Proposal: Check outlying precision while running linter

[service-paradis]

Issue: #3120, #3169

Description

Only a proposal and some adjustments might be needed.

I added a custom lint check to verify the average precision level. I report an error and update the ignored list if the average precision level is below or above a defined treshold. I used the values that were indicated in #3120. The error reported looks like this:
Precision level should be between ${iconMinFloatPrecision} and ${iconMaxFloatPrecision}; the average is currently ${precisionAverage}

Ex.

[ericcornelissen]

Thanks for the detailed review so far @mondeja and thanks for the suggestion @service-paradis. I definitely like the upper bound to ensure small file sizes (whether 5 is the correct value is debatable but that can be changed if needed). I'm not so font of the lower bound, if an icon can work with a precision of 0 or 1 that's great! Unfortunately, I think we will need to leave icons with a precision that is too low to manual reviews... Thoughts?

Two question

• Is "average precision" the right choice here? Shouldn't we be looking at the maximum precision?
• I'm surprised by the number of ignored cases in .svglint-ignored.json, in Review of icons with outlying precision levels #3120 (and The Great 2020 Icon Review #3169) there are only 38 icons listed. I'm not sure how we got the lists there, maybe that was done manually... Anyway, any idea why there is a discrepancy?

[mondeja]

I think we will need to leave icons with a precision that is too low to manual reviews... Thoughts?

I agree with you, @ericcornelissen. Discrepances in 0.01 pixel ranges and lower are easy to check reviewing manually and can be uploaded an icon with path fitting entirely in integer points.

• Is "average precision" the right choice here? Shouldn't we be looking at the maximum precision?
• I'm surprised by the number of ignored cases...

• I'm also inclined to try the maximum number of decimal places check instead of the mean.
• I was confusing about this discrepance because of Review of icons with outlying precision levels #3120 from Add linter to check svg path dimensions #3107 discussion. Perhaps the list, as you have pointed, was manually created.

[service-paradis]

Thanks for the feedback!

• I pinned svgpath version
• I use only the maximum precision level instead of the average
• I validate only if it is greater than the defined maximum

The number of ignored icon is way less this way!