[Qgis-developer] addFeatures without knowing new feature's id, bug?

[Denis Rouzaud]

Hi Jürgen,

Thanks for your reply.

Yesterday, I used the max+1 solution, but as you say this not optimal. I 
will try to define it safely.

But I already have a serial in my table: CREATE TABLE mytable (id serial 
NOT NULL);
I also tried to define on my own:
CREATE SEQUENCE mytable_id_seq ;
CREATE TABLE mytable (id integer DEFAULT nextval('mytable_id_seq') NOT 
NULL);
but it is strictly the same.

If you have any clue, it's great! I'm a little bit lost.

Isn't because QGIS is trying to write each field, ie:
INSERT INTO table (id,field1,field2,...) VALUES ('','val1','val2',...)
instead of doing
INSERT INTO table (field1,field2,...) VALUES ('val1','val2',...) ?

Btw, if you want to have a look on how I created the feature (I might 
have do something wrong), here at lines 95-113
https://github.com/3nids/Triangulation/blob/ba601c44282ba31c20c3484d123bff8241fbdc7e/place_arc.py 


Thanks again!

Greetings,
Denis




On 01/16/2012 04:54 PM, Jürgen E. Fischer wrote:
> Hi Denis,
>
> On Mon, 16. Jan 2012 at 13:04:01 +0100, Denis Rouzaud wrote:
>> I am trying to add a feature to a PostGIS table, in which there is a
>> serial primary key 'id'.
>> Without defining id, here is what I get when I try addFeatures:
>> ERROR: null value in column "id" violates not-null constraint
> You could setup a DEFAULT value for the id column in PostgreSQL (like SERIAL
> does it).
>
>> and I tried
>> iid  = self.layer.dataProvider().fieldNameIndex('id')
>> if iid != -1:
>> f.addAttribute(iid,'nextval(distribution.dimension_id_seq::regclass)')
>> which leaded to
>> ERROR: invalid input syntax for integer:
>> "nextval(distribution.dimension_id_seq::regclass)"
> QGIS interprets that as string and tries to stuff it into the integer column
> which obviously fails.
>
>> Is there a bug in the provider, or should I get first the new id (if
>> yes, how do I do this) ?
> You could use QgsDataProvider::maximumValue() to fetch it and increment it
> yourself.  But that might still produce concurrency problems, if multiple people
> work in parallel.  I'd stick with a DEFAULT in PostgreSQL.
>
>
> Jürgen
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/qgis-developer/attachments/20120117/741da982/attachment-0001.html