You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Put it in a separate crate, parallel to [optd-datafusion-bridge](https://github.com/cmu-db/optd/tree/main/optd-datafusion-bridge)
There exists a crate under the datafusion tree, datafusion-substrait, that claims to convert from datafusion to substrait representations. However, on the physical side (which is what we need to convert), it seems like it has the same problem that would be encountered if we were to send impl ExecutionPlan trait objects around, in that it doesn't seem to know the type or any information about each node. It seems like we wouldn't want to use it, in this case, and instead convert directly from optd nodes. I may be wrong on this---check for yourself here.
Bonus points: make sure serialization is working (can be a separate/later PR)
The text was updated successfully, but these errors were encountered:
@jurplel If you find anything that Substrait plans might be lacking on regarding a proper representation of Physical Plans, I would be interested in hearing them. We are definitely looking for opportunities to improve Substrait as we can learn a lot from actual usage of it.
[optd-datafusion-bridge](https://github.com/cmu-db/optd/tree/main/optd-datafusion-bridge)
datafusion
tree,datafusion-substrait
, that claims to convert fromdatafusion
tosubstrait
representations. However, on the physical side (which is what we need to convert), it seems like it has the same problem that would be encountered if we were to sendimpl ExecutionPlan
trait objects around, in that it doesn't seem to know the type or any information about each node. It seems like we wouldn't want to use it, in this case, and instead convert directly fromoptd
nodes. I may be wrong on this---check for yourself here.The text was updated successfully, but these errors were encountered: