pycsw administration is handled by the utility. is installed as part of the pycsw install process and should be available in your PATH.


Run -h to see all administration operations and parameters

Metadata Repository Setup

pycsw supports the following databases:

  • SQLite3
  • PostgreSQL
  • PostgreSQL with PostGIS enabled
  • MySQL


The easiest and fastest way to deploy pycsw is to use SQLite3 as the backend.


PostgreSQL support includes support for PostGIS functions if enabled


If PostGIS (1.x or 2.x) is activated before setting up the pycsw/PostgreSQL database, then native PostGIS geometries will be enabled.

To expose your geospatial metadata via pycsw, perform the following actions:

  • setup the database
  • import metadata
  • publish the repository

Supported Information Models

By default, pycsw supports the csw:Record information model.


See Profile Plugins for information on enabling profiles

Setting up the Database

$ -c setup_db -f default.cfg

This will create the necessary tables and values for the repository.

The database created is an OGC SFSQL compliant database, and can be used with any implementing software. For example, to use with OGR:

$ ogrinfo /path/to/records.db
INFO: Open of 'records.db'
using driver 'SQLite' successful.
1: records (Polygon)
$ ogrinfo -al /path/to/records.db
# lots of output


If PostGIS is detected, the script does not create the SFSQL tables as they are already in the database.

Loading Records

$ -c load_records -f default.cfg -p /path/to/records

This will import all *.xml records from /path/to/records into the database specified in default.cfg (repository.database). Passing -r to the script will process /path/to/records recursively. Passing -y to the script will force overwrite existing metadata with the same identifier. Note that -p accepts either a directory path or single file.


Records can also be imported using CSW-T (see Transactions).

Exporting the Repository

$ -c export_records -f default.cfg -p /path/to/output_dir

This will write each record in the database specified in default.cfg (repository.database) to an XML document on disk, in directory /path/to/output_dir.

Optimizing the Database

$ -c optimize_db -f default.cfg


This feature is relevant only for PostgreSQL and MySQL

Deleting Records from the Repository

$ -c delete_records -f default.cfg

This will empty the repository of all records.

Database Specific Notes


  • if PostGIS is not enabled, pycsw makes uses of PL/Python functions. To enable PostgreSQL support, the database user must be able to create functions within the database. In case of recent PostgreSQL versions (9.x), the PL/Python extension must be enabled prior to pycsw setup
  • PostgreSQL Full Text Search is supported for csw:AnyText based queries. pycsw creates a tsvector column based on the text from anytext column. Then pycsw creates a GIN index against the anytext_tsvector column. This is created automatically in pycsw.admin.setup_db. Any query against csw:AnyText or apiso:AnyText will process using PostgreSQL FTS handling


  • pycsw makes use of PostGIS spatial functions and native geometry data type.
  • It is advised to install the PostGIS extension before setting up the pycsw database
  • If PostGIS is detected, the script will create both a native geometry column and a WKT column, as well as a trigger to keep both synchronized.
  • In case PostGIS gets disabled, pycsw will continue to work with the WKT column
  • In case of migration from plain PostgreSQL database to PostGIS, the spatial functions of PostGIS will be used automatically
  • When migrating from plain PostgreSQL database to PostGIS, in order to enable native geometry support, a “GEOMETRY” column named “wkb_geometry” needs to be created manually (along with the update trigger in pycsw.admin.setup_db). Also the native geometries must be filled manually from the WKT field. Next versions of pycsw will automate this process

Mapping to an Existing Repository

pycsw supports publishing metadata from an existing repository. To enable this functionality, the default database mappings must be modified to represent the existing database columns mapping to the abstract core model (the default mappings are in pycsw/

To override the default settings:

  • define a custom database mapping based on etc/
  • in default.cfg, set repository.mappings to the location of the file:

Note you can also reference mappings as a Python object as a dotted path:


See the GeoNode Configuration, HHypermap Configuration, and Open Data Catalog Configuration for further examples.

Existing Repository Requirements

pycsw requires certain repository attributes and semantics to exist in any repository to operate as follows:

  • pycsw:Identifier: unique identifier
  • pycsw:Typename: typename for the metadata; typically the value of the root element tag (e.g. csw:Record, gmd:MD_Metadata)
  • pycsw:Schema: schema for the metadata; typically the target namespace (e.g.,
  • pycsw:InsertDate: date of insertion
  • pycsw:XML: full XML representation
  • pycsw:AnyText: bag of XML element text values, used for full text search. Realized with the following design pattern:
    • capture all XML element and attribute values
    • store in repository
  • pycsw:BoundingBox: string of WKT or EWKT geometry

The following repository semantics exist if the attributes are specified:

  • pycsw:Keywords: comma delimited list of keywords
  • pycsw:Links: structure of links in the format “name,description,protocol,url[^,,,[^,,,]]”

Values of mappings can be derived from the following mechanisms:

  • text fields
  • Python datetime.datetime or objects
  • Python functions

Further information is provided in pycsw/