Trying to get VALUE_ARRAY to work

[maaattes]

Hi,
I'm trying to replicate something like in this example https://www.ldproxy.nrw.de/topographie/collections/ax_bahnstrecke/items/DENWAT01D0007K35?f=html
Here, if there is more than one value for 'Streckennummer' the label shown multiple times.

In my PostgreSQL/PostGIS Database, the values for 'Streckennummer' are stored in an array like so: {16, 18, 19, 5}.
So my first educated guess was:

…
types:
  bahnstrecke:
    sourcePath: /bahnstrecke
    type: OBJECT
    properties:
      nummerderbahnstrecke:
        sourcePath: nummerderbahnstrecke
        type: VALUE_ARRAY
        valueType: STRING
…

But it's still showing the values as strings:

In some automatically generated config files I've seen the parameter

additionalInfo:
  multiple: "true"

but this isn't documented anywhere I think it should be true without quotations marks?
If I'm adding this to my config (without quotation marks) I'm getting

Could you give me a hint what I'm missing?

[cportele]

Hi @maaattes, I guess you are using PostgreSQL arrays? If that is the case, these are currently not supported. We have this capability on the radar, but so far all databases with multiple values used joins and separate tables, so for now only those are supported.

If you had a table bahnstrecke_nrb with a foreign key reference rid to the bahnstrecke feature and an attribute nrb with a ´single NRB value, the configuration would look as follows:

…
types:
  bahnstrecke:
    sourcePath: /bahnstrecke
    type: OBJECT
    properties:
      nummerderbahnstrecke:
        sourcePath: '[id=rid]bahnstrecke_nrb/nrb'
        valueType: STRING
        type: STRING
        label: Nummer der Bahnstrecke
…

If the table bahnstrecke_nrb does not have a standard sort key column (by default id), you would also have to specify a sort key in the source path, e.g. sourcePath: '[id=rid]bahnstrecke_nrb/nrb{sortKey=rid}'.

[maaattes]

Wow, that was quick. Thank's a lot!
Yes, 'nummerderbahnstrecke' is of type character varying[]. This is what the PostgreSQL Driver of 'ogr2ogr' chooses when importing the gml.

So basically I've downloaded the gml from the WFS:
curl "https://www.wfs.nrw.de/geobasis/wfs_nw_atkis-basis-dlm_aaa-modell-basiert?SERVICE=WFS&REQUEST=GetFeature&VERSION=2.0.0&TYPENAMES=adv:AX_Bahnstrecke" --output AX_Bahnstrecke.gml

Then imported it to my database using ogr2org (had to specify the layer because there is a layer called 'ALKIS_beziehungen')

ogr2ogr \
  -f "PostgreSQL" \
  PG:"host=localhost user=postgres port=5432 dbname=demo password=xyz" \
  AX_Bahnstrecke.gml \
  -overwrite \
  -progress \
  -nln "bahnstrecke" \
  -lco COLUMN_TYPES=beginnt=timestamptz \
  AX_Bahnstrecke

After that I just changed the primary key from 'ogc_fid' to 'gml_id'.

Since this is more or less for scientific pruposes, I don't think I will go through the trouble of setting up a data structure based on relationship and joins. But anyway thanks for the hint!
So maybe I will just try to alter the datatype of the column from array to string and call it a day.
Or is it possible to do this with https://docs.ldproxy.net/de/providers/details/transformations.html#string-template-filter ?
I've also tried to 'flatten' the array but this did not work either.

[cportele]

Thanks for the explanations and for raising the question. If GDAL converts this for PostgreSQL to an array datatype, then we probably should plan to support array datatypes PostgreSQL backends. I have created an issue for this: #671.

String template filters will not help here, they can transform strings, not arrays.