4 ms·

Google releases its own Javascript Lint tool

115 points by edd 16y ago

thechangelog 16y ago
Does anyone know how this would differ from jslint? I couldn't find anything immediately obvious.
- maw 16y ago
  It gives you a different set of ridiculous warnings about things that have nothing to do with your code's correctness. Ha ha, only serious But for real there are many warnings that pretty plainly seem to be about enforcing somebody's preferences -- gjslint wants you to write "function()" but jslint wants you to write "function ()", for example. Enforcing conventions is good, but I'm pretty sure it's orthogonal to flagging real, common problems (in javascript land, things such missing "var" or ";") Using both and suppressing the stupid warnings doesn't seem like a bad idea, if you can take the time to get it right.
  - btilly 16y ago
    In the example they gave, they had 9 errors. 1 was something that leads to browser incompatibility. (Trailing comma at the end of the array. IE doesn't implement the spec, Firefox does.) 1 was something that often points to logic errors. (Inconsistent indentation.) And the other 7 were someone's idea of correct style. That said, I suspect that this lint tool will catch some issues where technically correct code can make http://code.google.com/closure/compiler/ http://code.google.com/closure/compiler/ unhappy. If you're using the Closure library, that's kind of the whole point.
  - thezilch 16y ago
    Note: It would appear the coding-style assertions can be attributed to the "--strict" flag, used in the example. YMMV, but I get the above impression from the docs (How To) -- http://code.google.com/closure/utilities/docs/linter_howto.html http://code.google.com/closure/utilities/docs/linter_howto.h...
- warp 16y ago
  A very important difference for me is that jslint is not under a GPL-compatible license. jslint looks like open source, but isn't. Closure Linter is under Apache v2, which is GPLv3 compatible.
maw 16y ago
The installation instructions tell you to run "easy_install http://closure-linter.googlecode.com/files/closure_linter-latest.tar.gz http://closure-linter.googlecode.com/files/closure_linter-la.... All well and good, but install instructions that don't reference specific version numbers freak me out. I might have gotten a little too much reproducibility religion, with all that that entails, but it seems better than being so cavalier when setting up environments.
- rquirk 16y ago
  I'm not sure what easy_install does exactly, but I guess you could replace latest with a version number. On http://code.google.com/p/closure-linter/downloads/ http://code.google.com/p/closure-linter/downloads/ there's a 2.2.1 there, so instead of using "latest", you could use "2.2.1". If you care enough about this, then you probably could do some digging. Having a "latest" means they don't have to update the instructions every release, and for most people this will Just Work.
  - prodigal_erik 16y ago
    easy_install is yet another single-language package manager. It doesn't interop with the legitimate OS package manager's database of installed dependencies, and your admins probably have very little experience with it. Best to run it on one sacrificial box and then turn whatever it leaves behind into a real RPM package (or whatever your platform wants).
    - maw 16y ago
      You can set things up so that easy_install installs into a special root. And you nearly always should. The closest to an exception I can come up with is when you rely on image-based installs instead of package-by-package upgrades. And even then I can't manage to convince myself. No exceptions at all for individual developers' machines.
eclark 16y ago
It fails on file I have thrown at it with some complexity other than jquery.
kilian 16y ago
Ouch, they shipped with a broken fixjsstyle command: http://groups.google.com/group/closure-linter-discuss/browse_thread/thread/30fa86792d11b208 http://groups.google.com/group/closure-linter-discuss/browse... Luckily it didn't break my jsfile, but that's kind of clumsy.
- acdha 16y ago
  After fixing that, I ran into an even more surprising error which someone had already reported - turns out a simple JavaScript closure chokes it: (function() {})(); http://code.google.com/p/closure-linter/issues/detail?id=3 http://code.google.com/p/closure-linter/issues/detail?id=3
  - Hovertruck 16y ago
    I remember running into this issue with their minifier at some point as well.
kilowatt 16y ago
I was hoping I could run this realtime in vim, but it appears too slow on a 1000 line JS file.
stevelosh 16y ago
Here's a couple of lines to let you use this with :make (on a single file) and the QuickFix window in Vim: au BufNewFile,BufRead *.js set makeprg=gjslint\ % au BufNewFile,BufRead *.js set errorformat=%-P-----\ FILE\ \ :\ \ %f\ -----,Line\ %l\\,\ E:%n:\ %m,%-Q,%-GFound\ %s,%-GSome\ %s,%-Gfixjsstyle%s,%-Gscript\ can\ %s,%-G If any experts in Vim's makeprg/errorformat want to clean this up, that would be awesome. I just hacked it together and it seems to work well enough.

115 points by edd 16y ago

thechangelog 16y ago

Does anyone know how this would differ from jslint? I couldn't find anything immediately obvious.

maw 16y ago
It gives you a different set of ridiculous warnings about things that have nothing to do with your code's correctness. Ha ha, only serious But for real there are many warnings that pretty plainly seem to be about enforcing somebody's preferences -- gjslint wants you to write "function()" but jslint wants you to write "function ()", for example. Enforcing conventions is good, but I'm pretty sure it's orthogonal to flagging real, common problems (in javascript land, things such missing "var" or ";") Using both and suppressing the stupid warnings doesn't seem like a bad idea, if you can take the time to get it right.
- btilly 16y ago
  In the example they gave, they had 9 errors. 1 was something that leads to browser incompatibility. (Trailing comma at the end of the array. IE doesn't implement the spec, Firefox does.) 1 was something that often points to logic errors. (Inconsistent indentation.) And the other 7 were someone's idea of correct style. That said, I suspect that this lint tool will catch some issues where technically correct code can make http://code.google.com/closure/compiler/ http://code.google.com/closure/compiler/ unhappy. If you're using the Closure library, that's kind of the whole point.
- thezilch 16y ago
  Note: It would appear the coding-style assertions can be attributed to the "--strict" flag, used in the example. YMMV, but I get the above impression from the docs (How To) -- http://code.google.com/closure/utilities/docs/linter_howto.html http://code.google.com/closure/utilities/docs/linter_howto.h...
warp 16y ago
A very important difference for me is that jslint is not under a GPL-compatible license. jslint looks like open source, but isn't. Closure Linter is under Apache v2, which is GPLv3 compatible.

maw 16y ago

The installation instructions tell you to run "easy_install http://closure-linter.googlecode.com/files/closure_linter-latest.tar.gz http://closure-linter.googlecode.com/files/closure_linter-la.... All well and good, but install instructions that don't reference specific version numbers freak me out. I might have gotten a little too much reproducibility religion, with all that that entails, but it seems better than being so cavalier when setting up environments.

rquirk 16y ago
I'm not sure what easy_install does exactly, but I guess you could replace latest with a version number. On http://code.google.com/p/closure-linter/downloads/ http://code.google.com/p/closure-linter/downloads/ there's a 2.2.1 there, so instead of using "latest", you could use "2.2.1". If you care enough about this, then you probably could do some digging. Having a "latest" means they don't have to update the instructions every release, and for most people this will Just Work.
- prodigal_erik 16y ago
  easy_install is yet another single-language package manager. It doesn't interop with the legitimate OS package manager's database of installed dependencies, and your admins probably have very little experience with it. Best to run it on one sacrificial box and then turn whatever it leaves behind into a real RPM package (or whatever your platform wants).
  - maw 16y ago
    You can set things up so that easy_install installs into a special root. And you nearly always should. The closest to an exception I can come up with is when you rely on image-based installs instead of package-by-package upgrades. And even then I can't manage to convince myself. No exceptions at all for individual developers' machines.

eclark 16y ago

It fails on file I have thrown at it with some complexity other than jquery.

kilian 16y ago

Ouch, they shipped with a broken fixjsstyle command: http://groups.google.com/group/closure-linter-discuss/browse_thread/thread/30fa86792d11b208 http://groups.google.com/group/closure-linter-discuss/browse... Luckily it didn't break my jsfile, but that's kind of clumsy.

acdha 16y ago
After fixing that, I ran into an even more surprising error which someone had already reported - turns out a simple JavaScript closure chokes it: (function() {})(); http://code.google.com/p/closure-linter/issues/detail?id=3 http://code.google.com/p/closure-linter/issues/detail?id=3
- Hovertruck 16y ago
  I remember running into this issue with their minifier at some point as well.

kilowatt 16y ago

I was hoping I could run this realtime in vim, but it appears too slow on a 1000 line JS file.

stevelosh 16y ago

Here's a couple of lines to let you use this with :make (on a single file) and the QuickFix window in Vim: au BufNewFile,BufRead *.js set makeprg=gjslint\ % au BufNewFile,BufRead *.js set errorformat=%-P-----\ FILE\ \ :\ \ %f\ -----,Line\ %l\\,\ E:%n:\ %m,%-Q,%-GFound\ %s,%-GSome\ %s,%-Gfixjsstyle%s,%-Gscript\ can\ %s,%-G If any experts in Vim's makeprg/errorformat want to clean this up, that would be awesome. I just hacked it together and it seems to work well enough.