Superset ignore case-sensitivity of time column names with Postgresql

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
• A closed issue seems to be related: #2588 but was closed because inactivity.

Superset version

Current master: 85692612d6f1438d8b19f51aefae92a240bc3cf3

Expected results

A table with column name where case is important (for example invoiceDate) should be usable without any problem.

Actual results

When creating a chart with such a table, I get the following error:

column "invoicedate" does not exist LINE 1: SELECT DATE_TRUNC('day', invoiceDate) AT TIME ZONE 'UTC' AS ... ^ HINT: Perhaps you meant to reference the column "InvoiceLine.invoiceDate".

It seems to happens because of the Time Column case sensitivity in combination with a Time Grain (here day).

Steps to reproduce

Import a table with case-sensitive names.

[villebro]

TL;DR: I was talking more about how the query is generated in Superet, not what is valid SQL syntax. I believe that in this instance putting quotes around {col} for Postgres is probably the simplest solution, but I would not apply the same to all other engine types.

Long answer: I was referring to how Superset generates SQL queries. Currently most other parts of the query are generated using SQL Alchemy constructs that automatically detect the need for quotes. For example, selecting a column ColumnName with an alias col_alias is in fact roughly defined as column('ColumnName').label('col_alias'), which is then compiled into SELECT "ColumnName" AS col_alias. If either require quoting, ColumnName in this example, SQL Alchemy automatically puts quotes around it.

The problem here is that time grain expressions don’t contain a reference to a SQL Alchemy column object, but are static strings inside a literal_column object, which makes it impossible for SQL Alchemy to apply conditional quoting. See below where this happens:

https://github.com/apache/incubator-superset/blob/master/superset/connectors/sqla/models.py#L139-L143

I believe that Postgres is pickier than other engines in this case, i.e. most other engines probably wouldn’t require quotes around the column name in the time grain expression. Therefore I would not force quotes into the time grain expressions of other engines at this time.

[victornoel]

For the record, a second workaround was provided here: https://github.com/apache/incubator-superset/pull/5962#issuecomment-425623999.

If it is possible, I would like to go farther than workarounds so that this can be fixed seriously in superset (for the sake of correctness and for future users 😉. I can contribute if needed with proper guidance 😃