=== version history of XML::Compile

Unless noted otherwise, these changes where initiated and applied by
Mark Overmeer.

To be resolved:
 - length facet on base64/hex encoding about binary length
 - handle XML::LibXML::Error objects
 - useSchema() relation to ::Cache::declare()?
 - proposed new schema features:
     . alternative
       http://www.ibm.com/developerworks/library/x-xml11pt2/#N103D7
     . ::Instance cannot handle element and attribute declaration attribute
       "targetNamespace" (proposed schema feature)
 - namespace blocking option for templates; exclude deprecated elements

version 1.12: Mon Jan 11 11:46:43 CET 2010

	Fixes:

	- writer: optional block was not optional [Roman Daniel]

	- accept xsi:type with alternative in namespace which is not
	  used by the schema itself [Roman Daniel]

version 1.11: Fri Jan  8 12:37:23 CET 2010

	Fixes:

	- forgot to document the new xsi_type option to ::Schema::compile()

	- implement facets on lists correctly. Tests in new t/55facet_list.t
	  Pushed by rt.cpan.org #53392 [Николай Шуляковский]

version 1.10: Thu Dec 24 16:56:13 CET 2009

	Changes:

	- keyRewrite now also applies to attributes.

	Fixes:

	- writer sometimes destroyed the data-structure which was
	  passed in. Now automatically tested for each test_rw.

	- support sub-seconds in time format.

	- writer: do not produce an empty element when minOccurs is
	  combined with nillable in a complexType.
	  rt.cpan.org#51264 [Roman Daniel]

	- ::Iterator::currentType() could crash.

	Improvements:

	- writer: major readibility changes.

	- all: implemented xsi:type

	- template: do not produce superfluous quotes in Perl template
	  examples when anonymous HASH of ARRAY is included.

version 1.09: Fri Nov 13 10:56:45 CET 2009

	Fixes:

	- writer: do not produce an empty element when minOccurs is
	  combined with nillable in a simpleType
	  rt.cpan.org#51264 [Roman Daniel]

	Improvements:

	- template: use prefixes to avoid namespaces in types.

version 1.08: Sat Oct 24 20:11:10 CEST 2009

	Fixes:

	- reader: fix last parameter of replacement hook

	Improvements:

	- template: support for key_rewrite

	- use the new XML::LibXML::RegExp, to replace tricky XML::RegExp.

version 1.07: Mon Jun 22 11:35:22 CEST 2009

	Fixes:

	- mixed attribute on complexContent not detected.

	- anyType elements which have sub-elements will not stringify, but only
	  simple strings will get simplified.  Otherwise, an XML::LibXML::Node
	  is kept.

	Improvements:

	- reader: anyType element processing can be hooked via a new
	  compile(any_type) option.

	- partial support for attribute targetNamespace on element
	  and attribute declarations.  Currently produces a warning.

version 1.06: Thu May 28 10:28:45 CEST 2009

	Fixes:

	- prefer simpleType child over base type attribute in
	  simpleContent/restriction.  rt.cpan.org#46212 [Erich Weigand]

	Improvements:

	- ignore all references to an (unloaded) schema, for instance
	  with deprecated structures via ::compile(block_namespace) with
	  tests in t/76blocked.t

	- fixed tests in t/91noqual.t

	- include t/75type.t in the distribution.

	- added test for base64Binary to t/21types.t

	- reader and writer: replace hook also gets code-ref able to
	  process what is being replaced, implementing a wrapper.

version 1.05: di apr 28 13:04:42 CEST 2009

	Improvements:

	- writer: accept single element for tagged and mixed.

	- writer: accept scalar for mixed, upgraded to text node.

version 1.04: Fri Apr 24 16:08:43 CEST 2009

	Fixes:

	- the use of cached schema elements must be limited to single
	  ::Schema objects (for instance per WSDL object).
	  rt.cpan.org#44959 [Jozef Kutej]

	- complex extension of anyType types.

	Improvements:

	- document that key_rewrite only applies to elements, not
	  attributes.

	- add 'target_namespace' overrule option to importDefinitions()
	  and ::Schema::addSchemas()

	- have ::Schema::new() pass options to importDefinitions()

	- added various entries about fixing-up schema's to the FAQ.

version 1.03: Wed Mar 25 15:41:43 CET 2009

	Fixes:

	- template: protect output against combinatorial explosion, which
	  did hurt [Anton Berezin]

	Improvements:

	- new option ::Schema::compile(abstract_types)

	- template: compact the output of substitutionGroups, leaving the
	  abstract types out and use columns in the display.

	- template: add namespace details to XML and PERL output

	- template: add header to XML and PERL output.
	  New option  ::Schema::template(skip_header)

version 1.02: Thu Feb 12 11:37:49 CET 2009

	Fixes:

	- fix useSchema() deep recursion

	- writer: repair ARRAY complaint when scalar given on place
	  where an array of values is acceptable.  [Allan Wind]

version 1.01: Thu Feb 12 09:39:50 CET 2009

	Fixes:

	- do not use /bin/pwd in t/pod.t

	- writer: report warnings when accepting a difficult construction.

	- key_rewrite(PREFIXED) options now do stack, without the
	  possibility to get it rewritten twice.

	- reader: complain if no data was recognized. [Jozef Kutej]

	- handle schema's with targetNamespace but not qualified
	  elements.  http://www.w3.org/TR/xmlschema-0/#UnqualLocals
	  [Jozef Kutej]

	Improvements:

	- ::Schema::useSchema() let you share definitions between
	  various ::Schema extensions.

