Moved from crop to extent for filling images

[twolfson]

In #32, we received a report of 2 images failing to properly compare. We resolved that the issue was due to us improperly using crop which lead to 1 error and 1 of the images had a poor encoding. The combination of these 2 lead gm to create 1x1 images to diff.

Reference images here: #33 (comment)

To remedy this, we are moving off of the improper command crop (which truncates images) to the more accurate extent (which extends images). In this PR:

• Added test to verify we preserve transparent backgrounds
• Moved from crop to extent
• Stopped writing to damn gm prototype, even if we subclass it
• Updated CHANGELOG
• Fixes Diff failing on Windows #32, closes Parameterize ImageMagick library selection. #33

For the release verison, please release as 1.4.0:

./release.sh 1.4.0

/cc @mlmorg

[Qix-]

Why not more readable error messages? These parameters are read in using input from the user (eventually) so there's a chance the consuming code could have erroneous input. An assert doesn't really explain the issue all that well.

[twolfson]

These assertions are for preventing typos from developers. There's not really a case where the error would come from user error:

• If the file doesn't exist, then we will get an ENOENT error during the earlier fs.stat/fs.exists call and we won't be upscaling the image
  • https://github.com/uber/image-diff/blob/1.3.0/lib/image-diff.js#L122-L128
  • https://github.com/uber/image-diff/blob/1.3.0/lib/image-diff.js#L132
  • https://github.com/uber/image-diff/blob/1.3.0/lib/image-diff.js#L33-L52
  • https://github.com/uber/image-diff/blob/1.3.0/lib/image-diff.js#L137
  • https://github.com/uber/image-diff/blob/1.3.0/lib/image-diff.js#L170-L173
• Otherwise, we will always be receiving height/width from gm.size so if those don't exist, then we have a deeper problem in gm itself

[Qix-]

👍 gotcha

[x3au]

Thanks for the fix!