Intent

This PR introduces a large change to our *_join functions.

Before, we manually handled selection of columns and the translation from == to identical for columns with NA.

Now, we instead piggyback on dbplyr translations to handle this for us.

This should ensure we cover many more edge cases. In addition, we now also have good implementations of
full_join, anti_join and semi_join.

Approach

Instead of creating our overwritten joins via the sql_on argument and a subsequent select statement, we now create a special by argument that uses the a combination of na_matches = "never" (the default, non NA matching option) and na_matches = "na" to construct a join that selectively joins NA with NA.

This means many of our old helper functions (select helper and sql_on helper) are now replaced with helpers that handles dplyr::join_by() objects.

The new combined by argument does break the select statement of the dbplyr translation, so we have to manually fix it.

Known issues

This PR builds on un-merged PR

• fix(db_joins): Allow dplyr::join_by as by argument #156

Checklist

• The PR passes all local unit tests
• I have documented any new features introduced
• If the PR adds a new feature, please add an entry in NEWS.md
• A reviewer is assigned to this PR