dnwpark
dnwpark
@vpetrovykh I think you are correct that `array_set` should throw on out of bounds. I was thinking it should imitate `array_get` as their names are similar, but as a modifying...
@vpetrovykh I've updated `array_set` to raise an out of bounds exception. Given that the slicing works well to remove values, perhaps we should hold off on adding a remove function....
@vpetrovykh Agreed. I've reduced this PR to just `array_set` and `array_insert`.
@vpetrovykh Did some benchmarking on whether to "combine" branches in the implementation. Running each 10k times: - set_empty: `SELECT array_set([], 0, 1)` - set_out_of_bounds: `SELECT array_set([1,2,3,4,5,6,7,8,9], 25, -1)` - set_front:...
Needs lots of tests, especially for the schema. While compiler can handle nested arrays, the protocol can't ``` _localdev:main> select [[1]]; edgedb error: ProtocolEncodingError: array shape is invalid ``` `array_unpack`...
@aljazerzen I think it would still be necessary to point to the index itself. I think this should be done after enabling multiple spans in a single error.
The flag is a bit awkward, the end goal is to have two flags, `is_explain` and `use_type_inheritance_ctes` while removing the current `expand_inhviews`. Edit: flag is removed!
For example, given: ``` type Foo { n -> int64; } type Foo2 extending Foo; type Bar { foo -> Foo { n -> str } } type Bar2 extending...
Ollama support done for now. HF support probably needs to come after #8515
@anbuzin @elprans