Release History

Historically, the release number of PyXB indicated stability in a way similar to that of Linux several years ago, with even/odd of the middle digit indicating stability/development. That policy ended as of 1.1.3, and the relaxed policy is:

  • All releases consist of three non-negative integers
  • A release series refers to all releases with the same first two integers. The third integer indicates an incremental update to the system.
  • Should it be necessary to make wholesale changes that affect either stability or interface, the second or even first integer will be incremented
  • Critical bug fixes may result in new releases in older series, but this is expected to be extremely unusual. Such fixes will be made on a branch of the git repository.
  • All packaged releases are intended to be of comparable stability, with bug fixes and most enhancements causing an increment in the third integer

As examples, correcting the content model for weak determinism resulted in PyXB 1.2.0, and support of Python 3 was completed in PyXB 1.2.4. support for XML Schema 1.1 will not be addressed until PyXB 1.3.0 if at all.


The 1.2.x release series introduces a new content model which provides much more help in resolving validation errors. Concurrent with this support was dropped for Python 2.4 and 2.5. The 1.2.x release works with Python 2.6 and 2.7 and as of PyXB 1.2.4 also with Python 3.1 and later. Some unit tests are skipped under Python 2.6 due to insufficient support for validating exception attributes.

1.2.5 (18 Sep 2016)

This is a roll-up aggregating all fixes since the last major release nearly two years ago.

The following reported defects/enhancements have been addressed:

  • Deconflict class properties from module identifiers. issue 27
  • Correct handling of xsi:type on non-abstract types. issue 32
  • Fix generation of zero-span durations. issue 38
  • Document how to use file URIs on Windows. issue 34
  • Document need for common bindings when generating OpenGIS bindings. issue 28
  • Fix literal representation of not-a-number and infinite values. issue 43
  • Document that Python 3 is supported. issue 36
  • Fix processing of empty binary data tpyes. issue 51
  • Fix transition sort key to support Python 3. issue 52
  • Fix generation of modules that cannot be imported due to use of invalid characters in the module path. issue 48
  • Add a method to xsd.base64Binary to allow a heavy-weight validation step to be bypassed. issue 50

1.2.4 (19 Oct 2014)


The major feature of this release is a new approach to Python 3 support based on the six module. This resolves support for Python 3 and allows one release of PyXB to support Python versions from 2.6 through 3.4, excluding 3.0. The maintainer/2to3 script used in PyXB 1.2.3 is no longer supported. Both the core implementation and the bindings are compatible with Python 2 and 3.

A second important feature was correct handling of QName by changing the binding representation from a string to an expanded name that saves the namespace URI rather than the prefix from the original document.

Finally, with this release primary development and support moves from SourceForge to Github. Inter-release updates will be made first on github (on the next branch), and pushed to the SourceForge git repository aperiodically.

Other miscellanous changes:

  • Correct rich comparisons for _PluralBinding instances. SF ticket 218
  • Prevent adding non-element content as wildcard elements. SF ticket 219
  • Provide public API to _CF_enumeration helper methods. SF ticket 221
  • Specify requirements for an acceptable DOM implementation. SF ticket 224
  • Rebase xs:decimal on Python decimal.Decimal to provide high precision support. SF ticket 91
  • Support Python 3. SF ticket 150
  • Fix mis-ordered attribute groups use/resolution issue. SF ticket 230
  • Generate facet values that are valid Python expressions. SF ticket 231
  • Improve validation failure location for simple values. SF ticket 232
  • Disallow name attribute in internal complex types. issue 6
  • Avoid unbounded non-deterministic expansion. issue 5
  • Reduce namespace intrusion with enumeration types. issue 9
  • Add API to replace simple content complex type values. issue 7
  • Fix representation and use of QNames. issue 13
  • Fix generation of wildcard attributes. issue 14
  • Fix validation of wildcard types in DOM instances. issue 15
  • Reference github for bug reports. issue 16:
  • Eliminate a gratuitous log warning on a validation failure. issue 17
  • Eliminate improper assertion on extension binding use of imported bindings. issue 18
  • Support serialization of anonymous namespaces to allow Python copy.deepcopy module to work on binding instances. issue 19
  • Generate bindings that are compatible with both Python 2 and Python 3. issue 22

1.2.3 (18 Sep 2013)


This is primarily a bug-fix and clean-up release. One new feature is that starting with this release a Python 3 compatible version of PyXB will be available in the git repository. For this version, it may be obtained by:

git clone -b python3/pyxb-1.2.3 git:// pyxb3-1.2.3

The Python 3 code in this branch is created by automatic conversion of the Python 2 code using the maintainer/2to3 script. While not officially supported until PyXB 1.3.0, Python 3 appears to work correctly with this version.

Behavioral changes:

  • PyXB 1.2.1 automatically filled in the content of fixed elements. This generated invalid documents when the particular element should have been absent for some content. The feature has been removed so fixed elements initialize to absent just like every other element.
  • The former –pre-load-archives option to pyxbgen has been removed and replaced by –import-augmentable-namespace which has more consistent semantics. Schema import directives related to a non-augmentable namespace (one for which the referenced option has not been provided) will be ignored if content for the namespace can be located in any archive or builtin.
  • Generating DOM (or XML) expressions of bindings that were created by type constructors and not associated with elements now cause an exception to be raised rather than mis-using the type name as though it were a valid element tag.

Code cleanup:

  • Trailing whitespace throughout the source has been removed to decrease git complaints and non-reproducible patches.
  • Several coding practices that would not work under Python3 have been replaced with new solutions.

The following reported defects/enhancements have been addressed:

  • Fix DOM style handling of xsi:type attribute. SF ticket 166
  • Fix problems with nillable simple types. SF ticket 200
  • Fix problems inserting into plural element attributes. SF ticket 201
  • Correct namespace-qualified attributes. SF ticket 202
  • Eliminate automatic derivation of XML element tag from binding type when converting an element-free instance to DOM. SF ticket 203
  • Eliminate default to fixed value for elements. SF ticket 204
  • Re-word exception when element content was incompatible with fixed value. SF ticket 205
  • Fix the aslocal() method for xs:dateTime values with non-UTC timezones. SF ticket 206
  • Fix values for restricted duration ranges in pickled data such as namespace archives. SF ticket 207
  • Correct the interpretation of QName-valued base and similar attributes with respect to in-scope namespace declarations. SF ticket 208
  • Rework how to specify that imported namespaces should be processed to extend the content model. SF ticket 209
  • Removed redundant code in pyxbgen relevant to –wsdl-location. SF ticket 210
  • Refine diagnostics when an element with simple type appears in a context that expects non-element content. SF ticket 211
  • Add NonElementContent to simplify access to non-element content in a mixed-content instance. SF ticket 212
  • Convert assert failures to diagnostic exceptions when generating DOM expressions of complex types with simple content where the content is missing. SF ticket 213
  • Ensure attribute values are converted from lexical space to value space when assigned from an XML representation. SF ticket 216

1.2.2 (17 Apr 2013)


Almost all changes are bug fixes, including issues with element order when extending types, name conflicts between XML and Python entities in the binding, and unicode problems in diagnostics.

PyXB has never intentionally supported using bindings generated by different versions of PyXB. This will be enforced from version 1.2.2 onwards.

The following reported defects/enhancements have been addressed:

  • Correct management of changes to default namespace within documents. SF ticket 183
  • Do not interleave base and extension type elements. SF ticket 184
  • Incorrect character range management resulted in regex failure. SF ticket 186
  • Wide unicode mis-defines XML1p0e2 Char class. SF ticket 187
  • Improve contents of StructuralBadDocumentError. SF ticket 189
  • Generate exception details correctly with unicode content. SF ticket 190
  • Namespace must be a reserved symbol within CTD instances.
  • PyXB rejects restriction that changes element type. SF ticket 193
  • Multiple cases where names (from imported bindings, namespaces, and top-level classes and elements) resulted in conflicts with other top-level and class-level names. SF ticket 191, SF ticket 194, SF ticket 197
  • Avoid validating content of simple types marked with xsi:nil. SF ticket 195
  • Validate handling of attribute and element form attributes. SF ticket 196
  • Ensure all referenced namespaces are imported into a binding. SF ticket 198
  • Detect import of bindings generated by a different version of PyXB. SF ticket 199

1.2.1 (17 Dec 2012)



This release has an interface change: the content method on complex types is now deprecated in favor of the orderedContent method. The members in the new list have wrapper classes to correctly associate instance bindings with the element declaration and to distinguish instances with string values from non-element character data. The old method is supported for this release, but will produce a warning when used, and is likely to be removed in the next release.

