leptris 1.9.216.0-x86_64-linux → 1.9.220.0-x86_64-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: bd31f0587ebb9098ed318c732ffb69d6b38c4476b5e16c4dd8fc78bcc9f1eba2
-  data.tar.gz: cd6c4d000d2cff045c16ee809e5b359e0f344334e8c567d798d409857687b083
+  metadata.gz: 18237585fdb14e605fffad88e641b4030aee2d7db1f4e8bc3594077da61e090c
+  data.tar.gz: ad306389acd63683fd14fda7e39c874611812956631a85e852aafd1606aca2ae
 SHA512:
-  metadata.gz: '002850579eb358499474e6f8420046962b03d3ac2dd574251dec77928da4685c66e17bd08f0dbda8d4130cf350fb860c90813cdd3d35196e98db5bff1f45279b'
-  data.tar.gz: 3ccce56b40ee17ac63de60655ccc2972d463699a92f6500eb60e55d1dbf2d026604ef86ef30f205a591ba5006657035b55229f8da3d2e1134209581c9088d58a
+  metadata.gz: 807ceb316a225c9f832f080e998c3ddd1e2d1f2e2d9ccc0e042feaf20179b1fcf71157cfd8931747c28cc41f80781390df70c4e31d4d5db53822d9cd1535aa76
+  data.tar.gz: 7cb5959381e7a27940d4a3eb67d86416d2eabb41cd666bd25c38c0fa42ba9c73f5c5011fa1abb4c09f2155323b56502596d8366d829d749358e7f5a933f20caa

data/CHANGELOG.md

@@ -5,6 +5,17 @@ All notable changes to Leptris will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.9.220.0] - 2026-09-22
+
+### Performance — engine sync (1.9.217–1.9.220)
+
+- Vendored libleptris 1.9.220: four perf releases — the HTML
+  parse tag-id classifier stack + in-place borrowed text runs
+  (#1218 slices 1-2), the authoritative content_len text-join
+  pass (#1285 slice 1), and borrowed string-values for
+  `value-of select="."` (#682). No binding-facing surface
+  changes; audit 342/342, suite 770/0.
+
 ## [1.9.216.0] - 2026-09-21
 
 ### Added — the plan-ABI consumer contract (#298 fully bound; engine 1.9.214–216)

data/Rakefile

@@ -8,7 +8,7 @@ RSpec::Core::RakeTask.new(:spec)
 # Pin for `rake compile` and the platform-gem builds. Keep in lockstep
 # with .github/workflows/build.yml (which calls `rake compile`) and the
 # CHANGELOG when libleptris releases.
-LIBLEPTRIS_VERSION = "1.9.216"
+LIBLEPTRIS_VERSION = "1.9.220"
 # Vendored alongside libleptris for fn:normalize-unicode (TODO
 # .restructure/20): built per platform with a RELOCATABLE @rpath
 # install name, loaded by ffi.rb before libleptris so the

data/lib/leptris/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Leptris
-  VERSION = "1.9.216.0"
+  VERSION = "1.9.220.0"
 end

data/vendor-src/README.md

@@ -1,7 +1,7 @@
 Sources vendored in this gem (recompile rights; the ruby
 variant builds them at install via extconf.rb):
 
-- libleptris v1.9.216
+- libleptris v1.9.220
 - utf8proc v2.11.0
 
 Rebuild by hand:

data/vendor-src/libleptris/CHANGELOG.md

@@ -1,5 +1,65 @@
 ## [Unreleased]
 
+## [1.9.220] - 2026-09-21
+
+### Performance
+
+- in-place borrowed text runs (#1218 slice 2) (html)
+
+
+
+## [1.9.219] - 2026-09-21
+
+### Performance
+
+- authoritative content_len in the text-join passes (#1285 slice 1) (dom)
+
+
+
+## [1.9.218] - 2026-09-21
+
+### Performance
+
+- borrowed string-value for value-of select='.' (#682) (xslt)
+
+
+
+## [1.9.217] - 2026-09-21
+
+### Performance
+
+- tag-id classifier stack + single-lookup start tags (#1218) (html)
+
+
+
+## [1.9.217] - 2026-09-21
+
+### Added — plan-ABI additive v1 (descriptor materialization, PR #1275)
+
+Closes #1272, #1273, #1269 — the binding-facing plan upgrades:
+
+- **#1272 attribute-predicate rows**: child rows and attribute plans
+  accept `leptris_attr_predicate` pairs; same-wire siblings partition
+  by predicate (AND across pairs, exclusive routing only when
+  predicates are present — the #1115 multi-namespace behavior is
+  untouched).
+- **#1273 document-order identity**: every result value carries an
+  `order_index` cumulative across rows (strictly increasing in
+  document order); `LEPTRIS_PLAN_FLAG_EMIT_ORDER_SPINE` optionally
+  appends unmatched text runs so interleavings reconstruct without
+  fragment re-parses.
+- **#1269 in-pass types + fusion**: rows carry a `type_tag`
+  (1=int / 2=float / 3=bool) executed during the walk with soft-fail
+  to string; `leptris_plan_materialize` fuses parse→walk→free into
+  one call for the binding hot path.
+- New accessors: `leptris_plan_value_node_kind/_order_index/_int/
+  _float/_bool`. `LEPTRIS_PLAN_ABI_VERSION` stays 1 — all additions
+  are trailing fields, aggregate initializers keep compiling.
+
+Five new specs in `test/abi/test_descriptor.cpp` (Plan1272/1273/
+1269a/1269b); full ctest green across all 24 CI legs.
+
+
 ## [1.9.216] - 2026-09-21
 
 ### Fixed

data/vendor-src/libleptris/CMakeLists.txt

@@ -3,7 +3,7 @@
 cmake_minimum_required(VERSION 3.20)
 
 project(leptris
-    VERSION 1.9.216
+    VERSION 1.9.220
     DESCRIPTION "Fast XML parser and XPath evaluator in pure C"
     LANGUAGES C CXX
 )

data/vendor-src/libleptris/TODO.max-perf/7-validate-bindings.md

@@ -1,6 +1,16 @@
 # Lever 7: Lane 16.6 — leptris validate CLI + bindings
 
-Status: TODO. `leptris validate` subcommand (XML well-formedness
-+ DTD + RNG paths through the public API) and the binding-side
-validate wrappers. Small, closed scope; CLI command registration
-follows cli/CLI_ARCHITECTURE.md (new file in cli/commands/).
+Status: DONE (issue #1183). All three legs shipped and verified on
+main:
+
+- **CLI**: `cli/commands/validate.c` (286 lines) — RELAX NG +
+  Schematron validation, stdin/file input; registered in
+  `cli/main.c:97`, listed in the help text (`leptris validate`).
+- **leptris-ruby**: `RelaxNG#validate/#valid?/#validate_errors/
+  #validate_report`, `Schematron#validate/#valid?`,
+  `DTD#validate/#valid?` (lib/leptris/xml/{relaxng,schematron,dtd}.rb).
+- **leptris-py**: `RelaxNG`, `DTD`, `Schematron` exported from the
+  package root (leptris/__init__.py).
+
+Well-formedness validation is `leptris_parse_string` itself (line:col
+error contract); no separate wrapper required.

data/vendor-src/libleptris/benchmarks/html/bench_html_parse.c

@@ -67,7 +67,28 @@ static size_t gen_entities(char* out, int paras) {
     return len;
 }
 
+/* Table-of-rows shape (#1218): well-formed medium HTML — one big
+ * tbody, no error-recovery cases, matching the serialbench
+ * medium.html fixture class the issue reports 3.4-7x on. */
+static size_t gen_table_rows(char* out, int rows) {
+    size_t len = 0;
+    len += (size_t)sprintf(out + len,
+        "<!DOCTYPE html><html lang='en'><head><meta charset='utf-8'>"
+        "<title>Report</title></head><body>"
+        "<table class='data'><thead><tr><th>Id</th><th>Name</th>"
+        "<th>Value</th></tr></thead><tbody>");
+    for (int r = 0; r < rows; r++)
+        len += (size_t)sprintf(out + len,
+            "<tr class='r%d'><td>%d</td><td>row-%d</td>"
+            "<td data-v='%d'>%d.%02d</td></tr>",
+            r & 1, r, r, r, r, r & 99);
+    len += (size_t)sprintf(out + len,
+        "</tbody></table></body></html>");
+    return len;
+}
+
 int main(void) {
+    const char* only = getenv("HTML_BENCH_ONLY"); /* page|entities|table */
     size_t cap = 1u << 22;
     char* html = (char*)malloc(cap);
     if (!html) return 1;
@@ -80,6 +101,7 @@ int main(void) {
 
     int reps = getenv("HTML_BENCH_REPS") ? atoi(getenv("HTML_BENCH_REPS")) : 9;
     double best = 1e18;
+    if (!only || strcmp(only, "page") == 0) {
     for (int i = 0; i < reps; i++) {
         double t0 = benchmark_time_us();
         LeptrisDocument d = leptris_parse_html_string(html, len, &st);
@@ -89,8 +111,10 @@ int main(void) {
     }
     printf("leptris html parse: %zu bytes, %.0f us, %.1f MB/s (best of %d)\n",
            len, best, (double)len / best, reps);
+    }
 
     /* Entity-laden second shape (nokogiri reference: same page). */
+    if (!only || strcmp(only, "entities") == 0) {
     len = gen_entities(html, 20000);
     warm = leptris_parse_html_string(html, len, &st);
     if (!warm) { printf("entity parse failed\n"); return 1; }
@@ -106,6 +130,26 @@ int main(void) {
     printf("leptris html parse (entity-laden): %zu bytes, %.0f us, "
            "%.1f MB/s (best of %d)\n",
            len, best, (double)len / best, reps);
+    }
+
+    /* Table-of-rows ~2MB shape (#1218, serialbench medium class). */
+    if (!only || strcmp(only, "table") == 0) {
+    len = gen_table_rows(html, 32000);
+    warm = leptris_parse_html_string(html, len, &st);
+    if (!warm) { printf("table parse failed\n"); return 1; }
+    leptris_document_free(warm);
+    best = 1e18;
+    for (int i = 0; i < reps; i++) {
+        double t0 = benchmark_time_us();
+        LeptrisDocument d = leptris_parse_html_string(html, len, &st);
+        double us = benchmark_time_us() - t0;
+        leptris_document_free(d);
+        if (us < best) best = us;
+    }
+    printf("leptris html parse (table-rows): %zu bytes, %.0f us, "
+           "%.1f MB/s (best of %d)\n",
+           len, best, (double)len / best, reps);
+    }
     free(html);
     return 0;
 }

data/vendor-src/libleptris/src/leptris/dom/element.c

@@ -979,7 +979,12 @@ static size_t calculate_text_length_recursive(LeptrisNode* node) {
              * text->content_len directly. */
             const char* content = leptris_text_get_content(text);
             if (content) {
-                len += strlen(content);
+                /* #1285 slice 1: frozen (parse-created,
+                 * immutable) texts carry an authoritative
+                 * content_len — skip the scan. */
+                size_t clen = child->frozen ? text->content_len
+                                            : strlen(content);
+                len += clen;
             }
         } else if (child->type == LEPTRIS_NODE_TYPE_CDATA) {
             LeptrisCDATANode* cdata = (LeptrisCDATANode*)child;
@@ -1013,7 +1018,10 @@ static void copy_text_content_recursive(LeptrisNode* node, char* result, size_t*
             LeptrisTextNode* text = (LeptrisTextNode*)child;
             const char* content = leptris_text_get_content(text);
             if (content) {
-                size_t clen = strlen(content);
+                /* #1285 slice 1: frozen texts — authoritative
+                 * content_len (mirrors the length pass). */
+                size_t clen = child->frozen ? text->content_len
+                                            : strlen(content);
                 memcpy(result + *offset, content, clen);
                 *offset += clen;
             }

data/vendor-src/libleptris/src/leptris/dom/text.c

@@ -102,6 +102,16 @@ const char* leptris_text_get_content(LeptrisTextNode* text) {
     if (!text) return NULL;
 
     if (text->borrowed && text->pool) {
+        /* #1218 slice 2: an in-place terminated run (the parser
+         * wrote the NUL into the doc-owned input copy) is already
+         * materialized — serve it directly when it carries no
+         * entity. A run WITH '&' still takes the expansion path
+         * below (dp-borrowed nodes rely on it). */
+        if (text->content_len > 0 &&
+            text->content[text->content_len] == '\0' &&
+            memchr(text->content, '&', text->content_len) == NULL) {
+            return text->content;
+        }
         /* Entity expansion: if the borrowed content contains '&',
          * expand predefined XML entities (&, <, etc.) and
          * numeric character references (A, B) into a new

data/vendor-src/libleptris/src/leptris/html/html_parse.c

@@ -2121,17 +2121,8 @@ static int h_isalnum(char c) {
     return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
            (c >= '0' && c <= '9');
 }
-/* Void elements: never take children; their start tag is complete. */
-static const char* const k_void[] = {
-    "area", "base", "basefont", "bgsound", "br", "col", "embed",
-    "frame", "hr", "img", "input", "keygen", "link", "meta", "param",
-    "source", "track", "wbr", NULL,
-};
-static int h_is_void(const char* name) {
-    for (int i = 0; k_void[i]; i++)
-        if (strcmp(name, k_void[i]) == 0) return 1;
-    return 0;
-}
+/* Void / frameset-ok-clear classifications live in HTagInfo.
+ * fo_void (bit0 void, bit1 frameset-ok clear) — see h_tag_infos. */
 
 /* Raw-text elements: content runs to the case-insensitive close
  * tag, no markup inside. script/style take entities verbatim. */
@@ -2197,6 +2188,10 @@ typedef struct {
     uint8_t closes;    /* bit0: h_p_closes implied-end set; bit1:
                         * k_p_closers — replaced the per-close-tag
                         * strcmp scans (#1218 profile). */
+    uint8_t fo_void;   /* bit0: void element; bit1: start tag
+                        * clears frameset-ok (13.2.5.4.4) —
+                        * replaced the k_void/k_fo_clear strcmp
+                        * scans (#1218 profile). */
 } HTagInfo;
 
 static const HTagInfo* h_tag_lookup(const char* n);
@@ -2316,6 +2311,21 @@ static char* h_decode_ex(LeptrisMemoryPool* pool, const char* s,
  * returned length is the compacted one; the string is NUL-free). */
 static char* h_decode_body(LeptrisMemoryPool* pool, const char* s,
                            const char* e, int whatwg, size_t* out_len) {
+    size_t len = (size_t)(e - s);
+    /* #1218: entity-free, NUL-free runs — the dominant case on
+     * well-formed pages — skip the entity state machine and the
+     * NUL-strip pass: one bump alloc + memcpy. Identical output
+     * (decode_ex passes non-'&' bytes verbatim; the strip is a
+     * no-op without NULs). */
+    if (len && !memchr(s, '&', len) && !memchr(s, '\0', len)) {
+        char* fast = (char*)leptris_pool_alloc(pool, len + 1);
+        if (!fast) return NULL;
+        memcpy(fast, s, len);
+        fast[len] = 0;
+        if (out_len) *out_len = len;
+        return fast;
+    }
+    {
     size_t n = 0;
     char* d = h_decode_ex(pool, s, e, 0, whatwg, &n);
     if (!d) return NULL;
@@ -2327,6 +2337,7 @@ static char* h_decode_body(LeptrisMemoryPool* pool, const char* s,
     d[w] = 0;
     if (out_len) *out_len = w;
     return d;
+    }
 }
 static size_t h_utf8_encode(uint32_t cp, char* out) {
     if (cp < 0x80) { out[0] = (char)cp; return 1; }
@@ -2725,6 +2736,15 @@ typedef struct {
      * whole child chain, which is quadratic on table/text-heavy
      * pages. Slots are refreshed on push (NULL) and by hb_put. */
     LeptrisNodeRef open_tail[256];
+    /* #1218 slice 2: the doc-owned input copy — borrowed text
+     * nodes point into it. */
+    char* owned;
+    /* #1218: per-slot tag id (h_tag_infos index, 255 unknown) —
+     * the in-select / template-fence / table-context stack walks
+     * become integer compares instead of per-element name fetch +
+     * strcmp chains. Set at push; pops truncate depth, so the
+     * parallel array needs no pop logic. */
+    uint8_t open_id[256];
     size_t depth;
     LeptrisElement root;        /* first top-level element */
     LeptrisNodeRef top_head;    /* top-level chain: text, comments, root */
@@ -3064,90 +3084,90 @@ static int h_is_formatting(const char* n) {
  * no-formatting sets. */
 
 static const HTagInfo h_tag_infos[] = {
-    {"address", 7, 1, 1, 3},
-    {"applet", 6, 1, 0, 0},
-    {"area", 4, 1, 0, 0},
-    {"article", 7, 1, 1, 3},
-    {"aside", 5, 1, 1, 3},
-    {"base", 4, 1, 1, 0},
-    {"basefont", 8, 1, 1, 0},
-    {"bgsound", 7, 1, 1, 0},
-    {"blockquote", 10, 1, 1, 3},
-    {"body", 4, 1, 1, 0},
-    {"br", 2, 1, 0, 0},
-    {"button", 6, 1, 0, 0},
-    {"caption", 7, 1, 1, 0},
-    {"center", 6, 1, 1, 1},
-    {"col", 3, 1, 1, 0},
-    {"colgroup", 8, 1, 1, 0},
-    {"dd", 2, 1, 1, 1},
-    {"details", 7, 1, 1, 3},
-    {"dialog", 6, 1, 1, 3},
-    {"dir", 3, 1, 1, 3},
-    {"div", 3, 1, 1, 3},
-    {"dl", 2, 1, 1, 3},
-    {"dt", 2, 1, 1, 1},
-    {"embed", 5, 1, 0, 0},
-    {"fieldset", 8, 1, 1, 3},
-    {"figcaption", 10, 1, 1, 3},
-    {"figure", 6, 1, 1, 3},
-    {"footer", 6, 1, 1, 3},
-    {"form", 4, 1, 1, 3},
-    {"frame", 5, 1, 1, 0},
-    {"frameset", 8, 1, 1, 0},
-    {"h1", 2, 1, 1, 3},
-    {"h2", 2, 1, 1, 3},
-    {"h3", 2, 1, 1, 3},
-    {"h4", 2, 1, 1, 3},
-    {"h5", 2, 1, 1, 3},
-    {"h6", 2, 1, 1, 3},
-    {"head", 4, 1, 1, 0},
-    {"header", 6, 1, 1, 3},
-    {"hgroup", 6, 1, 1, 3},
-    {"hr", 2, 1, 1, 3},
-    {"html", 4, 1, 1, 0},
-    {"iframe", 6, 1, 0, 0},
-    {"img", 3, 1, 0, 0},
-    {"input", 5, 1, 0, 0},
-    {"keygen", 6, 1, 0, 0},
-    {"li", 2, 1, 1, 3},
-    {"link", 4, 1, 1, 0},
-    {"listing", 7, 1, 1, 1},
-    {"main", 4, 1, 1, 3},
-    {"marquee", 7, 1, 0, 0},
-    {"menu", 4, 1, 1, 3},
-    {"meta", 4, 1, 1, 0},
-    {"nav", 3, 1, 1, 3},
-    {"noembed", 7, 1, 0, 0},
-    {"noframes", 8, 1, 1, 0},
-    {"noscript", 8, 1, 0, 0},
-    {"object", 6, 1, 0, 0},
-    {"ol", 2, 1, 1, 3},
-    {"p", 1, 1, 1, 3},
-    {"param", 5, 1, 0, 0},
-    {"plaintext", 9, 1, 1, 1},
-    {"pre", 3, 1, 1, 3},
-    {"script", 6, 1, 1, 0},
-    {"search", 6, 1, 1, 1},
-    {"section", 7, 1, 1, 3},
-    {"select", 6, 1, 0, 0},
-    {"source", 6, 1, 0, 0},
-    {"style", 5, 1, 1, 0},
-    {"summary", 7, 1, 1, 1},
-    {"table", 5, 1, 1, 2},
-    {"tbody", 5, 1, 1, 0},
-    {"td", 2, 1, 1, 0},
-    {"template", 8, 1, 1, 0},
-    {"textarea", 8, 1, 1, 0},
-    {"tfoot", 5, 1, 1, 0},
-    {"th", 2, 1, 1, 0},
-    {"thead", 5, 1, 1, 0},
-    {"title", 5, 1, 1, 0},
-    {"tr", 2, 1, 1, 0},
-    {"track", 5, 1, 0, 0},
-    {"ul", 2, 1, 1, 3},
-    {"wbr", 3, 1, 0, 0},
-    {"xmp", 3, 1, 0, 1},
+    {"address", 7, 1, 1, 3, 0},
+    {"applet", 6, 1, 0, 0, 2},
+    {"area", 4, 1, 0, 0, 3},
+    {"article", 7, 1, 1, 3, 0},
+    {"aside", 5, 1, 1, 3, 0},
+    {"base", 4, 1, 1, 0, 1},
+    {"basefont", 8, 1, 1, 0, 1},
+    {"bgsound", 7, 1, 1, 0, 1},
+    {"blockquote", 10, 1, 1, 3, 0},
+    {"body", 4, 1, 1, 0, 0},
+    {"br", 2, 1, 0, 0, 3},
+    {"button", 6, 1, 0, 0, 2},
+    {"caption", 7, 1, 1, 0, 0},
+    {"center", 6, 1, 1, 1, 0},
+    {"col", 3, 1, 1, 0, 1},
+    {"colgroup", 8, 1, 1, 0, 0},
+    {"dd", 2, 1, 1, 1, 2},
+    {"details", 7, 1, 1, 3, 0},
+    {"dialog", 6, 1, 1, 3, 0},
+    {"dir", 3, 1, 1, 3, 0},
+    {"div", 3, 1, 1, 3, 0},
+    {"dl", 2, 1, 1, 3, 0},
+    {"dt", 2, 1, 1, 1, 2},
+    {"embed", 5, 1, 0, 0, 3},
+    {"fieldset", 8, 1, 1, 3, 0},
+    {"figcaption", 10, 1, 1, 3, 0},
+    {"figure", 6, 1, 1, 3, 0},
+    {"footer", 6, 1, 1, 3, 0},
+    {"form", 4, 1, 1, 3, 0},
+    {"frame", 5, 1, 1, 0, 1},
+    {"frameset", 8, 1, 1, 0, 0},
+    {"h1", 2, 1, 1, 3, 0},
+    {"h2", 2, 1, 1, 3, 0},
+    {"h3", 2, 1, 1, 3, 0},
+    {"h4", 2, 1, 1, 3, 0},
+    {"h5", 2, 1, 1, 3, 0},
+    {"h6", 2, 1, 1, 3, 0},
+    {"head", 4, 1, 1, 0, 0},
+    {"header", 6, 1, 1, 3, 0},
+    {"hgroup", 6, 1, 1, 3, 0},
+    {"hr", 2, 1, 1, 3, 3},
+    {"html", 4, 1, 1, 0, 0},
+    {"iframe", 6, 1, 0, 0, 2},
+    {"img", 3, 1, 0, 0, 3},
+    {"input", 5, 1, 0, 0, 3},
+    {"keygen", 6, 1, 0, 0, 3},
+    {"li", 2, 1, 1, 3, 2},
+    {"link", 4, 1, 1, 0, 1},
+    {"listing", 7, 1, 1, 1, 2},
+    {"main", 4, 1, 1, 3, 0},
+    {"marquee", 7, 1, 0, 0, 2},
+    {"menu", 4, 1, 1, 3, 0},
+    {"meta", 4, 1, 1, 0, 1},
+    {"nav", 3, 1, 1, 3, 0},
+    {"noembed", 7, 1, 0, 0, 2},
+    {"noframes", 8, 1, 1, 0, 2},
+    {"noscript", 8, 1, 0, 0, 0},
+    {"object", 6, 1, 0, 0, 2},
+    {"ol", 2, 1, 1, 3, 0},
+    {"p", 1, 1, 1, 3, 0},
+    {"param", 5, 1, 0, 0, 1},
+    {"plaintext", 9, 1, 1, 1, 2},
+    {"pre", 3, 1, 1, 3, 2},
+    {"script", 6, 1, 1, 0, 0},
+    {"search", 6, 1, 1, 1, 0},
+    {"section", 7, 1, 1, 3, 0},
+    {"select", 6, 1, 0, 0, 2},
+    {"source", 6, 1, 0, 0, 1},
+    {"style", 5, 1, 1, 0, 0},
+    {"summary", 7, 1, 1, 1, 0},
+    {"table", 5, 1, 1, 2, 2},
+    {"tbody", 5, 1, 1, 0, 0},
+    {"td", 2, 1, 1, 0, 0},
+    {"template", 8, 1, 1, 0, 0},
+    {"textarea", 8, 1, 1, 0, 2},
+    {"tfoot", 5, 1, 1, 0, 0},
+    {"th", 2, 1, 1, 0, 0},
+    {"thead", 5, 1, 1, 0, 0},
+    {"title", 5, 1, 1, 0, 0},
+    {"tr", 2, 1, 1, 0, 0},
+    {"track", 5, 1, 0, 0, 1},
+    {"ul", 2, 1, 1, 3, 0},
+    {"wbr", 3, 1, 0, 0, 3},
+    {"xmp", 3, 1, 0, 1, 2},
 };
 
 /* bucket[first_char] = start index; bucket[first_char+1] = end */
@@ -3157,7 +3177,11 @@ static const uint8_t h_tag_bucket[27] = {
     63, 70, 81, 82, 82, 83, 84, 84, 84,
 };
 
-/* Resolve a (lowercased) tag name's flags; NULL when unknown. */
+/* Resolve a (lowercased) tag name's flags; NULL when unknown.
+ * strncmp first: it stops at n's terminator, so the n[t->len]
+ * terminator probe below only runs when n really has t->len
+ * matching bytes (the old n[t->len]-first form read past short
+ * names — UB clang exploited under new inlining contexts). */
 static const HTagInfo* h_tag_lookup(const char* n) {
     if (!n) return NULL;
     unsigned char c = (unsigned char)n[0];
@@ -3165,13 +3189,47 @@ static const HTagInfo* h_tag_lookup(const char* n) {
     for (int i = h_tag_bucket[c - 'a'];
          i < h_tag_bucket[c - 'a' + 1]; i++) {
         const HTagInfo* t = &h_tag_infos[i];
-        if ((uint8_t)n[t->len] == 0 && t->name[t->len - 1] == n[t->len - 1] &&
-            memcmp(n, t->name, t->len) == 0)
+        if (t->name[1] == n[1] &&
+            strncmp(n, t->name, t->len) == 0 &&
+            (uint8_t)n[t->len] == 0)
             return t;
     }
     return NULL;
 }
 
+/* #1218: tag ids + per-id LUTs. The sampler pinned the strcmp
+ * scans (h_is_void 18 names, h_clears_frameset_ok 26 names) and
+ * the open-stack walks (name fetch + strcmp per element) at the
+ * top of WHATWG self-time on table-heavy pages. ids are
+ * h_tag_infos indexes, so one first-char bucket walk per start
+ * tag serves every later classification of that name. */
+#define H_ID_UNKNOWN 255u
+static uint8_t h_tag_id(const char* n) {
+    const HTagInfo* t = h_tag_lookup(n);
+    return t ? (uint8_t)(t - h_tag_infos) : (uint8_t)H_ID_UNKNOWN;
+}
+
+static uint8_t h_id_select, h_id_template, h_id_table, h_id_tbody,
+    h_id_thead, h_id_tfoot, h_id_tr;
+
+static void h_id_luts_init(void) {
+    static int done;
+    if (done) return;
+    done = 1;
+    h_id_select = h_tag_id("select");
+    h_id_template = h_tag_id("template");
+    h_id_table = h_tag_id("table");
+    h_id_tbody = h_tag_id("tbody");
+    h_id_thead = h_tag_id("thead");
+    h_id_tfoot = h_tag_id("tfoot");
+    h_id_tr = h_tag_id("tr");
+}
+
+static int h_is_void(const char* name) {
+    const HTagInfo* t = h_tag_lookup(name);
+    return t && (t->fo_void & 1);
+}
+
 static int h_is_special_ww(const char* n) {
     const HTagInfo* t = h_tag_lookup(n);
     return t && t->special;
@@ -3287,16 +3345,11 @@ static char* h_decode_foreign(LeptrisMemoryPool* pool, const char* s,
 }
 
