apex-ruby 1.0.29 → 1.0.31

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

@@ -1198,9 +1198,9 @@ char *apex_inject_table_attributes(const char *html, cmark_node *document, int c
             /* Check if this row should be in tfoot.
              * tfoot_rows are stored with AST row indices, so we need to check ast_row_idx.
              *
-             * CRITICAL: A row should only be in tfoot if it comes AFTER the === row in HTML.
-             * Even if a row is marked as tfoot in the AST, if it appears before the === row
-             * in the HTML output (because the === row is skipped), it must be in tbody. */
+             * A row is in tfoot only when the AST marked it and it comes AFTER the ===
+             * marker row. Do not use HTML-position fudge factors — those incorrectly
+             * forced the first footer row into tbody for normal GFM tables. */
             current_row_is_tfoot = false;
 
             /* First, find the === row's AST index */
@@ -1215,8 +1215,29 @@ char *apex_inject_table_attributes(const char *html, cmark_node *document, int c
                     }
                 }
                 if (eq_total > 0 && eq_total == eq_removed) {
-                    min_equals_row_idx = r;
-                    break; /* Found the first === row */
+                    /* Prefer rows that look like === markers when attrs are available */
+                    bool looks_like_equals = false;
+                    if (attrs) {
+                        for (cell_attr *a = attrs; a; a = a->next) {
+                            if (a->table_index != table_idx || a->row_index != r || !a->cell_text)
+                                continue;
+                            const char *text = a->cell_text;
+                            while (*text && isspace((unsigned char)*text)) text++;
+                            if (text[0] == '=' && text[1] == '=' && text[2] == '=') {
+                                looks_like_equals = true;
+                                break;
+                            }
+                        }
+                    }
+                    if (looks_like_equals || !attrs) {
+                        min_equals_row_idx = r;
+                        break;
+                    }
+                    /* Fully-removed non-equals row (e.g. em-dash helper): keep looking */
+                    if (min_equals_row_idx < 0) {
+                        /* fall through and continue search for a real === row */
+                    }
+                    continue;
                 }
             }
 
@@ -1229,113 +1250,13 @@ char *apex_inject_table_attributes(const char *html, cmark_node *document, int c
                 }
             }
 
