A literal result element and its attributes in a stylesheet namespace
bound by stylesheet-prefix are rewritten on output to the namespace and
prefix bound by result-prefix (#default supported), so a stylesheet can
emit literal xsl: elements. Prefixes resolve through the in-scope xmlns
declarations; aliases fold across include/import. Splits AVT parsing into
XSLTAVT.swift and trims the transformer to stay within length budgets.
format-number now honors named and default decimal formats: the decimal
separator, grouping separator, percent, minus sign, zero-digit (for
non-Latin digit sets), NaN, and infinity symbols come from the chosen
xsl:decimal-format, in both the picture and the output. FormatNumber is
parameterized over the symbol set and the formats thread through the
function table; folded across include/import.
An unrecognized XSLT element instantiates its xsl:fallback children in its
place, or is dropped when it has none; a fallback under a supported
instruction is ignored. The fallback content is handled in the instantiate
loop alongside variables, so the evaluate dispatch stays within budget.
Re-applies templates to the current node in the current mode, considering
only templates of lower import precedence than the one being instantiated,
falling back to the built-in rule when none match. Threads the current
mode and import precedence through the evaluation context, and moves the
transformer runtime types into XSLTRuntime.swift (namespaced, via private
aliases) to stay within the file-length and namespacing gates.
Only the static deduplicated helper moves to XSLTElements.swift (it needs
no private types); attributeSetAttributes stays in the transformer so the
evaluation context and result-item types remain file-private, satisfying
the one-file-scope-type namespacing gate.
Named attribute sets apply to a literal result element or xsl:element via
use-attribute-sets; sets may include other sets recursively with a cycle
guard, set attributes are lowest precedence so the element's own
attributes override them, and same-named attributes collapse to the last.
Folded across xsl:include/xsl:import. Splits the element-building helpers
into XSLTElements.swift to stay within the file-length budget.
xsl:message instantiates its body as diagnostic text and produces no
result-tree output; terminate=yes records the text and aborts the
transform with an XSLTError.terminated, halting further instantiation.
Splits the transform entry points into XSLTTransform.swift and moves the
message case to keep within the complexity and file-length budgets.
Whitespace-only text nodes are removed from the source tree for elements
named by strip-space before transformation, with preserve-space and a
more specific name test winning over *, and xml:space=preserve on a
source element or ancestor overriding the stripping. The name sets fold
across xsl:include/xsl:import. Refactors the stylesheet compile loop
through a Parts accumulator to absorb the new top-level elements within
the complexity budget.
A RELAX NG attribute failure now reports an undeclared attribute
("attribute @b is not allowed on <n>") separately from a declared
attribute carrying a bad value ("attribute @count on <n>: 'abc' is not a
valid integer"), quoting the value and naming its datatype by reusing the
text detail. The declared value pattern is found by matching the
attribute name through the residual.
A RELAX NG text/datatype mismatch now reports the rich datatype reason
(quoting the value and naming the type, e.g. 'abc' is not a valid
integer) and a value mismatch quotes both the actual text and the
required literal, instead of a generic "text content is not valid". The
detail is peeled from the content pattern at the failure point through
choice, group, list, and ref.
An element whose children break its content model now names each stray
child outside the model's alphabet individually, and reports a pure
order/count violation once with the allowed elements as a hint, instead
of a single opaque "do not match its content model". Adds an alphabet
extractor to the content-model matcher. This brings DTD to the same
located standard as the XSD, RELAX NG, and Schematron validators, so all
four schema languages give editor-grade diagnostics.
A content-model violation is now placed at the offending child with a
recovery hint naming what was expected, instead of one opaque "content
does not match the content model" per element. The sequence/choice
automaton is walked child by child via its follow-set to flag the first
child the model rejects, or the incomplete content at the end; an xs:all
group recovers past each stray child and reports every missing required
member. Well-placed children are still validated for their own content,
so an editor sees every problem at once. Brings XSD to the same located,
recovering standard as the RELAX NG, DTD, and Schematron validators.
