SELECT behavior

[pp86]

Please consider this query:

DATA_SET_VAR = SELECT(provider == "Campaner") HG19_BED_ANNOTATION;

MATERIALIZE DATA_SET_VAR into test_select;

It returns two samples.
Btw, no sample with provider = "Campaner" exists. The two output samples have the metadata original_provider = "Campaner".

What happens is that the SELECT checks is the attribute name ends with "provider".

I remember we decided to have a SQL-like approach, i.e. the names are splitted on "." (similar to the "mariage"/"age" example we discussed). So the results would be correct if the metadata was "original.provider", but not "original_provider".

Is anything changed in the specification, or should this be fixed?
@marcomass @sunbrn

[sunbrn]

According to what is written in the documentation, in the SELECT predicate on metadata only metadata exactly equal to what is specified should be matched, e.g., only metadata which are exactly "provider" (without prefix of any kind).

I think the matter with the separator (".") is only relevant when we talk about the default/EXACT/FULL options for semijoin/joinby_/_groupby predicates.
I mean that, when using SELECT, we should concern about the separator if we are writing something like
SELECT(...; semijoin: FULL(provider) IN DS)

Therefore, @pp86 I would fix the metadata predicate to only match the indicated metadata.
@marcomass could you please confirm? Am I missing something?

[marcomass]

I confirm; it must be fixed. Thanks @pp86 for spotting it

[marcomass]

Can this be fixed by @OlgaGorlova ?

[pp86]

In the case we have a sample with:
original.provider = "Campaner"
should the query
provider == "Campaner"
select the sample or not?

According to @sunbrn it should not, right?

@marcomass @OlgaGorlova I am taking care of it.