Add `browserslist.lint()`

[ai]

Add a method to find errors in current Browserslist config.

Possible checks:

1. missed-not-dead: on last N versions without not dead
2. limited-browsers: on queries only with browser names (and < 7 browser names)
3. country-was-ignored: on significant (≥20%) drop of coverage in some country (we can use all countries or only countries with ≥1M population).

[dangreen]

@ai

on significant (≥20%) drop of coverage in some country (we can use all countries or only countries with ≥1M population).

As I understood, if coverage for country is <80% then issue should be reported. But then "defaults" query will not pass this check (for countries with ≥1M population)

$ browserslist --coverage=AM "defaults"                  
These browsers account for 59.15% of all users in the AM

$ browserslist --coverage=MA "defaults"
These browsers account for 58.51% of all users in the MA

$ browserslist --coverage=RU "defaults"
These browsers account for 72.13% of all users in the RU

$ browserslist --coverage=SG "defaults"
These browsers account for 61.77% of all users in the SG

Is this ok?

[g-plane]

I think linter can be implemented in a separated pacakge, as https://github.com/browserslist/browserslist/issues/549#issuecomment-1044474071 said.

[ai]

I think linter can be implemented in a separated pacakge, as https://github.com/browserslist/browserslist/issues/549#issuecomment-1044474071 said.

I can move it by my own

As I understood, if coverage for country is <80% then issue should be reported. But then "defaults" query will not pass this check (for countries with ≥1M population)

Wow! It is a shame for our defaults.

But let add warning first and then we will think how to improve defaults (as part of another task).

[dangreen]

@ai it looks weird

$ browserslist --coverage=RU "last 100 versions"
These browsers account for 83.4% of all users in the RU

[ai]

Agree. Can I ask you to investigate? Maybe we have an error in coverage calculation?

[dangreen]

@ai 👌

[dangreen]

@ai In caniuse country data there is "total" value, it can be not 100%

https://github.com/Fyrd/caniuse/blob/main/region-usage-json/RU.json

"total":81.56795

my suggestion is to interpret this value as 100% and recalculate coverage relative to it

var total = 0

for (var j in data) {
  total += data[j] || 0
}

var tx = 100 / total

return browsers.reduce(function (all, i) {
  var usage = data[i] * tx
  if (usage === undefined) {
    usage = data[i.replace(/ \S+$/, ' 0')] * tx
  }
  return all + (usage || 0)
}, 0)

then coverage for "defaults" is

$ browserslist --coverage=AM "defaults"                  
These browsers account for 59.33% of all users in the AM

$ browserslist --coverage=MA "defaults"
These browsers account for 52.39% of all users in the MA

$ browserslist --coverage=RU "defaults"
These browsers account for 88.36% of all users in the RU

$ browserslist --coverage=SG "defaults"
These browsers account for 64.85% of all users in the SG

Example with "last 100 versions"

Before:

$ browserslist --coverage=AM "last 100 versions"                  
These browsers account for 97.63% of all users in the AM

$ browserslist --coverage=MA "last 100 versions"
These browsers account for 95.88% of all users in the MA

$ browserslist --coverage=RU "last 100 versions"
These browsers account for 81.5% of all users in the RU

$ browserslist --coverage=SG "last 100 versions"
These browsers account for 94.91% of all users in the SG

After:

$ browserslist --coverage=AM "last 100 versions"                  
These browsers account for 99.96% of all users in the AM

$ browserslist --coverage=MA "last 100 versions"
These browsers account for 99.98% of all users in the MA

$ browserslist --coverage=RU "last 100 versions"
These browsers account for 99.92% of all users in the RU

$ browserslist --coverage=SG "last 100 versions"
These browsers account for 99.93% of all users in the SG

[ai]

Seems like total ≠ 100% because of browsers which are not on Can I Use (like Yandex browser).

Let’s keep old logic for --coverage and use / total only in linter.

[dangreen]

@ai ok. But it solve problem only for RU.

Weak places in other regions:

AM 'firefox 52': 37.23163, - popular but not in "defaults"

MA 'ios_saf 6.0-6.1': 14.55397, - popular but not in "defaults"

SG 'android 4.4.3-4.4.4': 21.76044, - popular but not in "defaults"

[ai]

Yes, we will think about solving this problem later.

BTW, it will be nice to suggest in warning what browser to add to fix coverage in the region.

[nickmccurdy]

It seems like this hasn't been updated since browserslist-lint was created. Can this be updated or closed?