Change how lines are broken

pdftext/inference.py

@@ -133,7 +133,7 @@ def infer_single_page(text_chars, block_threshold=settings.BLOCK_THRESHOLD):
                 span = update_span(line, span)
                 line = update_line(block, line)
                 block = update_block(blocks, block)
-            elif prev_char["char"] in LINE_BREAKS and prediction_probs[1] >= .5: # Look for newline character as a forcing signal for a new line
+            elif prev_char["char"] in LINE_BREAKS: # Look for newline character as a forcing signal for a new line
                 span = update_span(line, span)
                 line = update_line(block, line)
             elif prev_font_info != font_info:

pdftext/pdf/utils.py

@@ -42,8 +42,7 @@ def get_fontname(textpage, char_index):
     # Re-interpret the type from char to unsigned short as required by the function
     buffer_ptr = ctypes.cast(buffer, ctypes.POINTER(ctypes.c_ushort))
     flag_buffer = ctypes.c_int()
-    flag_ptr = ctypes.pointer(flag_buffer)
-    font_info = pdfium_c.FPDFText_GetFontInfo(textpage, char_index, buffer_ptr, n_bytes, flag_ptr)
+    font_info = pdfium_c.FPDFText_GetFontInfo(textpage, char_index, buffer_ptr, n_bytes, flag_buffer)
     if font_info == 0:
         return None, None
     try:

pdftext/settings.py

@@ -8,7 +8,6 @@ class Settings(BaseSettings):
     MODEL_PATH: str = os.path.join(BASE_PATH, "models", "dt.joblib")
 
     # Fonts
-    FONT_BUFFER_SIZE: int = 1024 # How many characters to buffer when reading a font name
     FONTNAME_SAMPLE_FREQ: int = 10
 
     # Inference

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "pdftext"
-version = "0.3.3"
+version = "0.3.4"
 description = "Extract structured text from pdfs quickly"
 authors = ["Vik Paruchuri <vik.paruchuri@gmail.com>"]
 license = "Apache-2.0"