deeplook/svglib v2.2.0

svglib 2.2.0

on GitHub

Tooling and housekeeping

• Modernized type annotations across src/svglib/ and tests/ to use PEP
  585 builtin generics (list[...], dict[...], tuple[...], set[...])
  instead of typing.List/Dict/Tuple/Set, and simplified a few
  emptiness/membership checks (not points instead of len(points) == 0,
  a set literal instead of a list for in checks). No behavior change.
• Silenced a UserWarning from test_convert_pdf_png that came from
  ReportLab's own renderPM.drawImage, not from svglib: it reopens a
  <image> file referenced by path and converts it straight to RGB without
  normalizing palette-with-transparency PNGs first, unlike svglib's own
  handling of base64-embedded images. The referenced PNG (a dedicated
  tRNS-chunk test fixture) still renders correctly; there was nothing to fix
  in svglib's output.

Fixed

• font-size no longer has PX_TO_PT applied twice. The viewport group
  already carries the conversion, but String.fontSize and the advance widths
  used to position text fragments were converted to points as well. Text
  therefore rendered at 0.75x the size of the geometry in the same document,
  and multi-tspan runs were laid out in the wrong unit (issue #502). Both now
  stay in user units, like every other coordinate below the transform.
  String.fontSize in the shape tree is in user units accordingly; multiply by
  the enclosing group's transform to get the rendered size.