Adds RelaxNG.errors(in:) reporting every way a document fails the schema
as individual located errors with coding paths, recovering past each
failure so all of a faulty document's problems surface at once instead of
only the first. Errors carry recovery hints naming the element types
accepted at the failure point, and distinguish unexpected elements,
invalid or missing attributes, invalid text/datatype content, and missing
required content. RelaxNG.validation() exposes the schema as a
Validation<Node, Void>, composing with the same framework as the XSD,
DTD, and Schematron rules. The boolean derivative engine remains the
authority on validity; the diagnostic walk only places and explains the
failures, recovering each step so it never cascades. Splits the engine
and the diagnostics into separate files.
A prefer attribute on a <group> (or nested <catalog>) overrides the
catalog-wide preference for the <public> entries within it, inherited
through nested groups. The resolver records each public entry's effective
preference and consults it when falling back from an unmatched system
identifier, so two entries in one catalog can prefer differently. A system
match still always wins.
A <pattern abstract="true"> is a template whose rule queries carry $name
references; a <pattern is-a="…"> with <param> children instantiates it,
substituting each parameter into the template's context, test, <let>
value, and <value-of select=> before the XPath is compiled. The
substitution is a single left-to-right pass, so a substituted value is
never rescanned and a short name never captures a longer one; $name
references that are not parameters stay as written. The template itself
contributes no rules.
A string literal split with ~ ("a" ~ "b") is the single literal "ab",
the way long values and namespace URIs are wrapped across lines. The fold
applies wherever a string literal is read: value literals and the URIs of
namespace/datatypes/include/external declarations.
A <value> pattern now carries its datatype, so matching compares the
instance and the literal in that type's value space rather than purely
lexically. Numeric and date/time types compare by value (1 equals 01,
1.5 equals 1.50); booleans treat 1/true and 0/false as equal; string and
token compare by their whitespace-normalized lexical form, as before.
Untyped values default to the built-in token type, and a bare string
literal in the compact syntax is token-typed. Durations and binary types
fall under the lexical rule.
Adds a datatypeLibrary-aware datatype resolver to the XML-syntax parser:
the in-scope library is inherited from the nearest ancestor, the default
(empty) library defines only string and token, and the W3C XML Schema
library defines the full built-in set. A type the in-scope library does
not define is an unknown datatype, so the <data>/<value> pattern matches
nothing rather than silently validating.
Adds an nsNameExcept name-class case so nsName with an <except> child
(XML syntax) and prefix:* - name subtraction (compact syntax) narrow a
namespace wildcard. The compact lexer splits prefix:* into a trailing-
colon word plus a star symbol, which the parser now recognizes.
Existing RELAX NG facet tests declared XSD types without the library and
relied on lenient resolution; they now declare the XSD datatypeLibrary,
matching what a conformant processor requires.
The compact-syntax parser now skips [ ... ] documentation/grammar
annotations (balanced over nesting) wherever they may appear, before a
define or before a pattern, and treats div { ... } as transparent
grammar grouping that contributes its defines and start directly. A
define literally named div still parses when assigned. Splits the
name-class and token helpers into a companion file to stay within the
type and file length budget.
Resolver(sgml:) parses the legacy SGML catalog format: PUBLIC, SYSTEM,
DELEGATE, CATALOG, BASE, and OVERRIDE entries with -- comment --
delimiters. Replacement URIs resolve against BASE and the supplied
baseURI; OVERRIDE YES/NO maps to the prefer-public fallback. Unknown
keywords are skipped tolerantly. Feeds the same Resolver as the XML
catalog.
Part of #86
The compact-syntax name class * - nameClass now parses as anyName except
the subtracted class (the model already had anyNameExcept). ns:* - name
would need an nsName-except model case and is left as a refinement.
Part of #79
The RNC datatype parameter block xsd:int { minInclusive = "5" } is now
parsed into facets (previously skipped), matching the XML-syntax <param>
support. The facet mapping is shared by both parsers through a new
RelaxNGFacets helper.
Part of #79
<let> may now appear at schema and pattern level, not just in a rule.
Schema lets are evaluated once at the document root; pattern lets extend
them (also at the root); rule lets extend those at each context node. A
narrower-scope binding overrides a wider one of the same name.
