natural-pdf 0.1.30__py3-none-any.whl → 0.1.31__py3-none-any.whl

natural_pdf/core/element_manager.py

@@ -7,6 +7,7 @@ characters, words, rectangles, and lines extracted from a page.
 
 import logging
 import re
+from contextlib import contextmanager
 from itertools import groupby
 from typing import Any, Dict, List, Optional, Tuple, Union
 
@@ -47,6 +48,33 @@ HIGHLIGHT_DEFAULTS = {
     "color_value_min": 0.4,        # HSV V >
 }
 
+
+@contextmanager
+def disable_text_sync():
+    """
+    Temporarily disable text synchronization for performance.
+    
+    This is used when bulk-updating text content where character-level
+    synchronization is not needed, such as during bidi processing.
+    Fixes exponential recursion issue with Arabic/RTL text processing.
+    """
+    # Save original setter
+    original_setter = TextElement.text.fset
+    
+    # Create a fast setter that skips sync
+    def fast_setter(self, value):
+        self._obj["text"] = value
+        # Skip character synchronization for performance
+    
+    # Apply fast setter
+    TextElement.text = property(TextElement.text.fget, fast_setter)
+    
+    try:
+        yield
+    finally:
+        # Restore original setter
+        TextElement.text = property(TextElement.text.fget, original_setter)
+
 class NaturalWordExtractor(WordExtractor):
     """
     Custom WordExtractor that splits words based on specified character attributes
@@ -208,8 +236,7 @@ class ElementManager:
         yt = pdf_config.get("y_tolerance", 3)
         use_flow = pdf_config.get("use_text_flow", False)
 
-        # Define which attributes to preserve on the merged word object
-        # Should include split attributes + any others needed for filtering (like color)
+        # List of attributes to preserve on word objects
         attributes_to_preserve = list(
             set(
                 self._word_split_attributes
@@ -223,7 +250,7 @@ class ElementManager:
             )
         )
 
-        # -------------------------------------------------------------
+        # ------------------------------------------------------------------
         # NEW: Detect direction (LTR vs RTL) per visual line and feed
         #       pdfplumber's WordExtractor with the correct settings.
         # -------------------------------------------------------------
@@ -271,7 +298,9 @@ class ElementManager:
             # Build a WordExtractor tailored for this line's direction
             if is_rtl_line:
                 line_dir = "ttb"  # horizontal lines stacked top→bottom
-                char_dir = "rtl"  # characters right→left within the line
+                # Feed characters in right→left x-order; extractor can then treat
+                # them as left-to-right so that resulting text stays logical.
+                char_dir = "ltr"
             else:
                 line_dir = "ttb"
                 char_dir = "ltr"
@@ -288,9 +317,8 @@ class ElementManager:
             )
 
             # Prepare character sequence for the extractor:
-            #  • For LTR lines -> left→right order (x0 ascending)
-            #  • For RTL lines -> feed **reversed** list so that neighbouring
-            #    characters appear adjacent when the extractor walks right→left.
+            # Always feed characters in spatial order (x0 ascending)
+            # PDF stores glyphs in visual order, so this gives us the visual sequence
             line_chars_for_extractor = sorted(line_chars, key=lambda c: c.get("x0", 0))
 
             try:
@@ -324,15 +352,18 @@ class ElementManager:
                 # on the whole-line heuristic.
                 rtl_in_word = any(_is_rtl_char(ch.get("text", "")) for ch in char_list)
                 if rtl_in_word:
+                    # Convert from visual order (from PDF) to logical order using bidi
                     try:
                         from bidi.algorithm import get_display  # type: ignore
                         from natural_pdf.utils.bidi_mirror import mirror_brackets
 
-                        word_element.text = mirror_brackets(
-                            get_display(word_element.text, base_dir="R")
-                        )
+                        with disable_text_sync():
+                            # word_element.text is currently in visual order (from PDF)
+                            # Convert to logical order using bidi with auto direction detection
+                            logical_text = get_display(word_element.text, base_dir='L')
+                            # Apply bracket mirroring for logical order
+                            word_element.text = mirror_brackets(logical_text)
                     except Exception:
-                        # Fallback: keep original text if python-bidi fails
                         pass
 
         # ------------------------------------------------------------------
@@ -415,19 +446,6 @@ class ElementManager:
             f"Page {self._page.number}: Generated {len(generated_words)} words using NaturalWordExtractor."
         )
 
