Making "require" just work with some dynamic load path manip

The problem:

If I have a simple file, “a.rb”:

#!/usr/bin/env ruby
require 'b'

And it’s companion, “b.rb”:

puts "Hello, world!"

Running “a.rb” works fine if I start from the directory containing
those files. If I change to another directory, however, I get a
LoadError because Ruby doesn’t know how to find “b.rb”.

Workarounds:

  1. Set or modify the RUBYLIB environment variable so that Ruby knows
    which additional locations to search.
  2. Manipulate the load path ($:) from within the program itself.

These workarounds are fine for me because I understand the quirks of
Ruby’s load path. But I am looking for a way to make this stuff just
work automatically for other users who don’t necessarily understand
Ruby’s load path and will expect “a” to be able to find “b” because
they’re in the same directory, no matter what the current working
directory is at the time.

The solution:

So I already have a solution for this based on workaround “2” and I
wanted to get some feedback on whether this is a good idea or not. The
basic idea is to make a “custom_require” method which does the
following:

  1. Try requiring the file normally.
  2. On a LoadError, try adding the directory containing the file to the
    load path (only if it’s not already present).
  3. Now try requiring it again.
  4. If you get a second LoadError then put the load path back the way
    it was and re-raise the original error.
  5. If you succeed then put the load path back the way it was anyway.

So here’s the basic code to do this. I try to make it as conservative
as possible by making the modifications to the load path temporary,
and always adding absolute paths so as to minimize the likelihood of
the same file being evaluated twice. The security of dynamically
modifying the load path is not really a concern because the users
running this scripts already have full write access to them anyway and
can therefore already do whatever they want, and they (and nobody
else) have full control over the directories in which the scripts are
stored.

To use the new method you would do:

custom_require 'b', __FILE__

The second parameter is necessary (as far as I can tell) because I
know of no other clean way for the “custom_require” method (itself in
a different file) to know which file was being evaluated when the call
to “custom_require” was made.

What I’d like to know is, is there a better way?

def custom_require(file, fallback)
  begin
    require file
  rescue LoadError => e
    candidate = Pathname.new(fallback).dirname.realpath
    if $:.any? { |path| Pathname.new(path).realpath == candidate

rescue false }
raise e
else
begin
$:.push(candidate)
status = require file
rescue LoadError
raise e
ensure
$:.pop
status
end
end
end
end

Is there a reason you are not using setup.rb and/or gems?

T.

On 7 mar, 22:01, “Trans” [email protected] wrote:

Is there a reason you are not using setup.rb and/or gems?

T.

Neither of those is really appropriate because the files “a.rb” and
“b.rb” are dynamically generated and probably not very long lived (the
typical cycle will be: generate, use, throw away) and definitely not
worth installing permanently. The thing that generates those files is
actually distributed as a gem so there are no problems at that level;
I just want the generated files to be as easy to use for end users as
possible.

Cheers,
Greg

On Mar 7, 2007, at 6:00 PM, Greg H. wrote:

I just want the generated files to be as easy to use for end users as
possible.

Cheers,
Greg

IF you want to garauntee to require b in the same dir as a you can
do this:

a.rb

puts “in a.rb”
require File.join(FIle.dirname(FILE), ‘b’)

b.rb

puts “in b.rb”

Cheers-

– Ezra Z.
– Lead Rails Evangelist
[email protected]
– Engine Y., Serious Rails Hosting
– (866) 518-YARD (9273)