Part of #80
A <data type=> pattern now carries a SimpleType (base + facets) rather
than a bare built-in, and its <param name=>value</param> children
populate the facets (minLength, maxLength, length, pattern, min/max
inclusive/exclusive, total/fraction digits), reusing the XSD datatype
facet validation. So a constrained datatype like xsd:string with
minLength rejects too-short values.
datatypeLibrary resolution, typed <value> comparison, and the RNC
datatype parameter blocks remain.
Part of #79
The validation framework already conforms to the OpenAPIKit idiom: named
composable Validation values, a Validator with blank/validating/
withoutValidating, located errors, parameterized over both subject and
document. The DTD layer is already four named rules.
XSD content validation stays a single recursive Validation
(contentValidity) rather than a rule per constraint, because XSD is a
recursive type system: a child's type is resolved from its parent's
content model and position, then rewritten by xsi:type and the derivation
controls mid-walk, so the constraint categories are interdependent and
cannot be split into independent rules without changing semantics or
duplicating the recursion. That is the sanctioned multi-error form.
To keep each category verifiable on its own (the idiom's testability),
add XSDCategoryTests exercising ComplexValidator directly per category
(attributes, content, nillable, fixed); derivation, wildcards, and
namespaces are already isolation-tested through crafted schema strings.
Closes#101
- cdataAsText emits a CDATA section's content as escaped text instead of
a <![CDATA[ ... ]]> section.
- asciiOnly escapes every non-ASCII character as a numeric character
reference, so the output is pure ASCII (libxml2's character-reference
encoding).
Both apply to the Serializer and the incremental Writer through the
shared Escaping rules.
Closes#91
An assert/report may reference <diagnostic> messages by id through its
diagnostics= attribute; each referenced diagnostic is rendered (with the
same value-of/name/let substitution) and appended to the finding for
extra per-failure detail. <diagnostics> is collected schema-wide.
Schema/pattern-level let, abstract patterns, and key()/document() remain.
Part of #80
A <rule abstract="true" id="x"> is a template that does not fire on its
own; a concrete rule pulls in its let bindings and assertions with
<extends rule="x"/>, prepended so the concrete rule extends rather than
replaces them.
Part of #80
<rule> may now declare <let name= value=> bindings, evaluated at each
context node in order (so a later binding can reference an earlier one
through $name). The bindings are available to the rule's assert/report
tests and to <value-of select="$name"> in messages, using the XPath
engine's existing variable support.
Schema- and pattern-level let, abstract rules/extends, diagnostics, and
key()/document() remain.
Part of #80
Catalog replacement URIs (uri, rewritePrefix, delegate/next catalog) are
now resolved against the catalog's base URI and any in-scope xml:base,
reusing the existing RFC 3986 URIReference resolver. Resolver(_:baseURI:)
supplies the catalog's own location; xml:base on the catalog/group/entry
elements scopes nested entries. An empty base leaves relative URIs as
written, so catalogs without a base behave exactly as before.
The legacy SGML catalog format and group-level prefer override remain.
Part of #86
A define repeated with combine="choice" or combine="interleave" now
merges with the existing definition (as a choice or interleave pattern)
instead of the last one silently replacing it. A plain redefinition with
no combine still replaces. Applied consistently in a grammar and across
include merges.
<param> facets under <data>, datatypeLibrary resolution, typed <value>
comparison, and the RNC-syntax gaps remain.
Part of #79
- Eval-time prefix bindings: Query.evaluate(over:namespaces:) and
value(at:...:namespaces:) supply prefix to URI bindings, so a name test
like q:item resolves q to a URI and matches by namespace regardless of
the document's own prefix. The match is namespace-aware only when
bindings are present, so existing unprefixed/prefix-string queries are
unchanged.
- XPointer schemes: xpath1() evaluates an XPath expression (like
xpointer()), and xmlns(p=uri) binds a prefix for the expression parts
that follow it, threaded into the XPath evaluation.
The range() scheme (and so XInclude range inclusion) remains.
Part of #88
Three XInclude gaps from the audit:
- accept / accept-language / encoding are surfaced to the loader via a new
XIncludeRequest and a process(_:base:loading:) overload, so a loader can
honor content negotiation. The existing loadingURI form is kept.
