See #1715

See #107

This change should make tests easier to read and ensure that the cleanup-step cannot be forgotten.

If the test run fails, you can see where the file was written. The file is not directly deleted.

However, if the run does not fail (loudly) but you want to inspect the generated PDF, you have to 'assert False' or similar. That is  harder to figure out than commenting out a Path.unlink() line

…1729)

Fixes #1723

This is a shortcut. It is syntactic sugar to prevent users from having to use `clone_document_from_reader`.

Multiple fiters in streams are not using their decodParms

Closes #1737

Co-authored-by: Martin Thoma <info@martin-thoma.de>

Co-authored-by: pubpub-zz <4083478+pubpub-zz@users.noreply.github.com>

Fixes #1195

See #107

See #107

BUG: /Flags should have been /F - typo. Was never released, though.

The implication of this is that PDFs which were encrypted with pypdf before might be less secure than they should be.

It's unclear to me if "less secure" means "insecure" or if it's just a theoretical advantage.

Security (SEC):
-  Use Python's secrets module instead of random module (#1748)

New Features (ENH):
-  Add AnnotationBuilder.highlight text markup annotation (#1740)
-  Add AnnotationBuilder.popup (#1665)
-  Add AnnotationBuilder.polyline annotation support (#1726)
-  Add clone_from parameter in PdfWriter constructor (#1703)

Bug Fixes (BUG):
-  'DictionaryObject' object has no attribute 'indirect_reference' (#1729)

Robustness (ROB):
-  Handle params NullObject in decode_stream_data (#1738)

Documentation (DOC):
-  Project scope (#1743)

Maintenance (MAINT):
-  Add AnnotationFlag (#1746)
-  Add LazyDict.__str__ (#1727)

[Full Changelog](3.6.0...3.7.0)