spicyj noticed newer versions of babel-eslint seemed not to error on
no-unused-vars, and I was able to repro. It seems like something broke
between 3.1.9 and 3.1.10. (Smaller repro case and babel-eslint bug
report to come)
His commit failed on travis, but not on his local machine:
https://travis-ci.org/facebook/react/jobs/65468729
> The two callers of this function have different warning configs
> internally (static_upstream/core/createWarning.js) so we can't sync it
> like this without changing behavior. We should just split this out
> into two separate warning calls probably – this code is a little
> overabstracted.
https://github.com/facebook/react/pull/4021#discussion_r31690020
@spicyj
I think completely removing warnAndMonitorForKeyUse is a bit difficult, without
duplicating a ton of code. This at least ensures that the format string passed
to `warning` is unique. Plus, because the FB internal code in question only
matches the beginning of the format string, I think there should be zero
internal changes that need to be made to support this refactor.
The `warning` implementation checked that the message is long enough to
be useful. See commit f5038829d for more information.
It makes more sense to move this into a lint rule, and also to apply it
for both `warning` and `invariant`. We can safely remove stuff from the
`warning` implementation as we replace the function internally anyways.
https://github.com/facebook/react/pull/4021#issuecomment-108694976
@spicyj
The wording for an incorrect number of arguments based on counting %s
substitutions. The previous wording was backwards and a bit yoda-like
leading to possible ambiguity.
This should contain all the rules we probably want to use, except
react/wrap-multilines, which requires a larger codemod, and would
clutter this commit.
See #2869
Checks that the *second* argument of warning and invariant are a literal
string, or a concatination of literal strings, and that the number of
arguments is correct based on the number of %s substrings.
This commit also fixes a few places where the existing code had broken
error messages!
The rule itself is pretty straightforward, although adding the tests
ended up being a bit painful, as eslint-tester depends on mocha, and
therefore needs to be run in a separate grunt task.
This is a machine-generated codemod, but it's pretty safe since it was
generated by hooking into eslint's own report.
A few files had to be touched up by hand because there were existing
formatting issues with nested arrays/objects:
src/shared/utils/__tests__/OrderedMap-test.js
src/shared/utils/__tests__/Transaction-test.js
src/shared/utils/__tests__/traverseAllChildren-test.js
src/isomorphic/children/__tests__/ReactChildren-test.js
Introducing: a really lame version of composite components, right inside of ReactDOMComponent!
Now ReactDOMInput isn't an actual component. This brings us closer to exposing DOM nodes as refs.
Closes#3971.
> After #3968, the next thing we should do is start linting our tests.
> Historically we've ignored them due to lack of parser compatibility.
> But that shouldn't be a problem anymore. We may want to integrate
> https://www.npmjs.com/package/eslint-plugin-react to more aggressively
> lint our JSX in tests.
I understand this diff touches a lot of stuff, so I tried to keep it to
a near-minimal set of changes to make eslint happy.
Benjamin Woodruff
Paul O’Shannessy
Ben Alpert
Ben Alpert
Sebastian Markbåge
Edvin Erikson
Toru Kobayashi
Jim
Jim
Stefan Dombrowski
Wildhoney
Don Abrams