	- document that XML::Compile::dataToXML() can be used as
	  instance method as well.

version 1.00: Wed Jan 21 10:51:23 CET 2009

	Changes:

	- oops, rename elementFormDefault parameters into
	  element_form_default.  Breaks only XML::Compile::SOAP 2.00_01

	- substitutionGroup keys did not get prefixed, when key_rewrite
	  was enabled.

	Fixes:

	- writer: name-space qualified for NIL [Mark Blackman]

	- template: be more careful not to put trailing '}' on a line
	  with comment.

	- writer: contents of tagged elements were not processed, so
	  not checked or whitespace corrected.

	- report rewritten key when required value is missing, not the
	  full key.

	- compile(mixed_elements=STRUCTURAL) was ignored when the mixed
	  option was found with the complexContent.

	Improvements:

	- template: hook "COLLAPSE" to reduce output size for well-known
	  types.

	- writer: accept {_ => $value} as alternative for $value.

	- writer: optimize seq block with one block element.

	- added XML::Compile::Util::even_elements()

	- writer: show available tags in complex block with unused tags
	  when in debug mode.

	- show substitutionGroup members in debug mode.
	  Requires Log::Report v0.21

	- make addHooks() use addHook(), easier to override.

	- reader: elements with mixed="true" by default only unpacks the
	  ATTRIBUTES.  It is hard to figure-out why the node is not
	  translated into Perl.  Therefore, the resulting HASH now
	  contains a _MIXED_ELEMENT_MODE key as explanation.

	- with "Log::Report mode => 'DEBUG'", you may get extra help
	  on some errors.

version 0.99: Mon Dec 29 10:16:28 CET 2008

	Changes:

	- reset "used count" when include_namespaces

	- Qname will not complain about "unused namespace", because
	  that is often not true.  If true, than you have to provide
	  a the compile(prefixes) as HASH with used flag set.

	Fixes:

	- elements and attributes with non-default "form" attribute
	  were not recognized.

	- type-based replace hook selected too often.

	Improvements:

	- compile(include_namespaces) can have a code reference to
	  filter included namespace declarations to be included.

	- addSchemas/importDefinitions now have options
	  {element,attribute}FormDefault to overrule the (missing) info
	  in many (old) schemas (even those of soap and wsdl).

	- dataToXML() also finds filenames in the SCHEMADIR, not only
	  known namespaces.  That was always intended to happen.

	- optimize reading of complex with zero or one component.

	- new reader hook after => 'NODE_TYPE'

	- error when schema definition contains components of other
	  namespaces.

	- you now can compile types only, where the schema does not
	  contain the definition of the element which will be produced.
	  But I will not tell you how until the interface has settled ;-)

version 0.98: Tue Dec 16 08:32:04 CET 2008

	Fixes:

	- writer failed for choice with an element containing a false
	  value, like "0" or ''.  Reported by [Sander Hulst]

	- do not attempt to add namespace declarations to produced
	  attributes in the writer.

	- elements with minOccurs=0/maxOccurs=unbounded and optional
	  content failed.  rt.cpan.org#41725 [Osfameron]

	Improvements:

	- easier to understand error messages in block writer.

	- started XML::Compile::FAQ

version 0.97: Mon Nov 24 15:51:54 CET 2008

	Fixes:

	- dataToXML() with document node will return document element.

	- dateTime reversed minutes/seconds.  Fixed by [Allan Wind]

	- dateTime timezone part is optional.  Fixed in docs by [Allan Wind]

	- count error with array of elements writer, when maxOccurs is
	  an integer value larger than 1.  An error will be produced when
	  too many elements are provided.

	Improvements:

	- bigfloat support for xml2yaml [Slaven Rezic]

version 0.96: Fri Oct 10 16:02:09 CEST 2008

	Fixes:

	- warning condition on old libxml2 version was broken: warned too
	  often.  rt.cpan.org#39807 [Joerg Plate]

	- overruling "qualified" on top-level elements did not work anymore.

	- writer hook "replace" crashed on missing (optional) data.

version 0.95: Wed Oct  1 17:38:07 CEST 2008

	Fixes:

	- top-level elements could not be have as nillable, default, and
	  fixed.  rt.cpan.org#39215 [Tom Hukins]

	- do pattern match on unmodified float values.
	  rt.cpan.org#39224 [MIROD]

	- automatically defined xsi when any element is nillable.  You will
	  still need to specify "include_namespaces => 1" to have the
	  compiler add the name-space definition to the root namespace.
	  For the moment only supports SCHEMA2001

	- the applications of facets is explicitly ordered, to avoid
	  colissions between checks in lexical space and conversions
	  made in value-space.

	- accept but ignore minScale and maxScale facets (cannot be
	  represented in Perl objects yet)

	- removed facet totalFracDigits, which does not exist.  How did it
	  get in?

	- precissionDecimal was handled as int, should be float.

	- positiveInteger value check broken. [Sander Hulst]

	- accept the 'error' base-type.

	Improvements:

	- small (<1e9) float/decimal values will not become Math::Big*

	- facet lists and non-lists various optimizations in reader.

	- pre-lexical facet whiteSpace on list does not need to be checked
	  in writer.

	- handling of (+|-|)INF and NaN for precissionDecimals, floats and
	  doubles.

	- adapt names of test helpers to new X::C::Tester standard.  v0.03
	  of that module is now required.

version 0.94: Tue Aug 26 10:09:05 CEST 2008

	Fixes:

	- repair whitespace facets collapse and replace. [Allan Wind]

	- NMTOKEN has no blanks, but token may have them.
	  rt.cpan.org#38675 [Allan Wind]

	Improvements:

	- show more useful features of Log::Report in ::Schema SYNOPSIS

version 0.93: Mon Aug 11 21:13:22 CEST 2008

	Changes:

	- changed order of attributes back to situation of 0.90: the
	  base-class attributes are probably more important than the
	  attributes of the extension.

