Support pypdf 3.x and bump to 2.0.2

Guard the pypdf 4.0+ compress_identical_objects call with hasattr so it
is skipped on older pypdf rather than crashing. Lower the dependency
floor to pypdf>=3.17.0 and update CLAUDE.md (stale pikepdf references
left over from the 2.0.1 switch back to pypdf).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: jet52

CLAUDE.md

@@ -19,7 +19,7 @@ python splitmarks.py <input.pdf> [options]
 splitmarks <input.pdf> [options]
 
 # Build standalone executable (local: --onedir for fast startup; CI uses --onefile)
-pip install pyinstaller pikepdf
+pip install pyinstaller pypdf
 pyinstaller --onedir --name splitmarks splitmarks.py
 
 # Install local build: stash the bundle in ~/.local/share and symlink into ~/bin
@@ -44,7 +44,7 @@ Single-file tool: everything lives in `splitmarks.py` (no package structure).
 
 **Key components in `splitmarks.py`:**
 - `Bookmark` dataclass — tree structure preserving PDF outline hierarchy
-- `parse_outline_tree()` / `_parse_outline_items()` — recursive pikepdf outline parsing, handles both `/Dest` and `/A` (GoTo action) destinations
+- `parse_outline_tree()` / `_parse_outline_items()` — recursive pypdf outline parsing, handles both `/Dest` and `/A` (GoTo action) destinations
 - `calculate_page_ranges()` — maps each top-level bookmark to its page span
 - `split_pdf()` — main logic: open PDF, parse, filter, split, write output files
 - `add_bookmarks_to_writer()` — promotes child bookmarks to top level in split output files
@@ -55,7 +55,7 @@ Single-file tool: everything lives in `splitmarks.py` (no package structure).
 ## Key Details
 
 - **Python >= 3.10** required (uses `str | None` union syntax)
-- **pikepdf** is the sole external dependency (migrated from pypdf for smaller output files via `remove_unreferenced_resources()` and `ObjectStreamMode.generate`)
+- **pypdf** (`>=3.17.0`) is the sole external dependency. Output files are slimmed via `writer.compress_identical_objects(remove_identicals=True, remove_orphans=True)`, which only exists in pypdf 4.0+, so the call is guarded with `hasattr` and skipped on 3.x (splitting still works, just without that compression)
 - Version is tracked in **two places**: `__version__` in `splitmarks.py` and `version` in `pyproject.toml` — keep them in sync
 - Entry point registered in pyproject.toml: `splitmarks = "splitmarks:main"`
 

pyproject.toml

@@ -4,13 +4,13 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "splitmarks"
-version = "2.0.1"
+version = "2.0.2"
 description = "Split PDF files at top-level bookmarks into separate files"
 readme = "README.md"
 requires-python = ">=3.10"
 license = {text = "MIT"}
 dependencies = [
-    "pypdf>=4.0.0",
+    "pypdf>=3.17.0",
 ]
 
 [project.scripts]

splitmarks.py

@@ -16,7 +16,7 @@
     print("Error: pypdf is required. Install with: pip install pypdf", file=sys.stderr)
     sys.exit(1)
 
-__version__ = "2.0.1"
+__version__ = "2.0.2"
 
 
 @dataclass
@@ -427,10 +427,13 @@ def split_pdf(
                         writer, child, start_page, end_page
                     )
 
-            # Remove unreferenced objects (images/fonts from other sections)
-            writer.compress_identical_objects(
-                remove_identicals=True, remove_orphans=True
-            )
+            # Remove unreferenced objects (images/fonts from other sections).
+            # compress_identical_objects was added in pypdf 4.0; skip it on
+            # older versions rather than crashing.
+            if hasattr(writer, "compress_identical_objects"):
+                writer.compress_identical_objects(
+                    remove_identicals=True, remove_orphans=True
+                )
 
             try:
                 with open(output_path, "wb") as f: