cisaacstern commented on PR #27618: URL: https://github.com/apache/beam/pull/27618#issuecomment-1773849886
Ok, narrowing in on the issue here! Looks like this _does_ relate to the "to cast to list or not to cast to list" discussion I had with myself starting in https://github.com/apache/beam/pull/27618#discussion_r1357577926. As shown here, _not_ casting to list results in no computation actually happening: ```python # bag.py import dask.bag as db import distributed as ddist bag = db.from_sequence([1]) bag_dict = {"bag": bag} if __name__ == "__main__": with ddist.Client() as c: bag_result = c.gather(c.compute(bag)) print(f"{bag_result=}") dict_vals_result = c.gather(c.compute(bag_dict.values())) print(f"{dict_vals_result=}") dict_vals_to_unpacked_list_result = c.gather(c.compute(*list(bag_dict.values()))) print(f"{dict_vals_to_unpacked_list_result=}") ``` ```console ➜ python bag.py bag_result=[1] dict_vals_result=dict_values([dask.bag<from_sequence, npartitions=1>]) dict_vals_to_unpacked_list_result=[1] ``` So looks like casting to list is definitely the way to go. I think I backed off that change because it caused a serialization issue (which caused tests to fail). So I'll re-implement the list casting now and re-examine what that serialization issue was, and report back. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
