Code Architecture

(n.b. all descriptions below are intended to provide a high level overview of how pycsw is implemented. For full details, please refer to the codebase)

Overview

pycsw is a CGI based application written in Python, which accepts HTTP GET and POST requests as per OGC:CSW 2.0.2. The basic flow of events is:

client request --> pycsw (handle request, produce response) --> server response

pycsw is always called from csw.py, and always instantiates a server.Csw object and then uses its dispatch() method to handle the request and generate a response.

The server.Csw class sets up the server to be able to handle OGC:CSW requests accordingly:

• setup configuration (default.cfg)
• initialize the underlying repository (database) connection and queryables model
• set default HTTP properties (gzip compression)
• generate !GetDomain model
• load any profile code (e.g. as apiso)
• setup transactions (if specified)
• setup distributed search (if specified)
• setup logging (if specified)

At this point, pycsw is ready to handle the request, using server.Csw.dispatch(), which does the following:

• parse request (GET or POST or SOAP)
• do basic parameter checking (service, version, request)
• process the request accordingly

(server.Csw.exceptionreport() is always used when pycsw encounters an error and returns an OGC ExceptionReport)

All server.Csw methods return lxml.etree.Element objects, which are then processed by server.Csw.write_response() and returned to the client as XML.

server.Csw.getcapabilities()

• handle SECTIONS parameter if specified
• handle extra profile parameters if specified
• set / process updatesequence
• return response XML as lxml.etree.Element

server.Csw.describerecord()

• perform GET validation
• process the output of schemas as csw:SchemaComponent elements
• return response XML as lxml.etree.Element

server.Csw.getdomain()

• perform GET validation
• process parameter name
  • validate against internal domain model
• process property name
• validate existence of property against self.repository.queryables['all']
• query repository (SQL distinct query against XPath of queryable in records.xml
• return response XML as lxml.etree.Element

server.Csw.getrecords()

• perform GET validation
• query repository. SQL query, one of:
• spatial (util.query_spatial())
• aspatial (util.query_xpath())
• spatial + aspatial
• sorting (if specified)
• do distributed searching (if specified)
• write out results (based on outputschema)
• distributed search results are returned verbatim
• return response XML as lxml.etree.Element

server.Csw.getrecordbyid()

• perform GET validation
• query repository. SQL query by id (against records.identifier)
• write out results (based on outputschema)
• return response XML as lxml.etree.Element

server.Csw.getrepositoryitem()

• wrapper around server.Csw.getrecordbyid()
• gets raw XML record
• return response XML as lxml.etree.Element

server.Csw.transaction()

• validate XML document
• insert mode
• update mode
• delete mode

server.Csw.harvest()

• fetch XML from URL
• insert into repository, or update if identifier exists

Other Notes

• server.Csw._gen_soap_wrapper() is the generic SOAP wrapper private method
• server/config.py sets the server's operation model in config.MODEL. Any modifications are then made by calling code (e.g. to add more queryables, typenames, etc.)
• spatial query magic is via Shapely in server/util.py:query_spatial(), called via SQL function bound back to this method
• full text (e.g. '*:!AnyText') style queries are via server.util.py:query_anytext(), called via SQL function bound back to this method
• XPath style queries are via server.util.py:query_xpath(), called via SQL function bound back to this method