Pypdf

⚠️ We discovered that compress_content_stream has to be applied to a page of
the PdfWriter. It may not be applied to a page of the PdfReader!

Bug Fixes (BUG)
- compress_content_stream not readable in Adobe Acrobat (1698)
- Pass logging parameters correctly in set_need_appearances_writer (1697)
- Write /Root/AcroForm in set_need_appearances_writer (1639)

Robustness (ROB)
- Allow more whitespaces within linearized file (1701)

[Full Changelog](https://github.com/py-pdf/pypdf/compare/3.5.1...3.5.2)

Robustness (ROB)
- Some attributes not copied in DictionaryObject._clone (1635)
- Allow merging multiple time pages with annots (1624)

Testing (TST)
- Replace pytest.mark.external by enable_socket (1657)

[Full Changelog](https://github.com/py-pdf/pypdf/compare/3.5.0...3.5.1)

New Features (ENH)
- Add reader.attachments public interface (1611, 1661)
- Add PdfWriter.remove_objects_from_page(page: PageObject, to_delete: ObjectDeletionFlag) (1648)
- Allow free-text annotation to have transparent border/background (1664)

Bug Fixes (BUG)
- Allow decryption with empty password for AlgV5 (1663)
- Let PdfWriter.pages return PageObject after calling `clone_document_from_reader()` (1613)
- Invalid font pointed during merge_resources (1641)

Robustness (ROB)
- Cope with invalid objects in IndirectObject.clone (1637)
- Improve tolerance to invalid Names/Dests (1658)
- Decode encoded values in get_fields (1636)
- Let PdfWriter.merge cope with missing "/Fields" (1628)

[Full Changelog](https://github.com/py-pdf/pypdf/compare/3.4.1...3.5.0)

Bug Fixes (BUG)
- Switch from trimbox to cropbox when merging pages (1622)
- Text extraction not working with one glyph to char sequence (1620)

Robustness (ROB)
- Fix 2 cases of "object has no attribute \'indirect_reference\'" (1616)

Testing (TST)
- Add multiple retry on get_url for external PDF downloads (1626)

[Full Changelog](https://github.com/py-pdf/pypdf/compare/3.4.0...3.4.1)

NOTICE: pypdf changed the way it represents numbers parsed from PDF files.
pypdf<3.4.0 represented numbers as Decimal, pypdf>=3.4.0 represents them as
floats. Several other PDF libraries to this, as well as many PDF viewers.
We hope to fix issues with too high precision like this and get a speed boost.
In case your PDF documents rely on more than 18 decimals of precision you
should check if it still works as expected.
To clarify: This does not affect the text shown in PDF documents. It affects
numbers, e.g. when graphics are drawn on the PDF or very exact positions are
used. Typically, 5 decimals should be enough.

New Features (ENH)
- Enable merging forms with overlapping names (1553)
- Add 'over' parameter to merge_transformend_page & co (1567)

Bug Fixes (BUG)
- Fix getter of the PageObject.rotation property with an indirect object (1602)
- Restore merge_transformed_page & co (1567)
- Replace decimal by float (1563)

Robustness (ROB)
- PdfWriter.remove_images: /Contents might not be in page_ref (1598)

Developer Experience (DEV)
- Introduce ruff (1586, 1609)

Maintenance (MAINT)
- Remove decimal (1608)

[Full Changelog](https://github.com/py-pdf/pypdf/compare/3.3.0...3.4.0)

New Features (ENH)
- Add page label support to PdfWriter (1558)
- Accept inline images with space before EI (1552)
- Add circle annotation support (1556)
- Add polygon annotation support (1557)
- Make merging pages produce a deterministic PDF (1542, 1543)

Bug Fixes (BUG)
- Fix error in cmap extraction (1544)
- Remove erroneous assertion check (1564)
- Fix dictionary access of optional page label keys (1562)

Robustness (ROB)
- Set ignore_eof=True for read_until_regex (1521)

Documentation (DOC)
- Paper size (1550)

Developer Experience (DEV)
- Fix broken combination of dependencies of docs.txt
- Annotate tests appropriately (1551)

[Full Changelog](https://github.com/py-pdf/pypdf/compare/3.2.1...3.3.0)