apex-ruby 1.0.29 → 1.0.31

data/ext/apex_ext/apex_src/src/extensions/emoji.c

@@ -15,6 +15,10 @@
 static void normalize_emoji_name(char *name);
 static int is_table_alignment_pattern(const char *start, const char *end);
 static int is_inside_html_attribute(const char *pos, const char *start);
+static int emoji_shortcode_is_bounded(const char *text_start, const char *open_colon,
+                                      const char *close_colon);
+static int emoji_name_is_all_digits(const char *name, size_t len);
+static int emoji_fuzzy_max_distance(size_t name_len);
 
 /** True if content at p looks like a list marker (- , * , + , or digit+. ) */
 static int looks_like_list_marker(const char *p) {
@@ -40,6 +44,41 @@ static int line_is_indented_code_block(const char *read) {
     return *content && !looks_like_list_marker(content);
 }
 
+/**
+ * True if :name: is a bounded shortcode (not inside a word or digit run).
+ * Left of the opening colon and right of the closing colon must not be alphanumeric.
+ */
+static int emoji_shortcode_is_bounded(const char *text_start, const char *open_colon,
+                                      const char *close_colon) {
+    if (!text_start || !open_colon || !close_colon || open_colon < text_start) return 0;
+    if (open_colon > text_start && isalnum((unsigned char)open_colon[-1])) return 0;
+    if (close_colon[1] != '\0' && isalnum((unsigned char)close_colon[1])) return 0;
+    return 1;
+}
+
+/** True if name is non-empty and consists only of ASCII digits. */
+static int emoji_name_is_all_digits(const char *name, size_t len) {
+    if (!name || len == 0) return 0;
+    for (size_t i = 0; i < len; i++) {
+        if (!isdigit((unsigned char)name[i])) return 0;
+    }
+    return 1;
+}
+
+/**
+ * Cap fuzzy Levenshtein distance by name length. Distance 4 on 2-char tokens
+ * (e.g. "00" -> "100") is what corrupted timecodes. Names shorter than 3
+ * characters are exact-match only; longer names allow roughly one edit per
+ * three characters (so :smlie: can still reach :smile: at distance 2).
+ */
+static int emoji_fuzzy_max_distance(size_t name_len) {
+    if (name_len < 3) return 0; /* exact matches only for very short names */
+    int scaled = (int)((name_len + 2) / 3); /* 3-4 -> 1, 5-7 -> 2, ... */
+    if (scaled < 1) scaled = 1;
+    if (scaled > 4) scaled = 4;
+    return scaled;
+}
+
 /**
  * Find emoji entry by name
  * Returns pointer to emoji_entry or NULL if not found
@@ -286,6 +325,23 @@ char *apex_replace_emoji(const char *html) {
             continue;
         }
 
+        /* Skip emoji inside HTML comments (citation placeholders are
+         * <!--CITE:key-->; a key like "Wayne:" contains ":Wayne:" and would
+         * otherwise fuzzy-match to :wave: → <!--CITE👋-->). */
+        if (read[0] == '<' && read[1] == '!' && read[2] == '-' && read[3] == '-') {
+            const char *comment_end = strstr(read + 4, "-->");
+            if (comment_end) {
+                size_t comment_len = (size_t)(comment_end + 3 - read);
+                if (comment_len <= remaining) {
+                    memcpy(write, read, comment_len);
+                    write += comment_len;
+                    remaining -= comment_len;
+                    read = comment_end + 3;
+                    continue;
+                }
+            }
+        }
+
         /* Check if we're inside an index placeholder <!--IDX:...--> - if so, skip emoji processing */
         if (read >= html + 7 && strncmp(read - 7, "<!--IDX:", 8) == 0) {
             /* Find the end of the placeholder */
@@ -364,7 +420,7 @@ char *apex_replace_emoji(const char *html) {
 
             /* Look for closing : */
             const char *end = strchr(read + 1, ':');
-            if (end && (end - read) < 50) {  /* Reasonable emoji name length */
+            if (end && (end - read) < 50 && emoji_shortcode_is_bounded(html, read, end)) {
                 /* Extract emoji name */
                 int name_len = (int)(end - (read + 1));
                 const char *name_start = read + 1;
@@ -540,7 +596,7 @@ char *apex_replace_emoji_text(const char *text) {
         if (*read == ':') {
             /* Look for closing : */
             const char *end = strchr(read + 1, ':');
-            if (end && (end - read) < 50) {  /* Reasonable emoji name length */
+            if (end && (end - read) < 50 && emoji_shortcode_is_bounded(text, read, end)) {
                 /* Extract emoji name */
                 int name_len = (int)(end - (read + 1));
                 const char *name_start = read + 1;
@@ -696,12 +752,21 @@ static const char *find_best_emoji_match(const char *name, size_t name_len, int
     normalize_emoji_name(normalized);
     size_t normalized_len = strlen(normalized);
 