Key features of the release:

  • Full support for mixed content schema through a new method orderedContent on complex binding instances and flags that control when that list affects document generation. See Influencing Element and Mixed Content Order. This is particularly relevant to XHTML.
  • Immediate validation of values assigned to plural elements.
  • A first step to providing finer control of validation, using ValidationConfig

The following reported defects/enhancements have been addressed:

1.2.0 (07 Nov 2012)


This is a beta release intended to allow independent verification that the new content model works. It is released on SourceForge, but will not be made available through PyPI until such verification is satisfactory.

Key feature of the release:

  • Use of an advanced content model which provides more accurate validation and much better diagnostics when validation fails.
  • Elimination of support for Python 2.4 and 2.5 (necessary to ease the path to a future Python3 implementation of PyXB).
  • Complete rework of the exceptions that normally arise when using binding instances. These exceptions now provide information through attributes rather than a text message.

The following reported defects/enhancements have been addressed:

  • Improved diagnostics. SF ticket 26, SF ticket 75
  • Correct handling of schema where greedy processing fails. SF ticket 112, SF ticket 133
  • Elements and attributes in generated bindings are emitted in a deterministic order sorted by schema and location within schema, to ensure deterministic processing and reduce variation between bindings when generation is repeated. SF ticket 164
  • Make documentation annotations from schema available as documentation strings for the generated complex type bindings. SF ticket 79
  • Update OpenGIS bindings to use W3C XLink schema. SF ticket 165


The 1.1.x release series entered maintenance mode with release 1.1.5. It is unlikely that further releases will occur, though critical bugs fixes may be made available in the release branch. Bugs reported against 1.1.5 will generally be fixed in the 1.2.x series.

1.1.5 (01 Nov 2012)


This is a maintenance release to finalize the 1.1.x series before starting the 1.2.x series.

The following reported defects/enhancements have been addressed:

1.1.4 (15 Jun 2012)


Key features of the release:

  • A large number of bug fixes, especially in the area of internationalization. Naive uses of unicode have mostly been eliminated.
  • Corrections and improvements to date and time-related types, especially with respect to timezones.
  • More namespaces have been added in the common, WS-*, and OpenGIS bundles.
  • This release eliminates the separate packages for different bundles; all bundles and documentation except for OpenGIS are incorporated into the release file. OpenGIS is present but must be built manually; see OpenGIS

Examples have been added:

  • examples/customization shows advanced customization, including injecting behavior into binding classes corresponding to types from which other XML types are extended or restricted. It also demonstrates how to introspect the bindings within a module, automating much of the process of extending multiple bindings with the same functionality.
  • examples/unicode_jp demonstrates internationalization features including the ability to intercept PyXB’s construction of Python identifiers from XML identifiers to produce meaningful identifiers when the schema language is not compatible with the Python 2 requirement that identifiers be ASCII.

The following reported defects/enhancements have been addressed:

1.1.3 (09 Sep 2011)


General cleanup, finish up a few previously unsupported datatypes.

1.1.2 (30 May 2010)


Complete replacement of the model group portion of the content model. The NFA-to-DFA approach is gone. The resulting system does a better job in less space and significantly faster: 30% faster on the standard tmsxtvd test, orders of magnitude on documents with large sequences of optional elements.

Note: A couple unit tests fail due to incidental reasons that will be addressed when fixing SF ticket 75. Also, the architecture documentation for the validation portion of the content model is completely out of date.

  • Handle sequences of optional elements more effectively. SF ticket 33
  • Correct multi-step attribute restriction/extension. SF ticket 80
  • Support prefix rewrite for schema location URIs. SF ticket 81
  • Fix syntax error generating wildcard namespace constraints. SF ticket 84
  • Support whitespace validation bypass in simple type definitions. SF ticket 85

1.1.1 (28 Jan 2010, rev 2135)


Major feature is significantly improved ability to determine what content is not acceptable to the validating parser, and where it can be found in the document. See Locating Invalid Content.

Also, using keyword parameters in binding instance constructors will now raise an ExtraContentError in the case where the keyword does not correspond to the PyXB-generated identifier associated with an attribute or element of the class being instantiated.

