Add support for MySQL’s LIMIT with optional offset argument.

[temochka]

MySQL supports a LIMIT statement with an optional offset argument, i.e. LIMIT 10, 20 which stands for "fetch 20 records after skipping the first 10".

Noticed this while working on #18. I didn’t want to couple these changes, since I’m not sure if you’re interested in supporting such extensions of SQL standard. If you are, I can help with merging these.

[r0man]

If this can be done with offset, I don't think we need to change this. I use this library mostly with PostgreSQL and the unwritten plan is to support features of that dialect.

I'm not sure if I want to get into the business of adding vendor specific SQL (other than PostgreSQL) at this point. I think this needs some thought, proper tests against real databases and a design change in the compiler.

That said, I'm happy to make changes like the LIMIT ALL that allows people to run this against other databases.

Roman

[temochka]

I agree that it looks like it could be solved by adding more polymorphism into compiler, i.e. dispatch based on platform’s supported features (GCC back-end probably does something similar). However it’s hard to find the right balance and not introduce some sort of linter/optimizer for SQL syntax while going in that direction (i.e. replacing LIMIT 0,1 with LIMIT 1 OFFSET 0). Personally I’d like to see sqlingvo as a lightweight DSL that allows people to build and compose SQL queries while delegating validation to a target database. Considering this, it’s still not clear to me how to transparently support platform-specific extensions like MySQL’s LIMIT X,Y. I’m not even mentioning GROUP_CONCAT(DISTINCT column_name ORDER BY column_name DESC SEPARATOR ',') (oh, wait, I am) :-)

I’m totally fine with not merging this pull request, just interested in your opinion about this.

[r0man]

Hi Artem,

you can try this:

(require '[sqlingvo.compiler :refer [defarity compile-whitespace-args]])
(defarity compile-whitespace-args
  "group-concat")

(select [:student-name '(group-concat DISTINCT :test-score ORDER BY :test-score DESC SEPARATOR " ")]
  (from :student)
  (group-by :student-name))
;=> ["SELECT \"student_name\", group-concat(DISTINCT \"test_score\" ORDER BY \"test_score\" DESC SEPARATOR ?) FROM \"student\" GROUP BY \"student_name\"" " "]

Regarding platform specific extensions? The only thing I had in mind is dispatching not only on :op, but also on the database vendor, providing a default implementation for most of the stuff and let oeople override vendor specifc compilation ops.

Since I'm a happy PostgreSQL user and don't have to deal with anything else at the moment, this is not very high priority at the moment ...

If you have a better solution, I'm all ears.

Roman

[r0man]

I think something like this for dispatching http://stackoverflow.com/questions/9440253/defmethod-catch-all