-            /* CRITICAL: Before checking tfoot marking, verify HTML position.
-             * If this row appears before the === row in HTML, it MUST be in tbody,
-             * regardless of AST marking. This check runs even if the row is not marked as tfoot,
-             * to handle cases where the row might be incorrectly processed. */
-            if (min_equals_row_idx >= 0) {
-                /* Calculate how many HTML rows appear before the === row */
-                int html_rows_before_equals = -1;  /* Start at -1, will be 0 for header */
-                for (int r = 0; r < 100 && r <= min_equals_row_idx; r++) {
-                    bool has_non_removed = false;
-                    for (all_cell *c = all_cells; c; c = c->next) {
-                        if (c->table_index == table_idx &&
-                            c->row_index == r &&
-                            !c->is_removed) {
-                            has_non_removed = true;
-                            break;
-                        }
-                    }
-                    if (has_non_removed) {
-                        html_rows_before_equals++;
-                    }
-                }
-
-                /* If this row's HTML position is before the === row, force it to tbody.
-                 * Since the === row is skipped, rows with row_idx <= html_rows_before_equals + 1
-                 * appear before the === row in HTML.
-                 *
-                 * CRITICAL: We must set current_row_is_tfoot = false BEFORE the skip check,
-                 * so that rows forced to tbody are not skipped. */
-                if (html_rows_before_equals >= 0 && row_idx <= html_rows_before_equals + 1) {
-                    current_row_is_tfoot = false;
-                } else if (ast_row_idx >= 0 && ast_row_idx <= min_equals_row_idx) {
-                    /* Also check AST position as a fallback */
-                    current_row_is_tfoot = false;
-                }
-            }
-
-            /* If marked as tfoot, verify it actually comes after === in HTML */
-            if (is_marked_tfoot && min_equals_row_idx >= 0) {
-                /* Calculate how many HTML rows appear before the === row.
-                 * Count non-removed AST rows up to min_equals_row_idx.
-                 * This gives us the HTML row index of the last row that appears BEFORE the === row. */
-                int html_rows_before_equals = -1;  /* Start at -1, will be 0 for header */
-                for (int r = 0; r < 100 && r <= min_equals_row_idx; r++) {
-                    bool has_non_removed = false;
-                    for (all_cell *c = all_cells; c; c = c->next) {
-                        if (c->table_index == table_idx &&
-                            c->row_index == r &&
-                            !c->is_removed) {
-                            has_non_removed = true;
-                            break;
-                        }
-                    }
-                    if (has_non_removed) {
-                        html_rows_before_equals++;
-                    }
-                }
-
-                /* CRITICAL: The issue is that rows with AST index > min_equals_row_idx can still
-                 * appear in HTML before the === row (since === is skipped). So we need to check
-                 * if this row's HTML position (row_idx) is <= html_rows_before_equals.
-                 *
-                 * The key insight: html_rows_before_equals is the HTML row index of the last row
-                 * that appears BEFORE the === row. So if row_idx <= html_rows_before_equals,
-                 * this row appears before === in HTML, so it must be in tbody.
-                 *
-                 * But we also need to check AST position as a fallback, because the HTML position
-                 * calculation might be off if the row mapping is wrong. */
-                bool force_to_tbody = false;
-
-                /* First check: AST position */
-                if (ast_row_idx <= min_equals_row_idx) {
-                    /* AST says it's before or at ===, so it must be in tbody */
-                    force_to_tbody = true;
-                }
-
-                /* Second check: HTML position
-                 * Since the === row is skipped in HTML, rows with row_idx <= html_rows_before_equals + 2
-                 * appear before the === row in HTML. The +2 accounts for:
-                 * - html_rows_before_equals is the count of rows before === (including header)
-                 * - row_idx is 1-based (1=header, 2=first data, 3=second data, 4===, 5=footer)
-                 * - So row_idx <= html_rows_before_equals + 2 covers the first two data rows. */
-                if (html_rows_before_equals >= 0 && row_idx <= html_rows_before_equals + 2) {
-                    /* HTML position says it's before ===, so it must be in tbody */
-                    force_to_tbody = true;
-                }
-
-                /* Third check: If row_idx <= 3, it's definitely in tbody (header + first two data rows) */
-                if (row_idx <= 3) {
-                    force_to_tbody = true;
-                }
-
-                if (force_to_tbody) {
-                    current_row_is_tfoot = false;
+            if (is_marked_tfoot) {
+                if (min_equals_row_idx >= 0) {
+                    /* Footer content is strictly after the === marker row */
+                    current_row_is_tfoot = (ast_row_idx > min_equals_row_idx);
                 } else {
-                    /* AST and HTML both say it's after ===, so mark as tfoot */
-                    current_row_is_tfoot = true;
-                }
-            } else if (is_marked_tfoot) {
-                /* No === row found, but row is marked as tfoot - use AST marking */
-                /* BUT: Even if no === row is found, if this row's HTML position suggests
-                 * it should be in tbody (e.g., it's one of the first few rows), don't mark as tfoot.
-                 * This handles edge cases where the === row might not be detected correctly. */
-                if (row_idx <= 2) {
-                    /* If this is one of the first few rows, it's probably in tbody */
-                    current_row_is_tfoot = false;
-                } else {
-                    current_row_is_tfoot = true;
+                    /* === row not found as fully-removed; trust AST except header */
+                    current_row_is_tfoot = (ast_row_idx > 0);
                 }
             }
 
@@ -1420,25 +1341,13 @@ char *apex_inject_table_attributes(const char *html, cmark_node *document, int c
                 }
             }
 
