Changelog for splits is missing elements in `after_root_ids`

[bdpedigo]

Example

Requires caveclient, but I think is more an issue with what pychunkedgraph is returning

For splits, get_tabular_change_log only shows one "after_root_ids" per operation:

Code

import caveclient as cc
import pandas as pd

client = cc.CAVEclient("minnie65_phase3_v1")

cg = client.chunkedgraph

root_id = 864691136143786292

change_log = cg.get_tabular_change_log(root_id)[root_id]
change_log.set_index("operation_id", inplace=True)

splits = change_log.query("~is_merge")

splits["after_root_ids"].head()

Output

operation_id
244007    [864691136020603640]
244012    [864691135799937890]
244040    [864691135274393061]
244043    [864691135472223282]
244050    [864691136521807761]
Name: after_root_ids, dtype: object

Compare to get_operation_details

Code

details = cg.get_operation_details(splits.index.to_list())
details = pd.DataFrame(details).T
details["roots"].head()

Output

244007    [864691136309915226, 864691136020603640]
244012    [864691135737704049, 864691135799937890]
244040    [864691135700510587, 864691135274393061]
244043    [864691135182021634, 864691135472223282]
244050    [864691136444023171, 864691136521807761]
Name: roots, dtype: object

The latter properly returns both root ids that were created from a split operation.

[bdpedigo]

update: after talking to Forrest, I see that the only one returned in roots is the child of the split operation that was associated with the original root_id query. still think it might be nice to have both stored in the table somehow