Hey , we are trying use groupby to do list aggregation essentially we have a column for bbox for our data : [x1,x2,y1,y2] . We want to do groupby and list aggregate using a unique id to get a list of bboxes [[x1,x2,y1,y2],[x1,x2,y1,y2],[x1,x2,y1,y2]] ? however when we use groupby and use list aggregate it raise a list of list not implemented issue.

Seems like we just had not implemented this. Should be a simple fix @Kevin Wang wanna take a quick stab? @yashovardhan chaturvedi could you create an issue? Much appreciated!

sure awesome thank you 🙂 I will create an issue.

Did you try grouped_df.agg_list(col('bbox'))?

yes.

Maybe you could change the bbox into a string with list.join(',') then agglist then list.join with a different separator but it's very roundabout

yes for now that's what we are thinking of doing. But we loose the ability to do operations to derive insights from the bbox's to filter our data before we get it ready to ingest it for training workflows. The same goes for more complex structs.

Agreed.. i am also keen on having these and also more list operations on the grouped list of lists

yep we are exploring daft so that we can integrate it with ray datasets 🙂 .

😂 that string conversion seems highly unnecessary If you’re looking for a workaround for now as we work on the issue you might be able to use Python code in a

.map_groups

I see will keep in mind 🙂