-            /* CRITICAL SAFEGUARD: If this is one of the first few rows (row_idx <= 3), it's
-             * almost certainly in tbody, not tfoot. Do not skip via tfoot/=== heuristics.
-             * Fully-removed rows stay skipped (row_fully_removed).
-             * Note: row_idx is incremented before this check (line 894), so:
-             * - row_idx 1 = header
-             * - row_idx 2 = first data row
-             * - row_idx 3 = second data row (this is the one we were missing!)
-             * - row_idx 4 = === row (should be skipped)
-             * - row_idx 5 = footer
-             *
-             * This check must run FIRST, before any other skip logic, to ensure these rows
-             * are always rendered. */
+            /* Keep early data rows from being skipped by removal heuristics.
+             * Do not override tfoot classification here — a short table can have
+             * its footer at a low HTML row index once the === marker is omitted. */
             bool force_keep_row = false;
-            if (!row_fully_removed && row_idx <= 3) {
-                /* First few rows (header + first two data rows) are always in tbody - never skip them */
+            if (!row_fully_removed && row_idx <= 3 && !current_row_is_tfoot) {
                 force_keep_row = true;
                 should_skip_row = false;
-                /* Always set current_row_is_tfoot = false for these rows - they're definitely in tbody */
-                current_row_is_tfoot = false;
             }
 
             /* CRITICAL: Only check if row should be skipped if we haven't already determined

data/ext/apex_ext/apex_src/src/html_renderer.c

@@ -1291,10 +1291,14 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
     }
     header_map = reversed;
 
-    /* Process HTML to inject IDs */
+    /* Process HTML to inject IDs / optional permalink anchors.
+     * Anchor tags are much larger than id="" attributes; under-estimating
+     * capacity used to silently drop the HTML tail (For_Brett: lost items
+     * 144–145 when header_anchors was on and the Annotations pad was stripped). */
     size_t html_len = strlen(html);
-    size_t capacity = html_len + header_count * 100;  /* Extra space for IDs */
-    char *output = malloc(capacity + 1);  /* +1 for null terminator */
+    size_t per_header = use_anchors ? 220 : 100;
+    size_t capacity = html_len + header_count * per_header + 64;
+    char *output = malloc(capacity);
     if (!output) {
         /* Clean up */
         while (header_map) {
@@ -1309,7 +1313,27 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
 
     const char *read = html;
     char *write = output;
-    size_t remaining = capacity;  /* Reserve 1 byte for null terminator */
+    size_t remaining = capacity;
+
+    #define APEX_HEADER_ID_ENSURE(needed) do { \
+        size_t _need = (size_t)(needed); \
+        if (remaining < _need + 1) { \
+            size_t _used = (size_t)(write - output); \
+            size_t _new_cap = (_used + _need + 1) * 2; \
+            if (_new_cap < _used + _need + 64) { \
+                _new_cap = _used + _need + 64; \
+            } \
+            char *_new_out = realloc(output, _new_cap); \
+            if (!_new_out) { \
+                *write = '\0'; \
+                goto apex_inject_header_ids_done; \
+            } \
+            output = _new_out; \
+            write = output + _used; \
+            remaining = _new_cap - _used; \
+            capacity = _new_cap; \
+        } \
+    } while (0)
 
     while (*read) {
         /* Look for header opening tags: <h1>, <h2>, etc. */
@@ -1323,12 +1347,9 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
             while (*tag_end && *tag_end != '>') tag_end++;
             if (*tag_end != '>') {
                 /* Malformed tag, just copy */
-                if (remaining > 0) {
-                    *write++ = *read++;
-                    remaining--;
-                } else {
-                    read++;
-                }
+                APEX_HEADER_ID_ENSURE(1);
+                *write++ = *read++;
+                remaining--;
                 continue;
             }
 
@@ -1404,11 +1425,10 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
             if (use_anchors && header && header->id) {
                 /* For anchor tags: copy the entire header tag, then inject anchor after '>' */
                 size_t tag_len = tag_end - tag_start + 1;  /* Include '>' */
-                if (tag_len <= remaining) {
-                    memcpy(write, tag_start, tag_len);
-                    write += tag_len;
-                    remaining -= tag_len;
-                }
+                APEX_HEADER_ID_ENSURE(tag_len);
+                memcpy(write, tag_start, tag_len);
+                write += tag_len;
+                remaining -= tag_len;
                 read = tag_end + 1;
 