 
-/* 13.2.5.4.4: start tags that set frameset-ok to false. */
+/* 13.2.5.4.4: start tags that set frameset-ok to false — HTagInfo
+ * bit (see h_tag_infos). */
 static int h_clears_frameset_ok(const char* n) {
-    static const char* const yes[] = {
-        "pre", "listing", "li", "dd", "dt", "plaintext", "button",
-        "applet", "marquee", "object", "table", "area", "br",
-        "embed", "img", "keygen", "wbr", "input", "hr", "textarea",
-        "xmp", "iframe", "noembed", "noframes", "select", NULL};
-    for (int i = 0; yes[i]; i++)
-        if (strcmp(n, yes[i]) == 0) return 1;
-    return 0;
+    const HTagInfo* t = h_tag_lookup(n);
+    return t && (t->fo_void & 2);
 }
 
 /* 13.2.5.4.4 exception: <input> clears frameset-ok UNLESS its type
@@ -3552,22 +3605,20 @@ static int h_cur_ns(HBuilder* b) {
 
 /* An open <select> swallows svg/math start tags (in-select mode
  * ignores unknown start tags; the html4 entry keeps libxml2's
- * keep-everything shape). */
+ * keep-everything shape). #1218: open_id integer walk. */
 static int h_in_select(HBuilder* b) {
-    for (size_t i = b->depth; i > 0; i--) {
-        const char* on = leptris_element_name(b->open[i - 1]);
-        if (on && strcmp(on, "select") == 0) return 1;
-    }
+    h_id_luts_init();
+    for (size_t i = b->depth; i > 0; i--)
+        if (b->open_id[i - 1] == h_id_select) return 1;
     return 0;
 }
 
 /* #659: index of the nearest open <template> on the stack, or
  * -1. The in-template insertion rules fence on it. */
 static int h_template_idx(HBuilder* b) {
-    for (size_t i = b->depth; i > 0; i--) {
-        const char* n = leptris_element_name(b->open[i - 1]);
-        if (n && h_ieq_raw(n, "template")) return (int)(i - 1);
-    }
+    h_id_luts_init();
+    for (size_t i = b->depth; i > 0; i--)
+        if (b->open_id[i - 1] == h_id_template) return (int)(i - 1);
     return -1;
 }
 
@@ -3713,6 +3764,15 @@ static int h_is_table_context(LeptrisElement e) {
            h_ieq_raw(n, "tr");
 }
 
+/* #1218: stack-slot flavor — integer compare on the push-time id
+ * (identical membership: ids were resolved from the same stored
+ * lowercase names). */
+static int h_id_table_ctx(uint8_t id) {
+    h_id_luts_init();
+    return id == h_id_table || id == h_id_tbody ||
+           id == h_id_thead || id == h_id_tfoot || id == h_id_tr;
+}
+
 /* Insert n into parent's child chain BEFORE `before`. Same surgery
  * pattern as the head-content lift (first_child + sibling links +
  * element child_count). */
@@ -3944,7 +4004,7 @@ static void h_append(HBuilder* b, LeptrisNodeRef n) {
          * table" insertion mode - ordinary content still
          * fosters, hidden inputs insert at spot
          * (html5test-com:20). */
-        int tbl_ctx = h_is_table_context(top);
+        int tbl_ctx = h_id_table_ctx(b->open_id[b->depth - 1]);
         if (!tbl_ctx && b->depth >= 2 &&
             b->open_ns[b->depth - 1] == H_NS_HTML &&
             h_ieq_raw(leptris_element_name(top), "form") &&
@@ -3969,7 +4029,7 @@ static void h_append(HBuilder* b, LeptrisNodeRef n) {
                 int oi = -1;
                 for (int k = (int)b->depth - 1; k >= 0; k--)
                     if (b->open_ns[k] == H_NS_HTML &&
-                        h_is_table_context(b->open[k])) {
+                        h_id_table_ctx(b->open_id[k])) {
                         oi = k;
                     }
                 if (oi >= 0) {
@@ -4102,12 +4162,16 @@ static LeptrisElement h_open_foreign(HBuilder* b, const char* name,
         b->open[b->depth] = e;
         b->open_tail[b->depth] = NULL;
         b->open_ns[b->depth] = (uint8_t)ns;
+        b->open_id[b->depth] = h_tag_id(store);
         b->depth++;
     }
     return e;
 }
 
-static LeptrisElement h_open_element(HBuilder* b, const char* name) {
+/* Hot-path variant (#1218): the start-tag section resolves the
+ * HTagInfo once per token and passes the id down. */
+static LeptrisElement h_open_element_id(HBuilder* b, const char* name,
+                                        uint8_t tid) {
     LeptrisStringView nv = leptris_sv_from_cstr(name);
     LeptrisElement e = leptris_element_create_with_view(nv, b->pool);
     if (!e) return NULL;
@@ -4140,11 +4204,16 @@ static LeptrisElement h_open_element(HBuilder* b, const char* name) {
         b->open[b->depth] = e;
         b->open_tail[b->depth] = NULL;
         b->open_ns[b->depth] = H_NS_HTML;
+        b->open_id[b->depth] = tid;
         b->depth++;
     }
     return e;
 }
 
+static LeptrisElement h_open_element(HBuilder* b, const char* name) {
+    return h_open_element_id(b, name, h_tag_id(name));
+}
+
 /* Case-insensitive C-string compare (ASCII). */
 static int h_ieq_raw(const char* a, const char* bname) {
     if (!a) return 0;
@@ -4341,6 +4410,7 @@ static LeptrisElement h_afe_open_clone(HBuilder* b, LeptrisElement src) {
         b->open[b->depth] = c;
         b->open_tail[b->depth] = NULL;
         b->open_ns[b->depth] = H_NS_HTML;
+        b->open_id[b->depth] = h_tag_id(leptris_element_name(c));
         b->depth++;
     }
     return c;
@@ -4624,6 +4694,7 @@ static LeptrisElement h_open_named(HBuilder* b, const char* name,
         b->open[b->depth] = e;
         b->open_tail[b->depth] = NULL;
         b->open_ns[b->depth] = H_NS_HTML;
+        b->open_id[b->depth] = h_tag_id(name);
         b->depth++;
     }
     return e;
@@ -5303,6 +5374,43 @@ head_end = leptris_node_get_next_sibling(head_end);
 }
 
