Forum: Ruby Command-line option parsing

Greetings,

    I'm new to ruby (from Perl).  I notice there are several 
command-line
options parsing modules.  The two that stuck out were GetoptLong and
OptionsParser.  GetoptLong suffers from being "the same old Getopt".
OptionsParser claims to be "more advanced and more ruby-like", but has
extremely poor documentation.

    Are these the best two modules out there?  Is there an emerging
"standard" module? (In Perl, Getopt::Long is by far the most 
commonly-used
module).

(I apologize if this is a FAQ; I read the FAQ and did a google scan of 
past
articles.  Most of what I found was authors announcing their own 
options-
parsing modules).

Thanks in advance,

i like CommandLine::OptionParser
(http://rubyforge.org/docman/view.php/632/170/index.html)
because it's documentation is better than average as well
as its features.
it doesn't appear to be the ruby standard, however, but for
in-house projects there's a lot to like.

Eric J. Roode wrote:
> GetoptLong suffers from being "the same old Getopt".  

Well, what do you expect from an options parser?  It's pretty much a
solved problem, and using the standard getopt algorithms has the
advantage that users know what to expect.


mathew

Eric J. Roode <sdn.girths00869@zoemail.net> wrote:
> commonly-used module).
I used to use GetopLong in the beginning but switched to OptionParser.
Agreee, the documentation could be better but if you look carefully at 
the
example given in RDoc you get pretty much everything you need.  I'm not 
sure
about a standard parser but my impression was that OptionParser is more
widely used.  But I may be wrong here.

Kind regards

    robert

On 12/27/05, Lou Vanek <vanek@acd.net> wrote:
>
> i like CommandLine::OptionParser
> (http://rubyforge.org/docman/view.php/632/170/index.html)
> because it's documentation is better than average as well
> as its features.
> it doesn't appear to be the ruby standard, however, but for
> in-house projects there's a lot to like.


Yikes.
That is old documenetation and the link on Rubyforge was to be replaced 
eons
ago. The new docs are at:
  http://rubyforge.org/docman/view.php/632/233/posted-docs.index.html

There you will find the option parser docs after Application section. 
For
the most part, if you use the CommandLine::Application class, you don't 
need
to get
your hands that dirty with CommandLine::OptionParser.


If you have any questions, just post a question or send me
a private email. I plan to post a video on the usage this week. Maybe a
rails like video will generate more interest. :)

On Wednesday 28 December 2005 08:37 am, Robert Klemme wrote:
> > "standard" module? (In Perl, Getopt::Long is by far the most
>     robert
Does OptionParser come with standard Ruby?

SteveT

Steve Litt
http://www.troubleshooters.com
slitt@troubleshooters.com

Jim Freeze <jim@freeze.org> wrote:

> If you have any questions, just post a question or send me
> a private email. I plan to post a video on the usage this week. Maybe
> a rails like video will generate more interest. :)

Will there be stunt scenes with options and a high speed car chase?

    robert

On 12/28/05, Robert Klemme <bob.news@gmx.net> wrote:
>
>
> Will there be stunt scenes with options and a high speed car chase?


It will be action packed. There will be trains, rubies
and high speed typing!  ;)

mathew <meta@pobox.com> wrote in news:wjosf.9901$9e.8623
@tornado.texas.rr.com:

> Eric J. Roode wrote:
>> GetoptLong suffers from being "the same old Getopt".  
> 
> Well, what do you expect from an options parser?  It's pretty much a 
> solved problem, and using the standard getopt algorithms has the 
> advantage that users know what to expect.

Well... that's not *bad*, but I guess I had expected something a bit 
more
ruby-like.  Whatever that means.  :-)

"Robert Klemme" <bob.news@gmx.net> wrote in
news:41fiouF1e6diiU1@individual.net:

> I used to use GetopLong in the beginning but switched to OptionParser.
> Agreee, the documentation could be better but if you look carefully at
> the example given in RDoc you get pretty much everything you need. 
> I'm not sure about a standard parser but my impression was that
> OptionParser is more widely used.  But I may be wrong here.

I disagree about the example being all you need.

Here's the first option in the example:

    # Mandatory argument.
    opts.on("-r", "--require LIBRARY",
            "Require the LIBRARY before executing your script") do |lib|
      options.library << lib
    end

I am new to ruby, so I don't know what "options.library << lib" means. 
I
thought that that introduced a here-doc string.

The second example:

    # Optional argument; multi-line description.
    opts.on("-i", "--inplace [EXTENSION]",
            "Edit ARGV files in place",
            "  (make backup if EXTENSION supplied)") do |ext|
         ...

It took me a lot of staring at the code before it dawned on me that it
was the *brackets* that made it an optional argument.

The final example:

    # Another typical switch to print the version.
    opts.on_tail("--version", "Show version") do
      puts OptionParser::Version.join('.')
      exit
    end

I can't figure out how the on_tail method differs from the on method or
the on_head method.  Reading the source isn't helping.  The example
sucks.

The previous example has this comment:

    # No argument, shows at tail.  This will print an options summary.

"Shows at tail"??

At the end of the parse() method (and why is it self.parse, not just
parse?), there is this:

   opts.parse!(args)

I can't tell what function that performs for the class.

I can't tell how to make an option mandatory, how to configure it so 
that
some options require certain others or conflict with certain others.

It looks like a fine, powerful module.  But its lack of documentation
makes it pretty much useless for anyone who doesn't already know how to
use it.

I come from the Perl world.  You simply do not release a module to 
Perl's
CPAN unless it is thoroughly documented.  The vast majority of CPAN
modules have quite good documentation -- or at least, thorough
documentation.  I'm quite surprised to find that the situation is so
different in the ruby world.