                 /* Inject anchor tag after the header tag */
@@ -1417,21 +1437,19 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
                         "<a href=\"#%s\" aria-hidden=\"true\" class=\"anchor\" id=\"%s\"></a>",
                         header->id, header->id);
                 size_t anchor_len = strlen(anchor_tag);
-                if (anchor_len <= remaining) {
-                    memcpy(write, anchor_tag, anchor_len);
-                    write += anchor_len;
-                    remaining -= anchor_len;
-                }
+                APEX_HEADER_ID_ENSURE(anchor_len);
+                memcpy(write, anchor_tag, anchor_len);
+                write += anchor_len;
+                remaining -= anchor_len;
             } else if (!use_anchors && header && header->id) {
                 /* For header IDs: replace existing ID or inject new one */
                 if (has_id && id_attr) {
                     /* Replace existing ID: copy up to id=, skip old ID value, inject new ID, copy rest */
                     size_t before_id_len = id_attr - tag_start;
-                    if (before_id_len <= remaining) {
-                        memcpy(write, tag_start, before_id_len);
-                        write += before_id_len;
-                        remaining -= before_id_len;
-                    }
+                    APEX_HEADER_ID_ENSURE(before_id_len);
+                    memcpy(write, tag_start, before_id_len);
+                    write += before_id_len;
+                    remaining -= before_id_len;
 
                     /* Find the end of the old ID attribute value */
                     const char *old_id_end = id_attr + 3; /* After 'id=' */
@@ -1452,33 +1470,26 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
                     char id_attr_str[512];
                     snprintf(id_attr_str, sizeof(id_attr_str), "id=\"%s\"", header->id);
                     size_t id_len = strlen(id_attr_str);
-                    if (id_len <= remaining) {
-                        memcpy(write, id_attr_str, id_len);
-                        write += id_len;
-                        remaining -= id_len;
-                    }
+                    APEX_HEADER_ID_ENSURE(id_len);
+                    memcpy(write, id_attr_str, id_len);
+                    write += id_len;
+                    remaining -= id_len;
 
                     /* Copy rest of tag from after old ID until '>' */
                     read = old_id_end;
                     while (read < tag_end && *read != '>') {
-                        if (remaining > 0) {
-                            *write++ = *read++;
-                            remaining--;
-                        } else {
-                            read++;
-                        }
+                        APEX_HEADER_ID_ENSURE(1);
+                        *write++ = *read++;
+                        remaining--;
                     }
 
                     /* Copy closing '>' */
                     if (read < tag_end && *read == '>') {
-                        if (remaining > 0) {
-                            *write++ = *read++;
-                            remaining--;
-                        } else {
-                    read++;
+                        APEX_HEADER_ID_ENSURE(1);
+                        *write++ = *read++;
+                        remaining--;
                     }
-                }
-            } else {
+                } else {
                     /* No existing ID: copy tag up to '>', inject id attribute, then copy '>' */
                     const char *after_tag_name = tag_start + 3;
                     while (*after_tag_name && *after_tag_name != '>' && !isspace((unsigned char)*after_tag_name)) {
@@ -1487,11 +1498,10 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
 
                     /* Copy '<hN' */
                     size_t tag_prefix_len = after_tag_name - tag_start;
-                    if (tag_prefix_len <= remaining) {
-                        memcpy(write, tag_start, tag_prefix_len);
-                        write += tag_prefix_len;
-                        remaining -= tag_prefix_len;
-                    }
+                    APEX_HEADER_ID_ENSURE(tag_prefix_len);
+                    memcpy(write, tag_start, tag_prefix_len);
+                    write += tag_prefix_len;
+                    remaining -= tag_prefix_len;
                     read = after_tag_name;
 
                     /* Copy any existing attributes before injecting id */
@@ -1503,15 +1513,15 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
                     /* If there are existing attributes, copy them */
                     if (read > attr_start) {
                         size_t attr_len = read - attr_start;
-                        if (attr_len <= remaining) {
-                            memcpy(write, attr_start, attr_len);
-                            write += attr_len;
-                            remaining -= attr_len;
-                        }
+                        APEX_HEADER_ID_ENSURE(attr_len);
+                        memcpy(write, attr_start, attr_len);
+                        write += attr_len;
+                        remaining -= attr_len;
                     }
 
