Correction of multiline UPDATE statements in the Postgres driver

[therufa]

Added support for returning values in multi-line UPDATE query statements to the PostgreSQL database driver

[therufa]

The feature is provided by the PostgreSQL engine, and it would be great to be able to use this feature without any hack in CI.

Source:
http://www.postgresql.org/docs/9.2/static/sql-update.html#AEN79130

[narfbg]

Nice one, although it needs some additional work:

• Please check our styleguide, use tabs for indentation and remove the empty line at EOF.
• Add a changelog entry.
• I haven't checked, but if INSERT and UPDATE support the RETURNING clause, doesn't REPLACE do as well?
• Furthermore, is it SET that you're also adding to or UPDATE ... SET ?

[therufa]

I do hope, that I could accomplish everything properly.

By the way; there is no REPLACE statement in the PostgreSQL.
If I do understand your question right, yes I wanted the SET at that location :)

[therufa]

i hope everything is right now

[narfbg]

So, you're telling me that you can do something like this?

SET time zone 'Europe/Sofia' RETURNING '2';

This seems quite odd to me. :)
Otherwise ... the PR is not complete, but I'll do the afterwork so that it doesn't take days in explanations between the two of us. You'll see what I mean afterwards.

[therufa]

Well, the set part might make no sense with your example, but if you work with multi-line queries it can be quite annoying that the SET keyword turns the query to a non-reading type.

The example of @fema90 shows exactly the problem. Look at the second line; that SET there would break the whole query, and the result would be that you have no result, but that the query went well. Thats not what I want in such a case.

I hope this explains my step. :)

[therufa]

So I think the current solution fits everything, except what you've mentioned (to be honest I don't really understand what you've meant, with the PR, I hope that is not too much).

[narfbg]

@fema90's example shows an UPDATE query, the regular expression that I asked you about would validate a query that bebings with a SET. The one now, after your last commit is even messier. :)

See the above commit for the cleaner way. Sorry for making you add a changelog entry btw, I just realized it was the same bug that was fixed for INSERT, so no need to have 2 lines in there.

[therufa]

Well, this does not solve the base problem (at least not "mine"), but maybe I just don't understand something, and maybe its off topic:
Should I pass only single-line queries to the db->query() method?

In the Postgres community (as it is in the Oracle and in the MsSQL) its common practice to write long queries, and for readability these are split into multiple lines.

So wouldn't it make sense to support this practice, or is it considered as wrong? I'm just asking right now.

[narfbg]

The good old PCRE ignoring everything after the newline character, even without the 'm' modifier ... should be OK now. :)