[GitHub] [arrow] jorisvandenbossche commented on pull request #7240: ARROW-8792: [C++][Python][R][GLib] New Array compute kernels implementation and execution framework

[GitBox <gi...@apache.org>]

jorisvandenbossche commented on pull request #7240:
URL: https://github.com/apache/arrow/pull/7240#issuecomment-633196165


   I was playing a little bit with the python API:
   
   - Ran into the following issue:
   
       ```python
       >>> import pyarrow.compute as pc                                                                                                                                                                               
   
       # works nicely
       >>> pc.call_function("take", [pa.array([0, 1, 2, 3, 4, 5]), pa.array([0, 4, 1, None])])                                                                                                                        
       <pyarrow.lib.Int64Array object at 0x7f7cc8954a68>
       [
         0,
         4,
         1,
         null
       ]
   
       # segfaults
       >>> pc.call_function("filter", [pa.array([0, 1, 2, 3]), pa.array([True, False, True, None])])                                                                                                                  
       Segmentation fault (core dumped)
       ```
   
   * In array.pxi / table.pxi, some of the methods' implementation has been replaced with a `_pc.call_function(..)` call, and some are still using the C++ compute functions (eg `Take`). The goal is to eventually have them all go through `_pc.call_function()`?
   
   * `pc.call_function()` doesn't yet have a way to pass options. That's also something to be added in the future?
   


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org