                     /* Add space before id attribute if needed */
-                    if ((read > attr_start || *read == '>') && remaining > 0) {
+                    if (read > attr_start || *read == '>') {
+                        APEX_HEADER_ID_ENSURE(1);
                         *write++ = ' ';
                         remaining--;
                     }
@@ -1520,56 +1530,41 @@ char *apex_inject_header_ids(const char *html, cmark_node *document, bool genera
                     char id_attr_str[512];
                     snprintf(id_attr_str, sizeof(id_attr_str), "id=\"%s\"", header->id);
                     size_t id_len = strlen(id_attr_str);
-                    if (id_len <= remaining) {
-                        memcpy(write, id_attr_str, id_len);
-                        write += id_len;
-                        remaining -= id_len;
-                    }
+                    APEX_HEADER_ID_ENSURE(id_len);
+                    memcpy(write, id_attr_str, id_len);
+                    write += id_len;
+                    remaining -= id_len;
 
                     /* Copy closing '>' */
                     if (*read == '>') {
-                        if (remaining > 0) {
-                            *write++ = *read++;
-                            remaining--;
-                        } else {
-                            read++;
-                        }
+                        APEX_HEADER_ID_ENSURE(1);
+                        *write++ = *read++;
+                        remaining--;
                     }
                 }
             } else {
                 /* No ID to inject, just copy the tag */
                 size_t tag_len = tag_end - tag_start + 1;
-                if (tag_len <= remaining) {
-                    memcpy(write, tag_start, tag_len);
-                    write += tag_len;
-                    remaining -= tag_len;
-                }
+                APEX_HEADER_ID_ENSURE(tag_len);
+                memcpy(write, tag_start, tag_len);
+                write += tag_len;
+                remaining -= tag_len;
                 read = tag_end + 1;
             }
         } else {
             /* Copy character */
-            if (remaining > 0) {
-                *write++ = *read++;
-                remaining--;
-            } else {
-                read++;
-            }
+            APEX_HEADER_ID_ENSURE(1);
+            *write++ = *read++;
+            remaining--;
         }
     }
 
-    /* Ensure we have space for null terminator */
-    if (remaining < 1) {
-        size_t used = write - output;
-        size_t new_capacity = (used + 1) * 2;
-        char *new_output = realloc(output, new_capacity + 1);
-        if (new_output) {
-            output = new_output;
-            write = output + used;
-            remaining = new_capacity - used;
-        }
-    }
+    APEX_HEADER_ID_ENSURE(1);
     *write = '\0';
 
+apex_inject_header_ids_done:
+    #undef APEX_HEADER_ID_ENSURE
+
     /* Clean up */
     while (header_map) {
         header_id_map *next = header_map->next;
@@ -1769,6 +1764,9 @@ char *apex_convert_relaxed_table_headers(const char *html) {
     const char *read = html;
     char *write = output;
     size_t remaining = len * 2;
+    bool demote_next_thead = false;
+    static const char marker[] = "<!--apex-relaxed-table-->";
+    static const size_t marker_len = sizeof(marker) - 1;
 
     while (*read) {
         /* Expand buffer if needed */
@@ -1791,11 +1789,18 @@ char *apex_convert_relaxed_table_headers(const char *html) {
             remaining = new_capacity - written;
         }
 
+        /* Marker from relaxed-table preprocess: demote the following thead */
+        if (strncmp(read, marker, marker_len) == 0) {
+            demote_next_thead = true;
+            read += marker_len;
+            if (*read == '\n') read++;
+            continue;
+        }
+
         /* Check for <thead> */
         if (strncmp(read, "<thead>", 7) == 0) {
             const char *after_thead = read + 7;
             const char *thead_end = strstr(after_thead, "</thead>");
-            const char *tbody_start = strstr(after_thead, "<tbody>");
 
             if (thead_end) {
                 /* Check if thead contains only empty cells (dummy headers from headerless tables) */
@@ -1864,81 +1869,19 @@ char *apex_convert_relaxed_table_headers(const char *html) {
 
