Stephen Hoover

@John3-16 , thank you for the report. What change? If it's a generally applicable fix, you could contribute it as a pull request.

It sounds like you're trying to install from conda-forge. We don't yet have a conda binary for Windows on conda-forge. You could try to install via pip, but I see...

+1 ! You can tag me for review. I've dealt with similar issues in my own code.

The types will be populated without a full table scanner run. You can use something like ```python db_id = client.get_database_id(database) tables = client.tables.list(database_id=db_id, schema=schema, name=table) columns = client.tables.get(tables[0].id).columns dtypes =...

Hm. Good point. `pandas.read_csv` accepts a `dtype` argument, which is how we'd implement this. The `dtype` doesn't have to include all columns, and it can include columns which aren't present....

For me, the main point of confusion was in not using `table_columns`. I'd assumed that when appending to an existing table using `dataframe_to_civis`, the DataFrame column information would be used...

To be clear, this is an instance where the Civis API changed mid-notebook session and the user wanted to access the new API endpoint(s) from an existing notebook, correct?

I've discovered that the MLPClassifier does not encode string targets when y is 2D. Is that part of this issue, or should I file a new issue? It's not clear...

The reason to handle encoding for multilabel inputs is so that we can do multilabel-multiclass, so this issue sounds like it's still a blocker. @beckermr , I agree it's unclear....