Mutiple Fixes

Author: JorjMcKie

pdf4llm/pdf4llm/__init__.py

@@ -2,9 +2,9 @@
 from pymupdf4llm import *
 
 
-__version__ = "0.0.9"
-version = __version__
-version_tuple = tuple(map(int, version.split(".")))
+__version__ = pymupdf4llm.__version__
+version = pymupdf4llm.version
+version_tuple = pymupdf4llm.version_tuple
 
 
 def LlamaMarkdownReader(*args, **kwargs):

pdf4llm/setup.py

@@ -13,11 +13,11 @@
     "Programming Language :: Python :: 3",
     "Topic :: Utilities",
 ]
-requires = ["pymupdf4llm"]
+requires = ["pymupdf4llm>=0.0.18"]
 
 setuptools.setup(
     name="pdf4llm",
-    version="0.0.9",
+    version="0.0.18",
     author="Artifex",
     author_email="support@artifex.com",
     description="PyMuPDF Utilities for LLM/RAG",

pymupdf4llm/pymupdf4llm/__init__.py

@@ -1,6 +1,6 @@
 from .helpers.pymupdf_rag import IdentifyHeaders, to_markdown
 
-__version__ = "0.0.17"
+__version__ = "0.0.18"
 version = __version__
 version_tuple = tuple(map(int, version.split(".")))
 

pymupdf4llm/pymupdf4llm/helpers/multi_column.py

@@ -168,9 +168,9 @@ def join_rects_phase1(bboxes):
 
         Joins any rectangles that "touch" each other.
         This means that their intersection is valid (but may be empty).
-        To prefer vertical joins, we will ignore small horizontal gaps.
+        To prefer vertical joins, we will ignore small gaps.
         """
-        delta = (0, 0, 0, 2)  # allow this gap below
+        delta = (0, 0, 0, 10)  # allow this gap below
         prects = bboxes[:]
         new_rects = []
         while prects:
@@ -179,7 +179,7 @@ def join_rects_phase1(bboxes):
             while repeat:
                 repeat = False
                 for i in range(len(prects) - 1, 0, -1):
-                    if not ((prect0 + delta) & prects[i]).is_empty:
+                    if ((prect0 + delta) & prects[i]).is_valid:
                         prect0 |= prects[i]
                         del prects[i]
                         repeat = True
@@ -208,11 +208,11 @@ def join_rects_phase2(bboxes):
         prects.sort(key=lambda b: (b.x0, b.y0))
         new_rects = [prects[0]]  # initialize with first item
 
-        # walk through the rest, top to bottom, thwn left to right
+        # walk through the rest, top to bottom, then left to right
         for r in prects[1:]:
             r0 = new_rects[-1]  # previous bbox
 
-            # join if we have similar borders and are not to far down
+            # join if we have similar borders and are not too far down
             if (
                 abs(r.x0 - r0.x0) <= 3
                 and abs(r.x1 - r0.x1) <= 3
@@ -239,7 +239,7 @@ def join_rects_phase3(bboxes, path_rects):
                     # do not join across columns
                     if prect1.x0 > prect0.x1 or prect1.x1 < prect0.x0:
                         continue
-                    # do not join different backgrounds
+                    # do not join areas with a different background
                     if in_bbox(prect0, path_rects) != in_bbox(prect1, path_rects):
                         continue
                     temp = prect0 | prect1
@@ -297,7 +297,7 @@ def join_rects_phase3(bboxes, path_rects):
         sort_rects.sort(key=lambda sr: sr[1])  # by computed key
         new_rects = [sr[0] for sr in sort_rects]  # extract sorted rectangles
 
-        # move shaded text rects into a separate list
+        # move text rects with background color into a separate list
         shadow_rects = []
         # for i in range(len(new_rects) - 1, 0, -1):
         #     r = +new_rects[i]