                 /* If we found th cells and they're all empty, remove the entire thead */
                 if (found_any_th && all_cells_empty) {
+                    demote_next_thead = false;
                     read = thead_end + 8; /* Skip <thead>...</thead> */
                     continue;
                 }
-            }
-
-            if (thead_end && tbody_start && thead_end < tbody_start) {
-                /* Check if tbody contains a separator row (row with only em dashes) */
-                bool has_separator_row = false;
-                const char *tbody_end = strstr(tbody_start, "</tbody>");
-                const char *table_end = strstr(tbody_start, "</table>");
-
-                if (tbody_end && (!table_end || tbody_end < table_end)) {
-                    /* Look for rows with only em dashes in tbody */
-                    const char *search = tbody_start;
-                    while (search < tbody_end) {
-                        if (strncmp(search, "<tr>", 4) == 0) {
-                            const char *tr_end = strstr(search, "</tr>");
-                            if (tr_end && tr_end < tbody_end) {
-                                /* Check if this row contains only em dashes */
-                                bool row_is_separator = true;
-                                const char *cell_start = search + 4;
-                                while (cell_start < tr_end) {
-                                    if (strncmp(cell_start, "<td", 3) == 0 || strncmp(cell_start, "<th", 3) == 0) {
-                                        const char *tag_end = strstr(cell_start, ">");
-                                        if (!tag_end) break;
-                                        tag_end++;
-
-                                        const char *cell_end = NULL;
-                                        if (strncmp(cell_start, "<td", 3) == 0) {
-                                            cell_end = strstr(tag_end, "</td>");
-                                            if (cell_end) cell_end += 5;
-                                        } else {
-                                            cell_end = strstr(tag_end, "</th>");
-                                            if (cell_end) cell_end += 5;
-                                        }
-
-                                        if (cell_end && cell_end <= tr_end) {
-                                            if (!cell_contains_only_dashes(tag_end, cell_end - 5)) {
-                                                row_is_separator = false;
-                                                break;
-                                            }
-                                            cell_start = cell_end;
-                                        } else {
-                                            break;
-                                        }
-                                    } else {
-                                        cell_start++;
-                                    }
-                                }
-
-                                if (row_is_separator) {
-                                    has_separator_row = true;
-                                    break;
-                                }
 
-                                search = tr_end + 5;
-                            } else {
-                                break;
-                            }
-                        } else {
-                            search++;
-                        }
-                    }
-                }
-
-                /* If there's a separator row, it's a regular table - keep thead */
-                /* If there's no separator row, it's a relaxed table - convert thead to tbody */
-                if (!has_separator_row) {
-                    /* Convert thead to tbody */
+                /* Relaxed tables (marked during preprocess): demote thead to tbody */
+                if (demote_next_thead) {
+                    demote_next_thead = false;
                     memcpy(write, "<tbody>", 7);
                     write += 7;
                     remaining -= 7;
                     read += 7;  /* Skip <thead> */
 
-                    /* Convert <th> to <td> and skip </thead> */
                     while (read < thead_end + 8) {
                         if (strncmp(read, "<th>", 4) == 0) {
                             memcpy(write, "<td>", 4);
@@ -1955,7 +1898,6 @@ char *apex_convert_relaxed_table_headers(const char *html) {
                             write += 3;
                             remaining -= 3;
                             read += 3;
-                            /* Copy attributes until > */
                             while (*read && *read != '>') {
                                 *write++ = *read++;
                                 remaining--;
@@ -1965,9 +1907,7 @@ char *apex_convert_relaxed_table_headers(const char *html) {
                                 remaining--;
                             }
                         } else if (strncmp(read, "</thead>", 8) == 0) {
-                            /* Skip </thead> - we'll close tbody later if needed */
                             read += 8;
-                            /* Check if next is <tbody> - if so, skip opening tbody */
                             const char *next = read;
                             while (*next && (*next == ' ' || *next == '\n' || *next == '\t')) next++;
                             if (strncmp(next, "<tbody>", 7) == 0) {
@@ -1981,6 +1921,8 @@ char *apex_convert_relaxed_table_headers(const char *html) {
                     }
                     continue;
                 }
+
+                demote_next_thead = false;
             }
         }