Sort schema columns alphabetically

[susodapop]

What type of PR is this? (check all applicable)

• Feature

Description

This PR sorts the column names shown in the schema viewer alphabetically at the API level. It also moves table name sorting from the Model code into the get_schema() method of the schema resource. If the sort raises an exception then the API returns the columns in the order sent by the query runner.

Users will need to refresh their schemas before the change is visible in the front-end.

Related Tickets & Documents

https://github.com/redashlabs/product/issues/15.

Mobile & Desktop Screenshots/Recordings (if there are UI changes)

Before

After

[susodapop]

I'd like to add tests for this behavior. What's the best way to do so? I have a few ideas:

1. Inject a schema into the data source factory
2. patch the get_schema method to return an unordered list
3. Build an in-memory data-source and point SQLA there

[arikfr]

Re. tests, there are three options I can think of:

1. Mock get_schema to return some data.
2. Extract sorting logic into its own function and test it with dummy data.
3. Test this at the handler level with more of an integration test, and use the test Postgres database as the target.

[susodapop]

I used @arikfr's second testing option. Extracted the sort into its own function and tested with dummy data.

[susodapop]

Pending further feedback I think this is ready to merge.

[susodapop]

Not sure what's causing CircleCI e2e tests to fail. I didn't change any front-end files.

[gabrieldutra]

Not sure what's causing CircleCI e2e tests to fail. I didn't change any front-end files.

😪, looks like it kept flaky after #4545, for now it's a matter of re-trigger the build. I guess Cypress indeed keeps the chained element and we will need to make sure the results are fresh before asserting (cc @kravets-levko).

[kravets-levko]

@gabrieldutra I found another thing that makes pivot test flaky: cypress input is too fast. When test adds a new line to query - it immediately clicks Execute button. But query editor is debounced, so 50/50 it will execute query with old text - therefore test fails. Here is my attempt to fix this - ec2ede4 I have no other idea for how to detect when query object received a new text from editor.

[susodapop]

I'm continuing to test local. This isn't ready to merge yet. When connected to Redash's internal metadata database the schema sort doesn't work at all.

[susodapop]

Found the issue. When I refactored to _sort_schema() I returned the wrong variable from the sort method. And I trusted my test instead of testing manually.

          _ __            
   __  __(_) /_____  _____
  / / / / / //_/ _ \/ ___/
 / /_/ / / ,< /  __(__  ) 
 \__, /_/_/|_|\___/____/  
/____/                    
                          

[susodapop]

Alrighty. I added a second test case that would have caught this issue addressed in a31f901.

[susodapop]

On my second unit test, would it be better to use call_count to ensure that the sort_schema() method is called? It feels strangely duplicative using identical assertions on separate tests.

The first test verifies the sort behavior. The second test verifies that Redash uses that sort behavior. Seems like the call_count is the important bit.

Thoughts @arikfr or @LevkO?

[arikfr]

👍