  • Eliminate complaints about reference to deprecated sha module in Python 2.6.
  • Fix assorted problems with mixtures of –binding-root, –module-prefix, and –write-for-customization. SF ticket 64, SF ticket 65
  • Eliminate unnecessary attempts to download schema for builtin namespaces. SF ticket 63
  • Avoid assertion failure when provided schema location’s content looks a little, but not enough, like a schema document. SF ticket 72
  • References to declarations in other namespaces are now calculated locally to the module, rather than cached under the expectation the name is valid everywhere. SF ticket 73
  • Imposed consistency on behavior when document root element is not recognized. SF ticket 74
  • Partial implementation of improved error data stored in exceptions. SF ticket 75
  • Detect invalid keywords in instance constructor calls. SF ticket 76

1.1.0 (30 Nov 2009, rev 2092)

Mostly nit-picky things, but some changed interface, so we’re starting a new beta series.

  • Make handling of invalid xsi:type attributes more robust. SF ticket 56
  • Use validation flags to control validation (what a novel concept). SF ticket 57
  • Fix improper generation of content within elements marked as xsi:nil. SF ticket 58
  • Provide mechanism to control default handling of xsi:type attributes in instance documents. SF ticket 60
  • Fix bug where constraining facets were not inherited when deriving simple types by restriction. SF ticket 61

1.0.x (Stable)

1.0.0 (18 Oct 2009, rev 2068)

Nobody seems to have found anything critical for a while, and it does what I want it to, so we’ll let this thing run free.

  • Replace a reference to a non-existent variable with something that works. SF ticket 52
  • Consistently validate instances created from documents no matter which parser is used. SF ticket 53
  • Support caching message references in WSDL SOAP faults. SF ticket 54

0.7.x (Beta)

0.7.3 (15 Sep 2009, rev 2050)

A few bug fixes (one major), a couple new features, and separation of the various components to allow a smaller distribution.

  • Fixed a circular reference in the DFA state classes that caused memory bloat in long-running processes. SF ticket 49

  • When PyXB renames an XML element or attribute due to a collision with a reserved word, it will write a warning, and at runtime will raise an exception if an attempt is made to assign to the original name. SF ticket 50, SF ticket 51

  • Support for pattern restrictions for string-based simple types. SF ticket 47

  • Split distribution into multiple files to reduce download size:

    • PyXB-base-@VERSION@.tar.gz – Complete release, nothing pre-built
    • PyXB-doc-@VERSION@.tar.gz – Overlay with pre-built documentation
    • PyXB-common-@VERSION@.tar.gz – Overlay with XHTML bindings
    • PyXB-opengis-@VERSION@.tar.gz – Overlay with OpenGIS bindings
    • PyXB-wsspat-@VERSION@.tar.gz – Overlay with WS-* bindings
    • PyXB-full-@VERSION@.tar.gz – Complete release with all overlays

0.7.2 (25 Aug 2009, rev 1981)

Primarily to fix a problem with chameleon schemas, but also a couple other minor issues.

  • References to QNames with absent namespaces in chameleon schema included into schema with a target namespace could not be resolved. SF ticket 46
  • Documentation content was not correctly sanitized for use within a Python docstring. SF ticket 45.
  • Elements with complex types with simple content that derived from string could not be marked as nil. SF ticket 48

0.7.1 (23 Aug 2009, rev 1968)

None of the enhancements really got in, but lots of minor bug fixes.

  • Restored the ability to dump copies of all schemas retrieved from a remote location, so you can review them. SF ticket 28
  • Fixed assorted problems with simple types constructed by list and union. SF ticket 35, SF ticket 40
  • Provide link from content values to the element binding that was used to store them. Useful for quickly identifying the owner element when generating a valid child sequence for conversion to DOM. SF ticket 36
  • Fix validation of attributes with union types. SF ticket 37
  • Enumeration fixes: Only generate class-level constants for enumerations that restrict a string representation (not double, float, etc). Verify enumerations work correctly on list and union constructions. Duplicate the class-level constants for enumerations that are defined in union member types. SF ticket 38
  • Provide a facility to simplify the automatic creation of an instance of the appropriate type when storing values into an element. Makes life with anonymous local elements a lot easier. SF ticket 39
  • Enhance management of namespace prefixes by allowing a system-wide default in BindingDOMSupport and by inferring the base name of the module as the default prefix for namespaces defined in imported bindings. SF ticket 41
  • Provided crude facility to bypass validation when generating and parsing documents. SF ticket 44
  • Re-arrange the help for pyxbgen’s arguments, and add a script to automatically generate the Sphinx tables for it.
  • Support bundles of related schemas/namespaces, currently wssplat (for WS-*) and opengis (Geographic Information Systems).
  • Got the user-level documentation back up to where it might actually help somebody.

