================================================
generateDS.py -- Generate Python Data Structures
================================================

----------
What is it
----------

generateDS.py generates Python data structures from an Xschema
document.  It generates a file containing: (1) a Python class for
each element definition and (2) parsers (which use the Python
lxml library) for XML documents that satisfy the Xschema
document.  The class definitions contain:

- A constructor with initializers for member variables.

- Get and set methods for member variables.

- A 'build' method used during parsing to populate an instance.

- An 'export' method that will re-create (write out) the XML element
  in an XML document.

- Etc.

See ``generateDS.txt``/``generateDS.html`` for information about
more capabilities, how to use it, etc.


---------------------------
How to build and install it
---------------------------

De-compress the generateDS distribution file.  Use something like
the following:
 
    tar xzvf generateDS-1.5a.tar.gz

Then, the regular Distutils commands should work:

    python setup.py build
    python setup.py install


-------------
How to use it
-------------

See generateDS.html for documentation.

Produce class definitions and sub-class definitions with something
like the following:

    $ python generateDS.py -o people.py -s people_subs.py people.xsd

Here is a test using the enclosed sample Xschema file:

    $ python generateDS.py -o people.py people.xsd
    $ python people.py people.xml


----------------
More information
----------------

More information on generateDS.py is in generateDS.html
in the distribution or
`generateDS -- Generate Data Structures from XML Schema --
http://www.davekuhlman.org/generateDS.html
<http://www.davekuhlman.org/generateDS.html>`_.

There is also a tutorial.  See tutorial/tutorial.html
in the distribution or
`generateDS -- Introduction and Tutorial --
http://www.davekuhlman.org/generateds_tutorial.html
<http://www.davekuhlman.org/generateds_tutorial.html>`_.


-----------
Limitations
-----------

XML Schema limitations -- There are lots of things in Xschema that
are not supported.  You will have to use a restricted sub-set of
Xschema to define your data structures.  See the documentation
(generateDS.html) for supported features.  See people.xsd and
people.xml for examples.

Mixed content -- generateDS.py generates a parser and data
structures that do not handle or represent mixed content.  Here is
an example of mixed content:

    <note>This is a <bold>nice</bold> comment.</note>

My only, and some what feeble, excuse for this is that
generateDS.py is intended for structured data rather than marked
up text.  However, whether my excuse is a good one or a feeble
one, you should be warned that if you anticipate needing mixed
text, do *not* use generateDS.py.

Large documents -- The parser generated by generateDS.py uses
minidom.  This means that the entire XML document must be read and
a DOM tree constructed in memory.  In addition, the data
structures generated by generateDS.py must occupy memory.  This
means that generateDS.py is not well-suited for applications that
read large XML documents, although what "large" means depends on
your hardware.  Notice that the parsing functions (parse() and
parseString()) over-write the variable doc so as to enable Python
to reclaim the space occupied by the DOM tree, which may help
alleviate the memory problem to some extent.


-------
License
-------

Copyright (c) 2002 Dave Kuhlman

Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
"Software"), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:

The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

[MIT License -- http://www.opensource.org/licenses/mit-license.php]


--------------
Change history
--------------

Version 2.33.17 (11/01/2019)

- Fix to warning messages for simple type validation so that when
  line number is unknown, the line number is omitted rather than
  printing "undefined".

Version 2.33.16 (10/30/2019)

- Modified boiler-plate of ``__eq__`` method in ``GeneratedsSuper``.
  A set object is not hashable.  Caused exception.  Replaced with
  list.

Version 2.33.15 (10/28/2019)

- Fix to ``django``, ``django_etl``, and ``sqlalchemy_etl``.
  Replaced platform specific path separator with platform
  independent use of ``os.path.join``.  Switched to use of args as a
  string rather than a list/tuple and added ``shell=True`` for
  ``Popen.

Version 2.33.14 (10/21/2019)

- Fix to generation of ``validate_`` method so that children defined
  as simple types with ``maxOccurs="unbounded"`` will iterate over
  those children and call the relevant validate method on each one
  (i.e. in a ``for`` loop).

- Fix to function ``makeFile`` in ``generateDS.py`` so that, for
  Python 3 only, it opens/creates the file object with
  ``encoding="utf-8"``.

- Function to generate the ``validate_`` method in each complex type
  class was not being called when command line option
  "--disable-xml" was used.  Moved call to ``generateValidatorMethods``
  outside the ``if not XmlDisabled:`` statement.

- Modified the ``__eq__`` method that is included in the common
  superclass (``GeneratedsSuper``) so as to prevent a possible
  endless recursion through parent.  Also modified the ``__eq__``
  method so that the ``gds_collector_`` member is excluded from the
  test for equality.

Version 2.33.13 (10/14/2019)

- Add IDREF to list of string types so as to prevent it from being
  handled as a complex child.  Note, however that it may be that
  using IDREF to define a child is not even correct usage in a
  schema.  But, when it is used that way, this fix will prevent
  generation of code with undefined references (classes).

Version 2.33.12 (10/09/2019)

- Added ability to generate code so that, during the build phase,
  each instance of classes that represent ``xs:complexType`` objects
  will be given an instance variable ``self.gds_elementtree_node_``
  that is the ``ElementTree`` node.  Users can then add code that,
  e.g., prints ``self.gds_elementtree_node_.sourceline``.

- Added ability to generate code in which the simple type validation
  methods include the source line number of the containing node in
  validation warning messages.

- Fixed bug that occurs when an element as both a child and an
  attribute that have the same name.  We were creating a new
  ``XschemaAttribute`` object with a new name, but were not copying
  the other values in that old object.  So, for example, the data
  type of the attribute was forgotten.  Now, we copy the internal
  values into the new ``XschemaAttribute``.

Version 2.33.11 (10/02/2019)

- Added ability to generate a validate method ("obj.validate_") in
  each complex type class using the "--export" command line option
  (for example, --export="write validate").

Version 2.33.10 (10/01/2019)

- Modification to generated validator methods so that they return
  True or False: True means no validator warning messages issued;
  False means that there was at least one validator error and a
  message was issued.

Version 2.33.9 (09/30/2019)

- Fixes for simple type validation -- Removed use of the Python
  ``warnings`` module.  Added a "collector" class to generated
  modules.  Validation messages are added to this collector.  After
  parsing and building, these messages can be optionally written to
  stderr.  The generated parse functions (``parse``, ``parseEtree``,
  etc.) do this.  The collector class can be optionally replaced by
  an imported module.

Version 2.33.8 (09/26/2019)

- Fix to generation of ``Enum`` classes.  Added backslash escape to
  single quotes in enumeration values.  Thanks to Shane Rigby for
  catching and reporting this.

Version 2.33.7 (09/22/2019)

- Added ``xs:NCName`` and ``xs:QName`` to list of string types so
  that we treat them as strings.  Thanks to David Milner for
  catching and reporting this.

Version 2.33.6 (09/20/2019)

- Modified the type check in generated simple type validation
  methods so that we exit immediately if the test fails.

- Modification to generated simple type validation methods for
  enumeration restrictions: (1) converted items in the enumation to
  the target type; (2) generated simpler inclusion test using the
  Python ``in`` operator.

Version 2.33.5 (09/19/2019)

- Patch to handling of CDATA.  Thanks to huynhlv_54 for this fix.
  See his pull request at Bitbucket.  

- Modified definition of regular expression ``PRESERVE_CDATA_TAGS_PAT``
  (added flag ``re_.DOTALL``) so that it will capture CDATA
  containing a new line.

- Another fix to generation of simple type validation methods.  This
  one enables us to handle simple types whose name includes a name
  space prefix.

Version 2.33.4 (09/17/2019)

- Added type test to generated simple type validation methods.

Version 2.33.3 (09/13/2019)

- Fix to validation test on ``simpleType`` restriction
  ``totalDigits``.  Thanks to Buğra Gedik for catching this.

- Fix to parsing of simple types that are restrictions on float and
  decimal when they are used as attributes.  We need to convert them
  to Python numeric types.

- Fixed capture (build) of float types so that they use
  gds_parse_float.

- Fixed capture (build) of integer types so that they use
  gds_parse_integer.

- Fixes so that binary strings ("b'xxx'") are not displayed in
  validation warning messages.

Version 2.33.2 (09/04/2019)

- Added default hash method (__hash__) from class ``object``.
  Thanks to Mustafa Şenol Coşar and Henrique Andrade for guidance
  with this.

- When run under Python 3, ``generateDS.py`` was generating binary
  strings in MemberSpec.  Changed to produce plain strings.  Thanks
  again to Mustafa and Henrique.

- Validator methods generated by gDS caused an exception under
  Python 2 when the string being checked contained a non-ascii
  character.  The generated code contained an unnecessary cast
  (``str(xxx)``).  Removed the cast.  Thanks again to Mustafa for
  reporting and helping with this.

Version 2.33.1 (07/08/2019)

- Fix so that, for command line option ``--disable-xml``, we also
  disable the use of ``etree_`` in ``GeneratedsSuper.gds_build_any``.
  Thanks to Mustafa Şenol Coşar for identifying this issue.

Version 2.33.0 (06/28/2019)

- Added feature -- Capture the prefix of each element on input
  (during build), then use that during export.  There is an API to
  get and set the name space and prefix in instances of the
  generated classes: ``get_ns_prefix_``, ``set_ns_prefix_``,
  and for members (``XXX_nsprefix_``).

- Added feature -- Before export, the ``parse`` function scans the
  lxml XML element tree and creates all needed NS prefix definitions
  for export at the top level.  There is also a function (which is
  called by ``parse``) that collects all the name space prefixes and
  their definitions (URIs).

- Fix -- Added ``.encode('utf-8')`` for generation of Enum classes
  in order to prevent an exception when running under Python 2.

- Added programming language classifiers (Python 2 and 3) to
  ``setup.py``.  Thanks to Mustafa Şenol Coşar for this suggestion.

Version 2.32.1 (06/16/2019)

- Fix to ``generateToEtreeAttributes`` (generate export to_etree
  attributes) -- For simple types, we need to key on the base type
  so that for restrictions on simple types give us ``xs:integer``,
  ``xs:float``, etc.  Thanks to Van Huynh Le for reporting this and
  providing a test case.

Version 2.32.0 (05/23/2019)

- Added support for ETL/ELT (extract transform load) to SQLAlchemy
  database.  This support is experimental and is work in progress.
  With this support, you can (1) generate models for SQLAlchemy and
  (2) load data from an XML instance doc into your database using
  the ``exportSQLAlchemy`` methods generated by ``generateDS.py``.
  Note that you must run ``generateDS.py`` with command line option
  ``--export="write sqlalchemy"`` in order to generate the
  SQLAlchemy export methods.  For more information see
  ``./sqlalchemy_etl/README.txt``.

- There is an alternative implementation of support for SQLAlchemy.
  You can find that implementation in
  ``./sqlalchemy_etl_alternative``.

Version 2.31.4 (05/17/2019)

- Fixed (or maybe added an implementation of) the processing for
  ``xs:any``.  Now, for elements defined as type ``xs:any``,  we
  save an internal representation that is the Etree node/element
  converted with to a string.  We use
  ``etree_.tostring(node, encoding='unicode')`` to perform this
  conversion.  Then, during export, we write out this string as is.
  If the user wants to write code that manipulates the internal
  (string) representation of the node, then the user can use
  ``etree_.fromstring(self.anytypeobjs_)`` in order to produce an
  Etree Element, then use the Etree Element API to access and
  manipulate parts of the Element.

- Added ``vim`` mode line to ``generateDS.py`` and
  ``process_includes.py.``

- Changed generation of validator methods (``validate_xxx``)
  for simple types ``gYear`` and ``gYearMonth``.  We now use string
  comparison.  The old generate code was causing an exception when
  executed.  Thanks to Andrii Iudin for alerting me about this
  issue.

- Fixed several errors that occurred when ``--no-process-includes``
  option caused skipping of calling ``process_includes.py`` and,
  therefore, we do not return several values that are used later
  (for example, the Lxml element tree).

Version 2.31.3 (05/03/2019)

- A possible global (top-level) type definition caused a conflict.
  Use the type name, not the child/member name.

- Recursive (transitive?) ``substitutionGroup`` caused the wrong
  member name to be used in the ``build`` method.  Fixed so that it
  used the member name (instance variable name) for the
  ``substitutionGroup``.  Thanks to Paul Vajda for reporting this
  and for providing a detailed description and for helping to track
  down this problem and the one above.

Version 2.31.2 (04/25/2019)

- Fix to handling of ``nsmap`` in ``process_includes.py``.  Thanks
  to Jan Orel for reporting this problem and providing a hint on how
  to solve it.

Version 2.31.1 (04/22/2019)

- Modified the processing done in ``process_includes.py`` so that
  when there are definitions in different name spaces that have the
  same unqualified name, we rename some of them.  This enables us to
  generate separate classes for definitions (e.g.
  ``xs:complexType``) that have the same unqualified name but
  different qualified names.

- Added generation of a dictionary (``RenameMappings_``) in the
  generated module that maps qualified names to the renamed
  unqualified names.

- Comments: The above changes should enable uses to deal with
  schemas that have duplicate unqualified names that are actually
  unique qualified names (same name, different namespace).  It's not
  an elegant solution, but hopefully enables us to handle schemas
  (containing duplicate names) that we could not before.  For
  schemas that do *not* have duplicate unqualified names, behavior
  should be unchanged.

