ENH: Add parameter to select images to be removed

pypdf/__init__.py

@@ -14,6 +14,7 @@
 from ._reader import DocumentInformation, PdfFileReader, PdfReader
 from ._version import __version__
 from ._writer import ObjectDeletionFlag, PdfFileWriter, PdfWriter
+from .constants import ImageType
 from .pagerange import PageRange, parse_filename_page_ranges
 from .papersizes import PaperSize
 
@@ -31,6 +32,7 @@
 __all__ = [
     "__version__",
     "_debug_versions",
+    "ImageType",
     "mult",
     "PageRange",
     "PaperSize",

pypdf/_writer.py

@@ -77,6 +77,7 @@
     FieldFlag,
     FileSpecificationDictionaryEntries,
     GoToActionArguments,
+    ImageType,
     InteractiveFormDictEntries,
     PageLabelStyle,
     TypFitArguments,
@@ -132,12 +133,16 @@
 
 
 class ObjectDeletionFlag(enum.IntFlag):
+    NONE = 0
     TEXT = enum.auto()
-    IMAGES = enum.auto()
     LINKS = enum.auto()
     ATTACHMENTS = enum.auto()
     OBJECTS_3D = enum.auto()
     ALL_ANNOTATIONS = enum.auto()
+    XOBJECT_IMAGES = enum.auto()
+    INLINE_IMAGES = enum.auto()
+    DRAWING_IMAGES = enum.auto()
+    IMAGES = XOBJECT_IMAGES | INLINE_IMAGES | DRAWING_IMAGES
 
 
 def _rolling_checksum(stream: BytesIO, blocksize: int = 65536) -> str:
@@ -2193,33 +2198,42 @@ def remove_objects_from_page(
         if to_delete & ObjectDeletionFlag.ALL_ANNOTATIONS:
             return self._remove_annots_from_page(page, None)
 
-        if to_delete & ObjectDeletionFlag.IMAGES:
+        jump_operators = []
+        if to_delete & ObjectDeletionFlag.DRAWING_IMAGES:
             jump_operators = (
                 [b"w", b"J", b"j", b"M", b"d", b"i"]
                 + [b"W", b"W*"]
                 + [b"b", b"b*", b"B", b"B*", b"S", b"s", b"f", b"f*", b"F", b"n"]
                 + [b"m", b"l", b"c", b"v", b"y", b"h", b"re"]
                 + [b"sh"]
             )
-        else:  # del text
+        if to_delete & ObjectDeletionFlag.TEXT:
             jump_operators = [b"Tj", b"TJ", b"'", b'"']
 
         def clean(content: ContentStream, images: List[str], forms: List[str]) -> None:
-            nonlocal to_delete
+            nonlocal jump_operators, to_delete
             i = 0
             while i < len(content.operations):
                 operands, operator = content.operations[i]
-                if operator in jump_operators:
+                if (
+                    (
+                        operator == b"INLINE IMAGE"
+                        and (
+                            cast(ObjectDeletionFlag, to_delete)
+                            & ObjectDeletionFlag.INLINE_IMAGES
+                        )
+                    )
+                    or (operator in jump_operators)
+                    or (
+                        operator == b"Do"
+                        and (
+                            cast(ObjectDeletionFlag, to_delete)
+                            & ObjectDeletionFlag.XOBJECT_IMAGES
+                        )
+                        and (operands[0] in images)
+                    )
+                ):
                     del content.operations[i]
-                elif operator == b"Do":
-                    if (
-                        to_delete & ObjectDeletionFlag.IMAGES
-                        and operands[0] in images
-                        or to_delete & ObjectDeletionFlag.TEXT
-                        and operands[0] in forms
-                    ):
-                        del content.operations[i]
-                    i += 1
                 else:
                     i += 1
             content.get_data()  # this ensures ._data is rebuilt from the .operations
@@ -2242,23 +2256,25 @@ def clean_forms(
                 try:
                     content: Any = None
                     if (
-                        to_delete & ObjectDeletionFlag.IMAGES
+                        to_delete
+                        & ObjectDeletionFlag.XOBJECT_IMAGES
                         and o["/Subtype"] == "/Image"
                     ):
-                        content = NullObject()
+                        content = NullObject()  # to delete the image keeping the entry
                         images.append(k)
                     if o["/Subtype"] == "/Form":
                         forms.append(k)
                         if isinstance(o, ContentStream):
                             content = o
                         else:
                             content = ContentStream(o, self)
-                            content.update(o.items())
-                            for k1 in ["/Length", "/Filter", "/DecodeParms"]:
-                                try:
-                                    del content[k1]
-                                except KeyError:
-                                    pass
+                            content.update(
+                                {
+                                    k1: v1
+                                    for k1, v1 in o.items()
+                                    if k1 not in ["/Length", "/Filter", "/DecodeParms"]
+                                }
+                            )
                         clean_forms(content, stack + [elt])  # clean sub forms
                     if content is not None:
                         if isinstance(v, IndirectObject):
@@ -2269,6 +2285,8 @@ def clean_forms(
                             d[k] = self._add_object(content)  # pragma: no cover
                 except (TypeError, KeyError):
                     pass
+            for im in images:
+                del d[im]  # for clean-up
             if isinstance(elt, StreamObject):  # for /Form
                 if not isinstance(elt, ContentStream):  # pragma: no cover
                     e = ContentStream(elt, self)
@@ -2277,40 +2295,57 @@ def clean_forms(
                 clean(elt, images, forms)  # clean the content
             return images, forms
 
