coderaiser / putout-editor / 36719669410

100%

push

github

docs: memory: the two traps behind the check-documented-scripts fix

A finding closes into memory/ so the cause is not derived twice, and both
causes are reusable in any filesystem rule rather than in this one.

getFilename is absolute under redlint, so basename is the way to strip a
directory and not replace(/^\//, ''). And a crawlFile mask is matched against
basename, so 'package.json' returns every package.json in the tree - take the
sibling of the document by dirname, not the first.

The larger entry is the one that cost the time: parseFilesystem(['/', ...]) is
not the tree redlint produces, and a fixture built with it is wrong in the same
way the code is. It is root-relative and holds one file per name, so every
filesystem rule in this package has been tested against a shape the runner
never hands it. buildTree(process.cwd()) is the other source, and a rule needs
a fixture that came from the runner rather than from the helper the other
fixtures happen to use.

plugins.md gains the decision-table row for check-documented-scripts and two
sections: why matchFiles cannot be asked a question about a second file, with
the measurement rather than the argument, and when crawlFile beats trackFile.
It was asked for as a matchFiles rule, and the port would have reported zero
places with every test green - the inner plugin provably receives empty options,
no root and no trackFile.

The README section for the rule now says the same thing in two sentences, and
ideas.md takes the TypeScript migration write-up with the shape of the argument
it has to make: the rule contract is already typed upstream, so the claim is
that this package opts out, not that TypeScript is good - and the fixtures are
the part most likely to come out worse.

1081 of 1081 branches covered (100.0%)

Branch coverage included in aggregate %.

6670 of 6670 relevant lines covered (100.0%)

61.48 hits per line