At the very least, it would be nice to have aps-length throw a more intelligible error than ValueError: could not convert string to float: Unrecoverable
Better yet, identify works for nearly any image format, including .pdf. So, why not default to --figs identify? Is the logic that most users already have ghostscript (is it installed with most tex distros?) and they would need to install ImageMagick? In that case, perhaps the best thing to do is to see if the user has identify available on the command line, and, if so, just use that?
Similar to Issue https://github.com/jameskermode/aps-length/issues/1, I was also confused when I initially tried to use aps-length to count my manuscript, which uses .png figures.
At the very least, it would be nice to have aps-length throw a more intelligible error than
ValueError: could not convert string to float: Unrecoverable
Better yet,
identify
works for nearly any image format, including .pdf. So, why not default to--figs identify
? Is the logic that most users already have ghostscript (is it installed with most tex distros?) and they would need to install ImageMagick? In that case, perhaps the best thing to do is to see if the user hasidentify
available on the command line, and, if so, just use that?