We are using Metabase 0.44.6 with Clickhouse. We are using the unofficial metabase-clickhouse-driver as this is the only one available.
When Metabase does its "Scans", it also includes columns that are of
Map(String, Int64). That is, it runs following query on a
-- Metabase SELECT `my_db`.`my_table`.`my_column` AS `my_column_ex_544d1323` FROM `my_db`.`my_column` GROUP BY `my_db`.`my_table`.`my_column` ORDER BY `my_db`.`my_table`.`my_column` ASC LIMIT 1000 FORMAT TabSeparatedWithNamesAndTypes
This query is non performant and on average takes 1295secs (i.e. ~21mins)... per column of this type.
Our tables are HUGE.... > 1+ Billion rows, AND we have 30+ such tables and each table has 2-3 such columns.
My understanding is that Scanning is to sample values for drop downs, variable etc. It is unlikely that columns with such types every be used in 'Field Filters' or even other kind of variable inputs and thus Scanning these kinds of fields is wasteful.
Should this somehow be optimised such that only columns with simple types such as String, Integer, Bool, Enum, date, time, datetime etc should be sampled?