-    /* Check exact match first */
+    /* Check exact match first (allows :100: / :1234: via exact lookup only) */
     const emoji_entry *exact = find_emoji_entry(normalized, (int)normalized_len);
     if (exact) {
         return exact->name;
     }
 
+    /* Never fuzzy-match all-digit tokens (e.g. "00" -> "100") */
+    if (emoji_name_is_all_digits(normalized, normalized_len)) {
+        return NULL;
+    }
+
+    if (max_distance <= 0) {
+        return NULL;
+    }
+
     /* Find fuzzy matches */
     int best_distance = max_distance + 1;
     size_t best_length = SIZE_MAX;
@@ -795,12 +860,12 @@ char *apex_autocorrect_emoji_names(const char *text) {
         if (*read == ':') {
             /* Look for closing : */
             const char *end = strchr(read + 1, ':');
-            if (end && (end - read) < 50) {  /* Reasonable emoji name length */
+            if (end && (end - read) < 50 && emoji_shortcode_is_bounded(text, read, end)) {
                 /* Extract emoji name */
                 int name_len = (int)(end - (read + 1));
                 const char *name_start = read + 1;
 
-                    /* Validate: must have at least one character and no spaces */
+                /* Validate: must have at least one character and no spaces */
                 if (name_len > 0) {
                     /* Check for spaces in the name */
                     int has_space = 0;
@@ -862,8 +927,9 @@ char *apex_autocorrect_emoji_names(const char *text) {
                             read = end + 1;
                             continue;
                         } else {
-                            /* Try fuzzy matching */
-                            const char *best_match = find_best_emoji_match(name_start, (size_t)name_len, 4);
+                            /* Try fuzzy matching with length-scaled distance */
+                            int max_dist = emoji_fuzzy_max_distance(normalized_len);
+                            const char *best_match = find_best_emoji_match(name_start, (size_t)name_len, max_dist);
                             if (best_match) {
                                 /* Replace with corrected name */
                                 size_t match_len = strlen(best_match);

data/ext/apex_ext/apex_src/src/extensions/html_markdown.c

@@ -7,6 +7,7 @@
 #include "ial.h"
 #include "../html_renderer.h"
 #include "cmark-gfm.h"
+#include "cmark-gfm-extension_api.h"
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -672,7 +673,26 @@ static const char *find_closing_tag(const char *text, const char *tag_name) {
  * If img_attrs is non-NULL, image attributes (e.g. width/height from ref defs) are applied to images in markdown="1" regions.
  * full_doc is the original document used to resolve reference-style link definitions.
  */
-static char *apex_process_html_markdown_impl(const char *text, void *img_attrs, const char *full_doc) {
+static void attach_gfm_extensions(cmark_parser *parser, const apex_options *options) {
+    if (!parser || !options) return;
+    apex_ensure_cmark_extensions();
+    const char *names[4];
+    int count = 0;
+    if (options->enable_tables) names[count++] = "table";
+    if (options->enable_task_lists) names[count++] = "tasklist";
+    if (options->enable_strikethrough) names[count++] = "strikethrough";
+    if (options->enable_autolink &&
+        (options->mode == APEX_MODE_GFM || apex_mode_is_unified_family(options->mode))) {
+        names[count++] = "autolink";
+    }
+    for (int i = 0; i < count; i++) {
+        cmark_syntax_extension *ext = cmark_find_syntax_extension(names[i]);
+        if (ext) cmark_parser_attach_syntax_extension(parser, ext);
+    }
+}
+
+static char *apex_process_html_markdown_impl(const char *text, void *img_attrs, const char *full_doc,
+                                             const apex_options *options) {
     if (!text) return NULL;
     if (!full_doc) full_doc = text;
 
@@ -782,7 +802,7 @@ static char *apex_process_html_markdown_impl(const char *text, void *img_attrs,
 
