Updating schema for Vector Mosaic layer

GeoServer GeoServer User
,
1

Hello,

We are using the Vector Mosaic datastore and need to add a few new columns to the underlying FlatGeoBuf files, but we aren’t able to “Customize attributes”.

For one, there are 2 duplicate attributes, so when trying to add new attributes, we get this error:

  • Duplicate key tenant_id (attempted merging values AttributeDescriptorImpl tenant_id <tenant_id:String> nillable 0:1 and AttributeDescriptorImpl tenant_id <tenant_id:String> nillable 0:1 userData=( org.geotools.jdbc.nativeType ==> 1 org.geotools.jdbc.nativeTypeName ==> bpchar))

We don’t know why there are duplicates. The original source FlatGeoBuf files do not have duplicates, but the _index.fgb does also have those 2 attributes.

If we try to delete the duplicates, we get the same error.

Currently our only idea to successfully add new attributes is to rename the database table and recreate the store/layer from a new FlatGeoBuf file, then move all the old data into the new table.

We’ve obviously done something wrong, probably at the first step, but would like to know if there’s a way we can fix our store/layer to make it possible to manage this properly going forward.

Thanks,
Richard

2

The duplicate attributes are caused by having the same column names in the FGB file as in the Postgres index table. Our current plan is to rebuild the mosaics with different names for these columns, which will hopefully let us make changes more easily in the future.

3

This doesn’t seem possible. These column names must be the same or the mosaic cannot find any features. It seems like the validation of these attributes is too strict for this specific scenario.

4

Hi Richard,
I’ve read the emails but came up with no understanding of your situation.
Can you please share the schema of your index table, and the schema of the flatgeobuf files?
Also, the GeoServer version you’re using.

5

flatgeobuf Index file:

Field Type
params String
tenant_id String
h_id Integer

flatgeobuf file (leaving out all the data columns):

Field Type
tenant_id String
h_id Integer64

Index table:

CREATE TABLE IF NOT EXISTS theschema.thetable_index
(
ogc_fid integer NOT NULL DEFAULT nextval(‘vector_products.thetable_ogc_fid_seq’::regclass),
params character varying(254) COLLATE pg_catalog.“default”,
tenant_id character(21) COLLATE pg_catalog.“default”,
h_id bigint,
wkb_geometry geometry(Polygon,4326),
CONSTRAINT thetable_pkey PRIMARY KEY (ogc_fid)
)

GeoServer layer (also leaving out most of the data columns which mostly precede the first “tenant_id” but there are some that appear in between the duplicate columns):

Property Type Nillable Min/Max Occurrences
tenant_id String true 0/1
h_id Long true 0/1
elevation Double true 0/1
operator String true 0/1
tenant_id String true 0/1
h_id Long true 0/1