Detect duplicate colours

[lildude]

Description

As discovered in #4182, our current testing allows the reuse of colours as our test wasn't catering for the fact that in the event the two colours are the same, the result from past_threshold?() would fail and include the same language twice, one for the newly added language and one for the duplicate.

This PR addresses that.

With the change to the test I discovered we have four duplicates:

$ bundle exec ruby test/test_color_proximity.rb
Run options: --seed 55279

# Running:

F

Finished in 0.932629s, 1.0722 runs/s, 1.0722 assertions/s.

  1) Failure:
TestColorProximity#test_color_proximity [test/test_color_proximity.rb:21]:
The following 8 languages have failing color thresholds. Please modify the hex color.
- Glyph (#e4cc98) is too close to ["e4cc98", "e4cc98"]
- Harbour (#0e60e3) is too close to ["0e60e3", "0e60e3"]
- KRL (#28431f) is too close to ["28431f", "28431f"]
- NCL (#28431f) is too close to ["28431f", "28431f"]
- Perl (#0298c3) is too close to ["0298c3", "0298c3"]
- Ring (#0e60e3) is too close to ["0e60e3", "0e60e3"]
- Tcl (#e4cc98) is too close to ["e4cc98", "e4cc98"]
- VCL (#0298c3) is too close to ["0298c3", "0298c3"]

1 runs, 1 assertions, 1 failures, 0 errors, 0 skips
$

I've adjusted those slightly by picking the language that had its colour set most recently and then:

• loaded that colour into Google's colour picker,
• added or subtracted at least 16 (the minimum acceptable distance) from the colour,
• verified it looked similar to the original colour,
• updated languages.yml with the colour,
• ran just the proximity tests: bundle exec ruby test/test_color_proximity.rb,
• rinsed and repeated, adjusting the difference by a few here and there, until I got a non-clashing colour.

When that didn't work or was taking too long, I dragged the colour cursor ever-so-slightly and did the same thing again.

The results:

Glyph: #e4cc98 -> #c1ac7f
KRL: #28431f -> #28430A
Ring: #0e60e3 -> #2D54CB
VCL: #0298c3 -> #148AA8

I think those are close enough to the originals and certainly meet our proximity requirements.

Checklist:

• I am adding new or changing current functionality fixing a test
  • [N/A] I have added or updated the tests for the new or changed functionality.

[Alhadis]

I'm a little surprised that this remained under our noses for this long, haha.

LGTM! 👍

[pchaigno]

I'm a little surprised that this remained under our noses for this long, haha.

I mean, what's the chance two contributors will pick the exact same color? unless they copy paste...

[lildude]

I mean, what's the chance two contributors will pick the exact same color? unless they copy paste...

I’d normally say low, but Ring is a case of a very recent example and #4182 is an example of doing so just to keep the tests happy 😆

[aaronfranke]

You mentioned using Google's color picker to find alternative colors. Do you know if there's a tool for finding a "nearby" suitable/non-conflicting color? As you mentioned, blue is a fairly popular color, so this might not be easy just guess-and-checking unless there's a tool.

EDIT: I found a suitable color, but such a tool may be useful for others.

[lildude]

Do you know if there's a tool for finding a "nearby" suitable/non-conflicting color?

Unfortunately not. 😞