- Cycle detection tracks the visited (resolved-URI) set along the current
inclusion chain and raises XIncludeError.cycle, instead of relying only
on the depth backstop; a non-cyclic diamond include is not flagged.
- parse="text" combined with an xpointer is rejected
(XIncludeError.textWithFragment), since a fragment cannot select within
unparsed text.
xpointer range() inclusion still depends on the XPointer range() scheme
(tracked in the XPointer issue).
Part of #87
Two ISO Schematron gaps, the most validator-relevant first:
- Dynamic messages: an assert/report message is now a template, so
<value-of select=> renders the string value of its XPath at the context
node and <name> renders the node's name. A finding can report actual
values ("Age 120 must be under 100") instead of static text.
- Phases: <phase id><active pattern=> scopes which patterns run.
validate(_:phase:) honors a requested phase or the schema's
defaultPhase; nil or #ALL runs every pattern.
let variables, abstract rules/extends, diagnostics, and key()/document()
remain (the last depends on XSLT-style function injection).
Part of #80
Two OASIS catalog gaps from the audit:
- systemSuffix and uriSuffix entries: an identifier ending in the rule's
suffix maps to its URI, longest suffix winning, tried after exact and
rewrite matches.
- prefer=public|system on the catalog: a system match always wins, and
the public identifier is consulted as a fallback only when prefer is
public or no system identifier was given (the OASIS rule). The default
stays prefer=public.
xml:base / RFC 3986 relative-URI resolution of entry URIs and the legacy
SGML catalog format remain.
Part of #86
Broaden the serializer and writer options from the audit:
- attributeQuote selects single or double quotes; the active quote is
escaped in values (apostrophe as ' in single-quote mode).
- lineEnding controls the inserted newline (set to CRLF for \r\n output).
- xml:space=preserve suppresses pretty-print indentation for an element
and its subtree; xml:space=default re-enables it.
- Writer gains writeStartElementNS(prefix:localName:namespaceURI:) and
writeNamespace(prefix:uri:), so callers no longer hand-format prefixes
and xmlns attributes.
CDATA-vs-escape and entity re-escaping strategies are left as a
follow-up; the current escaping is the safe default.
Closes#91
Add the DOM/tree editing primitives from the audit that the editor use
case needs:
- Attribute helpers: attributeValue(_:), setAttribute(_:_:) (update in
place or append), removeAttribute(_:).
- shallowCopy() copies a node alone, complementing the deep copy().
- precedes(_:) compares two nodes in document order (ancestor before
descendant, earlier sibling before later; unrelated nodes compare false).
- normalize() coalesces adjacent text children, drops empty text, and
recurses into elements; CDATA is left intact.
First-class namespace/DOCTYPE/entity-reference nodes (an invasive
TreeNodeKind change) and cross-document adoption/ownerDocument (no
document-identity model) are left for a follow-up.
Closes#90
Close the two XSD-regex gaps from the audit:
- \p{...} and \P{...} now resolve Unicode general categories (one-letter
groups like \p{L} and two-letter codes like \p{Lu}/\p{Nd}) through
Swift's own scalar generalCategory, so no character tables are vendored.
A curated set of well-known \p{Is<Block>} blocks is supported; an
unknown category or block name is a compile error rather than a guess.
- Character-class subtraction [a-z-[aeiou]] removes the subtrahend from a
class, and nests ([a-z-[b-d-[c]]]) and composes with categories.
Categories and subtraction work both as standalone atoms and inside
character classes, so XSD pattern facets that use them now validate.
Closes#89
Normalize line endings in the Reader as the spec requires, which was not
done before:
- CR and CRLF fold to a single line feed (XML 1.0), so every downstream
scan sees line feeds only.
- When the declaration names version 1.1, NEL (U+0085) and LINE SEPARATOR
(U+2028), and a CR followed by NEL, also fold to a line feed; in 1.0
those characters are preserved as data.
The XML 1.1 name-character productions are identical to XML 1.0 Fifth
Edition, which the character classifier already implements, so names need
no change. The 1.1 character range (C0 controls permitted only as
references) is not yet enforced.
