition anymore.
In the upcoming branch for WHERE clause subqueries, I want to store
immutable references to WHERE clause expressions in `Resolver`, but this
is unfortunately not possible if we still use the aforementioned mutable
references.
Hence, we can temporarily make `WhereTerm::consumed` a `Cell<bool>`
which allows us to pass an immutable reference to `init_loop()`, and the
`Cell` can be removed once the virtual table constraint resolution is
moved to an earlier part of the query processing pipeline.
Closes#1597
Currently in the main translation logic after planning and optimization,
we don't _really_ need to pass a &mut Vec<WhereTerm> around anymore, except
for the fact that virtual table constraint resolution is done ad-hoc in
`init_loop()`. Even there, the only thing we mutate is `WhereTerm::consumed`
which is a boolean indicating that the term has been "used up" by the optimizer
and shouldn't be evaluated as a normal where clause condition anymore.
In the upcoming branch for WHERE clause subqueries, I want to store immutable
references to WHERE clause expressions in `Resolver`, but this is unfortunately
not possible if we still use the aforementioned mutable references.
Hence, we can temporarily make `WhereTerm::consumed` a `Cell<bool>` which allows
us to pass an immutable reference to `init_loop()`, and the `Cell` can be removed
once the virtual table constraint resolution is moved to an earlier part of the
query processing pipeline.
Currently we have some usages of LIMIT where the actual limit counter is
initialized next to the DecrJumpZero instruction, and then
`program.mark_last_insn_constant()` is used to hoist the counter
initialization to the beginning of the program.
This is very fragile, and already FROM clause subquery handling works
around this with a hack (removed in this PR), and (upcoming) WHERE
clause subqueries would also run into problems because of this, because
the LIMIT might need to be initialized once for every iteration of the
subquery.
This PR removes those usages for LIMIT, and LIMIT processing is now more
intuitive:
- limit counter is now initialized at the start of the query processing
- a function init_limit() is extracted to do this for
select/update/delete
Closes#1591
Currently we have some usages of LIMIT where the actual limit counter
is initialized next to the DecrJumpZero instruction, and then
`program.mark_last_insn_constant()` is used to hoist the counter
initialization to the beginning of the program.
This is very fragile, and already FROM clause subquery handling works
around this with a hack (removed in this PR), and (upcoming) WHERE clause
subqueries would also run into problems because of this, because the LIMIT
might need to be initialized once for every iteration of the subquery.
This PR removes those usages for LIMIT, and LIMIT processing is now more
intuitive:
- limit counter is now initialized at the start of the query processing
- a function init_limit() is extracted to do this for select/update/delete
1. allow calling op_null with Insn::BeginSubrtn
- BeginSubrtn is identical to Null, but named differently so that
its use in context is clearer
2. Insn::Return: add possibility to fallthrough on non-integer values as
per sqlite spec
Closes#1588
This pull request implements the `libsql_wal_get_frame()` API. To do
that, we also introduce a `wait_for_completion()` API in I/O dispatcher.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#1533
Found this when reviewing #1528 locally and this was crashing
```sql
INSERT INTO t SELECT * FROM generate_series(1,10,1);
```
Reason was that `op_vopen` was not replacing the already allocated
cursor slot, but using `.insert()`
Reviewed-by: Pere Diaz Bou <pere-altea@homail.com>
Closes#1583
If we don't reset the state of `IdxDelete`, next `IdxDelete` will start
in `Deleting` state which is completely wrong since it should seek from
the start.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#1584
Closes#1528 .
- Modified `translate_select` so that the caller can define if the
statement is top-level statement or a subquery.
- Refactored `translate_insert` to offload the translation of multi-row
VALUES and SELECT statements to `translate_select`
- I did not try to change much of `populate_column_registers` as I did
not want to break `translate_virtual_table_insert`. Ideally, I would
want to unite this remaining logic folding `populate_column_registers`
into `populate_columns_multiple_rows` and the
`translate_virtual_table_insert` into `translate_insert`. But, I think
this may be best suited for a separate PR.
## TODO
- ~Tests~ - *Done*
- ~Need to emit a temp table when we are selecting and inserting into
the Same Table -
https://github.com/sqlite/sqlite/blob/master/src/insert.c#L1369~ -
*Done*
- Optimization when table have the exact same schema - open an Issue
about it
- Virtual Tables do not benefit yet from this feature - open an Issue
about it
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#1566
Before #1570 , we never read anything from the on-disk WAL at startup,
so the `test_wal_frame_count()` test assertion `frame_count == 0`would
always pass, regardless of whether there was anything in the WAL or not
Now that the actual bug is fixed, the current test isn't idempotent
because the `clone_test_db.sh` command does not clear the wal, just the
db file
Closes#1581