 /* ---- tokenizer ---- */
+/* #1218 slice 2: build a text node for the run [s, e). Clean runs
+ * (no '&', no NUL) borrow the doc-owned input copy in place — the
+ * terminator byte (the '<' the dispatcher already consumed, or the
+ * buffer's own EOF NUL) becomes the content terminator: zero
+ * copies. Entity- or NUL-bearing runs take the decode path. */
+static LeptrisTextNode* h_text_node(HBuilder* b, const char* s,
+                                    const char* e) {
+    size_t len = (size_t)(e - s);
+    if (b->owned && len && !memchr(s, '&', len) &&
+        !memchr(s, '\0', len)) {
+        /* 13.2.6.4.7: non-whitespace text clears frameset-ok —
+         * the decode path scans the decoded string; the borrow
+         * path scans the raw run (identical for entity-free
+         * runs; FFFD in body counts as non-ws, :21). */
+        if (b->whatwg && b->frameset_ok) {
+            for (const char* c = s; c < e; c++) {
+                if ((unsigned char)c[0] == 0xEF && c + 2 < e &&
+                    (unsigned char)c[1] == 0xBF &&
+                    (unsigned char)c[2] == 0xBD) {
+                    b->frameset_ok = 0;
+                    break;
+                }
+                if (!h_is_ws(*c)) {
+                    b->frameset_ok = 0;
+                    break;
+                }
+            }
+        }
+        *(char*)e = '\0';
+        return leptris_text_create_borrowed(s, len, b->pool);
+    }
+    size_t dlen = 0;
+    char* dec = h_decode_text(b, s, e, &dlen);
+    if (!dec || !*dec) return NULL;
+    return leptris_text_create(dec, dlen, b->pool);
+}
+
 static LeptrisDocument html_parse_shared(
     const char* buf, size_t len, LeptrisStatus* status, int whatwg) {
     if (status) *status = LEPTRIS_OK;
@@ -5325,9 +5433,23 @@ static LeptrisDocument html_parse_shared(
     b.whatwg_head_set = whatwg;
     b.whatwg_foster = whatwg;
     b.whatwg_adopt = whatwg;
+    /* #1218 slice 2: one copy of the input into document memory.
+     * Clean text runs borrow this copy in place (the terminator
+     * byte becomes their NUL) instead of a pool alloc + memcpy
+     * per text node — the pugi-style in-place model. */
+    char* owned = (char*)leptris_pool_alloc(b.pool, len + 1);
+    if (!owned) {
+        if (status) *status = LEPTRIS_ERROR_MEMORY;
+        leptris_document_free(doc);
+        return NULL;
+    }
+    memcpy(owned, buf, len);
+    owned[len] = '\0';
+    b.owned = owned;
+    h_id_luts_init();
 
-    const char* p = buf;
-    const char* end = buf + len;
+    const char* p = owned;
+    const char* end = owned + len;
     const char* text = p;   /* pending text run start */
 