	- groups with maxOccurs > 1 renamed
	    from   gr_<label of first element>
	    into   gr_<name of group>
	  For example, when parsing schemas themselves (with PREFIXED)
	    was    gr_xs_minExclusive
	    now    gr_xs_facets
	  minExclusive is the name of the first element in the group
	    was    gr_xs_minExclusive
	    now    gr_xs_facets
	  minExclusive is the name of the first element in the group
	  block. xs:facets is the name of the group as used in the
	  schema schema and probably more consistent over releases.

	Fixes:

	- documentation on use of typemaps and hooks got mutulated
	  in 0.90

version 0.92: Wed Aug  6 11:33:27 CEST 2008

	Fixes:

	- minOccurs=0 in writer.  Reported by [Drew Taylor]

	Improvements:

	- WRITER "after" hook also gets the data, not only the produced XML.

	- extra tests for extension of empty complexType.

version 0.91: Fri Aug  1 13:25:26 CEST 2008

	Changes:

	- removed block_label from ::Util because it does not work with
	  key rewrite.  Probably not used by anyone.

	- "any" without maxOccurs>1 did still return an ARRAY,
	  which is inconsequent.  Needed to be fixed.

	- the output order of attributes has changed: the extended additions
	  before the base class attributes.  This does not change the meaning,
	  but may be painful for your regression tests.

	- writer adds prefixes always in the same order (alphabetic) to
	  ease regression-tests.

	Fixes:

	- Do not confuse un-used keys as 'any' elements in WRITER

	- key rewrite PREFIXED did not work with blocks.

	- QNAME in name-space-less schema's can go without prefix.

	- Fixed QName in list.  Tests in new t/74qname.t

	- undouble attributes when extending types.

	- ::Schema::compile(use_default_prefix) renamed into
	  use_default_namespace, which was already used in the
	  ::Translate docs.  Both names will work.

	- WRITER production errors did not arrive correctly to the
	  application.  Reported by [Gert Doering]

	- particle block handling in the WRITER was not as sophisticated
	  as in the READER, which cause problems reported by [Gert Doering]

	- groups which ref'd to an other name-spaces did not use the new
	  targetNamespace.

	- READER when choice/sequence/all with minOccurs=0 in the middle
	  of a list of elements failed if the data was not present.

	Improvements:

	- List existence of XML::Rewrite.

	- Collect prefix-table when reading.

	- better error when an any_(element|attribute) is not understood.

	- the writer will attempt to use the same prefixes as used in
	  the schema, unless these name-spaces are already pre-defined
	  in the 'prefixes' table.

	- be more careful that empty text fragments do not end-up as
	  text nodes in the WRITER output.

	- some error message did not show the path in the message where
	  the error was produced.

version 0.90: Fri Jul 18 21:52:52 CEST 2008

	Changes:

	- It was permitted to use an ARRAY with one element where a
	  single elements was needed in the WRITER.  This has been
	  removed because it did not work together with simpleType
	  lists.

	- Reworked the three back-ends from functional into clean OO.
	  Module renames:
	    ::Schema::Translate -> ::Translate
	    ::Schema::XmlReader -> ::Translate::Reader
	    ::Schema::XmlWriter -> ::Translate::Writer
	    ::Schema::Template  -> ::Translate::Template 
	  All functions renamed into methods from
	    element_wrapper  into  makeElementWrapper

	Improvements:

	- you can now register new translator back-ends.

version 0.89: Sat Jul 12 12:40:57 CEST 2008

	Fixes:

	- pattern restrictions on lists must be applied "late",
	  after the split.

	- repair facets on lists.

	- complexType/complexContent/restriction will only redefine the
	  elements, but will still take the attributes from the base-class.
	  Tests in new t/45ctcres.t

	- disabled a broken pattern optimization.

	Changes:

	- ::Schema::compile() options anyElement and anyAttribute renamed
	  to the consequent any_element and any_attribute.  The old names
	  are still supported.

	Improvements:

	- xml2yaml output filename '-' for STDOUT

	- xml2yaml exit when illegal options were detected

	- mixed_elements option STRUCTURAL now also works for writers.

version 0.88: Tue Jul  8 10:38:26 CEST 2008

	Fixes:

	- SubstitutionGroups do not require to be based on abstract
	  elements.

	- include name-space while collecting nodes to read, not only
	  local-names.	Now nodes with the same local-name but different
	  name-space can be used in one block or substituteGroup.

	- make rewrite work for substitutionGroups

	Improvements:

	- error when abstract elements are being used, tests
	  in t/34abstract.t

	- more speed at subst-group lookups.

	- template shows empty blocks explicitly.

version 0.87: Fri Jul  4 15:12:17 CEST 2008

	Changes:

	- removed /el() and #el(name) from location of error message,
	  which makes the location simpler and probably not less clear.

	Improvements:

	- implemented key_rewrite, tests in t/73rewrite.t

	- removed double check for existence of required value in writer.

	- renamed ::Schema::new() option 'output_namespaces' into
	  the nicer 'prefixes'.  Old name still available.

version 0.86: Mon Jun 23 17:01:43 CEST 2008

	Fixes:

	- fixed attributes are always optional, both in reader and
	  writer.

	- fixed elements are optional, but do return that value as
	  default in the reader.

	- class labels to group error messages should be spelled
	  _class in ::Translate.

	- fix choice with multiple substitutionGroups or any's.

	Changes:

	- mixed element default will always return a HASH, not a node
	  when there are no attributes defined.

