Allowing PROBABILITY OF * and GIVEN *

[MathieuHuot]

Similarly to the currently allowed

SELECT * 
FROM data

as sugar for

SELECT data.col1, data.col2,...
FROM data

We can add

SELECT PROBABILITY OF * AS Pr_Data UNDER model 
FROM data

as sugar for

SELECT PROBABILITY OF model.col1 = data.col1 and model.col2=data.col2 and ... AS Pr_Data UNDER model 
FROM data

for all columns which appear both in model and data.

Likewise, they could be compatible with the EXCEPT keyword (that I've seen used in the AutoML paper).

[MathieuHuot]

These features came up as possibly useful when trying to write compact high-level examples in the PLDI submission.

GIVEN * can be used for imputation or prediction, and as GIVEN Null should behave the same as not conditioning, it would work well in the presence of missing data.

PROBABILITY OF * could e.g. be used for anomaly detection to estimate the likelihood of rows from data under a model.