     while (p < end) {
@@ -5345,14 +5467,10 @@ static LeptrisDocument html_parse_shared(
             /* Flush pending text first. */
             if (text < p) {
                 size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
-                    LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                    if (t) {
-                        h_reconstruct(&b);
-                        h_append(&b, (LeptrisNodeRef)t);
-                    }
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) {
+                    h_reconstruct(&b);
+                    h_append(&b, (LeptrisNodeRef)t);
                 }
             }
             if (p + 3 < end && p[2] == '-' && p[3] == '-') {
@@ -5677,13 +5795,9 @@ static LeptrisDocument html_parse_shared(
                 const char* ns2 = p + 2;
                 if (b.whatwg && ns2 >= end) {
                     if (text < p) {
-                        size_t dlen = 0;
-                        char* dec = h_decode_text(&b, text, p, &dlen);
-                        if (dec && *dec) {
-                            LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                            if (t) h_append(&b, (LeptrisNodeRef)t);
-                        }
+                        LeptrisTextNode* t =
+                            h_text_node(&b, text, p);
+                        if (t) h_append(&b, (LeptrisNodeRef)t);
                     }
                     LeptrisTextNode* t =
                         leptris_text_create("</", 2, b.pool);
@@ -5695,13 +5809,9 @@ static LeptrisDocument html_parse_shared(
                 if (b.whatwg && !((*ns2 >= 'a' && *ns2 <= 'z') ||
                                   (*ns2 >= 'A' && *ns2 <= 'Z'))) {
                     if (text < p) {
-                        size_t dlen = 0;
-                        char* dec = h_decode_text(&b, text, p, &dlen);
-                        if (dec && *dec) {
-                            LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                            if (t) h_append(&b, (LeptrisNodeRef)t);
-                        }
+                        LeptrisTextNode* t =
+                            h_text_node(&b, text, p);
+                        if (t) h_append(&b, (LeptrisNodeRef)t);
                     }
                     const char* q2 = ns2;
                     while (q2 < end && *q2 != '>') q2++;
@@ -5715,13 +5825,8 @@ static LeptrisDocument html_parse_shared(
              * dropped too (same 13.2.5.44 rule). */
             if (b.whatwg && !memchr(p, '>', (size_t)(end - p))) {
                 if (text < p) {
-                    size_t dlen = 0;
-                    char* dec = h_decode_text(&b, text, p, &dlen);
-                    if (dec && *dec) {
-                        LeptrisTextNode* t =
-                            leptris_text_create(dec, dlen, b.pool);
-                        if (t) h_append(&b, (LeptrisNodeRef)t);
-                    }
+                    LeptrisTextNode* t = h_text_node(&b, text, p);
+                    if (t) h_append(&b, (LeptrisNodeRef)t);
                 }
                 p = end;
                 text = p;
@@ -5746,26 +5851,22 @@ static LeptrisDocument html_parse_shared(
              * whitespace-only run in a table context skips the
              * reconstruct (13.2.6.4.9 - tricky01:6). */
             if (text < p) {
-                size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) {
+                    const char* c = leptris_text_get_content(t);
                     int tbl_ws = b.whatwg_foster &&
                                  b.depth > 0 &&
-                                 h_is_table_context(
-                                     b.open[b.depth - 1]);
+                                 h_id_table_ctx(
+                                     b.open_id[b.depth - 1]);
                     if (tbl_ws)
-                        for (const char* w2 = dec; *w2; w2++)
+                        for (const char* w2 = c; *w2; w2++)
                             if (*w2 != ' ' && *w2 != '\t' &&
                                 *w2 != '\n' && *w2 != '\r') {
                                 tbl_ws = 0;
                                 break;
                             }
-                    LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                    if (t) {
-                        if (!tbl_ws) h_reconstruct(&b);
-                        h_append(&b, (LeptrisNodeRef)t);
-                    }
+                    if (!tbl_ws) h_reconstruct(&b);
+                    h_append(&b, (LeptrisNodeRef)t);
                 }
             }
             /* #659 tests26:17-20 - an HTML end tag with the foreign
@@ -6106,17 +6207,11 @@ static LeptrisDocument html_parse_shared(
                         if (!head_found2 && !b.head_tag_seen &&
                             !b.body_seen && !b.frameset) {
                             if (text < p) {
-                                size_t dlen = 0;
-                                char* dec =
-                                    h_decode_text(&b, text, p, &dlen);
-                                if (dec && *dec) {
-                                    LeptrisTextNode* t =
-                                        leptris_text_create(
-                                            dec, dlen, b.pool);
-                                    if (t)
-                                        h_append(&b,
-                                                 (LeptrisNodeRef)t);
-                                }
+                                LeptrisTextNode* t =
+                                    h_text_node(&b, text, p);
+                                if (t)
+                                    h_append(&b,
+                                             (LeptrisNodeRef)t);
                             }
                             b.head_tag_seen = 1;
                             b.head_end_seen = 1;
@@ -6446,14 +6541,10 @@ static LeptrisDocument html_parse_shared(
              * trailing '?' — content runs to the first '>'. */
             if (text < p) {
                 size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
-                    LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                    if (t) {
-                        h_reconstruct(&b);
-                        h_append(&b, (LeptrisNodeRef)t);
-                    }
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) {
+                    h_reconstruct(&b);
+                    h_append(&b, (LeptrisNodeRef)t);
                 }
             }
             const char* ts = p + 2;   /* skip "<?" */
@@ -6531,13 +6622,8 @@ static LeptrisDocument html_parse_shared(
          * parsing ends (webkit01:4: '<di' -> empty body). */
         if (b.whatwg && !memchr(p, '>', (size_t)(end - p))) {
             if (text < p) {
-                size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
-                    LeptrisTextNode* t =
-                        leptris_text_create(dec, dlen, b.pool);
-                    if (t) h_append(&b, (LeptrisNodeRef)t);
-                }
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) h_append(&b, (LeptrisNodeRef)t);
             }
             p = end;
             text = p;
@@ -6576,13 +6662,9 @@ static LeptrisDocument html_parse_shared(
             }
             if (!closed) {
                 if (text < p) {
-                    size_t dlen = 0;
-                    char* dec = h_decode_text(&b, text, p, &dlen);
-                    if (dec && *dec) {
-                        LeptrisTextNode* t = leptris_text_create(
-                            dec, dlen, b.pool);
-                        if (t) h_append(&b, (LeptrisNodeRef)t);
-                    }
+                    LeptrisTextNode* t =
+                        h_text_node(&b, text, p);
+                    if (t) h_append(&b, (LeptrisNodeRef)t);
                 }
                 p = end;
                 text = p;
@@ -6598,12 +6680,19 @@ static LeptrisDocument html_parse_shared(
          * reprocessed (tests1:90). */
         if (b.whatwg && strcmp(name, "image") == 0)
             name = h_pooled_lower(b.pool, "img", 3);
+        /* #1218: ONE classification per start tag — every
+         * downstream consumer (frameset-ok, void, open) reads the
+         * resolved entry instead of re-walking the bucket. */
+        const HTagInfo* tinfo = h_tag_lookup(name);
+        uint8_t tid = tinfo ? (uint8_t)(tinfo - h_tag_infos)
+                            : (uint8_t)H_ID_UNKNOWN;
         /* 13.2.5.4.4: the enumerated start tags clear frameset-ok;
          * <input type=hidden> does not (webkit01:51). */
         b.input_hidden_tag =
             b.whatwg && nlen == 5 && strcmp(name, "input") == 0 &&
             h_input_type_hidden(q, end);
-        if (b.whatwg && b.frameset_ok && h_clears_frameset_ok(name) &&
+        if (b.whatwg && b.frameset_ok && tinfo &&
+            (tinfo->fo_void & 2) &&
             !(nlen == 5 && strcmp(name, "input") == 0 &&
               h_input_type_hidden(q, end))) {
             b.frameset_ok = 0;
@@ -6676,13 +6765,8 @@ static LeptrisDocument html_parse_shared(
              * after-</head> whitespace must land on the chain to
              * become an html child (webkit01:35). */
             if (text < p) {
-                size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
-                    LeptrisTextNode* t =
-                        leptris_text_create(dec, dlen, b.pool);
-                    if (t) h_append(&b, (LeptrisNodeRef)t);
-                }
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) h_append(&b, (LeptrisNodeRef)t);
             }
             if (strcmp(name, "body") == 0) {
                 b.body_tag_seen = 1;
@@ -6750,13 +6834,9 @@ static LeptrisDocument html_parse_shared(
             } else if (!(b.frameset && b.depth > 0)) {
                 /* Dropped token: flush pending text first. */
                 if (text < p) {
-                    size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                    if (dec && *dec) {
-                        LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                        if (t) h_append(&b, (LeptrisNodeRef)t);
-                    }
+                    LeptrisTextNode* t =
+                        h_text_node(&b, text, p);
+                    if (t) h_append(&b, (LeptrisNodeRef)t);
                 }
                 while (q < end && *q != '>') q++;
                 p = (q < end) ? q + 1 : end;
@@ -6782,13 +6862,8 @@ static LeptrisDocument html_parse_shared(
              * (html5lib template.dat:64-67 — attrs do NOT merge
              * onto the outer elements). */
             if (text < p) {
-                size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
-                    LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                    if (t) h_append(&b, (LeptrisNodeRef)t);
-                }
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) h_append(&b, (LeptrisNodeRef)t);
             }
             while (q < end && *q != '>') q++;
             p = (q < end) ? q + 1 : end;
@@ -6808,21 +6883,16 @@ static LeptrisDocument html_parse_shared(
             strcmp(name, "noframes") != 0 &&
             strcmp(name, "html") != 0) {
             if (text < p) {
-                size_t dlen = 0;
-                    char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) {
+                    const char* c = leptris_text_get_content(t);
                     int ws = 1;
-                    for (const char* q2 = dec; *q2; q2++)
+                    for (const char* q2 = c; *q2; q2++)
                         if (!h_is_ws(*q2)) {
                             ws = 0;
                             break;
                         }
-                    if (ws) {
-                        LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                        if (t)
-                            h_append(&b, (LeptrisNodeRef)t);
-                    }
+                    if (ws) h_append(&b, (LeptrisNodeRef)t);
                 }
             }
             while (q < end && *q != '>') q++;
@@ -6833,31 +6903,27 @@ static LeptrisDocument html_parse_shared(
 
         /* Flush pending text before the element. */
         if (text < p) {
-            size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-            if (dec && *dec) {
+            LeptrisTextNode* t = h_text_node(&b, text, p);
+            if (t) {
                 /* 13.2.6.4.9: a WHITESPACE-ONLY run in a table
                  * context inserts into the current node WITHOUT
                  * the in-body reconstruct - no formatting clone
                  * wraps it (tricky01:6: the ws between
                  * </center> and <img> goes into the table). */
+                const char* c = leptris_text_get_content(t);
                 int tbl_ws = b.whatwg_foster && b.depth > 0 &&
-                             h_is_table_context(
-                                 b.open[b.depth - 1]);
+                             h_id_table_ctx(
+                                 b.open_id[b.depth - 1]);
                 if (tbl_ws) {
-                    for (const char* w2 = dec; *w2; w2++)
+                    for (const char* w2 = c; *w2; w2++)
                         if (*w2 != ' ' && *w2 != '\t' &&
                             *w2 != '\n' && *w2 != '\r') {
                             tbl_ws = 0;
                             break;
                         }
                 }
-                LeptrisTextNode* t =
-                    leptris_text_create(dec, dlen, b.pool);
-                if (t) {
-                    if (!tbl_ws) h_reconstruct(&b);
-                    h_append(&b, (LeptrisNodeRef)t);
-                }
+                if (!tbl_ws) h_reconstruct(&b);
+                h_append(&b, (LeptrisNodeRef)t);
             }
         }
 