	Improvements:

	- many more options to handle mixed elements.  See
	  ::Schema::compile option mixed_elements.

	- error for typemaps which specify a none-existing type, as
	  coding help.

	- error when both default and fixed are set.

version 0.85: Mon Jun 16 15:07:23 CEST 2008

	Fixes:

	- template with multiple name-spaces crashed on the handling of
	  prefixes.

	- permit empty substitutionGroups.

	Improvements:

	- template will show facets, when only simple enumerate or pattern
	  restrictions are used.

	- switch network access explicitly off for default parser, to
	  avoid unexpected network access.

	- fixes in manual-page footer.

	- reordering methods in documentation

	- few additional trace statements

version 0.84: Fri Jun  6 12:50:21 CEST 2008

	Fixes:

	- element and attribute qualification was incorrect when
	   - a complex was extended, based on a type in an other name-space
	   - substitutionGroup elements from different name-spaces
	     were combined

	- xs:pattern was missing

	- output of any and anyAttribute in example templates.

	- capability to write an empty list

	- accept default for list.  Reported by [Peter Hartmann].

	- interpret empty content as missing for elements with
	  a default, as the spec wants.

	Changes:

	- you cannot use an ARRAY with one element where one element
	  is required in the WRITER. (conflicts with list type needs)

	Improvements:

	- better template examples for ID and family.

	- improve documentation on how 'boolean' is handled.  Spotted
	  by [Allan Wind]

	- list XML::Compile::Cache as available module.

	- printIndex() got filter option 'kind'

	- simplified the prefix generation algorithm.

version 0.83: Wed May 14 22:22:41 CEST 2008

	Fixes:

	- builtin optimization needed fix.  cpan-tester [Slaven Rezic]

version 0.82: Tue May 13 21:04:02 CEST 2008

	Fixes:

	- qname should collapse whitespace.

	- qname prefix is blank by default, not an error. [Daniel Ruoso]

	- gYear has no month fragment, but has an optional timezone
	  rt.cpan.org#35699 [Daniel Carrol]

	- recursion detected where there wasn't. [Sander Hulst]

	Changes:

	- template output in Perl will not start with "type =>", so can
	  directly be used as test message.  Requested by [Alan Wind]

	- template output in Perl will not quote the objects, where
	  "mixed" elements need to be inserted.

	Improvements:

	- minor fix to template for substgroup

	- warn on useless type assertions (uncheckable types), not panic.

	- start of testing calling convensions (to be more flexible in
	  the future) in t/22call.t

	- implemented hooks for templates.

	- implemented typemaps in PERL example templates, not for XML
	  templates (yet).

	- weaken circular reference in Instance.

version 0.81: Tue Apr 29 18:57:48 CEST 2008

	Fixes:

	- pattern match facet did not anchor [Sander Hulst]

	- ::Translate::complextContentExtension with mixed produced
	  error message. [Ari Jolma]

	- re-use already compiled structures, which avoids combinatory
	  explosions.

	Changes:

	- requires XML::Compile::Tester

	Improvements:

	- add ::Schema::addTypemap() as synonym for ::addTypemaps()

	- extracted useful test-script tools from t/TestTools.pm
	  into new distribution XML::Compile::Tester.

	- converted test scripts to use XML::Compile::Tester.

	- ::Schema::compile(ignore_unused_tags) now accepts a regular
	  expression to indicate exactly which tags are to be ignored.
	  Added global ::new(ignore_unused_tags)

	- warn if people call the writer without document as first parameter.

version 0.80: Mon Apr 21 10:34:56 CEST 2008

	Improvements:

	- refer to mailinglist and IRC channal, set-up by [Matt S Trout]

	- implementation of typemaps in READER and WRITER, tests in
	  t/72typemap.t

	- refer to related distributions in PODTAIL.

version 0.79: Fri Apr 18 16:01:57 CEST 2008

	Fixes:

	- much better "mixed" support.

	Changes:

	- require Log::Report 0.17 to fulfil promissed of the doc (was 0.09)

	Improvements:

	- trace compiles of schema elements.

	- when debug is enabled, then <choice> will show the errors of the
	  special blocks involved.

	- optimize <sequence>, <choice>, and <all> with only one element
	  in reader and writer.

	- template Perl did quote keys with length 1.

	- template warning resolved in example value for nonNegativeInt

	- template Perl did not show '_' for a tagged element, and
	  did show an example twice.

	- renamed ::template(show) to clearer ::template(show_comments),
	  but old flag will stay supported.

	- template XML did not represent attributes correctly.

	- template XML did not understand '_'

version 0.78: Wed Apr 16 13:03:36 CEST 2008

	Fixes:

	- fix string length check in float facet [Sven Neuhaus]

	Improvements:

	- more trace with dataToXML.

version 0.77: Sun Apr 13 00:45:28 CEST 2008

	Fixes:

	- XML::RegExp version mix-up... lowered required release
	  from 0.03 to 0.02 [Alan Wind]

version 0.75: Sat Apr 12 09:57:29 CEST 2008

	Fixes:

	- fixed <choice> with nested blocks, substgroups, or any.

	- fixed <all> with nested blocks, substgroups, or any.
	  needed by [Sander Hulst]  Needs serious beating.

	- required is XML::RegExp for pattern to regexp translations.

	- NMTOKENS, IDREFS, and ENTITIES where not processed as list
	  type (not spilt into an ARRAY)

	Improvements:

	- remove annoying extra component in the location shown in
	  errors, which appeared with element refs.

version 0.74: Thu Apr 10 23:17:18 CEST 2008

	Improvements:

	- removed "try => GMP" from Math::BigInt invocation, because
	  Linux distributions currently have v1.77 which does not support
	  try.  Documented how you can get the performance improvements
	  yourself.  rt.cpan.org #34817 [Jesper Krogh]

	- lowered requiremet for Math::BigInt to 1.77.

version 0.73: Tue Apr  8 23:47:05 CEST 2008

	Fixes:

	- recursion detection is example template production.

