Importing a dataset
Importing a dataset is done in two stages, both tracked on the same record (an IntegrityLink, one per dataset): staging the raw data, then processing it into its published form.
1. Choose a source
The import wizard supports the following source types:
| Source | Description |
|---|---|
| File | CSV, GeoPackage, GeoJSON, zipped Shapefile, Parquet or GeoParquet upload |
| URL | Same file formats, fetched from a remote HTTP(S) URL |
| FTP | Same file formats, fetched from an FTP server |
| Database | A table or query read from one of the configured source databases |
| API | An OGC service: WFS or OGC API Features |
Once submitted, Datafeeder ingests the raw source into a staging table in PostgreSQL, with no transformation applied yet. This step lets you inspect the actual columns, geometry and data types before deciding how to publish them.
Re-staging
If the staging step needs to be redone (wrong file, source moved, credentials changed), you don't need to start over: editing the source configuration on an existing dataset re-triggers staging without losing the rest of the configuration.
2. Configure the transformation
Once staged, you configure how the data should be transformed into its final, published form:
- Column mapping: rename, drop or retype columns
- Projection: reproject the geometry to a target CRS (the available projections are configured platform-wide)
- Filters: restrict the rows carried over to the final table
3. Publish
Submitting the transformation writes the final table, then:
- creates the GeoNetwork metadata record for the dataset
- publishes the corresponding layer on GeoServer
If anything fails at this stage, the partial final table is dropped so the dataset never ends up in a half-published state.
Recurring imports
A dataset can be configured to automatically re-fetch from its original source and re-run the transformation on a schedule (every minute, hour, day, week, month or year). This is useful for sources that change over time, such as an OGC service or a periodically updated file at a fixed URL. See Managing datasets for how to configure and monitor a recurring import.
Empty datasets
It is also possible to create a dataset with no source at all (empty dataset): Datafeeder creates the
IntegrityLink record and a skeleton GeoNetwork metadata record immediately, with no data pipeline involved. This is
useful to reserve a metadata record ahead of publishing data manually, or for datasets that live entirely on
GeoServer already.