.

help_processor.py

@@ -494,7 +494,13 @@ def file_hash(path: Path) -> str:
 
 
 def _load_html_image(src: str, base_dir: Path) -> Optional[tuple[bytes, str]]:
-    """Връща (data, ext) или None. Пропуска HTTP/HTTPS."""
+    """Връща (data, ext) или None. Пропуска HTTP/HTTPS.
+
+    Относителните ``src`` се резолвират спрямо директорията на HTML файла.
+    URL-encoding (``%20``) се декодира — типично за Word/LibreOffice „Save as HTML“.
+    """
+    from urllib.parse import unquote
+
     if not src:
         return None
     s = src.strip()
@@ -511,14 +517,29 @@ def _load_html_image(src: str, base_dir: Path) -> Optional[tuple[bytes, str]]:
         return data, _ext_from_content_type(m.group(1))
     if s.startswith(("http://", "https://")):
         return None      # по правило пропускаме мрежови картинки
-    # локален път, относителен или абсолютен
-    p = (base_dir / s).resolve() if not Path(s).is_absolute() else Path(s)
+    if s.lower().startswith("file:"):
+        path = _file_url_to_path(s)
+        if path is None:
+            log.warning(f"    HTML image bad file URL: {src}")
+            return None
+        loaded = _read_image_file(path)
+        if not loaded:
+            log.warning(f"    HTML image not found: {src} → {path}")
+        return loaded
+
+    # локален път: %20 → space, ../ спрямо HTML папката
+    decoded = unquote(s).replace("\\", "/")
     try:
+        p = Path(decoded)
+        if not p.is_absolute():
+            p = (base_dir / decoded).resolve()
         if p.is_file():
             data = p.read_bytes()
-            ext  = p.suffix.lstrip(".").lower() or "png"
+            ext = p.suffix.lstrip(".").lower() or "png"
             return data, ext
-    except Exception:
+        log.warning(f"    HTML image not found: {src} → {p}")
+    except Exception as e:
+        log.warning(f"    HTML image unreadable: {src}: {e}")
         return None
     return None
 
@@ -622,13 +643,18 @@ _FIGURE_CAPTION_RE = re.compile(
 )
 
 
+def _normalize_inline_ws(text: str) -> str:
+    """Срива CR/LF/табове в един интервал (Word/LibreOffice soft breaks в <h2>)."""
+    return re.sub(r"\s+", " ", (text or "").strip())
+
+
 def _normalize_chapter_key(text: str) -> str:
-    return re.sub(r"\s+", " ", (text or "").strip().lower())
+    return _normalize_inline_ws(text).lower()
 
 
 def _is_numbered_chapter_heading(text: str) -> bool:
     """True for top-level numbered chapter titles (not TOC prose, figures, tables)."""
-    t = (text or "").strip()
+    t = _normalize_inline_ws(text)
     if not t or len(t) >= 120:
         return False
     if "|" in t:
@@ -831,24 +857,34 @@ def _strip_attrs(el):
             t["style"] = kept_style
 
 
+def _ingest_html_img(
+    img, base_dir: Path, sec_images: list, img_counter: list
+) -> Optional[str]:
+    """Извлича една <img> картинка; връща placeholder текст или None."""
+    from bs4 import NavigableString
+
+    src = img.get("src") or img.get("data-src") or ""
+    loaded = _load_html_image(src, base_dir)
+    if not loaded:
+        img.decompose()
+        return None
+    data, ext = loaded
+    if not _should_keep_image(data):
+        img.decompose()
+        return None
+    img_counter[0] += 1
+    ref = ImageRef(placeholder=f"img_{img_counter[0]:02d}", data=data, ext=ext)
+    sec_images.append(ref)
+    ph = f"[IMG: {ref.placeholder}]"
+    img.replace_with(NavigableString(ph))
+    return ph
+
+
 def _swap_imgs_in_block(el, base_dir: Path, sec_images: list, img_counter: list) -> None:
     """Намира всички <img> в подадения елемент, извлича данните и подменя с
     NavigableString placeholder ([IMG: img_NN])."""
-    from bs4 import NavigableString
     for img in el.find_all("img"):
-        src = img.get("src") or img.get("data-src") or ""
-        loaded = _load_html_image(src, base_dir)
-        if not loaded:
-            img.decompose()
-            continue
-        data, ext = loaded
-        if not _should_keep_image(data):
-            img.decompose()
-            continue
-        img_counter[0] += 1
-        ref = ImageRef(placeholder=f"img_{img_counter[0]:02d}", data=data, ext=ext)
-        sec_images.append(ref)
-        img.replace_with(NavigableString(f"[IMG: {ref.placeholder}]"))
+        _ingest_html_img(img, base_dir, sec_images, img_counter)
 
 
 def parse_html(path: Path) -> list[Section]:
@@ -909,14 +945,16 @@ def parse_html(path: Path) -> list[Section]:
     def start_section(title: str, level: int = 1):
         nonlocal current_title, current_level, sec_text, sec_html, sec_images
         flush()
-        current_title = title
+        current_title = _normalize_inline_ws(title)
         current_level = level
         sec_text, sec_html, sec_images = [], [], []
 
     for el in blocks:
-        txt = el.get_text(" ", strip=True)
+        # Soft breaks (\r\n в средата на Word/LO <h2>) → един ред за split/TOC
+        txt = _normalize_inline_ws(el.get_text(" ", strip=True))
         is_cover = _html_is_cover_title(el)
         heading_lvl = _html_heading_level(el)
+        el_name = (el.name or "").lower()
 
         # Корица (Word Title / class Title): заглавие на преамбюла, без нова секция
         if is_cover and txt:
@@ -934,6 +972,12 @@ def parse_html(path: Path) -> list[Section]:
             append_block_as_body(el)
             continue
 
+        # <ul>/<ol>/table в TOC: целият блок НЕ е една „1. …“ глава — край на TOC фазата
+        if toc_state.phase and el_name in _HTML_PLAIN_NL_TAGS:
+            toc_state.leave_phase()
+            append_block_as_body(el)
+            continue
+
         # Номерирана глава (H1/H2/bold/plain) — с TOC: първото срещане в съдържанието остава в преамбюла
         numbered_action = toc_state.handle_numbered(txt) if txt else "no"
         if numbered_action == "toc":
@@ -957,23 +1001,16 @@ def parse_html(path: Path) -> list[Section]:
             start_section(txt, heading_lvl)
             continue
 
-        if el.name == "img":
+        if el_name == "img":
             if toc_state.phase:
                 toc_state.leave_phase()
-            # самостоятелен <img> (не вътре в блок)
-            _swap_imgs_in_block(el.parent if el.parent and el.parent.name else el,
-                                base_dir, sec_images, img_counter)
-            # ако е заменен с placeholder, добавяме като текст
-            img_txt = el.get_text(" ", strip=True) if el.name else ""
-            if img_txt:
-                sec_text.append(img_txt)
-                sec_html.append(f"<p>{img_txt}</p>")
+            ph = _ingest_html_img(el, base_dir, sec_images, img_counter)
+            if ph:
+                sec_text.append(ph)
+                sec_html.append(f"<p>{ph}</p>")
             continue
 
-        if toc_state.phase and (el.name or "").lower() in _HTML_PLAIN_NL_TAGS:
-            # <ol>/<ul> TOC списък — остава в преамбюла, приключва TOC фазата
-            toc_state.leave_phase()
-        elif toc_state.phase and txt and not _is_numbered_chapter_heading(txt):
+        if toc_state.phase and txt and not _is_numbered_chapter_heading(txt):
             toc_state.leave_phase()
         append_block_as_body(el)