https://github.com/malloydata/malloy/pull/1325

[mtoy-googly-moogly]

Cleaned up and normalized aggregate expressions which hadn't been touched in a while ... this includes some things which will require docs changes ...

source is now a keyword , this is a breaking change.

Will warn when m4warnings are enabled ...

• COUNT(DISTINCT expression) is gone, will warn and suggestcount(expression)
• COUNT(*) is gone, will warn and suggest count()
• SUM/AVG(path) will warn and suggest path.SUM/AVG()
• SUM/AVG(expr) will warn and suggest source.SUM/AVG()

Here's the full matrix of support ...

Function	f()	f(expr)	source.f()	source.f(expr)	path.f()	path.f(expr)
min/max	err	ok	err	ok	ok	err
sum/avg	err	warn	err	ok	ok	err
count	ok	ok	ok	ok	ok	err

[christopherswenson]

In your matrix, can we clarify path with source_path and field_path? That is, I'm not sure with the current matrix how some_join.avg(some_field) behaves in comparison tosome_join.some_field.avg(). Also, it seems like the grammar doesn't have support for source.stddev(foo), which is a "custom" aggregate function. Should we have that? Maybe that can be a separate PR.

Function	f()	f(expr)	source.f()	source.f(expr)	srcpath.f()	srcpath.f(expr)	fldpath.f()
min/max	err	ok	err	ok	?	?	?
sum/avg	err	warn	err	ok	?	?	?
count	ok	ok	ok	ok	?	?	?
stddev	?	?	?	?	?	?	?

[mtoy-googly-moogly]

@christopherswenson i don't understand what a "source path" is as opposed to a "field path"

Yes, this totally ignores custom aggregate functions. We can delay fixing this until we have a solution for all things. This was a quick-ish attempt to address something @lloydtabb said was M4 blocking. We can for sure discuss if it makes sense to do this in two parts or wait for it all at once.

[mtoy-googly-moogly]

@christopherswenson is doing some new work based on this, closing this PR as his work starts from this branch

[mtoy-googly-moogly]

re-opening just for the checklist