- Fix to generation of enums from xs:simpleType -- Make sure that
  labels are unique within a single enum.

Version 2.30.24 (04/02/2019)

- Merged Marco Trevisan's enhancement to handle unicode blocknames.
  See https://bitbucket.org/dkuhlman/generateds/pull-requests/56 and
  https://www.w3.org/TR/xsd-unicode-blocknames.  Thanks Marco.

Version 2.30.23 (04/01/2019)

- Added Marco Trevisan's changes to improve enums and to add
  doc strings for enums.  See:
  https://bitbucket.org/dkuhlman/generateds/pull-requests/57.
  Thank you Marco.

- When running under Python 3, replaced use of ``imp`` module, which
  is deprecated in Python 3, with ``importlib``.  Note that the
  ``-u`` and ``--user-methods`` command line option now take a path
  and file name rather than a dotted module.

Version 2.30.22 (03/20/2019)

- Added ignored file list.

- Fixed Travis tests execution.

- Migrated tests to pytest.

- Deterministic rendering of dicts; strings without u'' prefix.

Thanks to Daniele Esposti for these improvements.

Running the unit tests in the ``tests/`` directory now requires
``pytest`` (``py.test``).  You can run the unit tests by going to
the ``tests`` directory and running ``$ pytest``.

Version 2.30.21 (03/18/2019)

- Modifications to use the ``requests`` package instead of plain
  urrlib.  Doing so will: (1) simplify the loading of resources from
  the network and (2) automatically send all the necessary headers.
  Thanks to Daniele Esposti for this fix.  Note that this makes
  installation of the ``requests`` module a requirement for running
  ``generateDS.py``.

Version 2.30.20 (03/18/2019)

- Fix to regular expression patterns generated for validators.
  Thanks to Raphaël Valyi for this fix.

Version 2.30.19 (03/07/2019)

- Modifications so that during export, the gds_format_xxx,
  gds_parse_xxx, and gds_validate_xxx methods in class
  ``GeneratedsSuper`` in either the generated superclass module or
  in module ``generatedssuper``, if that can be imported at
  run-time, are called.  That enables the user to provide a special,
  customized version of that module, which when importable can
  provide custom formatting during export.  A customized version of
  that module can be created by copying class ``GeneratedsSuper``
  and some required imports, function definitions, etc. from a
  generated superclass module.  Thanks to Gérard Yin for focusing me
  on this issue.

Version 2.30.18 (03/04/2019)

- Fixed export of attributes so that they are not omitted when
  use="required".  Thanks to Andrii Iudin for reporting this and
  providing a test case.

Version 2.30.17 (03/04/2019)

- Merged Raphaël's fix to improve capture of annotations.  Thanks,
  Raphaël.

Version 2.30.16 (03/04/2019)

- Merged Raphaël's fix in ``process_includes.py``.

Version 2.30.15 (02/25/2019)

- Fixed error that occurs when one schema includes another (with
  `xs:include`) and the included schema docs use a difference
  namespace prefix for the XML schema namespace than the main schema
  doc.  For example, one uses "xs:" and the other uses "xsd:".
- Merged fix to `process_includes.py` that copies and preserves
  annotations in `raise_anon_complestypes`.  Thanks to Raphaël Valyi
  for this fix.

Version 2.30.14 (02/12/2019)

- Bugfix: Avoid the TypeError which was caused if infile is a file
  object or BytesIO object.  Thank you, Andreas Brodtkor.

Version 2.30.13 (01/28/2019)

- Fix to order of arguments generated for call to superclass
  `exportChildren` method.  Thank you to Edwin Matthijssen for
  identifying and reporting this.
- Added generation of `set_xxxx_with_type` for children that are
  extensions and abstract and cardinality of zero or one.  Again,
  thanks to Edwin for describing this need and for his guidance
  with implementing it..
- Removed duplicate generation of `add_xxxx` method.
- Fix to generation of "xsi:type" attributes.  Added use of table
  `GenerateDSNamespaceTypePrefixes` from module
  `generatedsnamespaces.py` to specify the namespace prefix on a
  per type basis.
- Added ability to specify the namespace prefix for types that are
  extensions of (derived from) an abstract type and that are
  specified on export with the "xsi:type" attribute.

Version 2.30.12 (01/14/2019)

- Merged pull request that fixed the `parsexml_` generated function
  so that it now handles paths that are consistent with `pathlib`
  objects in addition to paths represented as strings.  Thanks to
  Chris Barnes for this fix.

Version 2.30.11 (12/19/2018)

- Fixes to export of `namespacedef_` so that exporting does not
  repeat the namespace prefix definition in nested elements.

Version 2.30.10 (11/30/2018)

- Fixes to logging.  Pass arguments to logging methods, instead of
  formatting before the call so the formatting will not be done
  unnecessarily.  Thanks to Mustafa Şenol Coşar for these fixes.

Version 2.30.9 (11/29/2018)

- Fix imported module is generated more than once.  Thank you
  Mustafa Senol Cosar for this fix.

Version 2.30.8 (11/14/2018)

- Added unit tests for enum import.  Thanks to Mustafa Şenol Coşar
  for adding these tests.
- Fix to generation of ``externalImports`` so that the generated
  import statements will be in a consistent, predictable order.
- Converted ``tests/EnumImport/test_generated_code.py`` to use the
  Python unit test framework.

Version 2.30.7 (11/12/2018)

- Applied patch with fix for subclass suffix "Sub".  Thanks to
  Lucius for this fix.
- Fixed lines in ``generateDS.py`` that were too long and exceeded
  the style guide recommendation (PEP 8).

Version 2.30.6 (11/09/2018)

- Merged additional namespace prefixes from François.
- Fixed one corner case with the new namespace prefix changes -- For
  mixed content (character content containing mark-up) the
  parameters in the call and definition of method ``export`` in
  class ``MixedContainer`` were out of sync.
- Question: Is it possible that we need an additional change to pass
  namespace prefixes and their definitions through mixed content and
  into complex content that it might contain?
- Added an additional unit test for the namespace prefix changes.
  See the ``ipo`` unit test.  Thanks again to François Guimond for
  help with this.

Version 2.30.5 (11/07/2018)

- Merged namespace prefix changes for export functions from François
  Guimond.  Thank you François.

Version 2.30.4 (11/06/2018)

- Another patch from Mustafa for enums.  Thanks Mustafa.

- A fix from François Guimond for passing namespaceprefix_ to
  ``self.exportChildren``.  Thanks François.

Version 2.30.3 (11/05/2018)

- Merged Mustafa Coşar's fix for enums.  Thank you Mustafa.

- Added several files in the ``test/`` directory to the repo.

Version 2.30.2 (11/01/2018)

- When dealing with a derived type, generate "set_xxx" and "add_xxx"
  methods that automatically set ``original_tagname_`` and
  ``extensiontype_``, so that when exported ``xsi:type`` is used to
  specify the type.  Thanks to Edwin Matthijssen for working with me
  on this.

Version 2.30.1 (10/18/2018)

- New feature -- Each generated data binding class has a new
  instance variable: ``parent_object_``.  It is automatically set to
  reference the parent (i.e. container) of this object.  Thanks to
  Florian de Boissieu for suggesting this enhancement.
- Added a comment to ``generateDS.py`` to help with customizing the
  code generated in getters and setters.  You can search
  ``generateDS.py`` for "add custom code here" and then add custom
  code there.  This is not a terribly convenient way for a user to
  add custom code, so if someone finds a need to use it, please
  contact me and we'll try to find a better way.

Version 2.29.25 (10/05/2018)

- Added a section to the documentation (generateDS.txt) for types
  derived by extension, i.e. types that use the xsi:type attribute
  in the XML instance document.  Thanks to Justin McManus for
  motivating me to learn about this and for providing guidance and
  pointers along the way.
- Added a unit test for types derived by extension.
- Converted unit tests so that we can use Python 3 not Python 2.
- Changes in ``generateDS.py`` so that it uses the ``six``
  compatibility library to handle the ``urllib`` name changes
  between Python 2 and 3.

Version 2.29.24 (08/27/2018)

- Changed name of parameter in the export method from "namespace_"
  to "namespaceprefix_" in an attempt to reduce confusion about its
  use.  It's value, if it has one, should be something like "abc:".
  Thanks to Bernd Zimmermann for his advice and encouragement on
  this.

Version 2.29.23 (08/16/2018)

- Added new command line option "--create-mandatory-children".  If a
  child is defined with minOccurs="1" and maxOccurs="1" and the
  child is xs:complexType and the child is not defined with
  xs:simpleContent, then in the element's constructor generate code
  that automatically creates an instance of the child.  Thanks to
  Vincent Helfre for analyzing this issue and providing guidance
  toward a solution.

Version 2.29.22 (08/03/2018)

- Fixed exception that occurs when (1) an xs:complexType is defined
  with an attribute and a child that have the same name and (2) the
  "-a" command line flag is used to specify a schema namespace
  prefix other than the default (e.g. generateDS.py -a "xsd:" ...).
  Thanks to Daniel Ramirez for tracking down and analyzing this
  issue.

Version 2.29.21 (08/02/2018)

- Fix to use of simpleType name.  Needed to do cleanupName() on the
  class name.  Thanks to Daniel Ramirez for reporting this and for
  suggesting a fix.

Version 2.29.20 (07/30/2018)

- Fix to generation of regular expression used to validate a
  simpleType.  Before this fix, we were replacing "|" with "$|^" to
  implement alternatives in a test.  By removing that replacement,
  we leave it up to the author of the schema to encode that test of
  alternatives into the regular expression.  Thanks to Bernd
  Zimmermann for reporting this issue.
- Fixes to formatting of xs:dateTime.  Thanks to Tim Hulst for this
  fix.

Version 2.29.19 (07/20/2018)

- Merged pull request that makes generated enums for each simpleType
  uppercase.  Thanks to Mustafa Şenol Coşar for this enhancement.
- Added new command line option "--mixed-case-enums" to control
  whether simpleType enums are changed to upper case.  The default
  is True.

Version 2.29.18 (07/19/2018)

- Fixes to the example code underneath `Demos/`.  Thanks to Witold
  Jarzynka for reporting this.

Version 2.29.17 (07/11/2018)

- Merged pull request that generates enum classes for each simpleType.
  Thanks to Mustafa Şenol Coşar for this enhancement.
- Fix to generatedssuper.py -- (1) Added `on_delete` to Django
  ForeignKey declarations.  (2) Added relative import (dot) to
  generated admin.py.

Version 2.29.16 (06/21/2018)

- Added dependencies to ``setup.py``.  Thanks to Laszlo for this
  suggestion.
- Fixes to specifying character encodings.  We now read XML schema
  files in binary mode.  (2) We now use the value of
  --external-encoding only when a generated module is run under
  Python 2 (not Python 3); we use it to encode the XML instance
  document; if --external-encoding was not used, then we use
  "utf-8" to encode exported XML.  Thanks to Laszlo for his
  guidance on this.

Version 2.29.15 (05/16/2018)

- Fixed issue related to use of command line flag
  --preserve-cdata-tags: when the generated regex pattern was not
  matched (returned None), created an exception.  Added a test for
  None.  Also, required conversion of bytes to str before the
  pattern match for Python 3.  Thanks to Lavanya Poondru for
  reporting this and helping with a fix.

Version 2.29.14 (05/08/2018)

- Fixes to logging in generateDS.py -- (1) Setup logging only when
  running as script.  (2) Use named logger instead of root one.
  Thanks to Andrei Fokau for this fix.

Version 2.29.13 (05/08/2018)

- Merged PR from Andrei Fokau.  Fix tests by striping varying lines.
  Also add config for Travis CI.  Thanks to Andrei for this
  enhancement.

Version 2.29.12 (04/23/2018)

- Patch to compute the fully qualified name of the simple types
  before resolving the built-in base type.  Thanks to Alim Gokkaya
  for this fix.

Version 2.29.11 (03/16/2018)

- Fix for the --no-namespace-defs command line option.  The work on
  namespaces in v. 2.29.6 appears to have conflicted with and
  deactivated this.  Thanks to Olof Kindgren for reporting this.
- Added unit test for --no-namespace-defs.

Version 2.29.10 (03/14/2018)

- Fix to resolution of child types -- Formerly, we were adding some
  unnecessary and unwanted entries to `fqnToElementDict`, which
  caused the look-up to get the wrong type.  Thanks to Olof Kindgren
  for guiding me through this.

Version 2.29.9 (03/02/2018)

- Added command line flag --always-export-default
  (AlwaysExportDefault).  When used, the generated module will
  always export attributes with a default value even when the
  current value is equal to the default value.  Thanks to Marc
  Capavanni for suggesting this.

Version 2.29.8 (03/02/2018)

- Added a change so that an attribute specified as `fixed` will be
  handled in the same way as one specified as `default`.  This
  leaves it to the user to validate and enforce the `fixed`
  restriction in some other way, e.g. through use of an XML
  validating parser such as `xmllint`.  Thanks to Sanja Abbott for
  suggesting this enhancement.
- Various fixes for string/unicode differences across Python 2 and
  Python 3.

Version 2.29.7 (02/05/2018)

- Fix for unicode error that occurs during simpleType validation
  under Python 2.7.  Thanks to Juha Tuomala for reporting this issue.
- Added a test for xs:simpleType validation that uses a pattern
  containing a multi-byte character.

Version 2.29.6 (01/22/2018)

- Fix to generation of namespace prefix in export methods.  With
  this fix, process_includes.py collects information about which
  xs:element and xs:complexType definitions are in which target
  namespaces.  Then generateDS.py uses that dictionary to generate
  export methods that produce the namespace prefix.  Thanks to Rob
  Calvert for identifying this problem and for helping me to
  understand it.

Version 2.29.5 (01/17/2018)

- Fix to prevent infinite recursion that happens when a simple type
  is defined whose name is the same as it's restriction base type
  except for the namespace prefix.  Thanks to Nicolas de Saint Jorre
  for reporting this problem and for providing a schema that
  reproduces it.

Version 2.29.4 (12/14/2017)

- Fix for exporting the child of an element, when that child is
  declared as an instance of an abstract type (abstract="true" in
  the schema).  When exporting, the type of the child needs to be
  determine at runtime through polymorphism.  This fix prevents the
  containing (parent) object from passing the name of the abstract
  class to the instance of the concrete class when calling its
  export function.  Thanks to Rob Calvert for reporting this and for
  his analysis that helped me understand the problem.

Version 2.29.3 (12/11/2017)

- Resolved an issue with a type casting problem that occurs when a
  numeric default value is provided as default for an xsd:attribute.
  Normally attributes were being type casted at the constructor to a
  Python type if the types are one of the builtins. However
  attributes derived from a builtin type via xsd:simpleType
  definitions couldn't be casted to a base type and were being
  treated as strings.  Refactored out the simple type resolution
  from the XsdElement class to a function so that we can reuse the
  same code for both elements and attributes.  Thanks to
  Alim Gokkaya for fixing this and for providing a pull request.

Version 2.29.2 (12/07/2017)

- Fix for use of StringIO in generated subclass modules.  Thanks to
  Rohan Dsa for alerting me on this.

Version 2.29.1 (12/07/2017)

- Fix to generation of code to export child elements that have
  default values.  When the element's value is equal to the default
  value, the export of the element should be omitted only if the
  element is optional (i.e. minOccurs=0).  Thanks to Andrii Iudin
  for reporting this.
- Several modifications to use the `six` module as a cleaner way to
  smooth over differences between Python 2 and Python 3.
- Added file generateds/django/README.txt containing instructions on
  running the Django code generation support.  Thanks to Christian
  González for reporting problems with this and for providing
  information that helped understanding the source of the
  difficulties.

Version 2.29.0 (11/28/2017)

- Fixes to export of namespace prefixes for schemas that are
  imported.  Thanks to Bob Barcklay for reporting this and for
  advise on fixes.

Version 2.28.2 (10/27/2017)

- Changed name/version number scheme to all numeric with dots.
  Required by setuptools or the Python Package Index.
- Accepted and merged pull request by Eugene Petkevich that fixes
  export of elements with mixed content.  Thank you Eugene.

Version 2.28d (10/25/2017)

- Fix for incorrect resolution of type for a child element type
  defined with <xs:element ref="Abc"/>.  The ref= can refer to a
  global (top level) xs:element rather than an xs:complexType, in
  which case we need to use the type= to determine the
  xs:complexType.  Thanks to Bob Barcklay and Olof Kindgren for
  reporting this issue and for working with me on this and helping
  to track down the fix.

Version 2.28c (10/17/2017)

- Fix for generation of GDSClassesMapping dictionary.  Formerly, we
  were generating entries in this dictionary for some xs:element
  items that were not at top level.  Fixed so that only xs:element
  items at top level (immediately under the root/schema node) are
  included in this dictionary.  Thanks to Christin Gunning for
  reporting this and for guiding me on this change.

Version 2.28b (08/22/2017)

- Fix for Django models and forms generation -- "float" data type
  was being mapped and was not treated as a simple data type.
  Thanks to Sriram Sundar for catching and reporting this.
- Sriram also requested that in the Django models and forms
  generation, we be able to omit the "_model" and "_form" suffix on
  generated class names.  There is now a "--no-class-suffixes"
  command line option accepted by both gends_run_gen_django.py
  and gends_generate_django.py to do that.  Thanks to Sriram for
  this suggestion.
- Added Python version to the information in the comments at the top
  of generated modules.

Version 2.28a (06/23/2017)

Significant work by Alim Gokkaya.  Thank you, Alim.

Here's a summary of what's been changed:

- Added new command-line options:

  - ``--disable-xml``: Toggles the generation of XML serialization
    related code
  - ``--disable-generatedssuper-lookup``: Disables generation of the
    try-except lookup for a `generatedssuper` module
  - ``--use-source-file-as-module-name``: Sets the source XSD file name as
    the target module name in the one file per XSD mode

- Retained ``xsd:choice`` related information in the generated class
  ``MemberSpec``.
- Retained the original XML schema attribute definitions in the generated
  class members.
- Fixed generation of ``import`` statements for the base classes.
- Fixed class not being generated when parent class is defined in
  another XML schema file.
- Fixed fqn-module mapping being unavailable for the classes defined
  in imported XML schema files.
- Fixed attribute names are sometimes not cleaned-up from the prefix
- Fixed ``xsd:simpleType`` validations methods are not being generated in
  ``one-file-per-xsd`` mode.
- Fixed equality check against objects defining extra attributes.
- Added unit test for command line options ``--disable-xml`` and
  ``--disable-generatedssuper-lookup``.

Version 2.27b (06/09/2017)

- Fixed a bug that occurred when an element definition contains a
  child defined as xs:any.  The member spec (``MemberSpec_``) was not
  generated with the correct name.  Also in the django support,
  added a temporary fix for xs:any child elements.  Thanks to Rémy
  Gibault for reporting this.
- Django support -- Models in django are case insensitive.  That
  means that if a schema defines multiple element types that differ
  only in case, and we generate two models that differ only in case,
  django says it's an error.  So, implemented a facility which,
  when multiple names differ only in case, adds a suffix so that
  those names will be unique even when case is ignored.  Again,
  thanks to Rémy for finding and reporting this.
- Django support: (1) Added a test and more explanatory error
  message for the case where gends_generate_django.py was failing to
  import the correct version of module generatedssuper.py.  (2)
  Created a mapping so that all generated model and form names are
  unique even when case is ignored.  (3) Added a name mapping to
  avoid clashes with Python keywords.

Version 2.27a (06/01/2017)

- Fixed bug in gends_extract_simple_types.py that caused an
  exception when the simpleType name has a namespace prefix.
  Thanks to Rémy Gibault for reporting this.
- Added two utilities that can be used to replace the capability
  invoked by the --one-file-per-xsd command line option.
  utils/collect_schema_locations.py can be used to collect and write
  out the top level schema locations.  batch_generate.py can be used
  to (read the output from collect_schema_locations.py and generate
  modules.  Use --help to obtain more information from each of
  these.  For instructions on this, see the docs and also the README
  in the utils/ subdirectory.
- Various fixes for the generation of namespace prefix definitions
  when the generated export functions are called.  Thanks to Eugene
  Petkevich for reporting and working with me on this.
- Added command line option --no-namespace-defs to force export
  functions to not added namespace prefix defintions.
- Added ability for generated modules to import a module
  (generatedsnamespaces.py) containing a dictionary
  (GenerateDSNamespaceDefs) that maps element type names to the
  namespace prefix definitions (or any XML attributes, actually)
  that are to be added to specific elements during export.  See the
  docs and also notes near where generatedsnamespaces.py is imported
  in a generated module.
- Fixed an error in gends_run_gen_django.py which caused it to fail
  when generateDS.py produced a warning message.  Thanks to Rémy
  Gibault for catching and reporting this.
- Added a utility to help with analyzing complex schemas.
  utils/show_schema_hierarchy.py can by used to show an indented
  hierarchy of schemas that are pulled in by xs:include and
  xs:import elements.  Type `utils/show_schema_hierarchy.py --help`
  for more info.  Also see the docs.

Version 2.26a (05/09/2017)

- Added command line options --no-collect-includes and
  --no-redefine-groups.  These options selectively turn off tasks
  performed in process_includes.py.  These options were added
  because the use of --no-process-includes (which omits all
  processing done in process_includes.py) was reported to cause
  errors.  See the documentation and the usage message (run
  `generateDS.py --help`) for more information.  Thanks to
  Florian Wilmshoever for reporting and working with me on this.
- Moved README to README.rst so that hopefully Bitbucket will
  render it as reStructuredText (with Docutils).  Also, fixed a
  number of reST/Docutils errors in README.rst.
- Another fix for unicode encoding in process_includes.py.
- A bug was uncovered when the "-o" command line option is omitted
  and the Python version is 3.  Added a check that forces the use of
  the "-o" option unless the one-per option is included.  Thanks to
  Oskari Petas for reporting this.

Version 2.25a (03/21/2017)

- Fixes to the Django support for Python 3.  Thanks to Shane Rigby
  for all his help with all of these changes to the Django code.
- Added `optional` to the MemberSpec so that when command line
  option "--member-specs" is "dict" or "list", the generated code
  specifies whether the member is optional or not.
- In the Django support, `django/gends_run_gen_django.py` now has
  new option "-s" ("--script") that can be used to write out the
  command lines used internally by `django/gends_run_gen_django.py`.
  This new flag can be used to generate a shell script that can be
  run instead of `django/gends_run_gen_django.py`.  (Note: The
  script might require a minor edit or two.)
- In the Django support, there is now some attempt to treat optional
  members specially and to generate "blank=True, null=True," in the
  `models.py` file.
- Fix to Django support so that we generate *unique* names for
  `related_name`.
- Added several date/time types for Django support: 'gYear',
  'gYearMonth', 'gMonth', 'gMonthDay', 'gDay',


Version 2.24b (01/02/2017)

- Added several fixes to generateDS.py and process_includes.py that
  are needed for the support for Python 3.  Thank you Ian Glover for
  catching this and for contributing the fixes.
- Fixed bug in generation of regular expression for validating
  pattern in a restriction on a simpleType.  In the pattern, we
  needed to replace "|" with "$|^", unless the vertical bar was
  escaped with a backslash.  This was necessary so that each regular
  expression separated by a vertical bar would be anchored at the
  left and right.  Thanks to Clint Pitzak for catching and reporting
  this.
- Modified the Django support (in ./django/) so that it will run
  under Python 3.  Thanks to Shane Rigby for reporting this problem.
- Fixed an error in encoding unicode ``valueOf_`` for  <xs:complexType 
  <xs:simpleContent> <xs:extension base="xs:string">.  Thanks to
  Andrii Iudin for catching this.

Version 2.24a (11/16/2016)

- Added entry_points to setup.py so that distutils will generate
  executable scripts for executable .py files (for example,
  generateDS.py and process_includes.py).  Thanks to Michael Jenny
  for suggesting this and for showing the way to do it.
- Fixed function call signature mismatch in MixedContainer call to
  export method.  Thanks to Lev Israel for catching this and
  providing the solution.
- Added "remove duplicate elements" fix to catch duplicate
  definitions of child elements with the same name inside a single
  parent element.  This fix does the following: (1) removes
  duplicate child; (2) makes the remaining child a Python list
  (effectively maxOccurs="unbounded"); (3) prints a warning message
  when it finds and removes a duplicate.  Thanks to Pietro Saccardi
  for catching and reporting this.
- More fixes for "remove duplicate elements".
- Removed command line option for "remove duplicate elements".  This
  behavior will now always be performed.
- Added unit test for "remove duplicate elements".
- Added command line option "--no-warnings" to turn off warning
  messages.  I needed it for the unit test for "remove duplicate
  elements".

Version 2.23b (09/26/2016)

- Added missing unit test files to build (MANIFEST.in).
- Fixed exception that occurs when character content is empty for an
  element defined as type xs:token.  Thanks to Andrii Iudin for
  reporting and checking this.

Version 2.23a (09/14/2016)

- Integrated Clayton Daley's fixes to the unit tests.  Thanks much,
  Clayton.
- Clayton's fixes to the unit tests uncovered several errors that
  had been masked and hidden.  Fixed those errors, for example: (1)
  eliminated generation of erroneous call to validation method; (2)
  added catalog file.

Version 2.22c (04/26/2016)

- Fixes to generation of validation methods for xs:date, xs:time,
  and xs:dateTime simpleType.  Thanks to Andrii Iudin for reporting
  this and for suggesting a solution.
- Added additional unit tests for validations of xs:date, xs:time,
  and xs:dateTime simpleType.

Version 2.22b (04/20/2016)

- Fixed endless recursion that occurred while attempting to replace
  attribute group names.  Thanks to Bing Wang for reporting this and
  for identifying and providing the XML schema that reproduced it.
- Fixed failure to clean up names containing special characters in
  function generateBuildStandard_1.  This error was uncovered
  when generating code from Bing Wang's schema.  Thanks again Bing.

Version 2.22a (04/11/2016)

- Added support for additional command line options to
  generateds_gui.py.  Added analogous support to generateDS.py for
  use of session files produced by generateds_gui.py.
- There is now a bit of documentation with a few usage notes on
  generateds_gui.py.  See generateds_gui_notes.txt and
  generateds_gui_notes.html.

Version 2.21a (04/01/2016)

- The GUI (graphical) front end to generateDS.py has been
  resuscitated and is now working again thanks to Aleksandr
  Dragunkin.  
  The GUI front end must be run under Python 3, and you must install
  Python support for Gtk.
  Aleksandr has also provided a Russian translation of the labels
  etc in the user interface.  You can run that with::

      $ cd /path/to/generateds/gui
      $ python3 generateds_gui.py --impl-gui=generateds_gui_ru.glade

  Note that the GUI interface still lacks support for a few of the
  command line options that were added most recently to
  generateDS.py.  If you need one or more of those missing options
  but would still like to use the GUI front end, you can consider
  using the "Capture CL" under the Tools menu, and then copy and
  paste the result into a shell script, add any needed options to
  that script, and run the script from the command line.


Version 2.20b (03/28/2016)

- Fixes to handling of simpleType with and without restrictions on
  another defined simpleType.  These were not being handled
  correctly when the name of the simpleType contained a dash.
  Thanks to Ryku for identifying this problem and for a very helpful
  description of what was wrong and for providing schemas to
  reproduce the problem.

Version 2.20a (02/25/2016)

- Another patch for Python 2 and 3.  We needed to protect against
  performing an encoding that caused an exception in generateDS.py
  and process_includes.py.  Thanks to Marcus Schäfer for catching
  this and for providing a fix.

Version 2.19b (02/16/2016)

- Modified generated code so that it will run under both Python 2
  and Python 3.  There is no longer any need to generate different
  code for Python 2 and Python 3.  If fact, the "--py3" command line
  option has been removed.

Version 2.19a (02/08/2016)

- Added the ability to generate code that can run under Python 3.
  Use the "--py3" command line option.  Note that if you generate
  code for Python 2 (the default), then you must run that generated
  code under Python 2.  And, if you generate code for Python 3,
  then you must run that generated code under Python 3.  There is
  currently no way to generate code that will run under both Python
  2 and Python 3.
- Modifications so that generateDS.py itself can be run with either
  Python 2 or Python 3.
- Fixed the template (TEMPLATE_HEADER) so that it uses the format
  function and keyword arguments.
- Added info on --py3 command line option to doc (generateDS.txt).
- Added new script (fix_subclass_refs.py) that can be used to fix-up
  (change) which subclass file (of two or more that were generated
  with the -s command line option) is used by the superclass file
  when parsing an XML instance document.  This will enable you to
  use the -s option to generate multiple subclass files, add
  different code to each of them, and then parse documents and
  create instances of classes from one then another during the same
  run.  But also, see next item.
- Added generation of code to lookup the subclass of a generated
  class using a global variable containing the subclass module.
  This provides an alternative and more convenient way to do the
  above (i.e., use fix_subclass_refs.py to select from multiple
  subclass files generated with the -s command line option).
  However, there may be tasks that can be performed with that script
  or a modified version of it that cannot be done with this approach
  using a global variable.  Here is a sample script that uses this
  option::

      import tmp01suba
      import tmp01subb
      def test():
          tmp01suba.supermod.CurrentSubclassModule_ = tmp01suba
          roota = tmp01suba.parse('test01.xml', silence=True)
          tmp01subb.supermod.CurrentSubclassModule_ = tmp01subb
          rootb = tmp01subb.parse('test01.xml', silence=True)
          roota.show()
          print '-' * 50
          rootb.show()
      test()

Version 2.18a (12/16/2015)

- Fixed quoting of simpleContent so that, e.g., "&amp;" is exported
  as "&amp;" and not as "&".  Thanks to Ardan Patwardha for
  reporting this and contributing a fix.
- Fix to generation of exportAttributes so that the test for already
  generated is properly quoted.  Thanks to Naresh Shenoy for
  reporting this and for contributing a fix.
- Another fix related to the unquoted constant in exportAttributes.
  A simple fix had a bad conflict.  Thanks to Christian Rasmussen
  for focusing my attention on this one.
- Fix for xs:simpleContent that extends type xs:float (or xs:integer
  or other numeric types).  When set to numeric zero (for example,
  after parsing the instance doc), the value was not being exported.
  Thanks to Ardan Patwardhan for diagnosing this and for
  contributing the fix.

Version 2.17a (08/17/2015)

- Modified setup.py so that process_includes.py is installed where
  it can be imported.
- Changed default settings for export -- Default is now to generate
  only the normal export methods, instead of both normal and
  literal.  See command line option --export.
- Fix to regex pattern used to capture "<![CDATA[ ... ]]>".  The old
  pattern was dropping ending characters when the content contained
  HTML/XML markup.  Thanks to Adrian Cook for this fix.
- Merged use of replacement patterns in cleanupName. With this fix
  users can specify patterns to look for and replacements strings to
  be used to clean up special characters and other patterns in
  names.  There are some notes in the document; search for
  "cleanup-name" in generateDS.html.  Thanks to Fedor Tyurin for
  suggesting and implementing this enhancement.
- Added unit test for enhanced cleanupName.  Added documentation to
  generateDS.txt.

Version 2.16a (05/28/2015)

- Added new command line option ("--preserve-cdata-tags") that
  causes generation of code that contains a special parser to retain
  CDATA tags.  Thanks to Adrian Cook for reporting this, for
  providing test data and test cases, and for help with testing and
  feed-back.
- Added ability for user to specify the names of classes generated
  from anonymous, nested xs:complexType definitions, rather than
  accept the names created in process_includes.py.
- Added a unit test for the anonymous, nested definition capability.
- Fix to error caused by check (in generated code) for whether lxml
  or ElementTree is being used.  We no longer support use of
  ElementTree.  Thanks to Emil Nordling for catching and reporting
  this.

Version 2.15b (04/07/2015)

- Fix to generation of simpleType validation code for list (unbounded)
  elements.  Thanks to wobanator for this fix.
- Fix to code for --one-file-per-xsd.  Added check to avoid an
  infinite loop schemas not suitable to --one-file-per-xsd.  Thanks
  Michael Vezie for catching this and for identifying relevant
  location in the code.  And, thanks to George David for providing a
  better fix than mine.
- Enhancement so that child elements defined with a default value
  will not export when the current value and the default value are
  the same.  Also added equivalent changes for attributes.  Thanks
  to Jan Biel for finding and reporting this.
- Added unit tests for the above default value enhancement.

Version 2.15a (02/18/2015)

- Modifications so that we generate code that can be used by Python 3.
  Thanks much to Richard Gerkin for this work.
- Removed possible use of ElementTree.  Lxml is now a requirement
  for both running generateDS.py itself and for running the
  generated code.
- Fixed exporting of text content so that, when it contains CDATA
  sections, the mark-up characters inside the CDATA sections are not
  escaped.  Thanks to George David for reporting this and for
  helping with a fix.

Version 2.14a (11/26/2014)

- Fixed export of simpleType lists (added "' '.join(data)".  Thanks
  to Per Rosengren for catching this.
- Added new style validation of simpleType data.  Validation
  requirements are captured from the XML schema definition of the
  simpleType, e.g. 'restriction base="..."' etc.  Thanks to
  azer gh for implementing much of this extended capability.
- Added unit test for simpleType validation, including test for
  proper detection of bad (invalid) data.
- Did some code cleanup with the help of the flake8 code checker.
- Added a fix so that attribute declarations that use ref= rather
  than type= will also be generated with the specific type.  Thanks
  to Florian Wilmshoever for catching and reporting this and for
  providing an XML schema as a test case.
- Added unit test for reference to simpleType.
- Fix to generation of names of substitutionGroup.  The namespace
  prefix was not being stripped in some cases.

Version 2.13a (09/09/2014)

- Minor fix to function generateToEtreeChildren.  Must generate
  call to get_valueOf only when defined (i.e. when element is
  simpleContent or isMixed).
- Fix to generation of class name prefixes added with the "-p"
  command line option.  This fix was added by Christian Ascheberg.
  Thank you Christian.
- Added unit test for class name prefixes command line option.

Version 2.12f (08/12/2014)

- Fix for substitutionGroup conflict with keyword name mapping.
  Thanks to Leonid Minchin for finding and helping with this
  problem.
- An exception occured when an element had a documentation string
  that was short (possibly 1 character).  Fixed.  Thanks to Matthias
  Zaake for finding this and for providing a patch.

Version 2.12e (06/16/2014)

- Fix for formatting error.  Thanks to Nikolay Lavrov for catching
  this and for providing a fix.
- Fix to gds_parse_datetime().  The Python datetime module's
  datetime object uses microseconds, but xs:dateTime uses fractions
  of a second (e.g. 0.123).  Converted from decimal fraction to
  microseconds.  Thanks to Mikki Weesenaar for catching this.
- Modified behavior and names for generated method insert_xxx(which
  are generated when, e.g., maxOccurs="unbounded"), so that now we
  generate insert_xxx_at and replace_xxx_at.  Thanks to Bart
  Wagenaar for pointing out this deviation from Pythonic style.
- Function transitiveClosure in generateDS.py was susceptible to
  infinite looping.  This seemed to occur when a substitutionGroup
  contains a member with the same name as the head of the
  substitutionGroup (but in a different namespace?).  Added a test
  to stop the recursion when this occurs.  Thanks to Stuart Chalk
  for finding and reporting this.
- Added explanation to the documentation explaining how the source                  
  distribution (generateDS-x.xxy.tar.gz or Bitbucket) is needed for                 
  use of the Django model generation capability.                                    

Version 2.12d (04/02/2014)

- Fix for an infinite loop caused by inconsistent use of
  mapped/clean names with list AlreadyGenerated.  Thanks to Jerome
  Allasia for catching this and for suggesting a fix.
- Added a unit test for the use of mapped/clean names, in particular
  when one xs:complexType is an xs:extension of another.
- Changed several lists to sets for faster look-up, for example
  AlreadyGenerated, AlreadyGenerated_subclass, DelayedElements, etc.
- Cleaned up the use of functions mapName() and cleanupName() to
  avoid duplicate transformations.

Version 2.12c (03/28/2014)

- Fix for "one module per XSD file" to handle an include or import
  element that refers to a *remote* schema on the Net (i.e. the
  location is "http:..." or "ftp:...") rather than a file on the
  local file system.  Added ability to access include/import file
  across the Net.  Thanks to Jinquan Liu for reporting this.
- Added schema to unit test for "one module per XSD file" that is
  read from remote site (http://www.davekuhlman.org).
- Fix to process_includes.py -- When run directly from the command
  line (as opposed to imported and called from another python
  module), the fixtypenames option was not being intialized.
- Fix for error in order of generation of classes that have
  superclasses.  When an anonymous simpleType occured, the name of
  the enclosing complexType was used, which caused generateDS.py to
  believe that the superclass had already been generated.  Thanks
  again to Jinquan Liu for reporting this issue.
- Fix for handling of xs:substitutionGroup -- Namespace prefix was
  causing gDS to fail to match on substitutionGroup name.
- Added code so that an instance of a generated class can remember
  the tag from which it was built.  This is needed for instances of
  a class that represents an element type that is a member of a
  xs:substitutionGroup.  But, in fact, generated code now uses this
  feature to remember and export the tag name of all complex
  elements.
- Enhanced command line option --root-element so that both the root
  tag and the root class can be specified (separated by a vertical
  bar).
- Added support for the ability of an element definition to inherit
  minOccurs and maxOccurs from the xs:sequence that contains it.
- The command line options and command line arguments used to
  generate modules are now included as comments near the top of the
  generated modules.  Also included in these generated comments is
  the command line used to generate the module.  This will help
  users later to determine which XML schema and what options were
  used to generate each module, and to re-generate the module, if
  needed.  Thanks to Mikki Weesenaar for suggesting this
  enhancement.

Version 2.12b (02/10/2014)

- Fix to the aliasing capability.  You should now be able to alias
  one element to another, and by doing so, only generate the
  targeted alias.  See notes on generateds_config.py in the
  documentation for more on this.  Thanks to Mikki Weesenaar for
  bring up the use case that needed this.
- Additional fixes for the "one module per XSD file".  Also,
  creation of a unit test for this capability.  See section "One
  Per -- generating separate files from imported/included schemas"
  in the documentation for more information.  Thanks again to
  George David for all his work on this.
- Fixes to process_includes.py -- Some uses of namespace prefix xs:
  were hard-coded, whereas some XML schemas use xsd: instead of xs:.
- Various fixes to unit tests so that all unit tests pass when using
  either the cloned Mercurial repository at Bitbucket
  (https://bitbucket.org/dkuhlman/generateds) or the tar achive.

Version 2.12a (10/29/2013)

- A name conflict issue caused by naming anonymous types.  An
  anonymous type is a complexType that does not have a name
  attribute and that is nested inside an element that does not have
  a type attribute.  Strengthened the code that generates new,
  unique names.  And, also fixed a problem or two in the surrounding
  code.  Thanks to Shahaf Abileah for reporting this and for
  providing test files to reproduce the problem behavior.
- Created unit test for anonymous types.
- Added command line option --fix-type-names.  This may be useful if
  there are name conflicts in your XML schema, for example, because
  the schema refers to two types with the same name but in different
  namespaces.
- Ability added to generate one Python module for each XML Schema
  (.xsd file) imported/included.  Added command line options
  --one-file-per-xsd, --output-directory=, and --module--suffix= in
  support of this.  Thanks much to George David for implementing
  this new feature.
- This change provided by Logan Owen. -- Return self from build
  function of generated classes, to allow easy chaining.  The main
  use case for this change is if you have a list of xml documents,
  and you want to change them into generateDS class instances.
  Thank you Logan.

Version 2.11a (08/16/2013)

- Added ability to use XML catalog to find included/imported
  schemas.  The -c command line option has been added to support
  this.  Thanks to George David for implementing this enhancement.
- Added unit test for the catalog capability.
- Added ability to pick up the target namespace and its prefix, then
  use them in calling the export functions from the parse functions.
  Thanks to George David for suggesting this.
- Several fixes to formatting date and floats during export.  Thanks
  to Domenico Mangieri for catching and fixing these.
- Added generation of an extra, optional "silence" argument to the
  parse functions so that export can be turned on or off at runtime.
  Domenico is the motivator on this one, too.
- The information about minOccurs and maxOccurs in the generateDS
  document (generateDS.txt) was misleading or wrong.  Edited it.
  Thanks to Rinat Yangurazov for catching this.

Version 2.10b (07/22/2013)

- Changed flag for generating getters and setters.  Removed flag
  --use-old-getter-setter.  Replaced it with new flag
  --use-getter-setter, which can have the following values:

     "old" - Name getters/setters getVar()/setVar().
     "new" - Name getters/setters get_var()/set_var().
     "none" - Do not generate getter/setter methods.

  The default is "new".  See the help (use --help option) or see the
  doc (generateDS.txt/generateDS.html) for more on this.  Thanks to
  Mike Vella for suggesting this.
- Changed suffix used to prevent name conflicts with Python keywords
  from "xx" to "_".

Version 2.10a (05/29/2013)

- Added ability to produce mapping (a dict) during call to
  to_etree() that maps gDS instances to their associated etree
  elements.  Also added convenience method gds_reverse_node_mapping
  to reverse the order of keys and values in a mapping/dict.  See
  function parseEtree in the generated code for hints about how to
  produce these mappings.  There is also a note on generating the
  Lxml Element tree in the docs (generateDS.txt/generateDS.html).
- Python datetime.date objects don't have tzinfo, so trying to
  access it in gds_format_date was throwing an error. According to
  http://stackoverflow.com/a/610923, the best way to avoid that type
  of error is to use a try/catch for AttributeError.  Thanks to
  Logan Owen for this fix.
- Fixed bug so that gDS will now handle a simpleType nested inside a
  restriction nested inside a simpleType.  Thanks to Christian
  Kaiser for finding this, for focusing my attention on it, and for
  providing the sample files to test it with.
- Fixed bug where gDS was failing to resolve defined a simpleType
  correctly.  It was failing to add the XSchema namespace (usually
  xs:).  Thanks again to Christian Kaiser for focusing me on this
  one.
- Fixes to handling of xs:dateTime when the XML schema specifies a
  default value and the XML instance document omits the value.
  Also, fixed formatting because datetime.strftime does not handle
  dates outside of range (e.g. earlier then 19000).  Attempts to use
  a consistent internal representation across xs:dateTime, xs:date,
  and xs:time, specifically instances of datetime.datetime,
  datetime.date, and datetime.time from the Python standard library.
  Thanks to Shahaf Abileah for reporting this and for providing an
  example of the schema.  *Caution*: Because this changes the
  internal representation of dates and times used by the generated
  code, this fix may break some existing applications.
- Various fixes to generation of method exportLiteral in generated
  classes.
- More code clean-up in generateDS.py to eliminate coding style
  warnings and errors reported by flake8.  Ditto for
  process_includes.py.  Also, made a few changes to reduce the
  warnings and errors reported by flake8 when run on code generated
  by gDS.

Version 2.9a (02/21/2013)

- Added support for exporting to an Lxml element tree.  The element
  tree can then be serialized to XML, e.g. using Lxml
  etree.tostring().  This innovation is by Logan Owen, who also did
  most of the work on it (but I helped some, too).  Note that this
  work is not yet complete; it's still "work in progress"; but it
  looks very promising.
- Added --export command line option.  This enables the user to
  selectively generate export methods for any or all of normal
  export, export to etree (lxml element tree), or export to literal
  python code.  This will enable users to reduce bulk in their
  generated files when any or all of these are not needed.  The
  default is "write literal", i.e. the normal export methods that we
  are used to.  Use the --help command line option or read the doc
  for a description of this option.
- Fixed a bug that occurs when a schema has an attributeGroup
  referenced with a name that includes a namespace prefix but the
  attributeGroup is defined with a name that does *not* have the
  namespace prefix.  Thanks to Mike Detecca for reporting this and
  for nudging me in the right direction when I, initially, made the
  wrong fix.
- Added unit test for export to etree.
- Various fixes to the to_etree (export to Lxml element tree)
  capability: (1) fix to preserve names that contain special
  characters (e.g. "-" and "."); (2) fix to preserve the type
  attribute (xsi:type) for abstract types that whose type is set
  explicitly.  Round turn (XML --> gDS object tree --> lxml element
  tree --> gDS --> lxml ...) now seems to work reasonably well,
  although I'm guessing that there are still bits missing (in
  particular, support for xs:anyAttribute).

Version 2.8c (provisional) (01/30/2013)

- Changed generated check for attributes that are already_processed
  to use a set object rather than a list.  Since sets are hashed, I
  believe that lookup is faster.

Version 2.8b (01/30/2013)

- Fixed missing underscore in reference to member names in
  generateExportLiteralFn_2.  Thanks to Sergii Chernysh for
  reporting this.
- Fixed use of NameTable for mapping names when an element has an
  attribute and a child with the same name.  Needed to use correct
  name (original name or mapped name) when doing (1)
  fix_dup_names, (2) exportAttributes, and (3) buildAttributes.
  Thanks to Mike Vella for reporting this.
- Fixed gds_parse_datetime so that it will handle fractional
  seconds.  Thanks to Matt Hughes for providing this fix.  Now,
  xs:dateTime values that include microseconds are successfully
  parsed and exported.
- Created a Mercurial repository for generateDS at Bitbucket:
  https://bitbucket.org/dkuhlman/generateds

Version 2.8a (01/10/2013)

  * Fix to process_includes.py so as to remove the limitation on the
    number of unique names it can generate when raising anonymous
    types to the top level.  Thanks to Daniel Browne for help with
    this.
  * Added support for multiple level attributeGroup, i.e. for
    attribute groups that themselves contain references to other
    attribute groups.  Thanks to Harley Green for pointing out the
    need for this.  Also added a unit test for attribute groups.
  * Added support for more date and time simple types, specifically
    gYear, gYearMonth, gMonth, gMonthDay, and gDay.  Thanks to
    Nicholas Krasney for catching this.  Added tests in the unit
    tests for these types.
  * Quite a bit of code clean-up with the help of the flake8 Python
    code checker.  This is predominantly code cleanup that does not
    affect behavior, most commonly splitting lines that are longer
    than 80 characters across multiple lines for readability.  (See:
    http://pypi.python.org/pypi/flake8 for info about the flake8
    Python code checker.  I use it with the Syntastic plugin for the
    Vim text editor.)
  * Added generation of a dictionary that maps element definition
    names to generated class names.  Thanks to Elena Dolinin for the
    original work on this one.
  * Added support for xs:date and xs:dateTime.  These are now
    captured as instances of class datetime.datetime from the Python
    standard library.  They are parsed and exported with the help of
    that class and using methods gds_format_date,
    gds_format_datetime, gds_parse_date, and gds_parse_datetime in
    class GeneratedsSuper (which is part of the generated module).
    Logan Owen did the work on this.  Thanks much to Logan for
    implementing this and contributing this patch.
  * Turned logging off.  I did not realize that generateDS.py had
    been creating a log file (generateDS.log).  Logging can be
    turned back on by modifying the logging calls near the top of
    generateDS.py.
  * Fixed exception that is thrown when the XML schema file (.xsd)
    only contains a simple type.  Now, the output is generated, but
    it contains no data representation classes.  Thanks to Daniel
    Browne for catching this.

Version 2.7c (08/06/2012)

  * Added xs:hexBinary to the list of string types in generateDS.py
    and django/generatedssuper.py.  Effectively, we are generating
    the same code for types xs:base64Binary and xs:hexBinary.  That
    leaves it up to the user to add code that converts into and out
    of these formats.  Thanks to Peter Kreinhöfer for finding this.
  * Added support for compressed export, that is, export without
    ignorable white space (indentation and new lines).  Normally the
    generated export methods produce pretty-printed (indented) XML
    output.  With this change, we generate modules which enable you
    to export in a way that omits ignorable whitespace.  It is
    anticipated that this feature will be useful to those who need
    to export XML documents that are machine readable but not human
    readable.  Thanks much to Logan Owen for doing the work on this
    one.  Compressed (non-pretty-print) output is produced by
    passing the keyword argument pretty_print=False to the export
    method.  There is a note in the document (generateDS.html) about
    it; see section titled "Exporting compact XML documents".

Version 2.7b (12/10/2011)

  * Fix for xs:any in buildChildren in an element defined with no
    other children so that we do not generate "else:" clause without
    an "if ...:".  Thanks to Keith Robertson for help with this.
  * Change for xs:any in buildChildren (when maxOccurs > 1) so that
    the gds_build_any() method always, consistently takes a single
    *child* node as input and returns a single built object.  Thanks
    Marcin Tustin for guidance with this.
  * Fix for element definition containing an anonymous
    xs:simpleType.
  * Added xs:time to list of handled simple (date, time) types.

Version 2.7a (11/04/2011)

  * Fix for case where a child is defined with a reference (ref="")
    to a complexType (rather than a simpleType) and the complexType
    is abstract.
  * Added minimal support for xs:any.  See section "Support for
    xs:any" in the documentation.
  * Added unit test for xs:any.

Version 2.6b (10/13/2011)

  * Fix for case where a child element is declared with a type that
    is a simpleType whose restriction base is another simpleType
    that is referred to with a namespace prefix.  With this fix we
    ignore the prefix, so that at least it will work when there are
    *not* two different simpleTypes whose qualified names have the
    same local name (qualified name minus the namespace prefix).
    Thanks to Thomas Nichols for finding and reporting this one.
  * Added a unit test for the above restriction base with namespace
    prefix.
  * Added a blank character when needed at the beginning and end of
    of doc strings inserted in generated classes to protect against
    the case where the doc string begins or ends with a double quote
    character.
  * Fixes to various files in the tutorial/Code/ directory and to
    the text files in the tutorial/ directory in order to make them
    more consistent and less confusing.  Added the individual sample
    code files to the distribution so that users will not have to
    find and unzip a zipped archive.
  * Fixes to files in tests/ and to the distribution config
    (MANIFEST.in) so that the distributed version would pass unit
    tests.  (Please let me know if it does not.)
  * Removed file generatedssuper.py from the distribution.  Added
    notes to the documentation on how to create this module by
    copying from a generated module for those who want to customize
    those methods in the common superclass.
  * Fix to django/generatedssuper.py -- Regularized and fixed the
    names generated in models and forms files.
  * Fix to the code that generates the ``member_data_item_/MemberSpec_``
    list/dict.  If the type of a child element is defined by a
    reference (ref="") to an element rather than, e.g.  a
    complexType, it was using the child's name not it's type.
  * Added xs:base64Binary and xs:language to the list of string
    types in generateDS.py and django/generatedssuper.py.  Also,
    xs:anyURI and xs:duration.

Version 2.6a (07/28/2011)

  * Fix to capture xs:/xsd:/etc namespace prefix from schema.  Was
    not setting global variable XsdNameSpace.  Thanks to Frank Liauw
    for focusing my attention on this one.
  * Fix for substitutionGroup -- Was not setting the correct
    instance variable during generation of build method when child
    is a member of substitutionGroup.  Thanks to Serge Dikic for
    finding this one and bringing it to my attention.
  * One more attempt to fix whether to call the exportChildren
    method when the complexType is an extension and not a
    restriction and not defined with simple content that extends a
    simpleType.  Thanks to Jaime Cepas for alerting me to this.
  * process_includes.py has a fix to the problem where there are
    more than one anonymous complexType that define elements with
    the same name.  The issue is that generateDS.py must generate a
    Python class for each complexType and cannot do so in this case.
    One solution, which is now implemented in process_includes.py,
    is to raise each complexType to top level in the schema DOM tree
    and to give it a name.  process_includes.py does this by
    appending "Type" to the name, and when there are duplicate
    names, appending "1" or "2" or ... to that.
    So far this change passes my tests, but it does not work for
    you, then comment out the call to raise_anon_complextypes().
    Thanks to Amal Khailtash for finding a schema that exhibits this
    problem and bringing it to my attention.
  * Fix for generation of export method that exports xs:anyAttribute
    when there is an xsi:type attribute.
  * Fix for use of ``valueOf_`` -- Should only be used when element is
    defined either with (1) mixed content or (2) simpleContent.
  * Question: The xsi:type attribute is being exported for any
    derived type.  Perhaps it's harmless, but it seems excessive.
    When should the export method have that code to export the
    xsi:type attribute?  Only for types derived from (an extension
    of) an *abstract* base type?  Only for the abstract base type
    itself?  Only when a derived type is substituted for a base type
    using the base type's tag and the xsi:type attribute to specify
    the derived type?  Need to investigate.
  * I've reworked the xsi:type attribute stuff.  It now operates on
    the following assumptions: (1) an instance of any complexType
    that has been extended can have an xsi:type attribute (which
    specifies one of the extending types) and (2) the generated code
    should export the xsi:type attribute only and always when (if
    and only if) the element in the input instance document has that
    attribute.
  * A patch to convert floats and ints etc to str during export.
    Thanks Jaime Cepas.
  * Fixes to ctor/initializers when there is a default value for a
    an child element defined as a complexType containing
    simpleContent.
  * librarytemplate -- (1) Renamed documentation files to
    librarytemplate_howto.html and librarytemplate_howto.txt for
    consistency with the name of the librarytemplate distribution
    file (currently librarytemplate-1.0a.zip).  (2) Added the
    documentation and distribution files for librarytemplate to the
    main generateDS distribution file.
  * Added xs:byte to the list of integer types.

Version 2.5a (06/06/2011)

  * Fix for generation of default value in parameters for the
    constructors.
  * Fix for lookup of attribute value in generated buildAttributes
    methods -- Formerly, attribute names having a namespace prefix
    were not being found.
  * Added some support for xs:group -- Named model groups (model
    group definitions) are now treated as definitions of blocks of
    elements to be copied/inserted where referred to.  This
    replacement has been added to the preprocessing done in
    process_includes.py.  So, this <xs:group ref="some_def"/> is
    replaced by the contents of <xs:group name="some_def"> ...
  * Fix to generation of calls to validator methods for child
    elements.  Before the fix, the validators were called in
    buildAttributes methods but not in buildChildren.  Also,
    generation of the validator method (stubs) was also missing in
    some cases.  Thanks to Béres Botond for alerting me to this.
  * Fixes to generateds_gui.py -- Now it can load a session again.
    Also a fix to the check for and warnings about the changes to
    current session on exit.
  * process_includes.py -- Fix for yet another problem with
    including the same file multiple times when included from
    different directories.

Version 2.4c (03/21/2011)

  * Added minimal support for unsignedLong, unsignedInt, and
    unsignedByte.
  * Made the retrieval of the parent (superclass) name and parent
    object for an element more consistent.  Fixed some cases where
    this was not handled correctly, in particular, the generation of
    arguments and paramenters for ctors (__init__) was inconsistent
    and caused errors.
  * Regularized the handling of ``fromsubclass_`` and added this
    handling to the exportChildren methods.  This is used to tell a
    superclass, during build and export, that the subclass has
    already performed certain operations.
  * Fix to process_includes.py so as to prevent it from loading
    schemas multiple times.  The check for already_processed was
    formerly incorrect.
  * Fix related to restrictions on complexType -- Do not generate
    call buildChildren in the superclass for restrictions (as
    opposed to extensions) of a complexType.  Ditto for
    exportChildren.  Note that restrictions must repeat (and
    restrict the value of) each sub-element that is to be included
    in the content of the restriction.  See:
    http://www.w3.org/TR/2004/REC-xmlschema-0-20041028/#DerivByRestrict

Version 2.4b (02/21/2011)

  * Fix to generation the superclass in the class statement header
    line.  Formerly we did not correctly pick-up the superclass
    name (from extension base="").  Thanks to Timo Sugliani for
    finding this bug.

Version 2.4a (02/16/2011)

  * A few fixes to format of some error messages.
  * Clean-up names in the exportableClassList (__all__).
  * Modify reading session object/doc to use lxml instead of
    minidom.
  * Fix to process_includes.py to protect against crash when an
    import element is missing a schemaLocation attribute.
  * Fix to parsing and exporting simpleTypes declared as lists
    (<xs:list>).
  * Added new methods to class GeneratedsSuper to validate (during
    build) and format (during export) for simpleTypes declared as lists
    (<xs:list>).
  * Fix for incorrect detection of type during generation of build
    method.
  * Added first cut at generating Django models and forms.
    Thanks to Derek Corbett for this suggestion.
  * Added "meta-app" that generates Django database models and
    Django forms.  See doc and files in subdirectory django/.
  * Fix to generation of __all__ list: converted non-word
    characters to "_" etc
  * Fix to process_includes.py so that it uses the entire path to a
    file when trying to determine whether it duplicates a previous
    import.  Perhaps this will avoid skipping an import when
    attempting to import two files with the same name from
    different directories.  Thanks to Mihai Ibanescu for pointing
    out this fix.

Version 2.3b (12/28/2010)

  * Fix for simpleTypes defined as a restriction whose (ultimate)
    base type is a pre-defined simple type which were not generating
    correct (type-specific) code in build method.  Thanks to
    Noel Merket for finding this problem.
  * Fix for simpleTypes defined as a xs:list with "itemType"
    attribute where the type was not being recognized.
  * Fix so that we recognize some other simple types as xs:string
    type (e.g. xs:NMTOKEN, xs:ID, xs:Name).
  * To do -- If a simpleType is a restriction on another simpleType and
    the base simpleType definition is declared as a list, we are
    not recognizing that it is a list.

Version 2.3a (12/02/2010)

  * Added generatation of code to handle attributes inherited by a
    restriction from its base type and the types that the base
    extends (i.e.  from a restriction base class and its
    superclasses).  Thanks for help from Jaime Cepas.
  * Fix to code that generate the references to the superclass from
    a type that is an extension: special characters (e.g. dash)
    were not being cleaned/mapped.  Reported by Koen Smets; thanks.
  * To do:

    - In a restriction, inherited attributes can be "prohibited". 
      It would be nice if gDS would do something to block their
      use.
    - When:

          AbstractElement mixed=false and
          Element1 mixed=true base=AbstractElement and
          Element2 mixed=FALSE base=AbstractElement 

      Incorrect parse code is generated for Element2.
      Reported by Jaime Cepas.
    - It might be desirable if the getter functions could be asked
      to return values encoded to utf-8 for xs:string types.
    - Code that is generated to export to python code needs
      updating, in particular we need to update encoding of
      exported strings.  Thanks to Kristoffer Kobosko for
      reporting this.
    - Update to the code that generates code that exports Python
      literals (exportLiteral ...).  In particular: (1) fix
      excoding of Python code and of string literals (unicode, utf-8).

Version 2.2b (11/10/2010)

  * Added generation of __all__ global variable containing a list
    of generated class names.  This enables you to do a reasonably
    safe "from mymodule import \*".  It's sorted, so it also gives
    you something in the way of an alphabetical table of contents
    of the generated classes.  Thanks to Jaime Cepas for this.
  * Added another fix so that the generated code for mixed content
    elements will not generate empty blank lines on export.  Thanks
    again to Jaime for this fix.
  * Added patch to sort mixed content in their class containers.
    Jaime contributed this one too.  Thanks again.
  * Added check for endless recursion while collection list of
    parent type element names.  When detected, raises exception
    that identifies the elements.  Thanks to Maximilian Holtzberg
    for finding this one.  One case that can cause this problem is
    when an element type definition extends a type definition of
    the same name in a different namespace.  Since generateDS.py
    ignores the namespace, this looks like a type that is extending
    itself.
  * Modified code generated to process token lists in order to
    prevent breakage processing some strings.
  * Updated the tutorial so that the examples use the new parsers
    (ElementTree or lxml).
  * The "Clear" buttons in generateds_gui.py is broken when run with
    GTK2.  generateds_gui.py is still usable, but, if you need to
    erase the contents of a text field, you will have to do so
    manually until I can figure out a fix.

Version 2.2a (9/14/2010)

  * Changes for coding consistency -- Used wrt() pervasively
    instead of outfile.write().
  * Re-write of process_includes.py -- It now handles
    xs:include/xs:import elements that include a file from another
    directory that incude a file relative to that directory that
    include a file across HTTP, and so on.
  * The command line option --search-path is no longer supported.  I
    don't think that behavior was standard for XML schema anyway.
    Removed support for search_path from generateDS.py,
    process_includes.py, and generateds_gui.py.
  * Added support for specifying additional name mappings in a config
    file: generateds_config.py.  That file, if it exists, must be
    located where it can be imported by generateDS.py and should
    contain a dictionary named NameTable.  For example, the following
    maps the name "range" to "rangeType" so that if the schema defines
    a complexType "range", generateDS.py will generate a class named
    "rangeType":

        NameTable = {
            'range': 'rangeType',
            }

    See the doc for more on this.
  * Instead of using the lower() function from the string module,
    added a function to the GeneratedsSuper class and used the string
    method.  Prepares for Python 3.0
  * Added ``gds_`` prefix to all methods in class GeneratedsSuper to
    make possible name clashes less likely.
  * Fixes to exporting elements with mixed="true" -- Reduced extra
    whitespace.
  * Fixes to building (capturing) attribute values for elements with
    anyAttribute="..." -- Eliminated capture of duplicate attribute
    values.
        
Version 2.1d (8/23/2010)

  * Fix to indentation of generated code in the build method for
    type checking of NonNegativeIntegerType.
  * Fix to generation parameters in call to superclass constructor. 
    Count of children was incorrect, triggering generation of
    ``valueOf_``.
  * Known issue -- If type B extends type A, and type A declares
    anyAttribute, then duplicate attributes with the same name may
    be produced during export.
  * Known issue -- Some namespaces ("{URI}") are not converted to
    namespace prefix during export.  The needed information is not
    available during export.

Version 2.1c (8/8/2010)

  * Fix to functions parse, parseString, and parseLiteral so that
    they start the build with the correct root class.  I believe
    that there yet is another case that this does not handle,
    specifically when element name is different from class/type
    name and the element definition is not the first definition in
    the schema.
  * Fix to generation of build method for derived elements (i.e.
    elements with "extension base=".  These were being treated as
    if they were abstract, i.e. 'abstract="true"'.
  * Fix to generation of the call to the superclass constructor in
    the generated subclass module.  Prevented the generation of
    duplicate arguments.
  * Added a comment to the generated superclass module at the top
    that specifies the utf-8 source code encoding:

        # -*- coding: utf-8 -*- 

Version 2.1b (8/2/2010)

  * Fix to generation of export functions.  If no children, must
    generate "pass" statement.
  * Changed generated get_all_text function so that it uses an "if"
    statement instead of a conditional (if) expression.  The
    conditional expression does not work with older versions of
    Python.

Version 2.1a (7/8/2010)

  * Added ability to capture annotation/documentation element text
    as doc-strings in the generated classes.  Thanks to
    Roy Williams for suggesting this and for guidance.

Version 2.0b (6/24/2010)

  * Fix to generation of export method so that ``valueOf_`` is exported
    when childCount == 0 and not isMixed.

Version 2.0a (6/21/2010)

  * Switched to use of lxml/ElementTree in generated files. 
    Thanks to Biswanath Patel and Jaime Huerta Cepas for
    encouraging me to implement the switch to lxml/ElementTree.
  * Modified the generation of functions parse(), parseString(),
    and parseLiteral() so that they automatically recognize the
    root element of an instance XML document and call the build
    method of the appropriate class.
  * Fix to ``hasContent_`` method so that in elements defined
    with extension-base, the superclass is checked also.
  * For classes that must call an overridden method m in the
    superclass, switched to use "super(superclassname, self).m(...)"
    instead of "m.(self, ...)".
  * Known issues -- (1) generateDS.py loops and crashes with
    "RuntimeError: maximum recursion depth exceeded" on some
    schemas (for example collada_schema_1_4.xsd).  (2) Failure in
    process_includes.py with import of remote file and nested
    imports (for example collada_schema_1_5.xsd).

Version 1.20g (5/21/2010)

  * Update to documentation -- Added a section on suggested ways to
    handle/recognize different top level (root) elements.

Version 1.20f (5/3/2010)

  * Fix to generation of export so that anyAttribute does not cause
    duplicate attributes to be exported.
  * Fix so that we do a better job of determining whether a
    reference to a type is a simple, builtin type in generation of
    constructor.
  * Fix to generation of constructors so that (1) ``valueOf_`` is
    intialized in subclass modules and (2) ``valueOf_`` is initialized
    to None (rather than '').
  * To do: Extend the --root-element flag so that we can specify both
    the tag name and the element/type name.  Sometimes they are
    different.

Version 1.20e (2/8/2010)

  * Fixed error that caused incorrect tag name to be exported
    when the tag name contains special characters and the tag name is
    different from the type name.
  * Fixed links so that latest versions are included in the install
    distribution file.

Version 1.20d (2/3/2010)

  * Updated version number/info in genereateds_gui.py.
  * Fix to process_includes.py -- Handle include elements and
    import elements in the same way.  In particular, allow both to
    reference schema files on either the local file system or
    remotely across the Net (via ftp or http).
  * Fix to generation of properties -- When the name of a member is
    mapped (e.g. a Python language keyword), wrong name for getter
    and setter was used.
  * Fix to generation of export methods: missing encoding.
  * Fix to selection of type for exportLiteral.
  * Added missing files in the tests/ directory to the distribution.

