First review #820
The function follows RFC 7386 JSON Merge Patch semantics:
* If the patch is null, the target is replaced with null
* If the patch contains a scalar value, the target is replaced with that
value
* If both target and patch are objects, the patch is recursively applied
* null values in the patch result in property removal from the target
Closes#821
Change JSON deserialization to enable json_patch implementation with
SQLite-compatible behavior:
* Preserves duplicate keys in JSON objects
* Applies patches only to the first occurrence of each key
* Trade-off: Changes key lookup from O(1) to O(n) to support duplicate
keys
* Have to be merged before json_patch() function
Closes#820
Json serialization logic was pulled from serde_json. Google's json5
serialization code was not flexible enough to allow for pretty printing
json, so I believe that the formatter design is a good layer to abstract
this logic. This refactor will trivially enable the implementation of
json_pretty function from sqlite. My other PR for json_quote, #763,
depends a tiny bit on a helper utility from the previous serialization
implementation. If this PR is considered first, I will change the code
in my other PR to account for this.
Reviewed-by: Diego Reis (@diegoreis42)
Reviewed-by: Kacper Madej (@madejejej)
Closes#771
## Purpose of this PR
- Automatically add labels to PRs for easy classification
- We can add useful rules as needed
## Changes
- Add 2 files to configure labeler
- For now, changes under certain directory will add labels
## ETC
- workflow repo: https://github.com/marketplace/actions/labelerCloses#817
This PR brings the Go database/sql driver to it's first working state
and adds a Go package to demonstrate.
The example pkg demonstrates (in it's most bare/naive form at this
point):
1. Open database (memory, in this case)
2. Create connection
3. Prepare statement (Create table)
4. `Exec`
5. Prepare statement (Insert, bind 3 arguments (int, string, blob) (
6. `Exec`
7. Prepare statement (Select *)
8. `Columns` -> print columns
9. `Query` -> print rows
10. Close db connection
still tons of work to do but I at least wanted to get it to a state
where it's not totally broken.
I'll add some actual tests tomorrow
Closes#796
This simple patch makes sure we can operate with a reference to the
string instead of being forced to transform it to a string, and makes
sure that the Arc doesn't have to be cloned (which can be expensive in
multi-core systems).
This doesn't really make a large difference in benchmarks, given how
expensive Parse::new() is.
Reviewed-by: Preston Thorpe (@PThorpe92)
Closes#808
This pull requests adds some libSQL vector extension functions such as
`vector()` and `vector_distance_cos()`, which can be used for exact
nearest neighbor search as follows:
```
limbo> SELECT embedding, vector_distance_cos(embedding, '[9, 9, 9]')
...> FROM movies ORDER BY vector_distance_cos(embedding, '[9, 9, 9]');
[4, 5, 6]|0.013072490692138672
[1, 2, 3]|0.07417994737625122
```
Note that libSQL also support approximate nearest neighbour search with
DiskANN indexing, which is something we eventually want to port to Limbo
as well.
Closes#798
This simple patch makes sure we can operate with a reference to the
string instead of being forced to transform it to a string, and makes
sure that the Arc doesn't have to be cloned (which can be expensive in
multi-core systems).
This doesn't really make a large difference in benchmarks, given how
expensive Parse::new() is.
This patch adds some libSQL vector extension functions such as
`vector()` and `vector_distance_cos()`, which can be used for exact
nearest neighbor search as follows:
```
limbo> SELECT embedding, vector_distance_cos(embedding, '[9, 9, 9]')
...> FROM movies ORDER BY vector_distance_cos(embedding, '[9, 9, 9]');
[4, 5, 6]|0.013072490692138672
[1, 2, 3]|0.07417994737625122
```
Now:
```sql
limbo> create table t (a,b,c); insert into t (a,b,c) values ("hello", 234, 432);
thread 'main' panicked at core/translate/expr.rs:1621:29:
internal error: entered unreachable code: Id should be resolved to a Column before translation
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
```
After fix:
```sql
limbo> create table t (a,b,c); insert into t (a,b,c) values ("hello", 234, 432);
× Parse error: no such column: "hello" - should this be a string literal in single-quotes?
limbo>
```
Closes#800
Reviewed-by: Diego Reis (@diegoreis42)
Reviewed-by: Preston Thorpe (@PThorpe92)
Closes#801
## Purpose of this PR
- Implement `close()` method for `LimboStatement`(+`JDBC4Statement`) and
`LimboResultSet`(+ `JDBC4ResultSet`)
## Changes
- Add `consumeAll` method in `LimboResultSet`
- Implement `close()` methods
- Because `JDBC4Statement` has longer lifecycle in compared to
`LimboStatement`, we manage different `close` fields(`LimboStatement` is
created when first `execute` method is called on `JDBC4Statemenet`)
## Reference
- [Issue](https://github.com/tursodatabase/limbo/issues/615)
Closes#799