Hi Eric

You may want to give CommandLine a try.
From the examples you give, this is how I would write an app:

require 'rubygems'
require 'commandline'

class MyApp < CommandLine::Application
  def initialize
    # Mandatory argument
    option :names => %w(--require -r),
               :opt_description => "Require the LIBRARY "+
                                            "before executing your 
script",
               :arg_description => "LIBRARY",
               :opt_found => get_arg,
               :opt_not_found => required

    option :names => %w(--inplace -i),
               :arity => [0,1],
               :opt_description => "Edit ARGV files in place",
               :arg_description => "[EXTENSION]",
               :opt_found => get_arg,
  end

  def main
    #put your code here
    p opts
  end
end#class MyApp

I don't know what #on_tail does either.

Eric J. Roode wrote:
> Well... that's not *bad*, but I guess I had expected something a bit more
> ruby-like.  Whatever that means.  :-)

That's a good expectation to have.

You mention OptionParser has poor documentation.  While it's not
complete, I find the example at [1] is quite adequate.

[1]
http://ruby-doc.org/stdlib/libdoc/optparse/rdoc/classes/OptionParser.html

Gavin

Eric J. Roode wrote:
>
> I disagree about the example being all you need.

Please disregard my other reply in this thread.  Your points are
perfectly valid and I'll try to improve the documentation.

> Here's the first option in the example:
>
>     # Mandatory argument.
>     opts.on("-r", "--require LIBRARY",
>             "Require the LIBRARY before executing your script") do |lib|
>       options.library << lib
>     end
>
> I am new to ruby, so I don't know what "options.library << lib" means.  I
> thought that that introduced a here-doc string.

The documentation's not at fault here: knowledge of Ruby is assumed :)

> The second example:
>
>     # Optional argument; multi-line description.
>     opts.on("-i", "--inplace [EXTENSION]",
>             "Edit ARGV files in place",
>             "  (make backup if EXTENSION supplied)") do |ext|
>          ...
>
> It took me a lot of staring at the code before it dawned on me that it
> was the *brackets* that made it an optional argument.

That will be easy to point out in the docs.  Thanks.

> sucks.
It means to show the "--version" option at the end (the "tail") of the
options list.

>
> At the end of the parse() method (and why is it self.parse, not just
> parse?), there is this:
>
>    opts.parse!(args)
>
> I can't tell what function that performs for the class.

Hopefully I can clear this up.  The difference between parse and parse!
is that parse! swallows ARGV.

> I can't tell how to make an option mandatory, how to configure it so that
> some options require certain others or conflict with certain others.

A "mandatory option" is a contradiction :)  You need to test for the
presence of an option in your own code.

> I come from the Perl world.  You simply do not release a module to Perl's
> CPAN unless it is thoroughly documented.  The vast majority of CPAN
> modules have quite good documentation -- or at least, thorough
> documentation.  I'm quite surprised to find that the situation is so
> different in the ruby world.

While not disagreeing in the least, I bet there are plenty of
badly-documented Perl modules, but they're not the often-used ones, so
they don't detract from the "Perl modules are well documented" meme.

Ruby is certainly worse, though!

Gavin

Gavin Sinclair wrote:

<snip>

> While not disagreeing in the least, I bet there are plenty of
> badly-documented Perl modules, but they're not the often-used ones, so
> they don't detract from the "Perl modules are well documented" meme.
>
> Ruby is certainly worse, though!
> 
> Gavin

How's that shell documentation coming along?

Dan

Eric J. Roode <sdn.girths00869@zoemail.net> wrote:
> I disagree about the example being all you need.
It seems to me that your posting pretty much confirms what I said: I 
didn't
claim it's "all you need" but that it gets you pretty far, if you "look
carefully". :-)  Admittedly it's not so easy for someone new to Ruby...

Kind regards

    robert

Daniel Berger wrote:

> How's that shell documentation coming along?

Not too well, just yet!

Gavin

> A "mandatory option" is a contradiction :)  You need to test for the
> presence of an option in your own code.

This is a fair point, but for complex command lines we sometimes really 
need options that are not optional. For example, I'm working on a ruby 
xmltv-like scraper, that has a bunch of regular options, plus essential 
arguments like --config CONFIG_FILE and --scraper SCRAPER_FILE for 
specifying files that are necessary.

I want to use options for these rather than arguments because they are 
both files and I prefer that the user specify the option --config or 
--scraper rather than having to get the order of them right on the 
command line.

So you see, I need an "option" that is mandatory in addition to it 
having a mandatory argument.

Sure I could do it in my own code - but given this is a common case I'm 
tempted to extend OptionParser to do it so that I can get at the option 
description when I report that its missing you see? This is why it 
should probably be a feature of OptionParser - to tie the error message 
to the switch that's missing. I don't think I can be bothered wading 
into that code myself, so I guess I'll do something a little hackier - 
test my options and repeat the option descriptions.

But how about it Gavin?

Also -  even if you don't  - you should probably clearly document the 
fact this isn't available. The frequent use of "MANDATORY" in the 
example led me and probably many others to think it was bug that my 
mandatory arguments weren't reported as missing. It took me a while to 
realize that the word "argument" was key here. It also took me a while 
to realize that the square brackets were important.

On 5/3/07, Rover Rhubarb <rover.rhubarb@gmail.com> wrote:
> both files and I prefer that the user specify the option --config or
> to the switch that's missing. I don't think I can be bothered wading
> to realize that the square brackets were important.
I agree completely.

The wording should be: mandatory option argument, not option.

And yes, it makes sense to /also/ make some options mandatory.