* Add option to insert @format pragma if absent
prependPragma --> insertPragma
add tst dir, move stripBom, add to README, update snapshots
remove noise from test + update snapshots
use jest@test new functions, update docs, rerun snapshots
fix accidental upgrade to jest@test
update jest-docblock again to latest with strip()
updated snapshot
* updated snapshots
* revert jest dep change
* apply a trim to parsed comments
* remove empty object
* Revert "apply a trim to parsed comments"
This reverts commit 467f71ce5f4b4f8a1b5c4474ee0484bc96c92141.
* rerun snapshots with proper jest version
* pin jest-docblock to latest @test release
* comment out docs for new feature from readme
* remove leading newlines
* ensure blank line between docblock and rest of file
* remove backtick string
* Make production tests always run against dist/
* Try to mock process.stdin instead of get-stream
* fix: mock `get-stream` from `index.js`
* refactor: remove unnecessary variable
* Try `transform: {}` in jest.config.js
* Add comment explaining the unusual get-stream mocking
* Fix Travis production builds
* Fix Travis production builds, attempt 2
* Remove `allow_failures` in .travis.yml
The reason we allowed the production builds to fail in the first place:
❯ git show 532ac0d3f
commit 532ac0d3f8
Author: Joseph Frazier <1212jtraceur@gmail.com>
Date: 2017-06-17 19:37:51 -0400
Build/test dist/ on Travis (#2176)
This helps us keep track of whether or not the dist/ build actually
works (see https://github.com/prettier/prettier/pull/2174). Note that
the dist/ jobs are allowed to fail, and do not affect overall build
time, so they shouldn't slow anything down (see
https://blog.travis-ci.com/2013-11-27-fast-finishing-builds/).
However, as we discovered in #2933 requiring the production builds to
succeed only adds about 1 minute of build time, so it is not worth the
risk of missing that the production builds fail.
* Account for empty lines in argument list in typical cases
* Fix build errors
* Fix one more build error
* Refactor
* Have better variable names for printing argument list
* Account for feedback on argument list empty lines
* Improve expanion argument logics
* Add a lot more tests
* Improve a test
* Make code cleaner
* Fix a lint error
* Refactor
* Add one last check
* Refactor
* Split Less and SCSS parsing into different parsers
Now, .less files are always parsed with postcss-less, and .scss files
areare always parsed with postcss-scss. This:
- Is less hacky.
- Is meant to avoid issues like #2829.
- Is probably more performant.
`parser: "postcss"` and `--parser postcss` continue to work like before:
First trying postcss-less, and if that fails, postcss-scss, unless a
regex says that we should try in the opposite order. The new values for
the parser option are "postcss-less" and "postcss-scss".
* Remove postcss from package.json since it is not used
* Rename parser-less to less and parser-scss to scss
* Deprecate parser:postcss in favor of parser:css
* Fix CSS tests