-        # --- Post-processing pass to ensure every word containing RTL characters is
-        #     stored in logical order and with mirrored brackets.  This is a
-        #     safeguard in case the per-line loop above missed some tokens.
-        try:
-            from bidi.algorithm import get_display  # type: ignore
-            from natural_pdf.utils.bidi_mirror import mirror_brackets
-
-            for w in generated_words:
-                if any(_is_rtl_char(ch) for ch in w.text):
-                    w.text = mirror_brackets(get_display(w.text, base_dir="R"))
-        except Exception:
-            pass  # graceful degradation – keep original text
-
         # 4. Load other elements (rects, lines)
         rect_elements = [RectangleElement(r, self._page) for r in self._page._page.rects]
         line_elements = [LineElement(l, self._page) for l in self._page._page.lines]
@@ -463,6 +481,8 @@ class ElementManager:
 
         logger.debug(f"Page {self._page.number}: Element loading complete.")
 
+        # If per-word BiDi was skipped, generated_words already stay in logical order.
+
     def _prepare_char_dicts(self) -> List[Dict[str, Any]]:
         """
         Prepares a list of character dictionaries from native PDF characters,

natural_pdf/elements/text.py

@@ -468,3 +468,32 @@ class TextElement(Element):
                 info[f"raw_{prop}"] = self._obj[prop]
 
         return info
+
+    @property
+    def visual_text(self) -> str:
+        """Return the text converted to *visual* order using the Unicode BiDi algorithm.
+
+        This helper is intentionally side-effect–free: it does **not** mutate
+        ``self.text`` or the underlying character dictionaries.  It should be
+        used by UI / rendering code that needs human-readable RTL/LTR mixing.
+        """
+        logical = self.text
+        if not logical:
+            return logical
+
+        # Quick check – bail out if no RTL chars to save import/CPU.
+        import unicodedata
+
+        if not any(unicodedata.bidirectional(ch) in ("R", "AL", "AN") for ch in logical):
+            return logical
+
+        try:
+            from bidi.algorithm import get_display  # type: ignore
+            from natural_pdf.utils.bidi_mirror import mirror_brackets
+
+            # Convert from logical order to visual order
+            visual = get_display(logical, base_dir="R")
+            return mirror_brackets(visual)
+        except Exception:
+            # If python-bidi is missing or errors, fall back to logical order
+            return logical

{natural_pdf-0.1.30.dist-info → natural_pdf-0.1.31.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: natural-pdf
-Version: 0.1.30
+Version: 0.1.31
 Summary: A more intuitive interface for working with PDFs
 Author-email: Jonathan Soma <jonathan.soma@gmail.com>
 License-Expression: MIT

