dylanpulver opened a new issue, #10938:
URL: https://github.com/apache/arrow-rs/issues/10938
### Describe the bug
`can_cast_types` returns `true` for `Interval(YearMonth) -> Int64` and
`Interval(DayTime) -> Int64`, but `cast_with_options` has no arm for either and
returns `CastError`.
`can_cast_types` at `arrow-cast/src/cast/mod.rs:324-330`; the only
`Interval`/integer arm in `cast_with_options` is `(Int32, Interval(YearMonth))`
at `mod.rs:2312`, so both fall through to the catch-all `Err`.
Both sides were added together in #1196. #5769 made `IntervalDayTime` and
`IntervalMonthDayNano` structured types and its description lists "Remove cast
support from the Int64Arrays to/from those types" — the `cast` side went, the
`can_cast_types` arm stayed.
This is reachable beyond the API check. `cast/union.rs::resolve_child_array`
picks the first union child that `can_cast_types` accepts, so a union with an
interval child and a `Utf8` child fails to cast to `Int64` even though the
`Utf8` child casts fine.
### To Reproduce
```rust
use arrow_array::{ArrayRef, IntervalYearMonthArray};
use arrow_cast::{can_cast_types, cast};
use arrow_schema::{DataType, IntervalUnit};
use std::sync::Arc;
assert!(can_cast_types(
&DataType::Interval(IntervalUnit::YearMonth),
&DataType::Int64
));
let a: ArrayRef = Arc::new(IntervalYearMonthArray::from(vec![12]));
cast(&a, &DataType::Int64).unwrap();
// CastError("Casting from Interval(YearMonth) to Int64 not supported")
```
Same for `Interval(DayTime)`. `Interval(MonthDayNano)` already reports
`false`.
### Expected behavior
`can_cast_types` and `cast` agree. No interval unit has an unambiguous `i64`
value, so making `can_cast_types` return `false` matches #5769's intent — but
implementing the cast instead is a defensible alternative if you would rather
have it.
### Additional context
`arrow/tests/array_cast.rs::test_can_cast_types` exists to keep the two in
sync and does not catch this, because `Int64` is missing from
`get_all_types()`: lines 505-514 read `Int8, Int16, Int32, UInt64, UInt8,
UInt16, UInt32, UInt64`, with `UInt64` in `Int64`'s slot and then again in its
own. Line 568 of the same file has the intended sequence. Restoring `Int64`
makes that test fail on main, and it is the only mismatch restoring it exposes.
Investigated with AI assistance (Claude Opus 4.8); the repro above was run
against `c44f8d4`.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