	- example template fix: add ',' after some ']'.

	- reader in list context should not return the label of the
	  element.  rt.cpan.org#34597 [Andreas Koenig]

	- removed test print when bigfloat was used.

	- nesting of substitutionGroups supported.
	  Requested by [Sander Hulst]

	Improvements:

	- reuse schema parsing results, when the same file or string is
	  passed multiple times.  Adds dependency to Digest::MD5.

	- added script schema2example

	- pass an ARRAY of schema sources at once to ::Schema::new()
	  and ::importDefinitions.

	- Math::BigInt warns us to upgrade Math::BigRat and bignum, so
	  included that as dependencies.

	- ::Namespaces::add() now accept multiple schema instances at
	  a time.

	- explain how XML::LibXML::Simple can be used in a READER
	  replace hook.

	- adding debug info about the parsing of XML information via
	  dataToXML().

        - timezoneFrag was not wrong (as stated by the 0.72 release).
	  Reverted the changes for simpler code.

	- more examples in ::Schema SYNOPSIS

version 0.72: Tue Mar 25 10:22:39 CET 2008

	Fixes:

	- [Drew Taylor] experienced that 'time' basic type was not
	  implemented.

	- ::Template::anyElement fix [Drew Taylor]

        - timezoneFrag is optional, wrongly implemented in many regexes.

	- ::Template::union incorrectly processed [Drew Taylor]

	Improvements:

	- document that XML::LibXML nodes can be specified on any location
	  for WRITERs. (feature was added in 0.70)

	- document how to handle mixed content.

version 0.71: Tue Mar 18 16:52:34 CET 2008

	Fixes:

	- require Math::BigInt 1.78, to understand 'try'

	- cannot run t/59big tests with BigInt GMP library.

version 0.70: Fri Mar 14 17:35:22 CET 2008

	Fixes:

	- document that ::Schema::compile(check_occurs) default value
	  is true, not false.

	- maxOccurs==0 now works.  Some tools use it, but it is silly.

	- more name-space switches.

	- missed an element component in simpleType facet related error
	  messages.  Spotted by [Allan Wind]

	- die when illegal float value [Sander Hulst]

	- template example for empty complexType incorrect.

	- template example for repeated blocks incorrect.

	- repeated particle blocks within particle blocks caused
	  problems. rt.cpan.org#34079 [Drew Taylor]

	Improvements:

	- minor typos in man-pages [Allan Wind]

	- ::Translate::compileTree() can start with a type.

	- path displayed in error-messages now use @ before attributes,
	  not at()

	- more docs on how substitutionGroups are represented.

	- new option ::Schema::compile(validation), which covers the
	  existing three flags for partial validations.

	- option check_values was not documented in ::Translate

	- template Perl did not use Data::Dumper; don't invoke it.

	- protect difficult characters in perl template keys.

	- The writer now accepts a prepared XML element, not only a HASH
	  or single value, for any simple or complex type element.

	- Math::BigInt uses Math::BigInt::GMP for more performance, if
	  installed.

version 0.69: Tue Feb 12 09:59:30 CET 2008

	Fixes:

	- name-space switch accidentally took 'qualified' on elements
	  to be used for attributes. Detected by [Sander Hulst]

version 0.68: Fri Feb  8 09:32:20 CET 2008

	Fixes:

	- ref element did not switch to settings of destination namespace.
	  Reported by [Sander Hulst]

	- ref attribute idem.

	- ref attributes should lookup namespace in destination schema,
	  not source schema.

	Improvements:

	- added ::Namespaces::printIndex() to list defined toplevel items.

	- XML::Compile::dataToXML() returns information about the source if
	  requested.

version 0.67: Mon Feb  4 09:31:17 CET 2008

	Fixes:

	- use of boolean 0 produced an error. [Allan Wind]
	  tests in t/21types.t

version 0.66: Thu Jan 31 09:32:00 CET 2008

	Fixes:

	- added a few tests to t/48subst.t to prove that min/maxOccurs
	  works with substitutionGroups. Had to remove two error checks.

	- ::Namespaces::findID() failed, because ::Instance::id()
	  was missing. [Philippe B]

	- ignore element when maxOccurs==0 [Allan Wind]

	Improvements:

	- added xml2yaml script, based on code by [Slaven Rezic]

	- XSD search path will split on ';' on Windows.

	- refrased start text of XML::Compile.  Suggested by [Sander Hulst]

	- looking for schema files with XML::Compile::findSchemaFile(),
	  will not check for readability, because using "-r" is useless.
	  So, the existence of the file suffices.

	- describe syntax of repetative substitutionGroups.

	- all errors in ::Schema are either marked in class 'usage', or
	  in class 'schema'.  Errors in other packages still need
	  classification.

	- ::Schema::Instance now reports include and import information.

	- reformulation in "{url}name" explanation, spotted by [Allan Wind]

version 0.65: Wed Jan 16 10:28:53 CET 2008

	Fixes:

	- fix an error "isa on unblessed" in XmlWriter.  An error will
	  be produced when an ARRAY or HASH is found where a simple
	  element value is expected.

	Improvements:

	- support schemas without targetNamespace.  Tests in t/71nons.t.

	- ::Schema::compile(interpret_nillable_as_optional) added, for
	  broken schemas, where people have interpreted the Schema
	  specs wrongly. Reported by [Christian Mueller]

version 0.64: Mon Jan  7 11:49:49 CET 2008

	Fixes:

	- ::BuiltInFacets requires Math::Big{Int,Float}

	- template() made mistake when a prohibited attribute was
	  encountered.

	- empty complexTypes are acceptable. [Kaare Rasmussen]

	- typo: implemented base64binary, but should have been
	  base64Binary.  Detected by [Gert Doering]

	Improvements:

	- perl template output with a long list of sequence element
	  names now split over multiple lines.

	- XML::Compile::Schema now explains how to use xsd files and
	  imports/includes found in them.

version 0.63: Sat Dec 15 15:15:43 CET 2007

	Changes:

	- Require XML::LibXML 1.65, which solves attribute creating
	  issues.

	Fixes:

	- dateTime did not format correctly: missing colon between
	  minute and seconds.

	- example template production failed on explicitly optional
	  attributes and any kind of arrays.  Reported by [Gareth Kirwan]

	Improvements:

	- Added XML::Compile::Util::type_of_node()

	- some error paths are cleaner, but still a few to be improved.

	- ::compile() new option 'ignore_unused_tags", requested by
	  rt.cpan.org#31539 [Slaven Rezic]

version 0.62: Tue Nov 27 12:23:38 CET 2007

	Fixes:

	- Remove requirement for Data::Dump::Streamer.

version 0.61: Tue Nov 27 11:43:08 CET 2007

	Changes:

	- Moved XML::Compile::Dumper from this distribution into
	  its own.

	Fixes:

	- ::compileClient(element_qualified => 'TOP') did remove the
	  existing qualification need.

version 0.60: Sat Nov 24 10:22:53 CET 2007

	Changes:

	- ELEMENT_ORDER hook returns packed types, not nodeName's,
	  because that would include a (non constant) prefix.

	Fixes:

	- ::Util::pack_type() with undef namespace returned undef.

	Improvements:

	- Warn people about out-dated libxml2 when running tests
	  in t/01use.t.  Let's see of CPANTesters like this.

	- when a READER is compiled with C<rpc_hrefs>, then elements
	  may have a href attribute, used for SOAP-RPC encoding.

version 0.59: Thu Nov  8 14:04:44 CET 2007

	Fixes:

	- Schema recursion detected too often.  It had to be based on the
	  schema-node, not the type name (which isn't unique enough)
	  Spotted by [Gert Doering].

version 0.58: Tue Nov  6 12:35:28 CET 2007

	Fixes:

	- could not handle empty sequences.

	- boolean 'true' and 'false' must be decoded into 1 and 0.

	- parsing of boolean was sensitive for blanks.

	- elements_qualified overrule extended too far.

version 0.57: Mon Nov  5 15:53:19 CET 2007

	Fixes:

	- basic schema elements (like SCHEMA2001 int) could not be
	  used as starting point for translation.

	Improvements:

	- XML::Compile::Util now exports some standard name-spaces as
	  symbolic names, like SCHEMA2001.

	- added 2001 XMLSchema instance xsd.

	- XML::Compile::Schema::importDefinitions ignores undef param.

	- compile(output_namespaces) now can be initialized with a
	  simple ARRAY, and explained a little better.

version 0.56: Thu Oct 18 09:25:16 CEST 2007

	Improvements:

	- XML::Compile::Util::pack_type() and ::unpack_type() now
	  understand name-spaceless types.

	- filled-in pre-declared type tables for older schema
	  definitions, but did NOT IMPLEMENT the differences in
	  SCHEMA RULES.  Some initial checks in t/81-2000.t

	- XML::Compile::SOAP::Server split-off from XML::Compile::SOAP;
	  documentation updated to reflect this.

	- added README.todo

	- implemented correct QNAME run-time "{ns}local" to "prefix:local"
	  translator, for writer.  May cause run-time error message.  

version 0.55: Wed Sep 26 10:32:49 CEST 2007

	Changes:

	- split-off the SOAP and WSDL things into a separate
	  XML::Compile::SOAP distribution, because they need more
	  complex dependencies and need more releases in the near
	  future.

	- install xsd's automatically in XML/Compile/xsd/
	  (XML::Compile::SOAP's xsd-files are in XML/Compile/SOAP/xsd/)

	Improvements:

	- added display of XML::LibXML version to t/01use.t

	- accept recursive schema's.  Reported by [Scott Peimann]
	  tests in t/62recurse.t

	- old bug in XML::LibXML shows up in its 1.64 release in combination
	  with the XML::Compile test-set.  XML::LibXML 1.65 contains the
	  patch.

	- skip t/pod.t when manuals are still in OODoc format.

	- XML::Compile::addSchemaDirs() effects are now global.

	- XML::Compile::knownNamespaces() can be used to define own
	  files for name-spaces, also global.

version 0.54: Fri Sep 14 15:35:53 CEST 2007

	Fixes:

	- XmlReader::block_handler called "last", where "return" was
	  ment, resulting in a warning. [Gert Doering]

	- Handle case where choice has minOccurs==1, but one (or more)
	  of the options has minOccurs==0... which makes the choice
	  optional implicitly.  Found by  [Gert Doering]

version 0.53: Thu Sep  6 21:15:14 CEST 2007

	Changes:

	- When calling SOAP operations, the elements do not use a Body
	  or Header anymore: these structures are flattened for the
	  user's convenience.

	Fixes:

	- totalDigits did only handle floats, not integers.  Reported
	  by [Gert Doering]

	- float values should not show trailing zeros.  The use of
	  sprintf "%lf" was superfluous, hence removed.

	- produce error when pre-installed schemas are not found.

	Improvements:

	- a bit of progress on WSDL/SOAP

	- YAPC::EU 2007 presentation available on the website.

version 0.52: Tue Aug 14 09:37:14 CEST 2007

	Fixes:

	- sloppy_integer did not pass some usage restrictions for
	  other than Integer->int translation.

version 0.51: Fri Aug 10 14:27:05 CEST 2007

	Changes:

	- the labels for blocks when maxOccurs >= 2 have changed from
	  f.i. 'xyz' into 'seq_xyz', to become clearer.

	Fixes:

	- both attribute with own simpleType definition, as attribute ref
	  to attribute with own simpleType were missing.

	- got direct compilation of top-level attributes to work, and
	  all available soap11 tests with it.

	Improvements:

	- much clearer docs about repetative blocks.

	- reread/corrected a few manual pages, but there is still a
	  lot to do.

version 0.50: Thu Aug  9 23:19:19 CEST 2007

	MAJOR REWRITE of the schema parser and code generators.

	Changes:

	- any maxOccurs >1 will produce/require an ARRAY.  Until now,
	  the sequence, all, choice, and group didn't work correctly.

	- removed a way to catch the errors produced while processing
	  a message: an error is an error (or use hooks).

	- convert error handling to Log::Report, which allows translations
	  and clean processing.

	- much improved error messages.

	- hooks are called differently, to be more useful.

	Fixes:

	- the reader now requires the correct order of the elements
	  inside the message.

	- "any" can appear everywhere, not only at the end of blocks.

	- "any" now correctly defaults as minOccur==maxOccur==1

	- accept attributes for complexType/simpleContent

	- floating-point restriction (facet) did not process well.

	- an empty XML container will be produced when an {} was
	  given: it will not be ignored.

	- improved template output

	Improvements:

	- added t/pod.t, which flagged some (produced) pod errors

	- added XML::Compile::Iterator, to ease the implementation of
	  XML::Compile::Schema::Translate

	- more checks whether the schema is sound.

	- added XML::Compile::Util

	- removed XML::Compile::parse() and parseFile() from the
	  public interface.

	- in doc: schema's should be schemas [Sven Neuhaus]

	- many more tests

	- no need for wsdl-patch.xsd anymore.

version 0.18: Wed Jun  6 22:46:44 CEST 2007

	Changes:

	- renamed XML::Compile::Schema::importData() into
	  ::importDefinitions()

	- renamed XML::Compile::WSDL::addSchemas() into
	  ::importDefinitions()

	- Schema::addSchema() will only add parsed XML.
	  Use Schema::importDefinitions() to added any (other) kind
	  of data.

	- reworked Namespaces::findElement() and ::findType() into
	  ::find(KIND, ...)

	- upgraded XML::LibXML requirement to 1.63

	Fixes:

	- Added dependency for IO 1.22 (IO::File)

	- Preference to address elements with a certain name, about
	  types.

	- Regular expression for all date types was not precise enough.
	  Now following the specs more closely.
	  Bug report from [Gerard Gerritsen]

	- Namespace of schema now separates attributeGroup and group
	  names from element, attributes and types.

	- include_namespaces parameter to Schema::compile() was always
	  set to 1.

	Improvements:

	- added xsd/2003-* for SOAP1.2

	- initial implementation for SOAP in lib/XML/Compile/SOAP.pm
	  and lib/XML/Compile/SOAP/SOAP1[12].pm

	- hooks with type/id/path will be invoked on all nodes.

	- removed Translate::reference(), in favour of a reference
	  lookup per kind of object.

	- ::Translate::element_by_node() renamed to ::element().

	- ::Schema::compile() and ::template() do not check provided
	  name anymore, but leave that to ::Translate::new()

	- ::Schema::compile(elements_qualified) now can limit ns-qual
	  to the TOP element only.

version 0.17: Tue Mar 20 12:21:57 CET 2007

	Improvements:

	- added XML::Compile::Dumper and t/21dump.t.  Added automatic
	  compatibility tests with non-dumped parsing.  However, this
	  results in "PmmREFCNT_dec: REFCNT decremented below 0! during
	  global destruction." errors in XML::LibXML, so is disabled by
	  default (see $skip_dumper in t/TestTools.pm)

version 0.16: Sun Feb 25 00:28:48 CET 2007

	Changes:

	- rename Schema::importSchema() into Schema::importData(), because
	  its parameter is very different from what addSchemas() expects
	  but the name was too close.

	- added patch to WSDL, to simplify the tOperation type, which
	  is overly complex.

	- now all four kinds of operations are detected.

	- build-in type anyURI is relative, hence cannot be checked at
	  parse time.  This means that XML::Compile doesn't need
	  Regexp::Common anymore (which was a forgotten dependency).

version 0.15: Fri Feb 16 16:21:59 CET 2007

	Changes:

	- removed "top()" from XML::Compile, because with multiple
	  XML sepcification files, there is no real "top" element.

	Fixes:

	- do not destroy caller's data when creating complex or tagged
	  elements.

	- added missing str2num() in BuiltInTypes for floats and doubles.

	Improvements:

	- move non-distributed xsd's to private dir.

	- added XML::Compile::WSDL and t/80wsdl.t, partial implementation

	- added XML::Compile::SOAP::Operation, partial implementation

	- added reader hooks and t/60hooks_r.t

	- added writer hooks and t/61hooks_w.t

	- implemented attribute by ref

version 0.14: Tue Jan 30 12:30:34 CET 2007

	Fixes:

	- Sequence of choices will multiply min/max.  Required to
	  understand WSDL.

	- Give error on unknown particles.

