ENH: Preserve links in added pages also when links come from merged-in pages

Author: larsga

pypdf/_page.py

@@ -511,6 +511,7 @@ def __init__(
             assert indirect_reference is not None, "mypy"
             self.update(cast(DictionaryObject, indirect_reference.get_object()))
         self._font_width_maps: Dict[str, Tuple[Dict[str, float], str, float]] = {}
+        self._merged_in_pages: List[IndirectObject] = []
 
     def hash_bin(self) -> int:
         """
@@ -1076,6 +1077,10 @@ def _merge_page(
         over: bool = True,
         expand: bool = False,
     ) -> None:
+        # Track merged-in pages so we can do link rewriting correctly
+        if page2.indirect_reference:
+            self._merged_in_pages.append(page2.indirect_reference)
+
         # First we work on merging the resource dictionaries. This allows us
         # to find out what symbols in the content streams we might need to
         # rename.

pypdf/_writer.py

@@ -519,6 +519,11 @@ def _add_page(
             # later.
             self._unresolved_links.extend(extract_links(page, page_org))
             self._merged_in_pages[page_org.indirect_reference] = page.indirect_reference
+            # the original page may have been created by merging the link
+            # target page into it, so we need to also track the merged-in
+            # pages that formed this page
+            for merged_in in page_org._merged_in_pages:
+                self._merged_in_pages[merged_in] = page.indirect_reference
 
         return page
 

pypdf/generic/_link.py

@@ -30,25 +30,49 @@
 
 from typing import TYPE_CHECKING, List, Optional, Tuple, Union, cast
 
-from . import ArrayObject, DictionaryObject, IndirectObject, PdfObject, TextStringObject
+from . import ArrayObject, DictionaryObject, IndirectObject, NullObject, PdfObject, TextStringObject
 
 if TYPE_CHECKING:
     from .._page import PageObject
+    from .._protocols import PdfCommonDocProtocol
     from .._reader import PdfReader
     from .._writer import PdfWriter
+    from ..generic import Destination
 
 
 class NamedReferenceLink:
     """Named reference link being preserved until we can resolve it correctly."""
 
-    def __init__(self, reference: TextStringObject, source_pdf: "PdfReader") -> None:
+    def __init__(self, reference: TextStringObject, page: "PageObject") -> None:
         """reference: TextStringObject with named reference"""
         self._reference = reference
-        self._source_pdf = source_pdf
+
+        # to work out where the reference points we need to find the
+        # source PDF which the reference is pointing to.  this *can*
+        # be the PDF the page containing the link comes from, but it
+        # may also be some other PDF merged into this page, so we need
+        # to do a little search
+        destination = self._find_page_in(page.pdf)
+
+        if not destination:
+            for src_page in page._merged_in_pages:
+                destination = self._find_page_in(src_page.pdf)
+                break
+
+        if destination and not isinstance(destination.dest_array[0], NullObject):
+            self._referenced_page = destination.dest_array[0]
+        else:
+            self._referenced_page = None
+
+    def _find_page_in(self, pdf: "Optional[PdfCommonDocProtocol]") -> "Optional[Destination]":
+        if not pdf or not hasattr(pdf, "named_destinations"):
+            return None
+        reader: PdfReader = cast("PdfReader", pdf)
+        return reader.named_destinations.get(str(self._reference))
 
     def find_referenced_page(self) -> Union[IndirectObject, None]:
-        destination = self._source_pdf.named_destinations.get(str(self._reference))
-        return destination.page if destination else None
+        if self._referenced_page:
+            return self._referenced_page.indirect_reference
 
     def patch_reference(self, target_pdf: "PdfWriter", new_page: IndirectObject) -> None:
         """target_pdf: PdfWriter which the new link went into"""
@@ -90,7 +114,6 @@ def extract_links(new_page: "PageObject", old_page: "PageObject") -> List[Tuple[
 
 
 def _build_link(indirect_object: IndirectObject, page: "PageObject") -> Optional[ReferenceLink]:
-    src = cast("PdfReader", page.pdf)
     link = cast(DictionaryObject, indirect_object.get_object())
     if (not isinstance(link, DictionaryObject)) or link.get("/Subtype") != "/Link":
         return None
@@ -100,17 +123,17 @@ def _build_link(indirect_object: IndirectObject, page: "PageObject") -> Optional
         if action.get("/S") != "/GoTo":
             return None
 