Version 1.20c (1/1/2010)

  * Replaced symbolic links in the distribution with hard links. 
    Symbolic links do not work on MS Windows.
  * Fix to the use of the subprocess module in generateds_gui.py,
    which had caused a problem on MS Windows.
  * Cosmetic fix in generateds_gui.py: labeled "Save" (session) button.
  * Fix so that File/Open action in generateds_gui.py will check
    for and warn user if the session data has been modified.
  * Fix to generation of code for simpleContent with restriction:
    now treats the restriction element as a superclass.  Thanks
    to Franis Sirkovic for catching this and for providing the
    patch.  Also added a unit test for this case.

Version 1.20b (12/14/2009)

  * Fix to process_includes.py so that it handles relative paths
    in include/import elements in the schema.
  * Various fixes and additions to the GUI front-end, e.g. added
    "Clear" buttons to erase some fields.
  * Fixed bug -- self.inRestrictionType was not initialized.
  * Added --session command line option to generateDS.py -- It
    can now use session files generated by the GUI front-end.
  * Fixes to the generation of the exportLiteral methods.  We can
    now export Python literal representation of an instance doc
    that can be read/imported by Python.
  * Added unit test for generation of Python literal representation.
  * With the help of Erica Tolbert, generateDS.py can now generate
    bindings for gcdml (Genomic Contextual Data Markup Language. 
    See http://gensc.org).  Thank you, Erica.
  * generateDS.py can now generate bindings for the following
    (rather large) schemas:

    - gcdml -- Genomic Contextual Data Markup Language --
      See http://gensc.org
    - Collada -- 3D Asset Exchange Schema 1.5 -- See
      http://www.khronos.org/collada/
    - Vcloud --Cloud computing -- See
      http://www.vmware.com/solutions/cloud-computing/vcloud.html
    - FpML -- Financial products Markup Language -- See
      http://www.fpml.org/

Version 1.20a (12/01/2009)

  * Added first version of the GUI front-end.  See the generateDS
    doc (generateDS.html).

Version 1.19a (10/21/2009)

  * Enhancement to the table of information generated in each class
    when the --member-specs=list|dict command line option is used. 
    For a complexType defined as a simpleType, we now generate a
    list of the simpleType and the simpleTypes it is based on using
    name ``valueOf_``.  Thanks to Ryan Leslie for much help and
    guidance with these changes.
    Example::

        'valueOf_': MemberSpec_('valueOf_', [u'RelationType', 
            u'RelationType2', u'xs:string'], 0),

    Note the following incompatible changes:

      - _MemberSpec changed to ``MemberSpec_`` -- We want avoid posible
        name conflicts, not make it "weakly hidden".  See the Python
        style guide for more on this
      - _member_data_items changed to ``member_data_items_`` -- Same
        reason.
      - Method ``MemberSpec_``.get_data_type() now returns the last item
        if the types is a list and the single type if not a list.
      - Method ``MemberSpec_``.get_data_type_chain() is a new method that
        returns the entire list of data types.

    The new tutorial (see tutorial/tutorial.html in the
    distribution) has an example of the use of the MemberSpec
    feature.
  * Fix to DecimalType -- In some cases treated as an integer. 
    Should be a float.  Thanks Ryan Leslie for catching this.
  * Removed last bits of the generation of a SAX parser.  It no
    longer worked and is not needed.
  * Several fixes to determination and handling of types.
  * Added unit test for extensions to simple types and for MemberSpec.
  * There is now a preliminary version of a tutorial.

Version 1.18f (9/14/2009)

  * Fixes to process_includes.py from Mihai Ibanescu.  These fixes
    address namespace and namespace prefix problems for XML tree
    that is copied into a document.  Thanks Mihai.
  * Added xs:anySimpleType to the list of OtherSimpleTypes.  This
    prevents anySimpleType from being used as a base type.
  * Change so that sub-classes are generated for types that do not
    have children or attributes.
  * Fixed crash that occurred when a simple type is nested in a simple
    type and use of memberTypes attribute.
  * Fix to GeneratedsSuper -- Inherit from "object".
  * Added command line option --no-versions,
    which, when used, tells generateDS.py not to insert the
    version in generated files.  This is useful when you want to
    be able to compare generated files and not detect version
    differences.
  * Patch to eliminate extra space after element tag.  Thank you
    Ryan Leslie.

Version 1.18e (9/1/2009)

  * Added patch from Mihai Ibanescu which handles and expands
    groups.  Also added Mihai's unit test for groups.  Thank you,
    Mihai.
  * Added patch, also from Mihai, that passes the node's text to
    the super-class constructor.  
  * Added patch that implements a --no-dates command line flag
    which, when used, tells generateDS.py not to insert the
    time-stamp in generated files.  This is useful when you want to
    be able to compare generated files and not detect date/time
    differences.  Thanks again to Mihai.

Version 1.18d (8/26/2009)

  * Automatic detection of the namespace prefix used in the schema
    document.  Thanks to Mihai Ibanescu for this enhancement.
  * Fix to deal conflicts with generateDS's internal function
    names, for example "build".  Thanks again to Mihai.
  * Upgrade to the unit test harness.  Replace popen (which is
    deprecated) with use of the subprocess module.  Thank you Mahai.
  * Fix in the class constructors (__init__) to cast XML primitive
    types (xs:integer, xs:float, etc) to Python built-in types
    (int, float, etc).  Thanks once more to Mahai.
  * Fix to add enumeration value resolution when the possible
    values are not declared in an explicit definition but in a "top
    level" type.  Also fix a bug with enumeration value population
    for elements where the unwound element stack contains more than
    one element.  Thanks to Chris Allan for this fix.

Version 1.18c (8/11/2009)

  * Small changes related to check for mixed content.
  * Enhancement to generation of hasContent_() method to check for items
    with maxOccurs > 1.
  * Fix for generation of test for ``valueOf_`` in hasContent() method.
  * Fix for generation of initializers in ctor -- children were being
    skipped when the element is mixed.

Version 1.18b (7/29/2009)

  * Fix for exception with simpleType that is an extension of 
    another simpleType.
  * Change to mixed extension chain -- Will now generate class.
  * Fix to generation of constructors -- Will now initialize to
    default value for simpleTypes.
  * Fixed generations of validator methods, validator bodies, 
    and call to validator bodies for attributes.
  * Command line option "--validator-bodies" now triggers check for
    option value is an existing directory.
  * Various cleanup, deleting commented-out debug code, etc.
  * Now writing help messages, error messages to stderr instead of
    to stdout.

Version 1.18a (7/14/2009)

  * Added command line flag --member-specs to generate the member
    specifications as described in "User Methods" section of the doc.
    The member specs can be a list or a dictionary.
  * Fix to export indentation.  Thanks Tim Marchelli.
  * Added a utility script: generate_coverage.py which generates a
    dictionary of class names and classes from a (superclass) module
    generated by generateDS.py.

Version 1.17d (7/2/2009)

  * Fix for generation of recursively defined simpleTypes, e.g.
    a simpleType defined as a restriction of another simpleType.
    (see fix_simpletype comment in generateDS.py)
  * Added version number to generated class files.
  * Fixes to/for process_includes.py -- DirPath/DIRPATH now initialized
    correctly and fixed failure to initialize a local variable.

Version 1.17c (6/24/2009)

  * Fix for error generating code for exporting related to
    simpleType.
  * Fix for syntax error in export of boolean types.
  * Fix for export of elements with type of attribute defined
    in-line.
  * Fix to generation of export function when the --silence
    command line option is used.

Version 1.17b (6/10/2009)

  * Fix so that generateDS.py will still work with Python 2.4. 
    Thanks to Dave Sugar for that.

Version 1.17a (5/20/2009)

  * Modified export of children of type xs:string so that
    (1) if None, not exported and (2) if not None but an
    empty string, exported (example "<aa></aa>").
  * Generated calls to format_string(), format_integer(), etc
    in the generated export methods.  Enables the user to
    override these methods to customize format of exported
    values.  See the "Overridable methods" section in the
    doc (generateDS.html) for more info and for an
    explanation of how to override these methods.
    Currently used to give the user control over formatting
    of values during export.
  * Fixes to generated build and export methods so that
    elements defined as xs:simpleType as handled as
    the specificsimpleType xs:restriction base, for example
    xs:string, xs:integer, etc.

Version 1.16e (4/28/2009)

  * Eliminated generation of SAX parser.  I'm sure it no longer
    worked, anyway.
  * Fix to export of CDATA characters, provided by Kerim Mansour.
    Thanks.
  * Added support for command line option --external-encoding.
    Exported character data is now encoded to sys.getdefaultencoding()
    or to the encoding specified by command line option
    --external-encoding.
  * Added attributes (in addition to children) to the list of
    data type specifications in _MemberSpec/_member_data_items.
    This fix was provided by Ryan.
  * Several fixes suggested by Kerim Mansour including one related
    to export of CDATA.  Thank you Kerim.
  * Removed generation of SAX parser.  It did not work
    any more anyway.

Version 1.16d (3/25/2009)

  * Fixes to generation of the exportLiteral functions.  We
    can now do exportLiteral, then import the resulting file 
    in Python.  See generated parseLiteral() for an example.
  * Added an additional parameter to the export() methods.
    Now, you can call export() as follows:

        rootObj.export(outfile, 0,
            namespacedef_='xmlns:abc="http://www.abc.com/namespacelo"')

    which will insert the namespace prefix definition in the
    exported root element.
  * Added new command line option ``--namespacedef=`` to specify
    the ``namespacedef_`` to be passed in by the generated
    parse() and parseString() functions.  Example use:

        generateDS.py --namespacedef='xmlns:abc="http://www.abc.com/"'
            -o out.py myschema.xsd

Version 1.16c (3/13/2009)

  * One more fix for abstract types -- When the implementation
    element/class for an abstract class exports itself, it adds
    the xsi:type="class_name" attribute.
  * A minor fix to handling namespace prefix and the -a command
    line option.
  * Additional fixes so that in constructors (__init__), all
    instance variables are initialized to None.
  * Some fixes to quoting and escaping quotes when exporting
    attribute values.  Thanks to Kerim Mansour for help with this.

Version 1.16b (3/9/2009)

  * Added support for restriction/list, i.e. a list of words
    separated by whitespace.

Version 1.16a (2/16/2009)

  * Generated export methods now check for empty content and 
    write out <xx ... /> rather than <xx ...></xx> if empty.
  * All generated constructors (__init__()) now initialize
    instance variables to None.
  * Generated export methods now check for None before attempting
    to write out attributes and children.
  * More consistent use of direct access to instance variables
    rather than calling getter methods with a class, that is
    use of self.xxx rather than self.get_xxx().

Version 1.15d (1/22/2009)

  * Fix to setup.py so that it also installs process_includes.py.
  * Enhancements to process_includes.py so that it can also
    retrieve included files via ftp and http.
  * Fixes for default values for attributes.
  * The above changes are all from Arne Grimstrup.  Thank you Arne.

Version 1.15c (11/26/2008)

  * Added switch (--silence) to cause generateDS.py to generate
    parsing functions that do not write output to stdout.  This fix
    contributed by Al Niessner.

Version 1.15b (11/24/2008)

  * Added Amnon Janiv's fixes for attribute groups and for logging.

Version 1.15a (11/20/2008)

  * Added support for abstract elements/types.  See:
    http://www.w3.org/TR/xmlschema-0/#abstract
    Thanks to Luigi Paioro for help with this.

Version 1.14g (10/17/2008)

  * Fix in generation of exportChildren (omitted "_" in "namespace".

Version 1.14f (10/06/2008)
  * Minor fix related to simple types in generateBuildStandard_1().

Version 1.14e (09/25/2008)
  * Minor fix for generation of syntax error (missing parenthesis).
  * Eliminated generation of specification of superclass
    (superclass =) for undefined types.
  * Fixed error setting value in SimpleElementDict.
  * Fixed error when getting type for building attributes.
  * Fixed and regularized exception reporting when building float
    and integer values.
  * Fixed error referring to simple types in build function.

Version 1.14d (08/28/2008)
  * Several fixes related to simple types.

Version 1.14c (08/16/2008)
  * One more namespace patch from Andre Adrian.
  * A fix to generated export methods for valueOf from Oscar (Oeg
    Bizz).
  * First attempt to fix the name_type problem, specifically an
    incorrect generation of the element name where it should
    generate the type name and vice versa.

Version 1.14b (06/17/2008):
  * More namespace patches from Andre Adrian.
  * Changed "lower()" to "str_lower()" in generated code so that
    we have a less common name in generated code.

Version 1.14a (06/03/2008):
  * In generateBuildFn, the generated code formerly would skip the
    children of a base class in an extension class if the
    extension class has children of its own.  This patch fixes
    that problem.  (The buildChildren call for the base class is
    inside a "if hasChildren == 0" block.)
  * The export functions formerly would output the attributes and
    children of the derived classes before those of the base
    class, where the XSL spec specifies that the base class
    elements are earlier than derived elements in a sequence.  This
    patch corrects the generation order.
  * This patch adds proper xs:boolean reading and writing to
    generateDS. "true" and "false" values in the XML will become
    True and False in Python, and will be written back out as
    "true" and "false", respectively.

Version 1.13a (05/26/2008):
  * Added support for generating namespace prefix during export
    if the XML Schema specifies the targetNamespace.  Thanks to
    Andre Adrian for implementing this feature. 

