Pypdf3

------------------------

- Patch from Hartmut Goebel to permit RectangleObjects to accept NumberObject
or FloatObject values.

- PDF compatibility fixes.

- Fix to read object xref stream in correct order.

- Fix for comments inside content streams.

------------------------

- Text strings from PDF files are returned as Unicode string objects when
pyPdf determines that they can be decoded (as UTF-16 strings, or as
PDFDocEncoding strings). Unicode objects are also written out when
necessary. This means that string objects in pyPdf can be either
generic.ByteStringObject instances, or generic.TextStringObject instances.

- The extractText method now returns a unicode string object.

- All document information properties now return unicode string objects. In
the event that a document provides docinfo properties that are not decoded by
pyPdf, the raw byte strings can be accessed with an "_raw" property (ie.
title_raw rather than title)

- generic.DictionaryObject instances have been enhanced to be easier to use.
Values coming out of dictionary objects will automatically be de-referenced
(.getObject will be called on them), unless accessed by the new "raw_get"
method. DictionaryObjects can now only contain PdfObject instances (as keys
and values), making it easier to debug where non-PdfObject values (which
cannot be written out) are entering dictionaries.

- Support for reading named destinations and outlines in PDF files. Original
patch by Ashish Kulkarni.

- Stream compatibility reading enhancements for malformed PDF files.

- Cross reference table reading enhancements for malformed PDF files.

- Encryption documentation.

- Replace some "assert" statements with error raising.

- Minor optimizations to FlateDecode algorithm increase speed when using PNG
predictors.

-----------------------

- Fix several serious bugs introduced in version 1.8, caused by a failure to
run through our PDF test suite before releasing that version.

- Fix bug in NullObject reading and writing.

-----------------------

- Add support for decryption with the standard PDF security handler. This
allows for decrypting PDF files given the proper user or owner password.

- Add support for encryption with the standard PDF security handler.

- Add new pythondoc documentation.

- Fix bug in ASCII85 decode that occurs when whitespace exists inside the
two terminating characters of the stream.

-----------------------

- Fix a bug when using a single page object in two PdfFileWriter objects.

- Adjust PyPDF to be tolerant of whitespace characters that don't belong
during a stream object.

- Add documentInfo property to PdfFileReader.

- Add numPages property to PdfFileReader.

- Add pages property to PdfFileReader.

- Add extractText function to PdfFileReader.

-----------------------

- Add basic support for comments in PDF files. This allows us to read some
ReportLab PDFs that could not be read before.

- Add "auto-repair" for finding xref table at slightly bad locations.

- New StreamObject backend, cleaner and more powerful. Allows the use of
stream filters more easily, including compressed streams.

- Add a graphics state push/pop around page merges. Improves quality of
page merges when one page's content stream leaves the graphics
in an abnormal state.

- Add PageObject.compressContentStreams function, which filters all content
streams and compresses them. This will reduce the size of PDF pages,
especially after they could have been decompressed in a mergePage
operation.

- Support inline images in PDF content streams.

- Add support for using .NET framework compression when zlib is not
available. This does not make pyPdf compatible with IronPython, but it
is a first step.

- Add support for reading the document information dictionary, and extracting
title, author, subject, producer and creator tags.

- Add patch to support NullObject and multiple xref streams, from Bradley
Lawrence.