Downloaded 381 times
![Current Developments Future Developments Use Cases Conclusion XPath XPath Example Simple example query: SELECT * FROM table1 WHERE (xpath(’//person/name/text()’, xdata))[1]::text = ’John Smith’; And using namespaces: SELECT * FROM table1 WHERE (xpath(’//person/@smpl:pid’, xdata, ARRAY[ARRAY[’smpl’, ’http://example.com’]]))::text = ’111’;](https://image.slidesharecdn.com/xml-090619094310-phpapp01/75/PostgreSQL-and-XML-25-2048.jpg)
Uploaded byPeter Eisentraut
PostgreSQL and XML
The document discusses PostgreSQL's current and future support for XML. It outlines current features like the XML data type, XML publishing functions, XML export, and XPath support. Future plans include adding validation against DTDs and XML schemas, annotated schema decomposition, XSLT transformations, improving performance, full-text search on XML data, and more advanced indexing. Several examples are provided to illustrate current XML functionality in PostgreSQL.
PostgreSQL and XML
- 1. PostgreSQL and XML Peter Eisentraut petere@postgresql.org Prague PostgreSQL Developers’ Day 2008
- 2. Current Developments Future Developments Use Cases Conclusion Outline 1 Current Developments 2 Future Developments 3 Use Cases 4 Conclusion
- 3. Current Developments Future Developments Use Cases Conclusion Outline 1 Current Developments 2 Future Developments 3 Use Cases 4 Conclusion
- 4. Current Developments Future Developments Use Cases Conclusion New Features Available in PostgreSQL 8.3: XML Data Type XML Publishing XML Export SQL:2003 conformance XPath
- 5. Current Developments Future Developments Use Cases Conclusion XML Data Type Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 6. Current Developments Future Developments Use Cases Conclusion XML Data Type XML Data Type CREATE TABLE test ( ..., data xml, ... ); Features: Input checking Support functions Issues: Internal storage format (plain text) Encoding handling
- 7. Current Developments Future Developments Use Cases Conclusion XML Data Type Using the XML Type Bizarre SQL way: INSERT INTO test VALUES ( ..., XMLPARSE (DOCUMENT ’<foo>...</foo>’), ... ); SELECT XMLSERIALIZE (DOCUMENT data AS varchar) FROM test;
- 8. Current Developments Future Developments Use Cases Conclusion XML Data Type Using the XML Type Bizarre SQL way: INSERT INTO test VALUES ( ..., XMLPARSE (DOCUMENT ’<foo>...</foo>’), ... ); SELECT XMLSERIALIZE (DOCUMENT data AS varchar) FROM test; Simple PostgreSQL way: INSERT INTO test VALUES (... , ’<foo>...</foo>’, ...); SELECT data FROM test;
- 9. Current Developments Future Developments Use Cases Conclusion XML Data Type XML Type Oddities No comparison operators To retrieve, use: Cast to text, or XPath, or Other key column To index, use: Cast to text, or XPath
- 10. Current Developments Future Developments Use Cases Conclusion XML Publishing Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 11. Current Developments Future Developments Use Cases Conclusion XML Publishing Producing XML Content The old way? SELECT ’<record id="’ || id || ’"><value>’ || ad_hoc_escape_func(value) || ’</value></record>’ FROM tab;
- 12. Current Developments Future Developments Use Cases Conclusion XML Publishing Producing XML Content The old way? SELECT ’<record id="’ || id || ’"><value>’ || ad_hoc_escape_func(value) || ’</value></record>’ FROM tab; The new way: SELECT XMLELEMENT(NAME record, XMLATTRIBUTES(id), XMLELEMENT(NAME value, value)) FROM tab;
- 13. Current Developments Future Developments Use Cases Conclusion XML Publishing XMLELEMENT Example SQL: XMLROOT ( XMLELEMENT ( NAME ’gazonk’, XMLATTRIBUTES ( ’val’ AS ’name’, 1 + 1 AS ’num’ ), XMLELEMENT ( NAME ’qux’, ’foo’ ) ), VERSION ’1.0’, STANDALONE YES )
- 14. Current Developments Future Developments Use Cases Conclusion XML Publishing XMLELEMENT Example SQL: Result: XMLROOT ( <?xml version=’1.0’ XMLELEMENT ( standalone=’yes’ ?> NAME ’gazonk’, <gazonk name=’val’ XMLATTRIBUTES ( num=’2’> ’val’ AS ’name’, <qux>foo</qux> 1 + 1 AS ’num’ </gazonk> ), XMLELEMENT ( NAME ’qux’, ’foo’ ) ), VERSION ’1.0’, STANDALONE YES )
- 15. Current Developments Future Developments Use Cases Conclusion XML Publishing XMLFOREST Example SELECT xmlforest ( "FirstName" as "FName", "LastName" as "LName", ’string’ as "str", "Title", "Region" ) FROM "Demo"."demo"."Employees"; might result in <FName>Nancy</FName> <LName>Davolio</LName> <str>string</str> <Title>Sales Representative</Title> <Region>WA</Region> . . . <FName>Anne</FName> <LName>Dodsworth</LName> <str>string</str> <Title>Sales Representative</Title> (1 row per record)
- 16. Current Developments Future Developments Use Cases Conclusion XML Publishing XMLAGG Example SELECT xmlelement (’Emp’, xmlattributes (’Sales Representative’ as "Title"), xmlagg (xmlelement (’Name’, "FirstName", ’ ’, "LastName"))) FROM "Demo"."demo"."Employees" WHERE "Title" = ’Sales Representative’; might result in <Emp Title="Sales Representative"> <Name>Nancy Davolio</Name> <Name>Janet Leverling</Name> <Name>Margaret Peacock</Name> <Name>Michael Suyama</Name> <Name>Robert King</Name> <Name>Anne Dodsworth</Name> </Emp> (1 row)
- 17. Current Developments Future Developments Use Cases Conclusion XML Export Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 18. Current Developments Future Developments Use Cases Conclusion XML Export XML Export Map table/schema/database contents to XML document Map table/schema/database schema to XML Schema Useful for: Downstream processing (e.g., SOAP, web services) Postprocessing using XSLT Backup??? Display formats (alternative to psql’s HTML mode)
- 19. Current Developments Future Developments Use Cases Conclusion XML Export XML Export Functions Data export: table_to_xml(tbl regclass, nulls boolean, tableforest boolean, targetns text) query_to_xml(query text, nulls boolean, tableforest boolean, targetns text) cursor_to_xml(cursor refcursor, count int, nulls boolean, tableforest boolean, targetns text) Schema export: table_to_xmlschema(tbl regclass, nulls boolean, tableforest boolean, targetns text) query_to_xmlschema(query text, nulls boolean, tableforest boolean, targetns text) cursor_to_xmlschema(cursor refcursor, nulls boolean, tableforest boolean, targetns text)
- 20. Current Developments Future Developments Use Cases Conclusion XML Export XML Schema Mapping Example CREATE TABLE test (a int PRIMARY KEY, b varchar(200)); is mapped to <xsd:complexType name="RowType.catalog.schema.test"> <xsd:sequence> <xsd:element name="a" type="INTEGER"></xsd:element> <xsd:element name="b" type="VARCHAR_200_200" minOccurs="0"></xsd:element> </xsd:sequence> </xsd:complexType> <xsd:complexType name="TableType.catalog.schema.test"> <xsd:sequence> <xsd:element name="row" type="RowType.catalog.schema.test" minOccurs="0" maxOccurs="unbounded" /> </xsd:sequence> </xsd:complexType>
- 21. Current Developments Future Developments Use Cases Conclusion XML Export XML Export Format Example <catalogname> <schemaname> <tablename> <row> <colname1>value</colname1> <colname2 xsi:nil=’true’/> ... </row> ... </tablename> ... </schemaname> ... </catalogname>
- 22. Current Developments Future Developments Use Cases Conclusion XPath Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 23. Current Developments Future Developments Use Cases Conclusion XPath XPath example Example table: CREATE TABLE table1( id integer PRIMARY KEY, created timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP, xdata xml );
- 24. Current Developments Future Developments Use Cases Conclusion XPath XPath Example Example data: INSERT INTO table1 (id, xdata) VALUES( 1, ’<dept xmlns:smpl="http://example.com" smpl:did="DPT011-IT"> <name>IT</name> <persons> <person smpl:pid="111"> <name>John Smith</name> <age>24</age> </person> <person smpl:pid="112"> <name>Michael Black</name> <age>28</age> </person> </persons> </dept>’ );
- 25. Current Developments Future Developments Use Cases Conclusion XPath XPath Example Simple example query: SELECT * FROM table1 WHERE (xpath(’//person/name/text()’, xdata))[1]::text = ’John Smith’; And using namespaces: SELECT * FROM table1 WHERE (xpath(’//person/@smpl:pid’, xdata, ARRAY[ARRAY[’smpl’, ’http://example.com’]]))::text = ’111’;
- 26. Current Developments Future Developments Use Cases Conclusion XPath XPath: Indexes Use functional indexes to avoid XPath evaluation at run time: CREATE INDEX i_table1_xdata ON table1 USING btree ( xpath(’//person/@name’, xdata) );
- 27. Current Developments Future Developments Use Cases Conclusion Outline 1 Current Developments 2 Future Developments 3 Use Cases 4 Conclusion
- 28. Current Developments Future Developments Use Cases Conclusion Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance issues Full-text search Advanced indexing (XLABEL) More, more, more
- 29. Current Developments Future Developments Use Cases Conclusion DTD and XML Schema validation Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 30. Current Developments Future Developments Use Cases Conclusion DTD and XML Schema validation DTD and XML Schema validation DTD validation: Implemented for 8.3, DTD is passed by URI Should be extended to allow passing DTD as text XML Schema (XSD) validation (XMLVALIDATE per SQL:2006): INSERT INTO messages(msg) SELECT xmlvalidate( DOCUMENT ’<?xml ...’ ACCORDING TO XMLSCHEMA NO NAMESPACE LOCATION ’http://mycompany.com/msg-schema’ ); (The result of XMLVALIDATE is a new XML value.)
