[Qgis-developer] DB Manager and other PostGIS issues: notes from a course

[Giovanni Manghi]

Hi all,



> * DB Manager
>   * create table: field types are limited; in particular, timestamptz is
> missing

old issue, old as db manager, it only shows a subset of available datatypes.



>   * create table: SRID has to be entered by hand; the projection widget
> should be added


see above: it has always been so in DB manager.


>   * edit table -> Table properties -> Field properties: the field
> `Length` appears wrong; for geom it shows `geomtype,srid`, for
> varchar(255) and others, adding a length will cause an error:
> ALTER TABLE "public"."buffer_union" ALTER "round" TYPE varchar(255) (24)
> length has to be entered by hand changing varchar(255) into e.g. varchar(20)


see above: old as db manager. If you choose a datatype with a
predefined length then manually changing the length causes an invalid
query to be run.


>   * we got frequent errors similar to [0]


confirmed. This does not seems to be the case in LTR, so this
configures as a regression.


>
> * Browser
>   * importing a shapefile of points results in a (generally undesirable)
> MULTIPOINT, without a possibility of choice for the user
>   * a table created with SRID=0
> CREATE TABLE pt
>  (id INTEGER, name text, geom GEOMETRY(Point,0,2));
> INSERT INTO pt (id, name, geom)
>  VALUES (1, 'Origin', ST_GeomFromText('POINT(0 0)', 0));
> results in two entries: one with mouseover= geom as Point in 0 and one
> geom as Point in -2147483648
>


confirmed down to 2.8.* so not sure is a regression.


> * Processing
>   * there are two Import into Processing (one QGIS and one OGR)


the OGR based tool is necessary because is the only one available that
is fast enough. It is 8+ times faster than any other tool available in
QGIS making it the only tool good enough to import into postgis large
layers.




>   * for the first (QGIS), batch processing fails with `Missing parameter
> value: Table to import to (leave blank to use layer name (row 1)`

the "table name" parameter has been incorrectly set as mandatory. This
is not the case in LTR, so this is a regression.



>   * for the second (OGR), it also fails because it tries to write in the
> spatial_ref_sys (CRS should default to source CRS) [1]

I will make the input/output CRS parameters mandatory.





>   * the gdal output is not interpreted correctly (html and encoding) [1]

cannot confirm.



cheers


-- G --