On Wed, Feb 6, 2013 at 11:20 AM, Seamus Abshere <seamus@abshere.net> wrote:
> merlin,
>
> Yes, you're correct, my phrasing was bad: all I meant was that it was a
> conflict, not a bug in Postgres or hstore.
>
> I personally don't know of any way around the conflict except changing JDBC
> or hstore, and I don't think JDBC is gonna change.
>
> Deciding not to accommodate JDBC on the Postgres side, though, is going to
> prevent hstore from being used properly with Java or any JVM-based language
> like JRuby.
>
> Please let me know if my assumptions are wrong.
This problem is not unique to ? character. Hibernate for example
reserves the use of : character for name parameter insertion with
similarly ridiculous results. Basically every language I know of
except for the java stack seems to understand that when embedding
constructs into a foreign language there must be some type of escaping
mechanism (note they may in fact allow this in some level: via
googling it isn't clear).
The point is that Postgres should not introduce language constraints
because of broken driver technology. To move forward in your
particular case, consider:
*) switching to 'hstore defined()' function:
*) hacking pg_operator (carefully look up and change oprname for the
specific hstore operator)
merlin