Closes#96
Broaden the TextReader pull cursor toward libxml2's xmlTextReader:
- Attribute nodes: moveToFirstAttribute/moveToNextAttribute/
moveToAttribute(_:)/moveToElement reposition the cursor onto each
attribute as an .attribute node and back.
- Namespace accessors: namespaceURI, localName, prefix from the current
element's or attribute's resolved name.
- xml:lang: inherited down the tree, overridable per element, reset when
it goes out of scope.
- Position: lineNumber/columnNumber report where each node begins, so a
streaming consumer can locate findings in the source. EventReader now
records each node's start mark.
DOCTYPE and entity-reference nodes are not surfaced (the DTD is exposed
through documentType, and entities are expanded inline); whitespace is
not split into significant/insignificant (needs the content model);
baseURI needs a document URI. The accurate positioning path for editor
findings remains readTree with source ranges.
Closes#98
Parse the XML declaration's pseudo-attributes instead of skipping the
<?xml?> processing instruction:
- version, encoding, and standalone are surfaced on a new XMLDeclaration,
returned from parseWithDocumentType and via PureXML.xmlDeclaration(_:).
- The pseudo-attributes are validated: fixed order (version, encoding,
standalone), known names only, and standalone must be yes or no. A
malformed declaration is rejected.
- The declared version is now readable, a prerequisite for XML 1.1.
parseWithDocumentType returns a ParsedDocument struct (node,
documentType, declaration) rather than a 2-tuple; the member names are
unchanged so existing call sites are unaffected.
Closes#95
Deepen DTD processing in the doctype scanner:
- Conditional sections are processed, not skipped: INCLUDE sections have
their declarations scanned, IGNORE sections are discarded, the keyword
may be a parameter-entity reference, and nested sections are handled.
- <!NOTATION> declarations are parsed and stored on DocumentType.
- Unparsed entities (<!ENTITY ... NDATA n>) are parsed and stored with
their notation, distinct from general entities.
- External parameter entities are loaded through the injected resolver so
their declarations expand; the default refusing resolver loads nothing,
keeping XXE closed.
- NOTATION attributes are a distinct type, validated against the declared
notations (value must be listed and the notation declared).
Tracking unparsed entities also lets ENTITY/ENTITIES attribute values be
validated against the declared NDATA entities, closing the gap left when
#73 could only check name syntax.
Closes#72
Complete the DTD attribute handling found in the audit:
- Default and #FIXED attribute values are applied to the parsed tree.
parseApplyingInternalDTDDefaults returns a tree where every element
carries the attributes its <!ATTLIST> defaults, so an editor sees the
same effective attributes a validating processor would.
- NMTOKEN, NMTOKENS, ENTITY, ENTITIES are validated: NMTOKEN(S) must be
XML name tokens, ENTITY/ENTITIES must be XML names. Previously all fell
through to unconstrained CDATA.
- DTD default values are normalized per XML 1.0: character and predefined
entity references are expanded, literal whitespace becomes a space, and
non-CDATA values are collapsed and trimmed.
ENTITY/ENTITIES check name syntax only; that the named entity is a
declared unparsed entity is not verified, since NDATA entity tracking is
not yet modeled.
Closes#73
Make XSD validation namespace-aware and enforce the form defaults and
xs:all structural constraints that were previously ignored:
- targetNamespace is tracked: a schema's global elements are in its
target namespace, and an instance's root element must be in it.
- elementFormDefault / attributeFormDefault (and per-declaration form)
qualify local element and attribute declarations. The content-model
matcher already compares namespace URIs, so qualified declarations
now match only same-namespace instance nodes; attribute matching
became namespace-aware too.
- xs:all groups are checked at compile time: every member must be an
element occurring at most once, the group may occur at most once, and
nested model groups inside an all are rejected (a SchemaError).
When a schema has no target namespace the stamping is a no-op, so
unqualified schemas validate exactly as before.
Cross-namespace type references through xs:import resolve by local name
only; full multi-namespace type tables are not modeled.
Closes#78
Read and enforce the XSD 1.0 derivation controls that were previously
ignored:
- abstract elements may not appear in an instance directly, only through
a substitution member; an element of an abstract complex type must
supply an xsi:type naming a concrete derived type.