- 31. Current Developments Future Developments Use Cases Conclusion Annotated schema decomposition Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 32. Current Developments Future Developments Use Cases Conclusion Annotated schema decomposition Annotated schema decomposition In some cases decomposition of XML Schema to relational data is better (no storing XML data, XML serves as transport only): When we need to store only small parts of the XML data Already developed tools might be designed only for relational data During decomposition the following capabilities could be used: Data normalization Foreign keys creation Conditional insertion of data chunks Insert parts of initial XML document as XML values
- 33. Current Developments Future Developments Use Cases Conclusion XSLT Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 34. Current Developments Future Developments Use Cases Conclusion XSLT XSLT The easiest way: adapt and expand contrib/xml2’s capabilities. Choose an approach: Move XSLT functionality to the core (and use --with-libxslt) Separate contrib/xslt
- 35. Current Developments Future Developments Use Cases Conclusion XSLT XSLT Crazy idea: PL/XSLT Define transformations as functions Version 0.0.0 exists :-)
- 36. Current Developments Future Developments Use Cases Conclusion Performance Issues Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 37. Current Developments Future Developments Use Cases Conclusion Performance Issues Performance Issues Ideas: Cache intermediate results to avoid redundant parsing and XPath evaluation Advanced physical storage to speedup access to arbitrary node in XML data Use PostgreSQL existing capabilities for full-text search Use additional structures/tables/indexes to avoid XPath evaluation at runtime Use slices (similar to array_extract_slice()) to avoid dealing with entire values (both in SELECT and UPDATE)
- 38. Current Developments Future Developments Use Cases Conclusion Full-Text Search Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 39. Current Developments Future Developments Use Cases Conclusion Full-Text Search Full-Text Search Simple way to create FTS index (available in 8.3): CREATE INDEX i_table1_fts ON table1 USING gist ( to_tsvector( ’default’, array_to_string(xpath(’//text()’, xdata), ’ ’) ) );
- 40. Current Developments Future Developments Use Cases Conclusion Full-Text Search Full-Text Search Proposal for overloading of built-in to_tsvector(): CREATE OR REPLACE FUNCTION to_tsvector(text, xml) RETURNS tsearch2.tsvector LANGUAGE SQL IMMUTABLE AS $$ SELECT to_tsvector( $1, array_to_string(xpath(’//text()’, $2), ’ ’) ); $$; CREATE INDEX i_table1_fts ON table1 USING gist (to_tsvector(’default’, xdata));
- 41. Current Developments Future Developments Use Cases Conclusion Full-Text Search Full-Text Search Further ideas for full-text search: Indexing parts of documents (available in 8.3, in some way) Element names in tsvector Relevance scoring (ranking) FTS parser for XML
- 42. Current Developments Future Developments Use Cases Conclusion Advanced Indexing Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 43. Current Developments Future Developments Use Cases Conclusion Advanced Indexing XLABEL Idea: Enumerate all XML node names in one database-wide table (xnames) Store shredded data in additional table (columnname_xlabel) Use numbering scheme to encode nodes (e.g., ltree) Use GiST/GIN indexes for numbering scheme column Rewrite XPath expression to plain SQL statement Implement partial updates support to avoid massive index rebuilding
- 44. Current Developments Future Developments Use Cases Conclusion Advanced Indexing XLABEL Enumerate all XML node names in the database: Table: xnames xname_id xname_name 1 person 2 dept 3 name 4 did 5 persons ... ...