                 /* Recursively process nested divs with markdown="1" BEFORE parsing */
                 /* This ensures nested divs are processed before cmark-gfm sees them */
-                char *processed_content = apex_process_html_markdown_impl(content, img_attrs, full_doc);
+                char *processed_content = apex_process_html_markdown_impl(content, img_attrs, full_doc, options);
                 if (processed_content) {
                     free(content);
                     content = processed_content;
@@ -801,6 +821,7 @@ static char *apex_process_html_markdown_impl(const char *text, void *img_attrs,
                 int cmark_opts = CMARK_OPT_DEFAULT | CMARK_OPT_UNSAFE | CMARK_OPT_FOOTNOTES;
                 cmark_parser *parser = cmark_parser_new(cmark_opts);
                 if (parser) {
+                    attach_gfm_extensions(parser, options);
                     cmark_parser_feed(parser, content, content_len);
                     cmark_node *doc = cmark_parser_finish(parser);
 
@@ -955,7 +976,7 @@ static char *apex_process_html_markdown_impl(const char *text, void *img_attrs,
     return output;
 }
 
-char *apex_process_html_markdown(const char *text, void *img_attrs) {
-    return apex_process_html_markdown_impl(text, img_attrs, text);
+char *apex_process_html_markdown(const char *text, void *img_attrs, const apex_options *options) {
+    return apex_process_html_markdown_impl(text, img_attrs, text, options);
 }
 

data/ext/apex_ext/apex_src/src/extensions/html_markdown.h

@@ -25,14 +25,17 @@
 extern "C" {
 #endif
 
+#include "../../include/apex/apex.h"
+
 struct image_attr_entry;
 
 /**
  * Process HTML tags with markdown attributes (preprocessing)
  * Returns newly allocated string with markdown content parsed.
  * If img_attrs is non-NULL, image attributes (e.g. width/height from ref defs) are applied to images in markdown="1" regions.
+ * GFM extensions (tables, task lists, strikethrough, autolink) enabled in options also apply inside those regions.
  */
-char *apex_process_html_markdown(const char *text, void *img_attrs);
+char *apex_process_html_markdown(const char *text, void *img_attrs, const apex_options *options);
 
 #ifdef __cplusplus
 }

data/ext/apex_ext/apex_src/src/extensions/insert.c

@@ -40,6 +40,32 @@ static const char *find_ial_after(const char *text, const char **ial_end) {
     return NULL;
 }
 
+/**
+ * If text starts an HTML tag (<name or </name), return a pointer just past its
+ * closing '>' (quote-aware), or NULL. Gives up at a blank line outside quotes
+ * so a stray '<' in prose doesn't swallow the rest of the document.
+ */
+static const char *html_tag_end(const char *text) {
+    if (text[0] != '<') return NULL;
+    const char *p = text + 1;
+    if (*p == '/') p++;
+    if (!isalpha((unsigned char)*p)) return NULL;
+
+    char quote = 0;
+    for (; *p; p++) {
+        if (quote) {
+            if (*p == quote) quote = 0;
+        } else if (*p == '"' || *p == '\'') {
+            quote = *p;
+        } else if (*p == '>') {
+            return p + 1;
+        } else if (*p == '\n' && (p[1] == '\n' || (p[1] == '\r' && p[2] == '\n'))) {
+            return NULL;
+        }
+    }
+    return NULL;
+}
+
 /**
  * Process ++insert++ syntax as preprocessing
  * Converts to <ins>text</ins> before parsing
@@ -61,6 +87,29 @@ char *apex_process_inserts(const char *text) {
     bool in_inline_code = false;
 
     while (*read) {
+        /* Copy HTML tags verbatim so ++ inside attribute values (e.g. base64
+         * data: URIs) is not converted */
+        if (!in_code_block && !in_inline_code && *read == '<') {
+            const char *tag_end = html_tag_end(read);
+            if (tag_end) {
+                size_t tag_len = (size_t)(tag_end - read);
+                if (tag_len >= remaining) {
+                    size_t written = write - output;
+                    capacity = (written + tag_len + 1) * 2;
+                    char *new_output = realloc(output, capacity);
+                    if (!new_output) goto cleanup;
+                    output = new_output;
+                    write = output + written;
+                    remaining = capacity - written;
+                }
+                memcpy(write, read, tag_len);
+                write += tag_len;
+                remaining -= tag_len;
+                read = tag_end;
+                continue;
+            }
+        }
+
         /* Track code blocks (skip processing inside them) */
         if (*read == '`') {
             if (read[1] == '`' && read[2] == '`') {