PLAYA-PDF 0.2.7: Definitive 0.2.x release

What's Changed

• Remove most uses of Typing.cast by @dhdaines in #37
• Optimize text placement (some dare call it "rendering") by @dhdaines in #38
• Fix font size and rotated/skewed bounding boxes by @dhdaines in #39
• fix: deprecate layout in CLI right away and do other useful stuff by @dhdaines in #40
• Correctly implement ToUnicode according to the PDF standard and not that bogus technical note (that the PDF standard refers to...) by @dhdaines in #41
• feat: support slices and tuples in page list by @dhdaines in #42
• Optimize text extraction a bit more by @dhdaines in #43
• Make text less Lazy 😥 by @dhdaines in #47
• Treat marked content sections (more) correctly
• fix: recognize junk before header and compensate (fixes: #46) by @dhdaines in #48

Full Changelog: v0.2.6...v0.2.7

PLAYA-PDF 0.2.6: New year, new acronym

What's Changed

• ci: test on windows and mac by @dhdaines in #33
• Support parallel operations over pages by @dhdaines in #36
• Partially correct the handling of some types of CMaps (not fully correct though)

Full Changelog: v0.2.5...v0.2.6

PLAYA-PDF 0.2.5: Bug fixes and improvements

What's Changed

• Fix various bugs in the lazy API
  • Add specialized __len__ methods to ContentObject classes
  • Clarify iteration over ContentObject
• Fix installation of playa-pdf[crypto]
• Fix attribute classes in structure tree elements
• Deprecate "user" device space to avoid confusion with user space
• Parse embedded CMaps (mostly)
• Update pdfplumber support
• Add parser for object streams and iterator over all indirect objects
  in a document

Full Changelog: v0.2.4...v0.2.5

v0.2.4

What's Changed

• Add (and fix) 3rd party test suites, primariy pdf.js by @dhdaines in #26
• Try much harder to read even very broken PDFs
• Try somewhat harder to not produce empty TextObject (still a work in progress)

Full Changelog: v0.2.3...v0.2.4

PLAYA-PDF 0.2.3: Release early and often (before vacation)

What's Changed

• Require a newline before EI to fix various inline images by @dhdaines in #25
• Refactoring the CMap parser missed a very important corner case (which somehow mypy did not flag?)
• structtree property did not actually exist on Document and Page (oops!)

Full Changelog: v0.2.2...v0.2.3

PLAYA-PDF 0.2.2: Make it go fast again

What's Changed

• Resolve filters before checking if it isn't a list by @dhdaines in #22
• Verify that we don't have pdfminer.six#1059 (and warn about it) by @dhdaines in #23
• Optimize cmaps by @dhdaines in #24

Full Changelog: v0.2.1...v0.2.2

PLAYA-PDF 0.2.1: Fix some bugs

What's Changed

• Fix the RLE implementation by @dhdaines in #19 (originally pdfminer/pdfminer.six#1055 by @helpmefindaname)
• Report the actual device space bounding box for rotated text by @dhdaines in #20
• Prevent endless looping on bogus stream length and other EOFs by @dhdaines in #21

Full Changelog: v0.2...v0.2.1

PLAYA-PDF 0.2: Break all the APIs

What's Changed

• Support TIFF predictor on image streams by @dhdaines in #18 (originally from pdfminer/pdfminer.six#1058 by @helpmefindaname)
• Support different "device spaces" (screen, page, and default user space)
• expose form XObjects on Page to allow getting only their contents
• expose form XObject IDs in LayoutDict
• make TextState conform to PDF spec (leading and line matrix) and document it
• expose more of TextState in LayoutDict (render mode in particular)
• do not try to map characters with no ToUnicode and no Encoding
• properly support Pattern color space (uncolored tiling patterns) the
  way pdfplumber expects it to work
• support marked content points as ContentObjects
• document ContentObjects
• make a proper schema for LayoutDict, document it, and communicate it to Polars
• separate color values and patterns in LayoutDict

Full Changelog: v0.1.2...v0.2

PLAYA 0.1.2: Initial release

Here's a first release, in case you want to use this. Reasons you might do so include:

• Faster than pdfminer.six (about 20% or so)
• Much friendlier APIs than PDFPageAggregator, PDFResourceManager, PDFPage, etc, etc.
• Many outstanding pdfminer.six bugs have been fixed

Why would you not want to use this?

• PyPI package name is not actually playa because somebody else took that name 13 years ago.
• May be more or less tolerant of broken PDFs than pdfminer.six, and has no "strict mode" to be absolutely intolerant.
• Doesn't let you extract image data (this is not always useful since PDFs tend to use compositing and thus you should use a real PDF renderer like pypdfium2 if you want to reliably extract images)
• Is not (or ain't) a layout analyzer, so no LAParams, TextBox, and so on.
• API subject to change and refinement.
• Does not have abstractions. You do not have the flexibility to subclass everything and build a PDF renderer on top of PLAYA.
• Probably contains bugs.
• Definitely lacks documentation.