[ANN] Better relative requires with Need

[Drew Olson]

Need makes ruby relative requires just work. Simply need a file with a
relative path
and the file will always be required correctly, regardless of what file
your application is
being launched through. Typically, ruby projects would unshift lib onto
$PATH or use the
File.dirname(__FILE__) trick. Using need means you don't have to worry
about either of these.

Assume you have two files, one directly in lib and the other in
lib/extensions. Let's assume that file_a in lib requires file_b, in
lib/extensions. Previously, you would doing some crazy load path
unshifting or use the __FILE__ trick to make these requires flexible
enough to work when your app is being accessed by rake, through a test
suite, or required as a gem. Now, just use need.

In file_a:
need{"extensions/file_b"}

Note that the block syntax is necessary. Need uses the binding of the
block to determine the
location of your file and correct perform your relative require for you.

== INSTALL:

* sudo gem install need

Hope you enjoy it!

- Drew

 

[ara howard]

Note that the block syntax is necessary. Need uses the binding of the
block to determine the
location of your file and correct perform your relative require for
you.

nice idea!

a @ http://drawohara.com/

 

[vinbarnes]

Kudos! I had worked on something similar although I didn't use the
block binding which is a better way to go. I alias_method_chained
require and was using the caller stack to figure out the correct file
path. But I kept running into the problem of having to require it
everywhere... Like I would be in a test file and you'd have to have,

require 'elreq' # name of mine
require '../../test_helper'

So I largely gave up b/c while it was cleaner, adding the extra
require everywhere seemed to negate the benefits. If only it was
globally available - like in core.

A suggestion, if you are concerned with portability you may want to
use File.join to concatenate the end result together. Also, you
probably want to stick it in the Kernel module, but that may just be a
design pet peeve of mine.

Glad something like this is out there.

Cheers,
Kevin

 

[Drew Olson]

Glad something like this is out there.

Cheers,
Kevin

Kevin -

Thanks so much for the response. I'm glad people may find it useful. As
far as joining the file path, what exactly does this buy me? I've seen
it used quite a bit, but never used it myself. Also, in regard to
extending Kernel rather than Object, isn't Kernel itself an Object? Or
is it seen as too intrusive to extend object?

Thanks for the advice and please feel free to submit a patch, I'll
definitely take a look at adding it to the source.

Thanks,
Drew

 

[Bob Hutchison]

Hi,

As
far as joining the file path, what exactly does this buy me?

You can avoid the problem (leaving it entirely in the hands of your
users) if you write it this way:

File.expand_path(block.call,
File.dirname(eval("__FILE__",block.binding)))

And it's shorter.

Anyway, thanks for the gem!

Cheers,
Bob

 

[ara howard]

Thanks so much for the response. I'm glad people may find it u
seful. As
far as joining the file path, what exactly does this buy me?

i always expand and join the path because otherwise doing

require '../a.rb'

followed by

require '.././a.rb'

will cause ruby to load the file twice - the hash is on the *exact*
filename

cheers.

a @ http://drawohara.com/

 

[Rob Sanheim]

I believe this should be handled by ruby's require in 1.9...so no more
double requires based on differing file paths.

- Rob

http://robsanheim.com
http://thinkrelevance.com

 

[Drew Olson]

A suggestion, if you are concerned with portability you may want to
use File.join to concatenate the end result together.

Kevin -

I read up on this and it makes quite a bit of sense. I'll patch the code
and release a new version.

Thanks!

- Drew