0.7.0 (03 Aug 2009, rev 1715)

This release is identical to 0.5.2 except that the binding interface has completely changed. All examples have been updated; the documentation will be updated later.

Note that the code in PyXB itself has not changed, and is not going to change. However, properties produce an interface more natural to most Python programs, so that has become the default style. See the –binding-style parameter to pyxbgen if you want to change this.

  • Change the method of accessing elements and attributes in bindings from accessor methods (e.g., elt.attr(), elt.setAttr()) to properties (e.g., elt.attr). SF ticket 18

0.5.x (Beta)

0.5.2 (03 Aug 2009, rev 1706)

Mostly changes to namespaces in support of profiles and application schemas which extend other namespaces with special components, or have their own copies of somebody else’s schemas, etc. See the pyxbgen private/public namespace switches.

  • Presence of nested choice model groups caused the plurality calculations to be wrong: some elements were dropped from the model, others were treated as lists when only one instance would be permitted. SF ticket 34
  • Content accessor would return wrong value for elements with simple type of list variety. SF ticket 32
  • Corrected handling of attributes when deriving complex types by restriction. SF ticket 27
  • Added a source module for XML namespace ( with bindings linked to the built-in component model. SF ticket 23
  • Fix inadequacies in namespace archive management. SF ticket 22

This is expected to be the last release in 0.5.x. 0.7.0 should follow soon thereafter, and only contain the change of default binding style from accessor to property. SF ticket 18

0.5.1 (16 Jul 2009, rev 1502)

Fixed various problems with OpenGIS schemas, specifically dependency loops induced by attempting to keep binding components within modules identified by the schema they came from, and loss of bindings for one namespace due to its being overwritten by those for another.

  • The presence of a wildcard in a sequence or choice model group caused other elements to be lost in the binding. SF ticket 29
  • Rework content verification to operate on binding instances rather than DOM nodes. SF ticket 24
  • Use value() to extract the value of a complex type with simple content; the content() method works only with element and mixed content. SF ticket 21
  • Add support to specify prefix to be used in QNames when generating DOM documents from binding instances. Provide facility to re-use the BindingDOMSupport instance for multiple conversions. SF ticket 15
  • Tightened requirements on content type matching: no more will strings and numbers automatically be converted to match the content model (probably for a branch you didn’t intend). SF ticket 14
  • Fix problems overwriting binding modules with data from a different set of namespaces. SF ticket 31.
  • Eliminated the schema group modules: too hard to avoid dependency loops. Now all components are written into a single module, which is a namespace group module if the components cross namespaces. A namespace-specific module still exists, and imports only the relevant components from the group module.

0.5.0 (07 Jul 2009, rev 1446)

Ability to parse many of the OpenGIS schemas (gml, iso19139, citygml, ogc, ogckml22, om, ows, sensorML, sos, swe, tml, ...).

  • Completely rework include and import directives to correctly handle relative paths in schemaLocation attributes and to only import from a given location once per namespace. SF ticket 9, SF ticket 8, SF ticket 5
  • Parse final attribute on type definitions (though this has no effect on the bindings). SF ticket 10
  • Complete missing implementations in complex types with simple content. SF ticket 6
  • Create new code generation model which supports inter-namespace and inter-schema dependency cycles. Namespace archives can now holds all namespaces that depend on each other.
  • Other miscellaneous. SF ticket 11
  • Did I mention the ability to handle OpenGIS? SF ticket 16


  • Most of the classes in the new code generation model are undocumented.
  • The interface to pyxbgen has changed completely. Look at examples/OpenGIS/ or other scripts in the examples directory.

0.4.x (Stable)

0.4.0 (20 Jun 2009, rev 1204)

First usable public release.

  • Use expanded names for all declarations and definitions. SF ticket 1
  • Use the finite automaton content model for generation as well as parsing. SF ticket 2
  • Support xsi:type and xsi:nil. SF ticket 3
  • Support substitution groups. SF ticket 4
  • Fix many problems in QName resolution
  • Implement the remaining date/time types
  • Rework content model so binding instances are always instances of the binding for the corresponding XML type, with no intervening element instance
  • Add support for SAX parser

0.2.x (Stable)

0.2.0 (30 May 2009, rev 818)

  • Initial public release