log a warning if dd is enabled but output product is `corrected_data`

[Athanaseus]

Describe the problem that the feature should address
This feature will provide more info for the user in the case ddcal is to be performed, but the output is not set appropriately.

Describe the solution you'd like
Give a warning that a direction-dependent term is enabled (dE.direction_dependent: true) but will not be applied since output product is corrected_data instead of corrected_residual.

Additional context
This applies only direction-independent terms.
https://gist.github.com/Athanaseus/c02cc9f309bcd381ce5b66944e813d7d?permalink_comment_id=4718996#gistcomment-4718996

[IanHeywood]

Relatedly, is there any use case for performing DD cal for N directions where subtract-dirs isn't of length N-1 or N? This might be another good scenario for triggering a warning if not

[JSKenyon]

Thanks for the report @Athanaseus. This is something that needs further work i.e. better informing users of oddities in configuration. For the sake of my own memory, a couple of the common pitfalls (which may change as I remember them) are:

• Failing to set output.subtract_directions correctly when doing direction-dependent calibration.
• Expecting corrected_data output to be corrected for direction-dependent effects.
• Failing to set term.direction_dependent when a direction-dependent model is specified.

[JSKenyon]

Relatedly, is there any use case for performing DD cal for N directions where subtract-dirs isn't of length N-1 or N? This might be another good scenario for triggering a warning if not

There might be an argument for a hybrid of peeling + subsequent application of DD solutions during imaging (which will eventually be available). I think in that case there may be a need for more selective peeling. That said, it would be a very special case so it is probably worth warning the user.

[IanHeywood]

Ah yes, I am looking forward to this functionality! 🙌