Writing visual unit tests to compare graphics code is powerful, but often brittle. Things like antialiasing can make the comparison unreliable.
Adapt the smeared comparison from the four-letter blocks project, so that comparison will ignore minor changes in text or antialiasing, but will detect the wrong text, the wrong font, or other failures.
[x] use Pillow's Gaussian blur, instead of numpy
[x] port the numpy image and smearing code
[x] probably fail if numpy isn't installed, because the smear calculation will be very slow
[x] fail if Pillow isn't installed
[x] increase the dtype size as the smear radius gets bigger?
Writing visual unit tests to compare graphics code is powerful, but often brittle. Things like antialiasing can make the comparison unreliable.
Adapt the smeared comparison from the four-letter blocks project, so that comparison will ignore minor changes in text or antialiasing, but will detect the wrong text, the wrong font, or other failures.
port the numpy image and smearing codeprobably fail if numpy isn't installed, because the smear calculation will be very slowincrease the dtype size as the smear radius gets bigger?