paleolimbot commented on PR #62:
URL: https://github.com/apache/arrow-nanoarrow/pull/62#issuecomment-1291997577
Yes, the pattern of an object responsible for calling the release callback
`struct ArrowWhatever` and doing a "move" whenever ownership changes is most
consistent with how Arrow C++ does it. The user-allocated cffi structs would
then be intermediaries (like they are when interacting with Arrow C++).
You almost certainly don't want to expose the Array/ArrayView difference to
actual users, but it's probably a useful concept in the internals (in the R
package you can create an external pointer to an ArrayView that keeps the
appropriate objects alive, but that's never exposed to the user). The interface
that I exposed in the R package is more like (apologies for what is almost
certainly invalid cython):
```cython
cdef class Schema:
cdef:
ArrowArray array
object parent
cdef class Array:
cdef:
ArrowArray array
ArrowSchema* schema_ptr # (might be NULL)
object schema # (might be None or point to a Schema)
object parent
```
You can do a `nanoarrow_array_set_schema()` to attach a schema to an array
(which I do internally when importing). You can in theory have an Array without
a schema but this mostly happens at intermediary phases of passing the objects
around.
--
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]