{natural_pdf-0.1.30.dist-info → natural_pdf-0.1.31.dist-info}/RECORD

@@ -28,7 +28,7 @@ natural_pdf/classification/results.py,sha256=Mcay-xLBHbYoZ8U7f4gMj2IhhH_yORNEkZH
 natural_pdf/collections/mixins.py,sha256=sj76Cn6EdBtb5f-bdAV-1qpdixX8tI4BzPccPiYLI1w,5117
 natural_pdf/collections/pdf_collection.py,sha256=HLlyakM--23ZOeHDPucoM6Tw3yUyMXm0SSoqJwxRc2E,30744
 natural_pdf/core/__init__.py,sha256=QC8H4M3KbXwMFiQORZ0pdPlzx1Ix6oKKQSS7Ib2KEaA,38
-natural_pdf/core/element_manager.py,sha256=96v_w3kXhSUqRsJlX5Bl6O6hJzpYRqDn4xoyRsdqZ7o,49260
+natural_pdf/core/element_manager.py,sha256=Mn4cYqPL_2LD_GK9lf2duExaJF1qhASCKsOdAZdQb00,49821
 natural_pdf/core/highlighting_service.py,sha256=WKDqRpex1yS8CWhkNitWtKhxbyRRCLu3Xsct_HTPsD4,40774
 natural_pdf/core/page.py,sha256=kQKKqsbOaNeLhW3ark6mueDS-4tsopJcGcoMmKPK6B8,125624
 natural_pdf/core/pdf.py,sha256=YfniZp54AyptzMyr7ZP8n617n4wlV28SPrajt32nNBk,74233
@@ -44,7 +44,7 @@ natural_pdf/elements/image.py,sha256=UjHNzCgDzOseQmLpkKshcxg51DPmWNIAVYxZ0TAMyUI
 natural_pdf/elements/line.py,sha256=aQm4pDdlQSDAAXqrdg4AU-oTl9JCXgYuaJN0EYls6E0,4920
 natural_pdf/elements/rect.py,sha256=kiVa3e377ZnqIOXc89d9ZSY4EcmDxtccdtUw-HOQzpw,3796
 natural_pdf/elements/region.py,sha256=v1PzWvQoGHGdn7SQiPf4Oq3hIGueIfYGwcZ05ZU6XPE,127692
-natural_pdf/elements/text.py,sha256=2neapKplef0FsAMYWr4OdICt-TmrZ3z9z0YBrX8FrSk,17738
+natural_pdf/elements/text.py,sha256=kw7u2KfHtDB905YawP7Hs89kcR8XnbtpkYQGEk6LNyk,18860
 natural_pdf/export/mixin.py,sha256=L1q3MIEFWuvie4j4_EmW7GT3NerbZ1as0XMUoqTS7gM,5083
 natural_pdf/exporters/__init__.py,sha256=g1WRPCDVzceaUUsm8dchPhzdHFSjYM0NfRyc8iN0mtE,644
 natural_pdf/exporters/base.py,sha256=XhR1xlkHOh7suOuX7mWbsj1h2o1pZNet-OAS5YCJyeI,2115
@@ -100,7 +100,7 @@ natural_pdf/utils/text_extraction.py,sha256=mDeN1_VevNi3RwvFe48PM5vBh-A5WeBlYgP6
 natural_pdf/utils/visualization.py,sha256=n3IZpbY5cf9LItzGavBcNyVZZrrUVxjYnmqZHYPa7NU,9386
 natural_pdf/widgets/__init__.py,sha256=QTVaUmsw__FCweFYZebwPssQxxUFUMd0wpm_cUbGZJY,181
 natural_pdf/widgets/viewer.py,sha256=2VUY1TzWMDe9I-IVNOosKZ2LaqpjLB62ftMAdk-s6_8,24952
-natural_pdf-0.1.30.dist-info/licenses/LICENSE,sha256=9zfwINwJlarbDmdh6iJV4QUG54QSJlSAUcnC1YiC_Ns,1074
+natural_pdf-0.1.31.dist-info/licenses/LICENSE,sha256=9zfwINwJlarbDmdh6iJV4QUG54QSJlSAUcnC1YiC_Ns,1074
 optimization/memory_comparison.py,sha256=XEHtjduSmzXzxnsJMvemTcq-OAlvGUBAm5wwnOXq8TY,6524
 optimization/pdf_analyzer.py,sha256=G3XWhsEqIYbohEgTqz6wzxkAnOx4MkbvbSspx577-8w,19145
 optimization/performance_analysis.py,sha256=vVlFDywEXxhJLd9n2KVVqqQnS6rwWoHV_jlogboGF2k,13784
@@ -115,8 +115,8 @@ tools/bad_pdf_eval/export_enrichment_csv.py,sha256=SMEm9WxFUN_RIf8AGfZfjGEmvBvrO
 tools/bad_pdf_eval/llm_enrich.py,sha256=PsFMymPc8BNck21T3vupTN18pLdum-A_OLoJEKr6f80,12234
 tools/bad_pdf_eval/reporter.py,sha256=LIhcguDZ5XKgb0WeJsyA7m0kcliebOohzveShvt_KmY,400
 tools/bad_pdf_eval/utils.py,sha256=FuxaPX6f26IjQXu1vP0a2i9h1jgJNbASb8mRyj5-elE,4849
-natural_pdf-0.1.30.dist-info/METADATA,sha256=4Jg-iXXt6zGNE4gSYE_nMF395JDzv1Dierh93x1Lklo,6711
-natural_pdf-0.1.30.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-natural_pdf-0.1.30.dist-info/entry_points.txt,sha256=1R_KMv7g60UBBpRqGfw7bppsMNGdayR-iJlb9ohEk_8,81
-natural_pdf-0.1.30.dist-info/top_level.txt,sha256=oZlRzSc3nZ9sV3L6kD_Di734Pp62ANrm46imFVa51qQ,58
-natural_pdf-0.1.30.dist-info/RECORD,,
+natural_pdf-0.1.31.dist-info/METADATA,sha256=tqimu2ZReyYu5pS0PsbCo-Z9fIzkpMj1ljGPNbaOFss,6711
+natural_pdf-0.1.31.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+natural_pdf-0.1.31.dist-info/entry_points.txt,sha256=1R_KMv7g60UBBpRqGfw7bppsMNGdayR-iJlb9ohEk_8,81
+natural_pdf-0.1.31.dist-info/top_level.txt,sha256=oZlRzSc3nZ9sV3L6kD_Di734Pp62ANrm46imFVa51qQ,58
+natural_pdf-0.1.31.dist-info/RECORD,,