+        if not isinstance(page, PageObject):
+            page = PageObject(self, page.indirect_reference)  # pragma: no cover
         if "/Contents" in page:
-            content = page["/Contents"].get_object()
+            content = cast(ContentStream, page.get_contents())
 
-            if not isinstance(content, ContentStream):
-                content = ContentStream(content, page)
             images, forms = clean_forms(page, [])
 
             clean(content, images, forms)
-            if isinstance(page["/Contents"], ArrayObject):
-                for o in page["/Contents"]:
-                    self._objects[o.idnum - 1] = NullObject()
-            try:
-                self._objects[
-                    cast(IndirectObject, page["/Contents"].indirect_reference).idnum - 1
-                ] = NullObject()
-            except AttributeError:
-                pass
-            page[NameObject("/Contents")] = self._add_object(content)
+            page.replace_contents(content)
 
-    def remove_images(self, ignore_byte_string_object: Optional[bool] = None) -> None:
+    def remove_images(
+        self,
+        to_delete: ImageType = ImageType.ALL,
+        ignore_byte_string_object: Optional[bool] = None,
+    ) -> None:
         """
         Remove images from this output.
 
         Args:
+            to_delete : The type of images to be deleted 
+                (default = all images types)
             ignore_byte_string_object: deprecated
         """
+        if isinstance(to_delete, bool):
+            ignore_byte_string_object = to_delete
+            to_delete = ImageType.ALL
         if ignore_byte_string_object is not None:
             warnings.warn(
                 "The 'ignore_byte_string_object' argument of remove_images is "
                 "deprecated and will be removed in pypdf 4.0.0.",
                 category=DeprecationWarning,
             )
+        i = (
+            (
+                ObjectDeletionFlag.XOBJECT_IMAGES
+                if to_delete & ImageType.XOBJECT_IMAGES
+                else ObjectDeletionFlag.NONE
+            )
+            | (
+                ObjectDeletionFlag.INLINE_IMAGES
+                if to_delete & ImageType.INLINE_IMAGES
+                else ObjectDeletionFlag.NONE
+            )
+            | (
+                ObjectDeletionFlag.DRAWING_IMAGES
+                if to_delete & ImageType.DRAWING_IMAGES
+                else ObjectDeletionFlag.NONE
+            )
+        )
         for page in self.pages:
-            self.remove_objects_from_page(page, ObjectDeletionFlag.IMAGES)
+            self.remove_objects_from_page(page, i)
 
     def removeImages(self, ignoreByteStringObject: bool = False) -> None:  # deprecated
         """
@@ -2319,7 +2354,7 @@ def removeImages(self, ignoreByteStringObject: bool = False) -> None:  # depreca
         .. deprecated:: 1.28.0
         """
         deprecation_with_replacement("removeImages", "remove_images", "3.0.0")
-        return self.remove_images(ignoreByteStringObject)
+        return self.remove_images()
 
     def remove_text(self, ignore_byte_string_object: Optional[bool] = None) -> None:
         """

pypdf/constants.py

@@ -8,7 +8,7 @@
 PDF Reference, sixth edition, Version 1.7, 2006.
 """
 
-from enum import IntFlag
+from enum import IntFlag, auto
 from typing import Dict, Tuple
 
 
@@ -585,3 +585,12 @@ class AnnotationFlag(IntFlag):
     TypArguments,
     TypFitArguments,
 )
+
+
+class ImageType(IntFlag):
+    NONE = 0
+    XOBJECT_IMAGES = auto()
+    INLINE_IMAGES = auto()
+    DRAWING_IMAGES = auto()
+    ALL = XOBJECT_IMAGES | INLINE_IMAGES | DRAWING_IMAGES
+    IMAGES = ALL  # for consistency with ObjectDeletionFlag

tests/test_writer.py

@@ -9,6 +9,7 @@
 import pytest
 
 from pypdf import (
+    ImageType,
     ObjectDeletionFlag,
     PageObject,
     PdfMerger,
@@ -1862,6 +1863,36 @@ def test_object_contains_indirect_reference_to_self():
     writer.append(reader)
 
 
+def test_remove_image_per_type():
+    writer = PdfWriter(clone_from=RESOURCE_ROOT / "reportlab-inline-image.pdf")
+    writer.remove_images(ImageType.INLINE_IMAGES)
+
+    assert all(
+        x not in writer.pages[0].get_contents().get_data()
+        for x in (b"BI", b"ID", b"EI")
+    )
+
+    with pytest.raises(DeprecationWarning):
+        writer.remove_images(True)
+
+    writer = PdfWriter(clone_from=RESOURCE_ROOT / "GeoBase_NHNC1_Data_Model_UML_EN.pdf")
+    writer.remove_images(ImageType.DRAWING_IMAGES)
+    assert all(
+        x not in writer.pages[1].get_contents().get_data()
+        for x in (b" re\n", b"W*", b"f*")
+    )
+    assert all(
+        x in writer.pages[1].get_contents().get_data() for x in (b" TJ\n", b"rg", b"Tm")
+    )
+    assert all(
+        x not in writer.pages[9]["/Resources"]["/XObject"]["/Meta84"].get_data()
+        for x in (b" re\n", b"W*", b"f*")
+    )
+    writer.remove_images(ImageType.XOBJECT_IMAGES)
+    assert b"Do\n" not in writer.pages[0].get_contents().get_data()
+    assert len(writer.pages[0]["/Resources"]["/XObject"]) == 0
+
+
 @pytest.mark.enable_socket()
 def test_add_outlines_on_empty_dict():
     """Cf #2233"""