Version 1.12b (05/20/2008):
  * Patches to escape special XML characters (entities) in valueOf
    and attributes.  Thanks to Darius Powell for this fix.

Version 1.12a (05/06/2008):
  * Fix to name used to generate validation method.
  * Embedded process_includes.py functionality into generateDS.py.

Version 1.11d (04/28/2008)
  * Added support for specifying default values in attributes
    and elements that are primitive types.

Version 1.11c (03/31/2008)
  * A fix in enumeration building code.

Version 1.11b (11/19/2007)
  * Fixed bug that caused an infinite loop when a class has a
    simple type as a base/super class.
  * Added additional simple types to the list of recognized simple
    types.  For a list of simple types, see:
    http://www.w3.org/TR/xmlschema-0/#SimpleTypeFacets
  * Added additional Python keywords to list of transformed names.
    See global variable NameTable.

Version 1.11a (10/11/2007)
  * Various added features contributed by Chris Allan.  For more
    information see:
    http://www.rexx.com/~dkuhlman/generateDS.html#additional-features

Version 1.10a (08/21/2007, again)
  * Added xs:int basic type.  Handle same as xs:integer.
  * Generate tests so that for elements declared with
    minOccurs="0" and maxOccurs="1" and empty value, then
    export does not generate output.

Version 1.10a (05/11/2007)
  * Added support for user methods.  See section "User Methods" in
    the documentation.

Version 1.9a (03/21/2007, again)
  * Added process_includes.py which can be used as a pre-processor
    to process include elements and create an XML Schema document
    containing all included content.
  * Modified generateDS.py so that it will read its input from a
    pipe when given the command line argument "-" (dash).

Version 1.9a (02/13/2007, again)
  * Changed naming of getter and setter methods.  Default is to
    use get_var() and set_var() instead of getVar() and setVar().
    The old behavior is available using the flag
    --use-old-getter-setter.

Version 1.9a (01/30/2007, again)
  * Fix so that validator methods for simpleType are also
    generated when the <xs:simpleType> occurs within an
    <xs:element>.

Version 1.9a (12/04/2006, again)
  * Fixed errors (occuring on import of superclass module) when 
    an element is defined as an extension of an element that is
    defined as a simpleType restriction on an xs:string.

Version 1.9a (11/27/2006, again)
  * Fix for elements that have attributes and no nested children.
    Eliminated writing out new line chars in export methods.

Version 1.9a (10/22/2006, again)
  * Fix to capture text content of nodes defined with attributes
    but with no nested elements into member varialbe ``valueOf_``.

Version 1.9a (10/10/2006)
  * Added minimal support for simpleType.
  * Generate stubs for and calls to validator methods for
    simpleType.
  * Retrieve bodies for validator methods for simpleTypes from
    files in a directory specified with the --validator-bodies
    command line flag.

Version 1.8d (10/4/2006, again)
  * Fixed several errors related to anyAttribute.  It was
    generating bad code if an element was defined with
    anyAttribute but had no other attributes.  And, in the same
    situation, it was not generating export code properly.

Version 1.8d (7/26/2006, again)
  * Allowed dot/period as special character in element tags/names.
  * Fixed several errors in generation of export and exportLiteral
    functions.  Special names (e.g. 'type', 'class') were not
    being mapped to special spellings (e.g. 'ttype', 'klass', ).
  * Fixed error in determining ExplicitDefine, which was
    preventing export of some objects.

Version 1.8d (7/19/2006, again)
  * Added support for empty elements, i.e. elements that have no
    children and no attributes.  Formerly, they were ignored due
    to a quirk in logic.

Version 1.8d (4/13/2006) 
  * Added support for the following simple types: duration, anyURI
    and unsignedShort.  They are coerced to (and treated the same
    as) xs:string, xs:string, and xs:integer, respectively

Version 1.8c (12/22/2005, again)
  * Fixed use of mapped names in generateExportLiteralFn().

Version 1.8c (12/20/2005, again)
  * Fix to generation of getters and setters for attributes.
    Formerly generating accessors that handled *lists* of attribute
    values.

Version 1.8c (12/15/2005, again)
  * Fix generated code so that it uses documentElement instead of
    childNodes[0] to get the root element.

Version 1.8c (5/10/2005, again)
  * Patch for <xs:attribute ref="xxxx"/> -- Use the value of ref
    as the name of the attribute.  I'm not sure whether this is
    correct in all situations.
  * Fix for generation of ctor for mixed type elements.  Before
    this fix, generateDS.py was failing to generate the
    initializers in the __init__ method signature.
  * Fix for generation of "class" declaration for extension
    classes whose base class name is qualified with a namespace
    (e.g., <xs:extension base="iodef:TextAbstractType">).  Removed
    the namespace.  This fix also corrected the order of
    generation of classes so that the base class is now correctly
    generated *before* the subclass.

Version 1.8c (4/26/2005)
  * Added support for several simple types: xs:token, xs:short,
    xs:long, xs:positiveInteger, xs:negativeInteger,
    xs:long, xs:nonPositiveInteger, xs:nonNegativeInteger,
    xs:date.
  * Fixed error produced when an element definition inherits from
    a simple type.

Version 1.8b (2/25/2005)
  * Added support for anyAttribute.

Version 1.8a (2/23/05, again)
  * Fixed incorrect generation of name and type for export
    functions for root element.
  * Fixed reference to root element type when root element name
    and type are different.

Version 1.8a (1/13/05, again)
  * Fixed incorrect handling of extension of in-line element
    definition.
  * Code cleanup in support of the above.

Version 1.8a (12/22/04)
  * Added support for attributeGroup.  Enables an XML Schema to
    define attribute groups and then include them in
    element/complexType definition.
  * Added support for substitutionGroup.  Enables use any of a set
    of element types as alternatives to another element type.
    Limitation: Does not work with simple element types.

Version 1.7b (11/15/04)
  * From an XML Schema, it is not possible to determine the
    outer-most element in instance documents.  generateDS.py now
    generates a parser (parseSelect) that first uses a small SAX
    parser to determine the outer-most element in the input
    document, then invokes the normal parser with that element as
    the root.

Version 1.7a (10/28/04)

  * Thanks very much to Lloyd Kvam for help with these fixes and
    improvements.  His ideas, suggestions, and work have been
    extremely valuable.
  * Implementd partial support for <xsd:extension base="">.
    Limitation: extension elements cannot override members
    defined in a base.
  * Refactored generated methods export and build, so that they
    can be called by subclasses.
  * The generated method exportLiteral has been left behind during
    recent work.  Brought it up-to-date.
  * For Python, a super-class must be defined before the
    sub-classes that reference it.  Implemented a delaying
    mechanism that enforces this ordering of generation of classes.
  * Fixed a bug that occurred when an element is defined with
    maxOccurs given a value other than "1" or "unbounded".

Version 1.6d (10/1/04)

  * Several bug fixes.
  * Added command-line flag --subclass-suffix="X".  Changes the
    suffix appended to the class name in subclass files.  Default
    if omitted is "Sub".
  * Added an underscore to certain local variables to avoid name
    conflicts.
  * Thanks to Lloyd Kvam for help with this release.  Lloyd found
    and fixed a number of these problems.
  * Added command-line flag "--subclass-suffix", which specifies
    the suffix to be added to class names in subclass files.
    Default is "Sub".
  * Added command-line flag "--root-element", which makes a
    specified element name the assumed root of instance documents
  * In some schemas, attributes on a nested <complexType> pertain
    to the containing <element>.  Added code to copy the
    attributes from the <complexType> to the <element>, if it is
    nested.

Version 1.6c (9/15/04)

  * generateDS.py was not walking lower levels of the tree data
    structure collected by the SAX parser that describes the
    classes to be generated.  Now, function generate() calls
    function generateFromTree() to recursively walk lower levels
    of this tree structure.
  * Fixed various errors that were introduced or uncovered by the
    above change.
  * Strengthen handling of mixed content.  When an element
    definition (<element> or <complexType>) contains the attribute
    "mixed=" with a true value, then we generate the code for text
    content, e.g. getValue(), setValue(), capture value in
    build(), etc.

Version 1.6b (9/10/04, yet again)

  * Still fixing bug related to generating all the sub-class
    stubs.  All sub-classes were not being generated when no
    superclasses were generated (-o flag omitted), because there
    are data structures that are created when superclasses are
    generated and which are needed during sub-class generation.
    Now we *always* write out super-classes, but write them to a
    temp file if they are not requested.

Version 1.6b (8/26/04, again)

  * Fixed bug -- complexTypes defined in-line were omitted from the
    sub-class file.  Now these sub-classes are being generated.

Version 1.6b (8/18/04)

  * Added ability to access the text content of elements that are
    defined but have *no* nested elements.  The member variable is
    ``valueOf_`` (note underscore which will hopefully avoid name
    conflicts) and the getter and setter methods are ``getValueOf_``
    and ``setValueOf_``.
  * Fixes to generation of exportLiteral methods.  Formerly,
    export of attributes was omitted.
  * Removed un-used function that contained "yield" statement,
    which caused problems with older versions of Python.

Version 1.6a (7/23/04, again)
  * Added optional generation of new style classes with
    properties.  This is experimental and, admittedly, not very
    useful, as the property functions are simple getters and
    setters.  Maybe someday ...  Use the "-m" flag to see the
    resulting code.

Version 1.6a (7/9/04, again)
  * Minor fixes.  Replaced dashes in names used as attributes (see
    cleanupName().

Version 1.6a (7/6/04, again)
  * For XMLBehaviors, implemented ability to retrieve
    implementation bodies for behaviors and for ancillaries
    (pre-conditions and post-conditions) from a Web address (URL).

Version 1.6a (6/30/04)
  * Added generation of behaviors.  An XML document can be used to
    specify behaviors (methods) to be added to generated sub-class
    files, including DBC (design by contract) pre- and
    post-condition tests.  See generateDS.html for more
    information on XMLBehaviors.

Version 1.5b (6/20/04, again)
  * Fixed handling namespace prefix in the XMLSchema file itself.
    generateDS.py now attempts to pick-up the namespace prefix
    (alias) from the "xmlns:" attribute on the "schema" element.

Version 1.5b (5/7/04)
  * Fixed several minor problems related to XML namespaces.
    Namespace prefix ignored when creating Python names (e.g. of
    classes and namespace prefix ignored during parsing.  That's
    about the best I know to do right now.
  * Fixed problems in generating code for names containing dashes.
    Now using underscore in place of dashes for Python names.

Version 1.5a (3/24/04)
  * Added keyword arguments to the generated factory functions.
  * Added generation of method "exportLiteral" and related support
    to export elements/instances to Python data structure
    literals.

Version 1.4c (3/10/04)
  * Element <complexType> in XSchema file not handled correctly.
    Fixed this so that when <complexType> is at top level, it will
    be handled the same way that an <element> is handled.  Note:
    We still have problems with <complexType> elements that are
    more deeply nested.

Version 1.4c (3/8/04)
  * Added ability to pass namespace abbreviation from the command
    line.  For example, the "-a" option enables you to replace
    "xs:" with "xsd:".

Version 1.4b (9/30/03, again)
  * Removed dependence on PyXML.  Will now import XML support from
    PyXML, if it is available, but if not, imports from the
    Python standard library.

Version 1.4b (9/30/03)
  * Fixed name conflict in factory function (added underscore).
  * Added generation of saxParseString function (parse string, not
    file/URL).
  * Fixed error -- ome constructors not using factory.

Version 1.4a (9/17/03)
  * Added generation of a SAX parser.

Version 1.3c (9/11/03)
  * Fixed problem caused by shared content model, i.e. when a
    field (content) is declared with a complex type and the name
    and the type are different.  The fix enabled the field name
    and the type of the object in that field to be different.

Version 1.3b (9/9/03)
  * Fixed error when a separate xs:element declaration is used for
    elements declared with a simple type.

Version 1.3a (8/18/03)
  * Removed YAML support.
  * Fixed error in name generation in generateBuildFn().
  * Various fixes and cleanup in tests/ and Demo/.

Version 1.2a (again, 5/16/03)
  * Fixed error in code generation for boolean attributes.
  * Fixed error in code generation for float values.
  * Added very simple unit tests in tests directory.  Can be run
    with:

        cd tests
        python test.py

Version 1.2a (3/14/03)
  * Added support for XML Schema xs:double and xs:boolean types.

Version 1.1a (8/13/02)
  * Added ability to generate subclass stubs for user method
    implementation.
  * A bit of clean-up to the command line options.

Version 1.0a (3/15/02)
  * Initial release


-----
To do
-----

The following enhancements and fixes remain to be done:

- The <sequence> element can have "minOccurs" and "maxOccurs"
  attributes.  I'm guessing, but am not sure that this specifies
  repeated groups.  For example, the following::

      <xs:sequence minOccurs="0" maxOccurs="unbounded">
          <xs:element name="description" type="xs:string"/>
          <xs:element name="size" type="xs:integer"/>
      </xs:sequence>

  specifies that we can have any number of pairs of elements
  "description" and "size".  A future enhancement to generateDS.py
  would enable us to specify and enforce this restriction.

- And so many more complexities in the XSchema specifications.


Dave Kuhlman
dkuhlman@davekuhlman.org
http://www.davekuhlman.org

.. A few dummy targets to eliminate docutils processing errors:

.. _namespace:
.. _namespaceprefix:
