CVS log for xmlschema/XMLSchema.py

Up to [Public] / xmlschema

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

make booleans be booleans

rebuild facet fixed properly

switch to reflectionOutMap

finish separation of reflection code from main modules

move reflmap stuff out to separate file

restructure in preparation for allowing either normal form to be reflected

fix reflection/rebuilding of attribute wildcard in attribute group

dump non-global, non-None scope as local, rebuild properly

conver num facets during rebuilding

rebuild list-valued facets properly

fix scope and variety rebuilding

remove debuggin print,
actually _use_ reloaded ind dump,
handle esspecial properly

individual normal form reload barely working, debugging prints still in

working on new restore

remove Schema, SSchema to XMLSchemaCore.py,
make new sources split actually work

move some parts out to XMLSchemaElt, AbInitio and FSM

vacuous, but establishes baseline on this branch

include variety in preparation
fix bug in restricted anon schema
move facets from prim type to restriction, knock-on effects

use name attr if avail. when undumping

allow richer ids

alternative dump format

fix bugs in previous (any subsumption fix) fix

chase scope during prepare,
test subsumption case of elt vs. wildcard better

give ur types sschema

clean up rebuilding,
add attr wildcard to urType,
dont inherit attr wildcard on restriction

repatriate alternate's dependance on PyLTXML

fix null ptr ref

work harder on redef, fixes bug with simpleContent redef

*** empty log message ***

correct bug in -D case introduced by defaultQualAttr arg to
layer.fromFile

remove unused type/decl args to validate
return targetNamespace from file sniff, use it to simplify namespace
 sync checks
Unify schemaLoc/namespace check on no-known-decl

fix one over-zealous not -> is None test

allow empty redefine

fix oob attr on schema bug
ignore oob attrs during component building

more robust wrt missing models,
plug some init gaps wrt layer

restructure FSM label again

flag attempt to derive a non-empty model from empty/text

typo in previous fix

fix residual element/qname possibility in fsm

fix corner-case bug wrt target NS and doSchemaLocs when validating schemas
flag disallowed complex-from-simple restriction case

work around last AttributeError traps -- perhaps not worth it?

big cleanup of all tests etc. to avoid unnecessary AttributeErrors

get rid of complexType.elementTable in favour of
 a) a simpler table on fsms for consistency checking;
 b) a different approach to type assignment in validation
Fixes bogus default with multiple local decls, etc

compute elementTable as side effect of building fsm
check subtype during elt-elt subsumption check

Add extensional restriction checking via FSM subsumption check

cleanup some error-handling bugs

always push schema state, fix an obscure bug exposed by datatypes/hfp

rework handling of namespace argument to checkinSchema & fromFile
provide wherewithall for applyschema to enforce <import> requirement

check schema loc on to-be-validated schema docs

replace Schema reuse hack with proper reuse
repair locations yet again
be sure to pop stack even when returning sForS w/o reading

proper fix for dbl ur init

re-introduce looping include cutout
temporary patch to defeat double urType initialisation????

add whiteSpace to checkFacetTable

make useDTD sticky

even more delicate probe of string's basetype at init
fold in mgd fix from main line

fix incl/excl bug
handle ##local vs ##other properly
detect clean state at init better

mark undumped schemas as prepared,
redo schema buggy/error variables again,
always return a schema if no error,
seems to work now, rebuilt regr test output

get pointer in attrs rebuild right,
handle model group defns properly at reflect/rebuild,
prepare ident constrs properly,
move initFromDump to factory initialisation,
working on schema error reporting

reflect model group defn properly

fix some pointer issues in rebulidng

settle on redundancy checking per namespace

fix some rebuilding bugs, including hacking class
of Group and KCons,
start actually using undumped sfors to validate schema docs

handle explict nil value during rebuilding
hack to rebuild occurs
index rebuilt components properly, ignore anyType

finish rebuild
actually reflect qname of named group
final isn't special?

more Rebuilds, stuck at PrimitiveTypeDefinitionRebuild. . .

part way through special rebuild definitions

improve handling of pointer/id,
fix list rebuilding,
actually USE converted value

start work on rebuilding after reading in sForS dump

first pass of reading dumped sFors complete

rename factory, derive from layer.factory,
begin work towards reading in and using dumped sForS instead of DTD

don't push schema stack for schemaLocs either

compensate for urlparse.urljoin bug wrt x:///

don't push schemaStack for new namespace in fromFile, but restore the
 sfors on the way out
identity constraints are now properly named and refered to by QNames
 this includes having a vtable for them and 'prepare'ing them

dont crash on empty <element>,
improve factory base handling

complete regularisation of schema doc processing

begin regularising all reading of schema files

misc changes

clean up isSubtype a bit, handling union, urSimpleType and fixing an
old bug

avoid prepare crashes

get element fixed/default into component

correct reflection map for Annotation

handle schema annotations

turn oob attributes into annotations

throw error on tns=""
fix bug in checkMinMax
start work on annotations, reflection thereof
use new namespaceDict mapper property

fix old (?) bug in home location for compiled version

fix extend-by-empty pblm

more reflection

towards reflection of identity constraint

(PSV)Infoset reorganisation

make sure no particle -> empty sequence when needed

cut all raw elements to allow pickling

fix some pblms with independent mode

cap min/max at 101 for performance

implement -i switch to do forced schema assessment independent of any instance

make home determination more robust

number back to decimal

fix xs: bug,
remove debugging print

actually get rid of max=0 components

merge v2001 back in to main line

simple type rename,
clear out facet dead wood,
actually process whiteSpace facet, which had been ignored (oops)

another clause in emptiable
shift to xs: as default in DTD

uriReference -> anyURI
implement block=substitution wrt substitution groups

detect chameleon include and treat all unprefixed QNames therein
specially

implement lazy checking of restriction wrt rere attribute groups
catch deletion of required attr in ct restriction

allow elt for choice restriction, to cover new SforS
do minimal name and type checking on elt for elt restriction
eliminate forbidden attrs from attribute groups

check that restriction doesn't relax/change fixed value constraint on attrs
restructure attribute group attrDecls to make them properly lazy in preparation
 for doing restrictive redefinition right, but remove that for the
time being

merge attr use changes back in to main line
implement attribute group redefinition

catch extension error

make AttrUse a full-fledged one

change NS to 2001, implement null->nil

parse XPaths at component build time
fix crash on subst subtype check error message

merged forceDTD+infoset back to mainline

fix bug in wildcard as edge label in reflection

support PSVI of attrGroupDefn,
protect better against missing types

fix spelling of whiteSpace,
make equivClass computation less inefficient,
catch loops in schema doc reading themselves

add hooks for whitespace

real facets

fix equiv name, add stubs for identity constraint reflection

put facets back in reflection, in the right place,
move assignUid out

fix conflicting message on local elt redef

bring urtype in line with spec,
reflection for particle, model group

fix bug in instanceList implementation,
treat primitiveType reflection as special (always pointer),
make non-global scope lazy,
derive CDATA from string and token from CDATA

fix a few minor properties,
add reflectedName and reflectionMap to many components

assign printable proper IDs to components on request,
provide reflectedNames for some components

fix type name in builtin -instance types

start adding built-in attr decls for xsi:type etc.

store whole declaration in typeTable
enforce new restrictions on element ref=
handle defaulting of nullable properly

Another order of magnitude speedup of eclosure, by not sorting the
results

speed up eclosure (by factor of 10 :-)

removed errors now caught by DTD checking

handle errors during DTD enforcement on schemas

Use a local XMLSchema.dtd to check schema documents if they dont
supply an external DTD themselves.  Some error cases not well-handled, yet.

fix bug in local scoping of defaults to <included> schemas

allow use on global elt attrs

add actual urSimpleType, use it appropriately
fix recording of 'fixed', 'default' for attrs
use subordinate simpleType in double-restriction case under
simpleContent

typo in last fix :-(

allow for empty model wiht 'mixed'

keep going if restriction missing

fix spelling of NotASchema
allow restriction to a member of a union for xsi:type
fix empty vs. mixed bug
add CDATA and token as synonyms of string, for now

actually put urType in type table under name anyType

always count errors

realModel tries to recover from no basetype

Add handy splitQName function and use in in class QName

Move checkString methods to applyschema.py, because they may need to look
at *instance* in-scope namespaces

can't restrict list with list

merge in CR branch

new NS URI

don't collapse as much wrt sequences when extending

implement redefine
back out restricition of union by union
change concrete syntax of field and selector

implement chameleon include

more urtype name changes

accommodate (partly) to new name for simple ur type

minimal support for restrictions of lists and unions

implement changes to content model of complexContent

a few more simpletype change leftovers

residual small bugs from simple type changes

fix a residual bug in SimpleType, implement List and Union

Change SimpleType component to agree with new design

convert to new XML repr of simpleType

New XML Schema NS for CR
implement equivClass -> substitutionGroup move

fix bug ignoring nested attributeGroup references

treat simple content as emptiable for now
provide null default for checkMinVals

don't die if no group in group, if no type for attr

handle no basetype in mergeContent
stub for min checking of strings

recover from no-ref internal group
handle missing base when checking isSubtype
check 'content' attribute value

move some methods from XMLSchema.py to applyschema.py

provide a printable name for Wildcards

check restriction occurs only if groups match;
handle one trivial case of emptiable;
allow textonly from emptiable mixed derivation

replace complex with simplistic check for circularity, give builtins a targetNamespace

deal with knock-on effect of group defn fix

catch LayerError
check model group defns for exactly one child
sort FSM edges on output

at least give better error messages for facets when derivedBy='list'

try once again to simplify/fix include/import/namespace processing

give float, double, decimal a common superclass to host CheckMinMax
give error if facet missing value attr

fix bug (circularity) introduced in previous fix for textOnly stuff

catch use of complexType as base for simpleType, type of attr
fix (?) indentation bugs in checking for textOnly in right places

cover empty <import>

fix bug in unqualified xsi:type

Make fixed not == required

give error if type attr and nested type defn

wildcard intersection fixes
any/any determinacy checking

typo in new allows code for Wildcard

typo in minInclusive fix

Fixed bugs in attr type error message, fullName/primitiveType
  loophole, minInclusive
Reworked implementation of wildcards and wildcard intersection

merge xmlout branches back in to main line

Content model determinism checking (not complete: any/any not done)

protect against more missing definitions

cope with missing attrDecl

use convertToNum throughout, raise error if not a numeral

register values, not instances
change enumeration to cope with above change

make fsm asXML label readable when wildcard

integrate basetype calculation patch from other branch

fix handling of complex type with simple model derived from complex
type (with simple model)

handle enumerations a bit
change import handling

manage unnamed top-level components without crashing
fix bug in handling of undefined *ed element ref in content model
allow undefined refed elements to match in fsm

merge QName checking from main branch,
fix QName error messages

change handling of prefix-lookup failure

Add method to QName to check basic validity

integrate import/null-schema patch from other branch

pop schema stack after fromFile even if it fails

convert error to log an element onto factory.resElt
convert where to add location attributes to an element
add asXML method to fsm for applyschema error logging
name the elt we want from layer.fromFile

protect against missing basetype,
handle Unicode names in QName

Check for strings in a way that works with or without 16-bit support

shift to using python's built-in url parsing

handle value and use on attributes

fix name error in group ref comparison in cm merger

various missing property initialisations fixed

oops, remove debugging printout

fix bug (in spec. too:-( wrt anyAttribute ##other)

catch and allow more cases of restrictive derivations from the ur-type

check that inputs are actually schemas,
remove schema arg to doImport, checkInSchema

replace stale use of anyElt in urType

handle undefed ref in note and translate

add copyright

cleanup residual bugs with massive rename,
rename Any to Wildcard,
replace AnyAttribute with Wildcard,
get validation of Wildcard working in both element and attribute contexts

wholesale name changes -- init. caps for all classes,
schema.py -> XMLSchema.py

get rid of AnyWrap, separate any and anyElt
structure sub-classes of any for validation use

move fsm translation to 'translate' methods on appropriate classes

add version string
fix typo in previous equivClass fix :-(

fixed race condition in type inheritance check

remove lots of dead wood

fix missing particle in urtype

value check on min/max occurs

removed dump, dumpForDTD and all related methods to separate file

move resolveURL around a bit to get it right

better handling of use of ns uri for loc

merge in private and comp branches

merge in comp branch

remove special casing of schema creation for no elt

Do imports inline
Use namespace URI for import if no schemaLoc
Hack textonly complextypes w/o base
Fix bug in hard-case for extension construction
A few defaults to allow DTD-free operation

completed basic conversion to components,
works for triv, tiny, tiny-nns and self

begin work on content model merger
fold in url normalisation
attribute groups improved

working on unifying particles and defrefs

much improved, minimal coverage almost everywhere,
groups properly handled,
facets and keys

get builtin facets right, finally
fix bug in numerical checkString, cover for non-numbers
propagate particle/term/group structure through translation into FSMs
get noteElement, noteParticle working again

begin work on model merger,

interpolate attributeUse between complexType and attributeDeclaration,
parallel to particle

bring built-in datatypes in to line with WD
add list types

fix facet initialisation of derived simple types
stub for string validation
minimal complex type support

working on complex type, attribute;
back out component.qname

odd attrs patch

add modest attempt at resolving relative URLs

residual bug in import

starting on complex type

Getting the basics sorted out:  bare minimum of simple type def and
element

beginning upheaval

lots of futzy fixes to effectiveStuff for Last Call schemas

lots of hacks to get effective type more involved

change to full import, no duplicates;
redo error handling, use line nums if available

allow 208 override,
some cleanup of complexType attr checking and defaulting

top-level attrs,
implement 208

convert to camelCase for builtin datatype names

remove abstract elements from equivClass expansion for 'dump'

trivial move to new def'n of restriction

convert * to unbounded for maxOccurs

exact->block, tentatively

merge private branches back into public after 20000225 release

fix any bug

source->base, restrictions goes away, uri -> uri-reference

convert dump to use simple/complexType
expand equivClass in dumped models (should have a switch)

fix minor datatypes initialisation bug

handle #any --> anyAttribute in dumping

fork branck for non-public changes
lots of changes to accommodate datatype/type -> simple/complexType,
group -> all, choice, sequence
builtins in separate namespace
got rid of subtypes and leaves, not used

info->documentation; fix bug wrt dump of empty type with attrs

implement inheritance for attribute restrictions

fix long-standing serious bug in implementation of restriction for
complex types

simple indentation
key etc. handled
namespace prefixes are broken for all but schema for schemas

dump now works, I think, except for keys etc. and indentation

fix bug ignoring inline types for attributes
part-way into normalisation 'dump' method on schema

minor changes for xsi:schemaLocation

towards support for xsi:schemaLocation

QNames using python-level NS dictionary

remove debugging prints
turn references to element class exemplars into disjunctions

fix bug in setting up occurs;
working on dumpToDTD: equivClasses

Got DumpToDTD working again, minimal support for QNames

define error to raise a SchemaError exception

minor initialisations for key processing at schema levels

better message on import failure

missing defaults

init keys etc.

fix anyattr merge
groups for key types

start anyAttribute support

Make urType more real, use it for ultimate default
Fix another attr ns bug, uncover incoherence in use of attrGroupRef to
get ns-qualified attr names in instances!

correct check of equivClass type derivation from exemplar

allow for qualified attr names

fix (?) some attribute type bugs

stub for anyAttr

began getting dumpTo/ForDTD back in shape

more stubs

stubs for keys etc.

fixed ns bug in noteElement

save and restore factory.schema in fromFile

implement import statements

pull equivalence class building up front in preparation phase

Save ALL table & type initialisation stuff until user calls 'prepare',
to avoid deadly embraces.  Handle abinitio/builtin stuff lazily.

handle targetNamespace earlier and better
build stub schema for schemas properly if necessary

lots of changes for qnames

attempting to start QName processing

merged in new-design branch

recursive handling of equiv. classes, equiv. classes into fsm

slowly working through type derivation

more work on new-design

working on new-design

begun to work on new-design upgrade

fix info

fix bug in translating min/max occurs to fsm
a little more <any> support

fix typo, add/correct some facets

handle more classes, rename source to basetype

add some facets

handle <any> (not complete yet)

use realorder not order

fix case where start node of determinsed fsm is also end node

resurrect dumpTo/ForDTD, preliminary

change handling of default attribute values -- make them class vars
which get shadowed by assignments to instances

fix basetype window bug

minor bugfixes, make rootName a self-computing property

minor bugs

fix some bugs in interaction between complex types and new approach to
simple types
give textonly complex types a coreType to carry their datatype part

merge in branch which switched to separate classes for ab initio types

convert to classes for primitive types, use them as effectiveType for
all simpleTypes

determinise FSMs

more content models to FSMs
add fsm method to types

start translation of content models to FSMs

classes for ab initio types

min/max facet comparison

more work on facet checking

start checking facets for simple types, some built-ins now have facets

get simple more up to speed with complex, start on effective type
computation for simple, change representation of facets

effectiveType now a magic attr, bug fixes

use effectiveType() in type.elementTable

attr merging works better, prints better

*** empty log message ***

elementTable() now a method

give group realmodel and realorder attributes via __getattr__

elementTable

use __getattr__ on .realtype to do type lookup for elements and
attributes

make attr printing work again

added <any>

just testing

minimal model merging (not tested)
stub facet merging
attr merging doesn't break, but doesn't work either

a bit more on merging

attribute groups in place, minimal merging

framework for effective type computation
stuck at interface from complex to simple

starting work on effective type computation

min/max for attributes

more attributes printed and conflists resolved

basetype now converted to real thing
plug embedded types in to their supertype
check some content constraints

fix unnoticed version merge

min/max occurs for groups

more printing

attribute printing

print more things, indented

get rid of extend, not in python before 1.5.2

begin work on groups

scoped patterns, named groups, rename type to Type, preliminary
__str__

lots of fixes, schema init first pass

fix schema/topSchema name conflict

radical overhaul underway, using layer methodology

partway to moving to Simplified syntax

special case handling of 'import' tag to avoid name collision with
built-in

added a fair bit of bomb-proofing to allow some output in
the face of missing archetypes

* schema.py: debugged model group support, only trivial example
checked

more work on model groups, need to implement plugInModelGroup
computation

preliminary support for model groups

* schema.py: check refined archetypes are refinable/open; finish
dumpForDTD, including expanding refined names

add tracking of elementTypeRefs, ignore multiple
definitions of archetypes and elementTypes, warn about undefined
archetypes and elementTypes;
build full content model, implement computation of
effective content model;
Add dumpToDTD

* schema.py: fix bug in rootName computation, add a further term
to the rootCompare sorting predicate

* graph.py: Attempt to compute scrollregion from schema properties

switched to tk file open widget

fix bug in higherHighest
add stubs from more top-level schema elements

Made some things really embedded, instead of pointed to by name
Sort the leaves before we start, by complexity of their root set

No more Gato, own simple roots-at-right layout, graph in separate file
now

Baseline, uses Gato, has roots at top, no subgraphs