[MRG] Switch from {pdfminer.six, pypdf} to PAVÉS

camelot/backends/base.py

@@ -6,6 +6,6 @@ def installed(self) -> bool:  # noqa D102
         raise NotImplementedError
 
     def convert(  # noqa D102
-        self, pdf_path: str, png_path: str, resolution: int = 300
+        self, pdf_path: str, png_path: str, resolution: int = 300, page: int = 1
     ) -> None:  # noqa D102
         raise NotImplementedError

camelot/backends/ghostscript_backend.py

@@ -6,7 +6,9 @@
 class GhostscriptBackend(ConversionBackend):
     """Classmethod to create GhostscriptScriptBackend."""
 
-    def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
+    def convert(
+        self, pdf_path: str, png_path: str, resolution: int = 300, page: int = 1
+    ) -> None:
         """Convert a PDF to a PNG image using Ghostscript .
 
         Parameters
@@ -17,6 +19,8 @@ def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
             [description]
         resolution : int, optional
             [description], by default 300
+        page: int, optional
+            Single page to convert.
 
         Raises
         ------
@@ -35,6 +39,8 @@ def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
             "gs",
             "-q",
             "-sDEVICE=png16m",
+            f"-dFirstPage={page}",
+            f"-dLastPage={page}",
             "-o",
             png_path,
             f"-r{resolution}",

camelot/backends/image_conversion.py

@@ -104,7 +104,7 @@ def implements_convert():
 
             return backend
 
-    def convert(self, pdf_path: str, png_path: str) -> None:
+    def convert(self, pdf_path: str, png_path: str, page: int = 1) -> None:
         """Convert PDF to png_path.
 
         Parameters
@@ -113,6 +113,8 @@ def convert(self, pdf_path: str, png_path: str) -> None:
             Path where to read the pdf file.
         png_path : str
             Path where to save png file.
+        page: int, optional
+            Single page to convert.
 
         Raises
         ------
@@ -122,13 +124,13 @@ def convert(self, pdf_path: str, png_path: str) -> None:
             [description]
         """
         try:
-            self.backend.convert(pdf_path, png_path)
+            self.backend.convert(pdf_path, png_path, page=page)
         except Exception as f:
             if self.use_fallback:
                 for fallback in self.fallbacks:
                     try:
                         converter = BACKENDS[fallback]()
-                        converter.convert(pdf_path, png_path)
+                        converter.convert(pdf_path, png_path, page=page)
                     except Exception as e:
                         msg = f"Image conversion failed with image conversion backend {fallback!r}\n error: {e}"
                         raise ImageConversionError(msg) from e

camelot/backends/pdfium_backend.py

@@ -20,7 +20,9 @@ def installed(self) -> bool:  # noqa D102
             return True
         return False
 
-    def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
+    def convert(
+        self, pdf_path: str, png_path: str, resolution: int = 300, page: int = 1
+    ) -> None:
         """Convert PDF to png.
 
         Parameters
@@ -29,6 +31,8 @@ def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
             Path where to read the pdf file.
         png_path : str
             Path where to save png file.
+        page: int, optional
+            Single page to convert.
 
         Raises
         ------
@@ -39,5 +43,5 @@ def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
             raise OSError(f"pypdfium2 is not available: {PDFIUM_EXC!r}")
         doc = pdfium.PdfDocument(pdf_path)
         doc.init_forms()
-        image = doc[0].render(scale=resolution / 72).to_pil()
+        image = doc[page - 1].render(scale=resolution / 72).to_pil()
         image.save(png_path)

camelot/backends/poppler_backend.py

@@ -22,7 +22,9 @@
 class PopplerBackend(ConversionBackend):
     """Classmethod to create a poplerBackendBackend class."""
 
-    def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
+    def convert(
+        self, pdf_path: str, png_path: str, resolution: int = 300, page: int = 1
+    ) -> None:
         """Convert PDF to png.
 
         Parameters
@@ -31,6 +33,8 @@ def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
             Path where to read the pdf file.
         png_path : str
             Path where to save png file.
+        page: int, optional
+            Single page to convert.
 
         Raises
         ------
@@ -39,13 +43,24 @@ def convert(self, pdf_path: str, png_path: str, resolution: int = 300) -> None:
         ValueError
             [description]
         """
-        pdftopng_executable = shutil.which("pdftopng", path=path)
+        pdftopng_executable = shutil.which("pdftocairo", path=path)
         if pdftopng_executable is None:
             raise OSError(
-                "pdftopng is not installed. You can install it using the 'pip install pdftopng' command."
+                "pdftocairo is not installed. Please install `poppler-utils`."
             )
 
-        pdftopng_command = [pdftopng_executable, pdf_path, png_path]
+        png_stem, _ = os.path.splitext(png_path)
+        pdftopng_command = [
+            pdftopng_executable,
+            "-png",
+            "-singlefile",
+            "-f",
+            str(page),
+            "-l",
+            str(page),
+            pdf_path,
+            png_stem,
+        ]
 
         try:
             subprocess.check_output(

camelot/core.py

@@ -24,6 +24,7 @@
     from typing_extensions import TypedDict, Unpack
 
 from .backends import ImageConversionBackend
+from .image_processing import undo_rotation
 from .utils import build_file_path_in_temp_dir
 from .utils import get_index_closest_point
 from .utils import get_textline_coords
@@ -548,6 +549,7 @@ def __init__(self, cols, rows):
         self.filename = None
         self.order = None
         self.page = None
+        self.rotation = ""
         self.flavor = None  # Flavor of the parser that generated the table
         self.pdf_size = None  # Dimensions of the original PDF page
         self._bbox = None  # Bounding box in original document
@@ -618,8 +620,8 @@ def get_pdf_image(self):
                     os.path.basename(self.filename), ".png"
                 )
                 backend = ImageConversionBackend(use_fallback=True)
-                backend.convert(self.filename, self._image_path)
-            self._image = cv2.imread(self._image_path)
+                backend.convert(self.filename, self._image_path, page=self.page)
+            self._image = undo_rotation(cv2.imread(self._image_path), self.rotation)
         return self._image
 
     def set_all_edges(self):