<@U030M2AL48M> I'm running into something that pre...
# integration-snowflake
@jamie I'm running into something that previously didn't happen and I wonder if it has to do with the snowflake-python connector or how you are feeding it table names. Previously, I could define:
Copy code
def ABCD_my_asset(...) -> pd.DataFrame:
    return df
The table created would be called
in Snowflake, in all caps. Now, however, the table created is called
but upon trying to save the DataFrame (and during a validation step? or maybe it's at the start of the next step?), dagster is looking for the table by its name in all caps:
and not finding it. Is it possible something changed with how the asset definition function names are converted into the Snowflake table names? Maybe you're sending them in double-quotes instead of without quotes? That's the only thing I know of that would cause Snowflake to create them not in all caps.
i don’t think anything changed on our end, but i’ll look into it! can you share the version of dagster, dagster-snowflake, and snowflake-connector-python you have?
sure. I'm thinking maybe it was a problem with cached metadata? Or that dagster was pre-compiling SQL but did not recompile it when I updated the module with the asset/job?
I did a full rebuild of all of the containers and tossed out the postgres database and the problem went away but I can see a remnant of it hanging out in Snowflake
That's the only time I've seen dagster create tables that weren't all caps
huh very strange. i’ve also never seen it make lower case tables, but we aren’t doing anything on our end to convert asset names to all caps or anything, we just pass it along to the pd.DataFrame.to_sql call
weird. I'll see if I can track it down but I'll be out for a few days. Thanks anyhow!
yeah i’ll dig a bit too, but it’ll be hard to track down if i can’t replicate haha. if you see it again definitely let me know!