[postmodern-devel] Dash in Postgresql Column name - "query-dao"

[Marijn Haverbeke]

Hello Phil,

Postmodern works this way because I wasn't even aware that one could
use column names like that -- though actually, seeing the way quotes
work, it's rather obvious. Currently the assumption 'dashes have to be
converted to underscores' is present all through the system. Is
support for such column names very important for you? My first
recommendation would be to just rename the thing to use an
underscore... but if you have a good reason to use a dash, I guess
we'll have to look into another solution.

Best,
Marijn


On Wed, Dec 3, 2008 at 5:59 PM, Phil <zaries at global.co.za> wrote:
> On Wed, 2008-12-03 at 17:16 +0100, Maciej Pasternacki wrote:
>> On 2008-12-03, at 16:28, Phil wrote:
>>
>> >>> query-dao does not seem to like column names with a dash in them
>> >>> like
>> >>> column-name and I suspect it would have the same problem with
>> >>> spaces.
>> >>>
>> >>> I know that the preffered for postgresql would be column_name but I
>> >>> though that being able to use the dash would simplify going from
>> >>> db to
>> >>> dao object to html and back, well at least for me.
>> >>>
>> >>> Is column names with dashes just not supported, a bug or on the wish
>> >>> list?
>> >>
>> >> Everything is supported with :RAW sql-op:  (select-dao 'foo (:= (:raw
>> >> "\"foo-bar\"") 23)) calls query (SELECT * FROM foo WHERE ("foo-bar" =
>> >> 23)) (SELECT-DAO is a shortcut for QUERY-DAO, where you supply only
>> >> the WHERE clause).  There doesn't seem to exist any other way now,
>> >> but
>> >> it may be possible to add a bit of macro sugar.
>> >
>> > Ok maybe I should have been a bit more precise in my description. I
>> > was
>> > not using the columns names in the query.
>> >
>> > Here follows a quick repeatable test:
>> >
>> > CREATE TABLE test
>> > (
>> >  id integer NOT NULL,
>> >  "some" character varying NOT NULL,
>> >  "test-dash" character varying
>> > )
>> > WITH (OIDS=FALSE);
>> > ALTER TABLE test OWNER TO postgres;
>> >
>> > insert into test (id,"some","test-dash") values(1, 'some','dash-1');
>> >
>> > (defclass test ()
>> >  ((id :col-type string :initarg :id
>> >   :accessor biographical-id)
>> >   (some :col-type  (or db-null string) :initarg :some
>> >     :accessor biographical-some)
>> >   (test-dash :col-type  (or db-null string) :initarg :test-dash
>> >    :accessor biographical-test-dash))
>> >  (:metaclass postmodern:dao-class)
>> >  (:keys id))
>> >
>> > (postmodern:query-dao 'test (:select '* :from 'test))
>> >
>> > That last bit produces the following error:
>> >
>> > No slot named test-dash in class TEST. DAO out of sync with table, or
>> > incorrect query used.
>> >   [Condition of type SIMPLE-ERROR]
>>
>>
>> I see.  It was SELECT-DAO that made me think there's problem in the
>> query.  What you write about is not supported at the moment, and while
>> it's easy to add this, it's not trivial.
>>
>> I just looked into postmodern/table.lisp and I see that in class
>> DIRECT-COLUMN-SLOT there is a slot SQL-NAME.  It would seem trivial to
>> add an initarg, e.g. :COL-SQL-NAME, for this slot, and modify SHARED-
>> INITIALIZE of DIRECT-COLUMN-SLOT to set this slot only if it's not
>> already bound by an initarg.  This way it would be possible to
>> add :COL-SQL-NAME initarg to the slot to explicitly set column name,
>> with upside being it wouldn't need to be a symbol.
>>
>> However, this slot seems to be rarely used, and there are many cases
>> where simple TO-SQL-NAME is called on slot name, which would need to
>> be fixed.  Also, column name would need to be properly double-quoted
>> everywhere.
>>
>> Your problem seems interesting and it looks like it would be useful to
>> support your case.  If all that was needed would be adding the initarg
>> to :COL-SQL-NAME, the patch would be probably already attached to this
>> post ;) but the way it is, I can only give these starting points so
>> that you can fix it yourself.  It shouldn't be hard, but testing the
>> patch may be a bit time-consuming.
>>
>> Regards,
>> Maciej.
>>
>
> Maciej thanx for the reply. Unfortunatley what I know about LISP is
> dangerous and what I know about CLOS is non existant (think I read to
> many of Paul Grahams articles, not a fan of objects...lol). I had to
> re-read your explanation a couple of times just to get the jist of
> it...lol
>
> I would be more than willing to do the grunt work if it would be
> something that could be incorperated into postmodern, but I would need a
> lot more guidence. I dont want to go hacking postmodern for myself (with
> bad lisp style) and then end up with a version that is broken if and
> when postmodern is updated.
>
> Regards
> Phil
>
>
>
>
> _______________________________________________
> postmodern-devel mailing list
> postmodern-devel at common-lisp.net
> http://common-lisp.net/cgi-bin/mailman/listinfo/postmodern-devel
>