Knob to allow alternative formatting of call chains

In addition to the standard method of getting call chains formatted like this:

rawKPIs = KPIsSession.query(
    KPI.name, KPI.value, KPI.created
).filter(KPI.entryId == entryID).filter(KPI.subsetId is None).filter(
    KPI.created >= fromUnixTimestamp(fromTimestamp)
).filter(KPI.created <= fromUnixTimestamp(toTimestamp)).filter(
    func.unix_timestamp(KPI.created) % moduloSelector == 0
).order_by(KPI.name, KPI.created).all()

Allow formatting as:

rawKPIs = \
    KPIsSession.query(
        KPI.name, KPI.value, KPI.created
    ).filter(
        KPI.entryId == entryID
    ).filter(    
        KPI.subsetId is None
    ).filter(
        KPI.created >= fromTimestamp
    ).filter(
        KPI.created <= toTimestamp
    ).filter(
        KPI.created % moduloSelector == 0
    ).order_by(
        KPI.name, KPI.created
    ).all()

or:

rawKPIs = \
    KPIsSession.query(KPI.name, KPI.value, KPI.created) \
    .filter(KPI.entryId == entryID) \
    .filter(KPI.subsetId is None) \
    .filter(KPI.created >= fromTimestamp) \
    .filter(KPI.created <= toTimestamp) \
    .filter(KPI.created % moduloSelector == 0) \
    .order_by(KPI.name, KPI.created) \
    .all()

This makes the code more readable and stack traces clearer. (due to line number)

[qsorix]

Hi. I’m confused by the comments here. Assuming I already have ( ) surrounding my chained call, is it possible to configure yapf to do this?

rawKPIs = (
    KPIsSession.query(KPI.name, KPI.value, KPI.created) 
    .filter(KPI.entryId == entryID) 
    .filter(KPI.subsetId is None) 
    .filter(KPI.created >= fromTimestamp) 
    .filter(KPI.created <= toTimestamp) 
    .filter(KPI.created % moduloSelector == 0) 
    .order_by(KPI.name, KPI.created) 
    .all()
)

Right now my yapf (0.20) turns that code into this, which is far from what I want 😉.

rawKPIs = (
    KPIsSession.query(KPI.name, KPI.value, KPI.created)
    .filter(KPI.entryId == entryID).filter(KPI.subsetId is None)
    .filter(KPI.created >= fromTimestamp).filter(KPI.created <= toTimestamp)
    .filter(KPI.created % moduloSelector == 0).order_by(KPI.name,
                                                        KPI.created).all()
)

I have a lot of chained calls in my pyspark application, and poor formatting of those is the one reason preventing me from using yapf. Ideally I’d like something like in the first quoted fragment.

[advance512]

Another idea is adding parenthesis, but this is obviously more problematic:

rawKPIs = (
    KPIsSession.query(KPI.name, KPI.value, KPI.created) 
    .filter(KPI.entryId == entryID) 
    .filter(KPI.subsetId is None) 
    .filter(KPI.created >= fromTimestamp) 
    .filter(KPI.created <= toTimestamp) 
    .filter(KPI.created % moduloSelector == 0) 
    .order_by(KPI.name, KPI.created) 
    .all()
)