Query planner should validate distinct is passed a field

[pauldix]

When the user does a select distinct query, it should validate that the argument passed to distinct is a field. If it's a tag it should return an error like: select distinct must be passed a field. To query the unique values for a tag use "show tag values from measurement_name with key = 'key_name'"

Where measurement_name and key_name are replaced with whatever was in the query.

[corylanou]

Currently, they fail as expected:

> select distinct host from mem
ERR: host isn't a field on measurement mem
> select distinct(host) from mem
ERR: host isn't a field on measurement mem
> select count(distinct host) from mem
ERR: host isn't a field on measurement mem

However, I suspect once we merge #2598 that it will stop working. If that is the case, I have to wait for that PR to get finished before I can tackle this one?

[pauldix]

I think you can get to this one before that PR gets merged. The validation would still happen in the query planning stage when looking at what is passed into distinct.

The only thing I would say is that the error message should be more helpful since there is a way to get at what they're asking for i.e. show tag values from foo with key = 'bar'

[corylanou]

Ok, I was thinking more about testing. I can write a unit test which should be fine, but I won't be able to finish out integration tests which will leave the issue open until the other PR is merged.

[pauldix]

Ah right you are

[toddboom]

Still seems to be working after #2598:

> select distinct host from "mydb"."default"."cpu_load_short";
ERR: host isn't a field on measurement cpu_load_short
> select distinct(host) from "mydb"."default"."cpu_load_short";
ERR: host isn't a field on measurement cpu_load_short
> select count(distinct host) from "mydb"."default"."cpu_load_short";
ERR: host isn't a field on measurement cpu_load_short

[corylanou]

There is a distinction between a SELECT DISTINCT foo and SELECT count(DISTINCT foo) As they are different intentions, you will get a custom error for the first, and the general for the second. The thinking is that you can't do aggregate functions on tags, but you can do select distincts, just not with that syntax (as the error will inform you).

So now, if you issue either of these queries (where host is a tag, not a field):

SELECT distinct(host) FROM cpu
SELECT distinct host FROM cpu

You will get an error like:

host isn't a field on measurement cpu; to query the unique values for a tag use SHOW TAG VALUES FROM cpu WITH KEY = "host"

If you issue a query like this:

SELECT count(distinct host) FROM cpu
SELECT count(distinct(host)) FROM cpu

You will get the generic message like:

host isn't a field on measurement cpu

[toddboom]

that seems fine to me

[pauldix]

@corylanou that looks right. Eventually we'd want them to be able to do a count(distinct ...) on a tag value. Maybe update the error to be host isn't a field on measurement cpu. count(distinct) on tags isn't yet supported