FilterBox visualisation runs one extra query

[IanSavchenko]

Make sure these boxes are checked before submitting your issue - thank you!

• I have checked the superset logs for python stacktraces and included it here as text if any
• I have reproduced the issue with at least the latest released version of superset
• I have checked the issue tracker for the same issue and I haven't found one similar

Superset version

0.22.1

Expected results

For FilterBox visualization (widget) Superset should run one query per filter to get possible filter values.

Actual results

Superset makes two queries (actually, N + 1, where N - number of filters in the widget)

Steps to reproduce

Create FilterBox, and run the query. You will see one extra query on DB end. Easy to reproduce with default dashboard "World's Banks Data" and existing FilterBox.

This is an issue for us because for one of the queries FilterBox does take quite long time (around a minute), but in fact, this time gets doubled in the end, because there are at least two queries run.

I managed to find the source of the issue: file superset/viz.py lines 276, 278. See code and my comments:

# first query executed here by design, 
# but it's results are actually ignored in FilterBoxViz subclass method `get_data`
df = self.get_df()    
  if not self.error_message:
    # N queries are executed here
    data = self.get_data(df) 

Here, in the subclass FilterBoxViz (lines 1532-1564):

# df not used here!
def get_data(self, df): 
        qry = self.query_obj()
        filters = [g for g in self.form_data['groupby']]
        d = {}
        for flt in filters:
            qry['groupby'] = [flt]
            
            # N "legit" queries are executed in the loop here
            df = super(FilterBoxViz, self).get_df(qry)
            d[flt] = [{
                'id': row[0],
                'text': row[0],
                'filter': flt,
                'metric': row[1]}
                for row in df.itertuples(index=False)
            ]
        return d

I'm not a Python dev and not really sure how to fix this. I would override some other methods in FilterBoxViz subclass, but since get_df is used in more methods like get_csv (it must be also broken now for FilterBox, btw), I don't know what is the right design. If nobody steps in, I will try to make a PR, but this fix should be trivial for those who know this code.

[mistercrunch]

Thanks for reporting this and providing all the detail. Fix is here: #4276