	Improvements:

	- Implemented 'any' element

	- fields of type QName are translated from "prefix:type"
	  into "{namespace}type".

	- Makefile.PL requires 5.8.0, which is probably still a mild
	  setting.  XML::LibXML is probably more demanding.

version 0.13: Mon Jan 29 00:11:12 CET 2007

	Fixes:

	- One place with nested Complex extensions, but more may need
	  fixing.

	Improvements:

	- Added WSDL and SOAP schema's

	- fixed importSchema() where addSchema() missed an 's'
	  -> addSchemas().

	- XML::Compile::new() now also accepts a filename, a ref
	  Scalar, and pre-defined name-spaces.

	- skip key, keyref, unique, selector, fields elements; just
	  like notation and annotation are lost.

	- defined top-level element "attribute", but not yet used.
	  Only useful in anyAttribute, which is not checked.

	- check all top-level elements for their name, to find typo's.

	- implemented anyAttribute

	- XML::Compile::dataToXML is used to accept user spec of XML
	  on many places.

version 0.12: Tue Dec 19 15:22:59 CET 2006

	Fixes:

	- fixed attributes can be optional. [Igor Shynkarev]

	- correct dateTime values were disapproved. [John R.]

	- converted to use oodist script (new feature of OODoc)

version 0.11: Wed Oct 25 08:08:20 CEST 2006

	Fixes:

	- Need to use "getChildrenBy..." in stead of "getElementsBy" in
	  XmlReader, because otherwise we search the whole tree.
	  Problem reported by [Igor Shynkarev]

	Improvements:

	- When the reader is called with a document, its document-element
	  is taken automatically.  In response to [Igor Shynkarev]

	- First attempt for XML template output.

version 0.10: Tue Oct 24 11:40:09 CEST 2006

	Fixes:

	- Some minor cleanups in XmlWriter

	- Now depends on XML::LibXML 1.61, which has a lot of improvements
	  and fixes.  As disadvantage does it require a recent libxml2,
	  which is only included with very recent Linux distributions.

	- t/49big.t test failed when installed on 64bit Perl, because
	  one of the test-values was smaller then MAX_INT in that case.

	- Default attribute type anyType.

	- anyType is correct namespace.

	- Type resolving when schemaNamespace is not the default was
	  broken.

	- Complex body permitted to be empty.

	- Default for "check_occurs" set to `true', because default should
	  give best results.

	- Missing element with default gave error when check_occurs.

	Improvements:

	- Added Template.pm to produce a Perl template example of a certain
	  type.  See XML::Compile::Schema::template()

	- Implemented type definitions contain an example which is used
	  for the templates.

version 0.09: Mon Sep 25 08:40:24 CEST 2006

	Fixes:  [thanks to cpan-testers work by Slavan Rezic]

	- MANIFEST not correct: contained old files

	- Removed dependency for Regex::Common, added Math::BigInt

	- Upgraded to XML::LibXML 1.60, which changed from Attribute
	  handling.

	Improvements:

	- Reorganized order in some man-pages.

version 0.08: Fri Sep 22 15:52:49 CEST 2006

	Improvements:

	- Turned XML::Compile::Schema::Translate into a singleton object,
	  and reshaped functions into methods.  The arguments hash intern
	  in object.  Together much cleaner.

	- Replaced use of "croak" for errors during translation into a
	  call to the error() method, with nicer and more consequent
	  error text layout.

	- Split XML::Compile::Schema::BuiltinStructs into lazy loading
	  XML::Compile::Schema::XmlReader and ::XmlWriter.

version 0.07: Thu Sep 21 15:06:05 CEST 2006

	- implemented substitution-groups

	- implementation of ref-element was bogus.

	- improved documentation (more about Schemas)

	- $schema->addSchemas from text

	- $schema->importSchema() with support of search-path.  Schema's not
	  automatically installed (yet)

	- added other schema's to the distribution.

version 0.06: Thu Sep 14 10:10:31 CEST 2006

	- The internals of the translator (and the produced code) are
	  fully rewritted to support complexType/simpleContent in a
	  clean way.

	- better support for list.

	- major improvements in the documentation.

version 0.05: Sun Aug 27 12:00:00 CEST 2006

	- Release as OODoc processed package :(

version 0.04: Sun Aug 27 11:49:54 CEST 2006

	- Rewrote the particle handling to better process nesting and
	  min/maxOccurs.

	- Many rewrites to implement complexType/simpleType processing.

	- Extended list describing missing features (and resolved quite
	  a few of them).  Parser is much stricter now, closer to the
	  specs.

version 0.03: Tue Aug 15 09:13:06 CEST 2006

	- Doc about what to pass as type to compile() was not updated.

	- Added attribute default, fixed and use=forbidden.

	- Added simpleType within list.

	- Added memberTypes with union.

version 0.02: Mon Aug 14 13:42:02 CEST 2006

	- Perl internal integer dependent size of INT_MIN and INT_MAX
	  to avoid use of BigInt.  Idea by [Merijn H Brand]

	- Added type assertions to translator, to check known attributes
	  to follow the spec.

	- All constructing routines are called with ($path, $args, ...)
	  where before, the $args option had different locations in the
	  parameter lists.

	- Moved element fixed tests to t/47fixed.t, reimplemented it to
	  make it work with BigInt.

	- Implemented element default and t/48default.t

	- Non-sloppy integers implemented in BigInts, tests t/49big.t

	- Element ref() should not specify own name.

	- XML::Compile::Schema::Instance collects info about one single
	  schema.

	- XML::Compile::Schema::Namespaces collects schema instances.

	- Under fly adding of schemas with ::Schema::addSchema

	- understands qualified and unqualified processing

	- and many more...