Module chvalid 

XML character-class validation (upstream chvalid.c / parserInternals.c).

The exported xmlIs* family and xmlCharInRange use the generated character-class tables from unicode_tables.rs (extracted verbatim from upstream codegen/ranges.inc; see tools/archaeology/gen_chvalid_tables.py).

UPSTREAM-PARITY

The semantics mirror upstream chvalid.h macros exactly:

• xmlIsBaseCharQ: (c < 0x100) ? xmlIsBaseChar_ch(c) : xmlCharInRange(c, &xmlIsBaseCharGroup)
• xmlIsBlankQ: (c < 0x100) ? (c==0x20 || 0x9<=c<=0xa || c==0xd) : 0
• xmlIsCharQ: (c < 0x100) ? (0x9<=c<=0xa || c==0xd || 0x20<=c) : (0x100<=c<=0xd7ff || 0xe000<=c<=0xfffd || 0x10000<=c<=0x10ffff)
• xmlIsCombiningQ: (c < 0x100) ? 0 : xmlCharInRange(c, &xmlIsCombiningGroup)
• xmlIsDigitQ: (c < 0x100) ? (0x30<=c<=0x39) : xmlCharInRange(c, &xmlIsDigitGroup)
• xmlIsExtenderQ: (c < 0x100) ? (c==0xb7) : xmlCharInRange(c, &xmlIsExtenderGroup)
• xmlIsIdeographicQ: (c < 0x100) ? 0 : (0x4e00<=c<=0x9fa5 || c==0x3007 || 0x3021<=c<=0x3029)
• xmlIsPubidCharQ: (c < 0x100) ? xmlIsPubidChar_tab[c] : 0
• xmlIsLetter: xmlIsBaseCharQ(c) || xmlIsIdeographicQ(c) (parserInternals.c)
• xmlIsBlankNode: tree.c — text/CDATA node whose content is empty or all blank.

Courts

CHVALID-* differential tests compare against the oracle DSO for the whole BMP + representative supplementary-plane code points.

Upstream contract

Mirrors upstream chvalid.c / xmlunicode.c / chvalid.h (SRC-LIBXML2-2.15.0-CHVALID-C et al., parity target libxml2 2.15.3 oracle): xmlCharInRange, the exported xmlIs* predicates and the xmlIsPubidChar_tab data table.

Conceptual behavior

Implements the upstream Q-macro semantics verbatim (each xmlIs*Q expansion is listed above): sub-0x100 code points are answered from tables/linear tests, larger code points from the generated range groups via binary search. xmlIsLetter (parserInternals.c) and xmlIsBlankNode (tree.c) complete the surface.

Ownership & safety invariants

The tables are immutable static data (extracted from upstream codegen/ranges.inc); xmlCharInRange only reads its group argument (SAFETY: NULL group returns 0, valid group must cover nbShortRange/nbLongRange entries). Nothing here allocates.

Historical quirks & epochs

R-000135: the seven char-class tables were extracted verbatim from upstream ranges.inc by tools/archaeology/gen_chvalid_tables.py (sha256-bound, oracle sha256 e7575963…) and the DATA-GLOBALS-001 court fingerprints FNV-1a hashes of all nine xmlIs* functions over the BMP — the tables are stable across the 2.7.8 → 2.15.3 oracle span.

Deliberate oddities

xmlIsPubidChar only accepts < 0x100 (per the Q-macro); the Latin-1 linear cases in xmlIsBaseChar etc. are kept exactly as upstream encodes them rather than merged into the range groups.

Proving courts

DATA-GLOBALS-001 (tools/abi/data_globals_probe.py + committed C probe) compiles the probe against the system libxml2 and the candidate DSO and requires byte-identical output; CHVALID-* differential tests cover the whole BMP; cargo test runs the unit assertions.

Tempting simplifications that would break parity

Do not regenerate the tables from Unicode data files: upstream tables carry historical drift (e.g. ideographs bounded at 0x9fa5, the pubid table) that byte-parity requires. Do not replace the binary search with a hash set: the group ranges are what the C ABI exposes through xmlChRangeGroup.

Functions

xmlCharInRange^⚠

Binary search over the short/long range tables (upstream xmlCharInRange, chvalid.c — the tables are sorted, so the search is exact).

xmlIsBaseChar^⚠

xmlIsBaseChar(unsigned int ch) — XML 1.0 BaseChar production.

xmlIsBlank^⚠

xmlIsBlank(unsigned int ch) — space, tab, LF, CR.

xmlIsBlankNode^⚠

xmlIsBlankNode(const xmlNode *node) — text/CDATA node with empty or whitespace-only content (tree.c 2.15).

xmlIsChar^⚠

xmlIsChar(unsigned int ch) — XML 1.0 Char production.

xmlIsCombining^⚠

xmlIsCombining(unsigned int ch) — XML 1.0 CombiningChar production.

xmlIsDigit^⚠

xmlIsDigit(unsigned int ch) — XML 1.0 Digit production.

xmlIsExtender^⚠

xmlIsExtender(unsigned int ch) — XML 1.0 Extender production.

xmlIsIdeographic^⚠

xmlIsIdeographic(unsigned int ch) — XML 1.0 Ideographic production.

xmlIsLetter^⚠

xmlIsLetter(int c) — BaseChar or Ideographic (parserInternals.c).

xmlIsPubidChar^⚠

xmlIsPubidChar(unsigned int ch) — PubidChar production (ASCII table).