[
https://issues.apache.org/jira/browse/ARROW-8927?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Antoine Pitrou updated ARROW-8927:
----------------------------------
Component/s: GPU
C++
> [C++] Support dictionary memos when reading/writing record batches using cuda
> IPC
> ---------------------------------------------------------------------------------
>
> Key: ARROW-8927
> URL: https://issues.apache.org/jira/browse/ARROW-8927
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++, GPU
> Reporter: Alex Baden
> Priority: Major
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Currently, the cuda IPC calls for `ReadRecordBatch` do not accept a
> dictionary memo as a parameter, building an empty memo before calling
> `ipc::ReadRecordBatch`. As such, adding a test which duplicates
> `TestCudaArrowIpc, BasicWriteRead` but calls `MakeDictionary` to create the
> record batch fails because the dictionaries for the original record batch are
> not found. The cuda IPC calls should be modified to take a dictionary memo,
> just like the standard IPC API.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
