I'm not sure if this is a problem with textmate2 or with the bundle, but "console" calls are showing up in the function browser dropdown (and at the wrong indentation, no less).

Is there an easy fix?

Here's a small amount of code to cause this:

class Test
    constructor: ->
        console.log()

I installed the tmbundle following the commands and I can successfully reload the bundles, however, why I don't see the change of syntax highlighting? How do I make syntax highlighting work?

Using the option key in combination with the arrow keys moves the cursor to word boundaries, where a boundary is any one in the set of word separators for a language. In the tm bundles for most other languages, the period is treated as a word separator. This is not the case in the current implementation of coffee-script-tmbundle.

There is no such thing as an Application dir in any Linux distro. Please don't state "tested in Ubuntu 11.04" if you didn't.

E.G: in Ubuntu 11.10, the bundle can be copied, without bein zipped, directly in .config/sublime-text-2/Packages

I'm trying to recreate TextMate's Support/lib/ui.rb's #menu helper which calls out to the tm_dialog cocoa app, but my attempt is hanging textmate. I don't know NodeJS well enough to know if I'm calling out correctly (see "Show menu (demo).tmCommand" in my fork, but reproduced below too).

My attempt (which hangs TM):

dialog: process.createChildProcess dialog_cmd, ["-u"]
dialog.addListener "exit", (code) ->
  dialog.close()
dialog.addListener "output", (data) ->
  res: data.strip()

Ruby's version:

res = ::IO.popen("#{TM_DIALOG} -u", "r+") do |io|
  Thread.new do
    plist = { 'menuItems' => options }.to_plist
    io.write plist; io.close_write
  end
  OSX::PropertyList::load(io)
end

It includes the RubyCocoa call OSX::PropertyList::load(io) which I cannot find in the Xcode docco :/

Any thoughts on how I can reproduce the Ruby call in CoffeeScript/NodeJS?

I "test" my command by opening it into TextMate (Ctrl+Cmd+E, and pressing Cmd+R; that hangs TM. Also activating the command hangs TM).

On most setups (all?), coffee script uses soft tabs (just like Ruby). I couldn't manage to enable soft tabs in this bundle, setting environment variables in the bundle's config didn't work:

    {   name = 'TM_SOFT_TABS';
        value = 'YES';
    },
    {   name = 'TM_TAB_SIZE';
        value = '2';
    },

I added to my coffee script bundle a new command which runs CoffeeLint for making sure my coffee script is nice and clean

• Scope Selector: source.coffee
• Key Equivalent: ^⌘L
• Save: Current Document
• Output: Show in New Window Format: Html

#!/bin/bash
coffeelint "$TM_FILEPATH" | ansifilter -H;

AnsiFilter i found on homebrew and i converts shell outputs

http://www.coffeelint.org/
http://www.andre-simon.de/doku/ansifilter/en/ansifilter.html

I suppose someone can figure out if adding this into this bundle is a good thing or not... I just wanted to show other people what i have setup

A few minor things:

1. When I type # inside of a """ string, it should auto-expand to #{_} the same way that it does inside of a " string.
2. Since Cmd+" generates """\n_\n""", why not make Cmd+' generate '''\n_\n'''? The shortcuts could be labeled "Generate Heredoc Quotes (double)" and "Generate Heredoc Quotes (single)." Typing heredoc strings without a shortcut in TextMate is a bit nasty, because three ' keypresses result in '''_'...
3. When I type ###, it ought to expand to ###_###, don't you think? Or perhaps ###\n_\n### as with the heredoc quote keyboard shortcut mentioned above.

It would appear that this change in CoffeeScript assignment operators:

"0.9.0 introduces a massive backwards-incompatible change: Assignment now uses =, and object literals use :, as in JavaScript."

It hasn't bubbled back into the TextMate bundle snippets.

Just a heads-up, it's a small change to the following snippets:

Function, Class, Require

and the following Commands:

New Function

I believe that's it, thanks!
-Kris

Hello, I am wondering if this is possible or am I doing something wrong?

the following code:

Mustache = require 'mustache'

person =
  name : "john"
  email: "[email protected]"

template = """
  <p>Hello {{name}}</p>
  <p>your email is: {{email}}</p>
"""

output = Mustache.to_html template, person

console.log output

works fine when using cat test.coffee | coffee -s through terminal, but if running with cmd+b the Run command within textmate it returns:

Error: Cannot find module 'mustache' at Function._resolveFilename (module.js:289:11) at Function._load (module.js:241:25) at require (module.js:317:19) at Object. (.:3:14) at Object. (.:11:4) at Module._compile (module.js:373:26) at Object.run (/usr/local/lib/node/.npm/coffee-script/1.0.1/package/lib/coffee-script.js:55:19) at /usr/local/lib/node/.npm/coffee-script/1.0.1/package/lib/command.js:115:29 at Socket. (/usr/local/lib/node/.npm/coffee-script/1.0.1/package/lib/command.js:149:14) at Socket.emit (events.js:39:17)

Is there any way to use require modules using textmate bundle?
It seems require 'sys' is possible though...

symbols are interpreted as comments in extended regex even if they are escaped, which works when run

Here's a minor syntax highlighting glitch:

foo: bar

highlights foo differently than

foo : bar

Relately, foo= bar highlights foo differently than foo = bar.

These three snippets lack a UUID which prevents them from being loaded in TextMate. Didn't do a pull request as I am unsure of their purpose and where they would fit in the menu structure. (You can generate a UUID with uuidgen in terminal.)

I'm not getting proper highlighting of anything after the "@something" in "class @something extends SomethingElse"

I'm not using TextMate; I'm using SublimeText, which is compatible with tmLanguage files, just so you know.

Clicking the arrow to collapse text is a no-op in TM2 (build 8971).

The only bad part about Coffee Script is that it reports it's errors as line numbers in the compiled result.

Is there anyway the Compile and Display JS window could show line numbers? (or even paste itself into a new TextMate document) This way it would be easier to see where the error occurred in the compiled JS from a stack trace.

I've started to find the "Go to Symbol" prompt to be extremely useful. Unfortunately, there are a lot of problem cases with it. It seems to treat : and = interchangeably everywhere (not just in a class), so

a = -> b: ->

gives you

a
  b

but

c = -> {d: ->}

gives you

c
d

Neither seems very sensible to me; functions in objects (other than class prototypes) should be ignored, the way that functions given as callbacks are. Don't you think? Or is it more important to list every function that's assigned a symbol than to present a clean hierarchy? Keep in mind that the current behavior means that if you write

$('#item1').bind click: -> false
$('#item2').bind click: -> false
$('#item3').bind click: -> false
$('#item4').bind click: -> false
$('#item5').bind click: -> false

then you'll see "click" 5 times in the "Go to Symbol" list.

It would be helpful if super was highlighted when defining classes (unless it already is and I just have an old version). http://coffeescript.org/#classes

Thanks!

Hello!

Well, the title tells everything.

Have a nice day! :)

Loupi

see examples 2 & 4

IcedCoffeeScript is an extension of CoffeeScript, with two new keywords (await and defer) that look just like function calls. So the syntax highlighting for CoffeeScript is fine for IcedCoffeeScript as well.

Therefore, it would be nice if coffee-script-tmbundle recognized .iced as an extension that should be treated as CoffeeScript.

I don't know if this is the right place to do this. The TM2 bundle repository doesn't have an issue tracker attached to it.

From Bartłomiej Kubiak on the Textmate mailing list:

CoffeeScript bundle has broken syntax highlighting 
when using division operator ("/") - badly interprets it 
as start of regexp…

This might be the culprit.

This is a enhancement request for string interpolation expressions (not sure if it's the correct name). Basically I would like the string interpolation expressions behave like they do in the Ruby bundle.

1. Syntax highlighting in string interpolation expressions
2. If the cursor is in a string and I type # the bundle would add {} and place the cursor between the braces
3. Selecting a word in a string and pressing # would wrap the word in #{}

Currently,

class Foo.Bar

highlights Foo as a class name, but not Bar. I believe the most correct behavior would be for the entire string, Foo.Bar, to be highlighted. Or, perhaps just Bar should be highlighted, since that's the name given to the constructor function.

Note that the entire string Foo.Bar is highlighted in

new Foo.Bar

Extra bonus bug: While reporting the bug above, I discovered that single-char class names aren't highlighted at all. So class A.B doesn't highlight A, while class AA.B does highlight AA.

Say the object obj has the property return. (Hey, I didn't choose it.) Then the line

obj.return = 1

compiles just fine, but the word return is highlighted like a keyword. Likewise for any other keyword. Obviously one could type obj['return'], but it would be nice to have accurate highlighting for the dot syntax.

I added

|^\s*class\s.*$

to foldingStartMarker. I'm not sure if it covers all cases, but it's better than nothing. At least it my classes fold. ^_^

Add this line at the end to do so:

osascript -e 'tell app "TextMate" to reload bundles'

Folding doesn't work on coffeescript functions, only on code that looks like javascript. I changed my folding markers as follows, and while it no longer folds on {}'s (which I prefer anyway) it works properly for functions...

foldingStartMarker = '.*(->|=>)\s*$';
foldingStopMarker = '^\s*$';

Currently all the indentation in snippets is hardcoded to two spaces. If you change from using spaces to tabs in the snippets then you configure how large you want the indentation to be using soft tabs. The previous behaviour is available if you set Tab Size: 2 and Soft Tabs. If you want to indent 4 spaces instead you can set Tab Size: 4 and Soft Tabs.

Hitting apple-B to compile& show a coffeescript file fails if it contains a string constant with a unicode character such as "ẚ".

For instance, if the entire file is:

'ẚ'

Hitting apple-b will fail, saying

*** Error while highlighting: UnicodeDecodeError: 'ascii' codec can't decode byte 0xe1 in position 1: ordinal not in range(128) (file "/Library/Python/2.6/site-packages/Pygments-1.4-py2.6.egg/pygments/lexer.py", line 148, in get_tokens)

According to rtomayko/rocco#10 if you add "-O encoding=utf8" to the call of pygmentize it resolves this issue.

Compiling a coffee script file with TextMate (Compile and Display JS) gives a different result compared to compiling with coffee on the command line. This happens if I insert raw javascript with backticks and the javascript code is invalid. For example:

<% if @autoload %>

The above will be compiled into an empty expression ending with a semicolon.

49a55aa removed line numbers from the compiled output - I really prefer them for tracking down browser-side errors.

Any chance it was accidental?

Looks like the colorization was been removed? I checked the issues and did not see anything regarding this.

This code in textmate:

sys: require 'sys'
http: require 'http'

server: http.createServer (req,res) ->
res.writeHeader 200, {'Content-Type':'text/plain'}
res.write 'Hello, World!'
res.end()

server.listen 3000

sys.puts "Server running at http://localhost:3000/"

Produces this:
({
sys: require('sys'),
http: require('http'),
server: http.createServer(function(req, res) {
res.writeHeader(200, {
'Content-Type': 'text/plain'
});
res.write('Hello, World!');
return res.end();
})
});
server.listen(3000);
sys.puts("Server running at http://localhost:3000/");

Which as expected doesn't work:
ReferenceError: http is not defined

Changing to equals fixes it:

sys = require 'sys'
http = require 'http'
server = http.createServer (req,res) -> ...

Slight bug in Literate Coffeescript when code follows a markdown list

The line

str.replace(/\\/g, '')

is perfectly valid CoffeeScript, removing all backslashes from a string. However, the regex /\\/ doesn't get highlighted. Worse, if there's a slash later on the line, then regex highlighting goes all the way up to it, e.g.

str.replace(/\\/g, '\\\\').replace(/\\"/g, '"')

highlights

/\\/g, '\\\\').replace(/

In short, the syntax highlighter isn't currently distinguishing between a single backslash (which escapes the closing / of a regex) and a double backslash (which doesn't). More precisely, it needs to ignore any even number of backslashes preceding the closing \, since the current behavior is correct for a regex like

/\\\/foo/

I'm an inexperienced TextMate user, so be nice. ;)

With Check Spelling While Typing enabled, spelling mistakes in string literals are underlined but those in comments are not. According to the 4.10 Spell Checking…

By default spell checking is disabled for source code except strings and comments

Any idea as to how I'd go about having spell-checking apply to comments as well as strings? ack spell returns no results – it would appear that the spellChecking preference mentioned in the TextMate documentation is not being specified. Why then, I wonder, does the behaviour differ from the (quoted) default?

Seems like the increaseIndentPattern should have an entry for -> so that function bodies get indented correctly

At least one TextMate command ('Continue Line Comment') relies on the scope specifying the character used to start a line comment. The coffeescript grammar is currently not following that convention.

It's a quick fix (change 2 occurrences of comment.line.coffee to comment.line.number-sign.coffee) that I have ready. I'll merge it asap.

by adding the following to README.markdown:

osascript -e 'tell app "TextMate" to reload bundles'

Thanks!

Coffee script seems to work OK in the terminal

coffee -v #=> CoffeeScript version 1.1.1

npm -v # => 1.0.6

node -v # => v0.4.8

When I try to compile to JS in TextMate I get this error:

module:238
throw new Error("Cannot find module '" + request + "'");
^
Error: Cannot find module 'vm'
at loadModule (module:238:15)
at require (module:364:12)
at Object. (/usr/local/lib/coffee-script/lib/coffee-script.js:5:8)
at Object. (/usr/local/lib/coffee-script/lib/coffee-script.js:110:4)
at Module._compile (module:384:23)
at Module._loadScriptSync (module:393:8)
at Module.loadSync (module:296:10)
at loadModule (module:241:16)
at require (module:364:12)
at Object. (/usr/local/lib/coffee-script/lib/command.js:7:18)

TextMate Shell variables:

http://cl.ly/1Y2o2W3n3V2M3U0H183I

Paths:

http://cl.ly/2T2p2A0u3d0C1h3F042K

Related to #99: When I'm editing an IcedCoffeeScript file (.iced extension), it would be nice if await and defer were highlighted as keywords (or in some way had a different color from ordinary identifiers).

In plain old CoffeeScript (.coffee), await and defer are just ordinary identifiers. So it might make sense (if it's easy to do so -- I know very little about authoring TextMate bundles) to only highlight the extra keywords if the filename is .iced. Then again, IcedCoffeeScript can run plain old .coffee files too (and runs them as IcedCoffeeScript, await/defer and all), so it's debatable whether it should be extension-sensitive.

For some reason, the line

a /= 255

is colored incorrectly, as are lines that follow. The _=, += and -+ operators do _not* have this problem. Looking at the .tmLanguage file, I see that /= is defined differently (line 381), preceded in the regex by

(?<!\()

I don't understand the reason for this. Explanation?

Is this intentional, or would a patch be merged?

It would be nice if escape sequences could be highlighted in regular expression literals, just as in the JavaScript and Ruby bundles. For example:

a = /\t/

In the above code, both the JavaScript and Ruby bundles highlight "\t" with a slightly brighter green than the rest of the regular expression literal.

foo = (bar = false) ->

"false" is not highlighted in the above example.

Would be nice for the coffeescript language to be detected when being used inline with HTML:

<script type="text/coffeescript">

http://sublimetext.userecho.com/topic/55217-/

if you name a variable something like:

formData = 1
# gets indented after newline

When you hit enter, it will automatically indent even though its not a for loop.