- 45. Current Developments Future Developments Use Cases Conclusion Advanced Indexing XLABEL For an XML column implicitly create additional table (using xlabel.register_column() function): Table: table1_xdata tid xlabel node_type xname_id value 1 a 1 (elem.) 2 NULL 1 a.b 2 (attr.) 4 DPT011-IT 1 a.c 1 (elem.) 3 NULL 1 a.c.a NULL NULL IT ... ... ... ... ... 1 a.d.a.b 1 (elem.) 3 NULL 1 a.d.a.b.a NULL NULL John Smith ... ... ... ... ... CREATE INDEX i_table1_xdata_xlabel ON table1_xdata USING gist (xlabel);
- 46. Current Developments Future Developments Use Cases Conclusion Advanced Indexing XLABEL Rewrite XPath expression to plain SQL statement: SELECT * FROM table1 WHERE array_dims(xpath(’//person/name’, xdata)) IS NOT NULL; . . . becomes . . . SELECT * FROM table1 WHERE EXISTS( SELECT 1 FROM table1_xdata AS t1, table1_xdata AS t2 WHERE t1.xname_id = 1 AND t2.xname_id = 3 AND t3.xlabel <@ t1.xlabel ); . . . where <@ means “is a child of”.
- 47. Current Developments Future Developments Use Cases Conclusion Advanced Indexing XLABEL Current thoughts: Separate table is problematic (déjà vu: fti vs. tsearch2) It would be great if one structure solves 2 problems at once: access to arbitrary node SELECTs with XPath
- 48. Current Developments Future Developments Use Cases Conclusion More Ideas Outline 1 Current Developments XML Data Type XML Publishing XML Export XPath 2 Future Developments DTD and XML Schema validation Annotated schema decomposition XSLT Performance Issues Full-Text Search Advanced Indexing More Ideas 3 Use Cases 4 Conclusion
- 49. Current Developments Future Developments Use Cases Conclusion More Ideas More, more, more Inline ORDER BY for XMLAGG (SQL:2003) ... XMLAGG(XMLELEMENT(...) ORDER BY col1) ... XMLCAST (SQL:2006) XML Canonical Pretty-printing XML Registered XML Schemas (SQL:2006) Schema evolution Improve Data Model (XDM) XQuery support (SQL:2006) Updatable XML views (over relational data) Relax-NG validation
- 50. Current Developments Future Developments Use Cases Conclusion More Ideas And even more! Bulk loader for XML data (parallelize the XML parsing) XML-awareness in APIs and PLs Additional contribs/projects (web services, ODF, DocBook utilities, etc.) New tools and applications, integration with existing ones
- 51. Current Developments Future Developments Use Cases Conclusion Outline 1 Current Developments 2 Future Developments 3 Use Cases 4 Conclusion
- 52. Current Developments Future Developments Use Cases Conclusion Use Cases Use Case 1: Document Management System Use Case 2: Store Logs in the Database Use Case 3: Heterogeneous Catalog
- 53. Current Developments Future Developments Use Cases Conclusion Use Case 1: Document Management System The primary goal: to store documents in the RDBMS as is
- 54. Current Developments Future Developments Use Cases Conclusion Use Case 2: Store Logs in the Database Table: action action_id SERIAL action_type_id INT4 action_status_id INT4 action_person_id INT4 action_data XML The primary goal: to achieve flexibility, avoid database schema changes (schema evolution)
- 55. Current Developments Future Developments Use Cases Conclusion Use Case 3: Heterogeneous Catalog Task: to build heterogeneous catalog (items of different types, a lot of properties)
- 56. Current Developments Future Developments Use Cases Conclusion Use Case 3: Heterogeneous Catalog Task: to build heterogeneous catalog (items of different types, a lot of properties) How?
- 57. Current Developments Future Developments Use Cases Conclusion Use Case 3: Heterogeneous Catalog Ugly way
- 58. Current Developments Future Developments Use Cases Conclusion Use Case 3: Heterogeneous Catalog Entity-Attribute-Value model
- 59. Current Developments Future Developments Use Cases Conclusion Use Case 3: Heterogeneous Catalog Semi-structured data approach
- 60. Current Developments Future Developments Use Cases Conclusion Use Case 3: Heterogeneous Catalog Metadata Query Interface for Heterogeneous Data Archives (International Virtual Observatory): http://alcor.sao.ru/php/search/
- 61. Current Developments Future Developments Use Cases Conclusion Outline 1 Current Developments 2 Future Developments 3 Use Cases 4 Conclusion
- 62. Current Developments Future Developments Use Cases Conclusion Credits J. Gray et al. for contrib/xml2 Pavel Stehule for initial patch for SQL/XML publishing functions Nikolay Samokhvalov for Google Summer of Code 2006 project and part of this presentation me :-) PostgreSQL developer community for fixing our bugs
- 63. Current Developments Future Developments Use Cases Conclusion More Information SQL:2006, Part 14: XML-Related Specifications PostgreSQL documentation XML Development Wiki Page: http://developer.postgresql.org/index.php/XML_Support N. Samokhvalov, “XML Support in PostgreSQL”, Proceedings of SYRCoDIS, Moscow, Russia, 2007, http://samokhvalov.com/syrcodis2007.ps pgsql-hackers@postgresql.org