@@ -7015,13 +7081,8 @@ static LeptrisDocument html_parse_shared(
          * template.dat:64-67) and in-frameset. */
         if (b.whatwg && strcmp(name, "html") == 0 && b.html_seen) {
             if (text < p) {
-                size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
-                    LeptrisTextNode* t = leptris_text_create(dec, dlen,
-                                                             b.pool);
-                    if (t) h_append(&b, (LeptrisNodeRef)t);
-                }
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) h_append(&b, (LeptrisNodeRef)t);
             }
             h_stash_attrs(&b, q, end, b.html_attrs, &b.html_attr_n);
             while (q < end && *q != '>') q++;
@@ -7367,13 +7428,8 @@ static LeptrisDocument html_parse_shared(
         }
         if (tmpl_act == 1) {
             if (text < p) {
-                size_t dlen = 0;
-                char* dec = h_decode_text(&b, text, p, &dlen);
-                if (dec && *dec) {
-                    LeptrisTextNode* t = leptris_text_create(
-                        dec, dlen, b.pool);
-                    if (t) h_append(&b, (LeptrisNodeRef)t);
-                }
+                LeptrisTextNode* t = h_text_node(&b, text, p);
+                if (t) h_append(&b, (LeptrisNodeRef)t);
             }
             while (q < end && *q != '>') q++;
             p = (q < end) ? q + 1 : end;
@@ -7640,7 +7696,7 @@ static LeptrisDocument html_parse_shared(
 
         LeptrisElement e = (elem_ns != H_NS_HTML)
                                ? h_open_foreign(&b, name, elem_ns)
-                               : h_open_element(&b, name);
+                               : h_open_element_id(&b, name, tid);
         if (!e) goto done;
         if (strcmp(name, "html") == 0) {
             b.html_seen = 1;
@@ -7867,7 +7923,7 @@ static LeptrisDocument html_parse_shared(
          * their entry; applet/object/marquee/td/th/caption open a
          * marker scope (13.2.4.3). */
         if (b.whatwg_adopt && elem_ns == H_NS_HTML &&
-            !(self_closing || h_is_void(name))) {
+            !(self_closing || (tinfo && (tinfo->fo_void & 1)))) {
             if (h_is_formatting(name)) {
                 h_afe_push(&b, e);
             } else if (strcmp(name, "applet") == 0 ||
@@ -7884,7 +7940,7 @@ static LeptrisDocument html_parse_shared(
          * HTML elements (parse error, div stays open -
          * webkit01:46). Only foreign self-closing elements pop.
          * html4 keeps libxml2's honored-slash shape. */
-        if (h_is_void(name) ||
+        if ((tinfo && (tinfo->fo_void & 1)) ||
             (self_closing &&
              (elem_ns != H_NS_HTML || !b.whatwg)))
             b.depth--;
@@ -7939,16 +7995,16 @@ static LeptrisDocument html_parse_shared(
 
     /* Trailing text. */
     if (text < end) {
-        size_t dlen = 0;
-        char* dec = h_decode_text(&b, text, end, &dlen);
-        if (dec && *dec) {
+        LeptrisTextNode* t = h_text_node(&b, text, end);
+        if (t) {
             /* #659 after </html> with a frameset body, non-ws text
              * reprocesses into the frameset and drops (13.2.6.4.18,
              * tests19:41/42). Whitespace stays an html child after
              * the frameset (13.2.6.4.19, tests6:46). */
+            const char* c = leptris_text_get_content(t);
             int drop = 0;
             if (b.whatwg && b.after_html && b.frameset) {
-                for (const char* w = dec; *w; w++)
+                for (const char* w = c; *w; w++)
                     if (*w != ' ' && *w != '\t' &&
                         *w != '\n' && *w != '\r') {
                         drop = 1;
@@ -7956,12 +8012,8 @@ static LeptrisDocument html_parse_shared(
                     }
             }
             if (!drop) {
-                LeptrisTextNode* t =
-                    leptris_text_create(dec, dlen, b.pool);
-                if (t) {
-                    if (!b.post_table_text) h_reconstruct(&b);
-                    h_append(&b, (LeptrisNodeRef)t);
-                }
+                if (!b.post_table_text) h_reconstruct(&b);
+                h_append(&b, (LeptrisNodeRef)t);
             }
         }
     }

data/vendor-src/libleptris/src/leptris/xslt/xslt_exec.c

@@ -1209,6 +1209,30 @@ static int op_value_of(XsltExec* ex, const XsltInstr* in,
      * kind, markers stripped). */
     char* sv = NULL;
     if (in->select_is_dot) {
+        /* #682 borrow fast path: a TEXT context node, or an ELEMENT
+         * whose entire content is one text child, has its string-
+         * value ALREADY materialized — append that pointer instead
+         * of get_node_text's malloc + two-pass join + copy. op_text
+         * copies synchronously, so the borrow never outlives this
+         * call. */
+        int ty = ((LeptrisNode*)node)->type;
+        const char* borrowed = NULL;
+        if (ty == LEPTRIS_NODE_TYPE_TEXT) {
+            borrowed = leptris_text_get_content((LeptrisTextNode*)node);
+        } else if (ty == LEPTRIS_NODE_TYPE_ELEMENT) {
+            LeptrisNodeRef fc = leptris_elem_first_child(node);
+            if (fc && ((LeptrisNode*)fc)->type == LEPTRIS_NODE_TYPE_TEXT &&
+                !leptris_node_get_next_sibling(fc))
+                borrowed =
+                    leptris_text_get_content((LeptrisTextNode*)fc);
+        }
+        if (borrowed) {
+            op_text(ex, &(XsltInstr){ .kind = XSLT_INSTR_TEXT,
+                                      .text = (char*)borrowed,
+                                      .doe = in->doe },
+                    node);
+            return 0;
+        }
         extern char* get_node_text(void* n);
         sv = get_node_text(node);
     } else if (in->select_attr_name) {

data/vendor-src/libleptris/vcpkg.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://raw.githubusercontent.com/microsoft/vcpkg/master/scripts/vcpkg.schema.json",
   "name": "leptris",
-  "version": "1.9.216",
+  "version": "1.9.220",
   "description": "High-performance XML parser and XPath 1.0 engine in C (libleptris)",
   "homepage": "https://github.com/leptris/leptris",
   "license": "MIT",

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: leptris
 version: !ruby/object:Gem::Version
-  version: 1.9.216.0
+  version: 1.9.220.0
 platform: x86_64-linux
 authors:
 - Ribose Inc.