Add process_lhs() method for Transform

[Anssi Kääriäinen]

The Lookup class has two methods, process_rhs() and process_lhs(). Typical usage is:

def as_sql(self, compiler, connection):
    lhs_sql, lhs_params = self.process_lhs(compiler, connection)
    rhs_sql, rhs_params = self.process_rhs(compiler, connection)
    # return the sql

The typical usage for transforms is:

def as_sql(self, compiler, connection):
    lhs_sql, lhs_params = compiler.compile(self.lhs)
    # return the sql

I think it could make sense to add process_lhs() to transform base class. This way one could write transforms using the same process_lhs() method that is used for Lookups, too. At least for me it is hard to remember when to use process_lhs() and when to use compiler.compile.

[Josh Smeaton]

Could I suggest going the other way and standardising on compiler.compile rather than process_X? I'm not sure how much work process_lhs does in the normal lookups, so it may not be feasible. But I'd like to standardise on the expressions way to ease the gap between them.

[Samriddha Kumar Tripathi]

Hi, I’ve been exploring this ticket and would like to try working on it. I noticed it’s been open for a while, so I’ll first check whether the proposed change still applies to the current codebase. I’m still learning the internals, but I’ll start by reviewing how Transform and Lookup handle SQL compilation and see if adding process_lhs() improves consistency. I’ll share any progress or questions here as I go.

[Jacob Walls]

​PR

Thanks for the PR. Don't forget to uncheck Patch Needs Improvement on this ticket when tests are passing to signal you're ready for a review.

[Samriddha Kumar Tripathi]

Got it, I’ll uncheck the box once the tests pass. Thanks!