A critique of commander for node.js
Because someone (@mgan) asked here is my critique of commander.js, a popular option parsing library for node.js, and comparison to node-dashdash, an option parsing library for node.js that I wrote. This isn’t meant as a “you suck, I rock” post. I wrote dashdash to fill my needs because other libs didn’t. I’d be happy if dashdash worked well for you, but I’m not selling it. :)
Pros
-
Commander.js involves less typing and more compact, which helps for quick readability.
-
[Update 2014-03-28: As of dashdash@1.5.0, custom option types are supported by dashdash.]
Commander.js supports custom parsing of option arguments. That’s good and something that dashdash doesn’t have. Dashdash has a number of canned types that so far have suited my needs. Adding custom option types would be worthwhile.
-
program.command(...)
support in commander.js is neat. I use node-cmdln for that, but the handling in commander.js is certainly more lightweight. -
Commander.js supports an option having an optional argument, e.g.:
node mycmd.js -f ARG node mycmd.js -f # no argument to option '-f'
Dashdash doesn’t support this, so if you need it, then boom.
-
Having many examples is good. [Update: Dashdash is growing some examples of its own.]
Cons
-
How to get rid of
-h|--help
and-V|--version
? I definitely advocate that all CLI tools have a those options (sometimes just “–version” and not the “-V” shortcut), but I don’t think an option lib should require it. -
That the lib uses
process.exit()
(handling –help, on unknown options) makes it difficult to integrate option processing in tools that aren’t a small command, e.g. a shell. -
Commander.js doesn’t allow
interspersed=false
. This is useful in node-cmdln, but I’ll grant most may not need this. -
No support for multiple uses of the same arg to build up an array (as supported by dashdash’s
arrayOf*
option types). Commander.js’slist
option type is another take on this:node mycmd.js --list a,b,c # program.list === ['a', 'b', 'c']
-
Both commander.js and dashdash have automated formatting of help output. However, dashdash intentionally only generates the option block and not the usage string and “options:” section prefix. That’s a balance in favour of being usable for other help output styles vs. requiring the user to write a little bit more themselves.
-
If order of options does matter then dashdash’s
<parsed>._order
can be helpful. Again a limited use case. -
Mixing parser and the parsed results. Obviously this is minor, but I like the ability to be able to log the parsed options and just have data logged. Also the sharing of namespace means that you can’t have an option named
Option
,Command
orargs
. In fairness in dashdash you can’t have an option named_args
or_order
. -
No equivalent to
allowUnknown=false
. There are PRs for this on commander.js -
Dashdash’s help output is more configurable. Minor.
There are lots of PRs and tickets open on commander.js, including one for an issue that I noticed when playing a bit. This can be good (lots of interested, issues you might have probably have code available). Far be it for me to complain about keeping up with issues and PRs on ones own projects.
Overall, I think commander.js looks fine for option parsing. Most of my “Cons” are either minor or a result of my just having had particular use cases that perhaps commander.js’s authors have not. I haven’t used commander.js other than briefly looking at it today, so I could very likely have made mistakes in grokking commander.js. Feedback is welcome.