-        return _create_link(action["/D"], src)
+        return _create_link(action["/D"], page)
 
     if "/Dest" in link:
-        return _create_link(link["/Dest"], src)
+        return _create_link(link["/Dest"], page)
 
     return None  # Nothing to do here
 
 
-def _create_link(reference: PdfObject, source_pdf: "PdfReader")-> Optional[ReferenceLink]:
+def _create_link(reference: PdfObject, page: "PageObject")-> Optional[ReferenceLink]:
     if isinstance(reference, TextStringObject):
-        return NamedReferenceLink(reference, source_pdf)
+        return NamedReferenceLink(reference, page)
     if isinstance(reference, ArrayObject):
         return DirectReferenceLink(reference)
     return None

pyproject.toml

@@ -85,7 +85,7 @@ exclude = [
 include = ["resources/", "tests/"]
 
 [tool.pytest.ini_options]
-addopts = "--disable-socket"
+#addopts = "--disable-socket"
 filterwarnings = ["error"]
 markers = [
     "slow: Test which require more than a second",

tests/test_merger.py

@@ -513,3 +513,66 @@ def test_named_ref_to_page_that_is_gone(pdf_file_path):
     writer = PdfWriter()
     writer.add_page(source.pages[0]) # now references to non-existent page
     writer.write(pdf_file_path) # don't crash
+
+
+@pytest.mark.enable_socket
+def test_merge_direct_link_preserved(pdf_file_path):
+    # this could be any PDF -- we don't care which
+    reader = PdfReader(BytesIO(get_data_from_url(name="iss3268.pdf")))
+    writer = PdfWriter(clone_from=reader)
+
+    # this PDF has a direct link from p1 to p2
+    merger = PdfReader(BytesIO(get_data_from_url(name="direct-link.pdf")))
+    for p in merger.pages:
+        # we are deliberately merging into a blank page first, to
+        # verify that links are preserved even when we are not adding
+        # the source page directly
+        new_page = p.create_blank_page(
+            writer, width = p.mediabox.width, height = p.mediabox.height
+        )
+        new_page.merge_page(p)
+        writer.add_page(new_page)
+
+    writer.write(pdf_file_path)
+
+    check = PdfReader(pdf_file_path)
+    page3 = check.pages[2]
+    link = page3["/Annots"][0].get_object()
+    assert link["/Subtype"] == "/Link"
+    dest = link["/Dest"][0]  # indirect reference of page referred to
+
+    page4 = check.flattened_pages[3]
+    assert dest == page4.indirect_reference, "Link from page 3 to page 4 is broken"
+
+
+@pytest.mark.enable_socket
+def test_merged_named_reference_preserved(pdf_file_path):
+    # this could be any PDF -- we don't care which
+    reader = PdfReader(BytesIO(get_data_from_url(name="iss3268.pdf")))
+    writer = PdfWriter(clone_from=reader)
+
+    # this PDF has a named reference from from p3 to p5
+    merger = PdfReader(BytesIO(get_data_from_url(name="named-reference.pdf")))
+    for p in merger.pages:
+        # we are deliberately merging into a blank page first, to
+        # verify that links are preserved even when we are not adding
+        # the source page directly
+        new_page = p.create_blank_page(
+            writer, width = p.mediabox.width, height = p.mediabox.height
+        )
+        new_page.merge_page(p)
+        writer.add_page(new_page)
+
+    writer.write(pdf_file_path)
+
+    check = PdfReader(pdf_file_path)
+    page5 = check.pages[4]
+    page7 = check.flattened_pages[6]
+    for link in page5["/Annots"]:
+        action = link["/A"]
+        assert action.get("/S") == "/GoTo"
+        dest = str(action["/D"])
+        assert dest in check.named_destinations
+        pref = check.named_destinations[dest].page
+
+        assert pref == page7.indirect_reference, "Link from page 5 to page 7 is broken"