- block on a type forbids the disallowed xsi:type substitutions; block on
an element drops substitution-group members that derive by a blocked
method.
- final on a type rejects, at compile time, any type that derives from it
by a final method.
- xs:redefine requires the redefined type to derive from itself.
Track each named complex type's base and derivation method as the
backbone these checks walk. Also compose complexContent extension and
restriction content models, which were previously dropped: an extension
now inherits its base's attributes and content (base first, then the
extension, in sequence).
Complex-type restriction is not verified to be a structurally valid
subset of its base (Particle Valid (Restriction)); that limit is
documented at its sites.
Closes#76
Add xs:any and xs:anyAttribute wildcards with namespace and
processContents constraints (skip/lax/strict), and resolve the
list built-in datatypes (IDREFS, ENTITIES, NMTOKENS) to their
item types so each list item validates.
WildcardNamespace models ##any/##other/enumerated URI sets;
Wildcard.admits gates undeclared attributes and wildcard-matched
children. ContentNFA.follow exposes the follow set for completions.
Closes#77
Slice 4 of #94 (completes the epic). Quick-fixes are derived from the structured
schema completions, never by parsing a diagnostic message:
- TreeNode gains contentRange (the span between an element's start and end tags),
captured by the ranged reader, so an inserted attribute or child is placed
exactly.
- QuickFix and TextEdit model an editor edit (a replacement over a source range;
an insertion is a zero-width range).
- QuickFixEngine turns a Completions into fixes: a required-but-absent attribute
becomes an insertion before the start tag's >, and a still-expected required
child becomes an insertion before the end tag. Schema.Document.quickFixes(at:in:)
is the entry over a ranged tree.
Tests apply a fix and re-validate: adding the required id, or inserting the
required qty, yields XML the schema then accepts; a complete element offers none.
#94 (editor integration) is complete: ranged recovering read, unified lint,
schema completions, and quick-fixes.
Slice 3 of #94. From the compiled schema, compute exactly what is allowed at an
element, for autocomplete and "what's missing".
- ContentNFA.follow(after:) returns the precise follow-set: the element-name
labels the content automaton can accept next after the children already
present, plus whether the content may legally end there.
- Schema.Completions reports the allowed next elements, the complete flag (false
means a required child is still expected), a wildcard flag, and each declared
attribute with its required and present status.
- CompletionEngine resolves an element's declared type along a coding path (the
global declaration, then each parent's content model) and computes the
completions; Schema.Document.completions(at:in:) is the entry over a ranged
tree.
Tests pin the follow-set as the children accumulate (name then qty then the
optional note, complete flipping true), the attribute required/present status,
and resolution for a nested element by its coding path.
Slice 2 of #94. One call for an editor: PureXML.lint reads the document once into
the ranged, best-effort tree (never throwing), then merges the parse diagnostics
and the validation findings into a single source-ranged, severity-tagged list in
document order.
- LintDiagnostic carries message, severity, and the SourceRange to highlight.
- lint(_:validate:) takes a node-based validate closure that runs over the
RECOVERED tree, so validation works even when the document is not well-formed:
a truncated, schema-invalid document yields both a parse diagnostic and the
validation finding, each with a range.
- Schema.Document gains validate(_ node:) so a caller validates the same tree it
is editing; validate(_ xml:) delegates to it.
Tests cover a clean lint, well-formedness-only lint, a range-located validation
finding, and the both-truncated-and-invalid editor case.
Slice 1 of #94 (editor integration). The keystone: a located validation finding
can now be mapped back to a source range.
- TreeNode gains a sourceRange (start/end Mark); SourceRange is a half-open span.
- PureXML.readTree builds a mutable, parent-aware, best-effort tree whose every
node carries its span (an element from the < of its start tag to past its end
tag; leaves over the text they consumed). Never throws, recovers in place,
returns located diagnostics.
- TreeNode.node(at:) navigates a validation coding path to its node using the
same canonical convention the validators build paths with (a sibling index
only when a name repeats), and sourceRange(at:) returns the span. So
Document.validate's ValidationError.codingPath resolves to characters in the
document.
Tests prove the spans, the navigation, the end-to-end bridge (an XSD finding to
a source range), and that ranged reading of invalid input never crashes.
