Rich-T-kid opened a new issue, #11116: URL: https://github.com/apache/arrow-rs/issues/11116
**Is your feature request related to a problem or challenge?** Part of https://github.com/apache/arrow-rs/issues/7761 `MapArray` goes through `GenericInProgressArray` → `concat_maps`. A MapArray is structurally identical to `List<Struct<key, value>>` — it has an offset buffer and an entries `StructArray` child. A specialized implementation mirrors the proposed `InProgressListArray` (see the companion List issue) but with the entries child dispatched to `InProgressStructArray`. **Describe the solution you'd like** ```rust pub(crate) struct InProgressMapArray { entries_field: FieldRef, ordered: bool, source: Option<ArrayRef>, batch_size: usize, nulls: NullBufferBuilder, offsets: Vec<i32>, // Map always uses i32 offsets entries: InProgressStructArray, // entries is always a StructArray } ``` **Implementation ideas:** 1. **Reuse `InProgressListArray` logic**: the offset accumulation for Map is identical to `List<i32>` — factor out a common `InProgressOffsetArray` trait or helper 2. **Entries specialization**: the entries child is always a `StructArray` with key and value fields; use `InProgressStructArray` for the entries, which recursively specializes key and value fields 3. **Combined benefit**: a `Map<Utf8, Int64>` gets: - Incremental i32 offsets (vs. scan-then-reconstruct in `concat_maps`) - Entries struct with `InProgressPrimitiveArray` for the Int64 values - Keys as Utf8 (through GenericInProgressArray for now, or specialized if List PR lands) **`create_in_progress_array` dispatch:** ```rust DataType::Map(field, ordered) => { // entries field is always a StructArray containing key+value fields Box::new(InProgressMapArray::new(Arc::clone(field), *ordered, batch_size)) } ``` **Benchmarks** ``` cargo bench --bench coalesce_kernels --features test_utils -- map ``` -- 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]
