The colon bites us again, see also the previous issue fiboa/specification#36
@cholmes wrote on Slack:
I'm having a slight bit of doubt about colon as standard for our extensions. My tooling will add an admin boundary, then partition on it, including into hive partitions in folders, which theoretically leads to folder names like admin:country_code=CO. Doing so made for failing tests in windows, and just realized that on mac it renames to admin/country_code=CO. This may be fine if hive readers interpret it right, but I'm sorta doubting they will...
I found this: https://hive.apache.org/development/desingdocs/dynamicpartitions/#design-issues
Under point 2 it raises the issue about the field names in the directory and says this is to be solved by the tooling, but I guess the implementors didn't test on Windows so didn't really care about the colon? Anyway, it lets me think that it's an implementation issue, not a (vecorel) spec issue. Maybe we can clarify this in the hive spec? Not sure...
DuckDB also has this short but related chapter: https://duckdb.org/docs/stable/data/partitioning/partitioned_writes#handling-slashes-in-columns
The colon bites us again, see also the previous issue fiboa/specification#36
@cholmes wrote on Slack:
I found this: https://hive.apache.org/development/desingdocs/dynamicpartitions/#design-issues
Under point 2 it raises the issue about the field names in the directory and says this is to be solved by the tooling, but I guess the implementors didn't test on Windows so didn't really care about the colon? Anyway, it lets me think that it's an implementation issue, not a (vecorel) spec issue. Maybe we can clarify this in the hive spec? Not sure...
DuckDB also has this short but related chapter: https://duckdb.org/docs/stable/data/partitioning/partitioned_writes#handling-slashes-in-columns