[GitHub] [arrow] wesm commented on issue #6578: ARROW-7371: WIP: [GLib] Add GLib binding of Dataset

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

wesm commented on issue #6578:
URL: https://github.com/apache/arrow/pull/6578#issuecomment-617515815


   I haven't looked at the details of this binding too much, but I wanted to let you know that I'm taking a closer look at the way that filter expressions work in the datasets API in the context of being able to support more general purpose expression evaluation -- beyond the scope of just the datasets API -- i.e. with many more functions. 
   
   In particular, I am concerned about having significant bindings for the `Expression` subclasses until we feel confident in the C++ API that we have an array-expression API that can accommodate the expanded scope of general purpose query processing. In general, the expressions here are not specific to datasets -- we should be working toward an expression API (closely tied to a kernel/function catalog) that can also be used in projections, hash aggregations, join predicates, and other query processing uses. 
   
   So until then, I would recommend that you make minimal bindings of the factory functions needed to be able to form filters in the datasets API and avoid wrapping the expression subclasses if you can. This will save you work now and potentially spare us painful refactoring or API breaks later.


----------------------------------------------------------------
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