Getting the version number for rubygems?

[Gregory Brown]

Hi,

In ruport I have a tool that generates a bunch of boilerplate code for
folks, and in it, I use require_gem to lock the files down to a
specific version of Ruport.

I'd like to make this friendly to RubyGems 0.9.1 by using the gem
method rather than require_gem when people have 0.9.1, but want to
generate require_gem for older versions.

So basically, is there a constant or method I can call that'll give me
back the version of rubygems that i'm running?

 

[Jonathan Allen]

Well, from the command line you can call "gem -v" to get the version.
Perhaps there is something you can do with that.

Jonathan

 

[Devin Mullins]

I'd like to make this friendly to RubyGems 0.9.1 by using the gem
method rather than require_gem when people have 0.9.1, but want to
generate require_gem for older versions.

So basically, is there a constant or method I can call that'll give me
back the version of rubygems that i'm running?

For 0.8.11 and 0.9.0, this should work:
require 'rubygems/rubygems_version'
p Gem::RubyGemsVersion
but no promises on any other version. Check the SCM history.

That said, why not rescue NoMethodError?

Devin

 

[Eric Hodel]

In ruport I have a tool that generates a bunch of boilerplate code for
folks, and in it, I use require_gem to lock the files down to a
specific version of Ruport.

I'd like to make this friendly to RubyGems 0.9.1 by using the gem
method rather than require_gem when people have 0.9.1, but want to
generate require_gem for older versions.

Just use #gem, and don't bother being backwards-compatible. RubyGems
older than 0.9.1 has a serious security exploit. (About 20% of
rubyists are running a version of RubyGems without #gem (prior to
0.9.1's release).)

So basically, is there a constant or method I can call that'll give me
back the version of rubygems that i'm running?

Don't think so hard:

require 'rubygems'
Kernel.respond_to? :gem

And now for the real answer to your question:

The rubygems constant is in Gem::RubyGemsVersion, which you can get
from 'rubygems/rubygems_version'. Kernel#gem first appeared in 0.9.0.

 

[Gregory Brown]

Just use #gem, and don't bother being backwards-compatible. RubyGems
older than 0.9.1 has a serious security exploit. (About 20% of
rubyists are running a version of RubyGems without #gem (prior to
0.9.1's release).)

I'll probably do this by the next release, but the one coming up in a
week or two i'd like to give people a bit of grace period.

Don't think so hard:

require 'rubygems'
Kernel.respond_to? :gem

Yep, that's better. thanks.

 

[Jeremy McAnally]

require 'rubygems'

unless Kernel.respond_to? :gem
alias gem require_gem
end

gem 'ruport'

That should work I believe...

--Jeremy

I'll probably do this by the next release, but the one coming up in a
week or two i'd like to give people a bit of grace period.


Yep, that's better. thanks.

--
My free Ruby e-book:
http://www.humblelittlerubybook.com/book/

My blogs:
http://www.mrneighborly.com/
http://www.rubyinpractice.com/

 

[Gregory Brown]

require 'rubygems'

unless Kernel.respond_to? :gem
alias gem require_gem
end

gem 'ruport'

That should work I believe...

instead of alias i'll use alias_method. But I bet that would work.

 

[Devin Mullins]

instead of alias i'll use alias_method.

Why? (Curious, not argumentative.)

Devin

 

[Gregory Brown]

Why? (Curious, not argumentative.)

alias is really a somewhat scary function. I can't think of a good
example right now, but i've seen a couple different surprising things
with alias, wheras alias_method is simple and has normal behaviour.
Perhaps someone on the list can help me out with this.

 

[Gregory Brown]

alias is really a somewhat scary function.

well, keyword. alias_method is actually a method, alias is not.

 

[Eric Hodel]

require 'rubygems'

unless Kernel.respond_to? :gem
alias gem require_gem
end

gem 'ruport'

That should work I believe...

NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO!

#gem and #require_gem do different things (autorequire).

You'll screw over other gems by aliasing things around.

Instead, just fall back to require_gem when gem isn't loaded.

if Kernel.respond_to? :gem then
gem blah
else
require_gem blah
end

DO NOT go aliasing methods on top of each other. #require and
#require_gem do different things, much like #sub and #sub! do
different things, respect that.

 

[Jeremy McAnally]

Maybe you should tell the Rails team because I'm fairly sure that's
how they planned on handling it...

http://dev.rubyonrails.org/ticket/6886

--Jeremy

NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO!

#gem and #require_gem do different things (autorequire).

You'll screw over other gems by aliasing things around.

Instead, just fall back to require_gem when gem isn't loaded.

if Kernel.respond_to? :gem then
gem blah
else
require_gem blah
end

DO NOT go aliasing methods on top of each other. #require and
#require_gem do different things, much like #sub and #sub! do
different things, respect that.

--
My free Ruby e-book:
http://www.humblelittlerubybook.com/book/

My blogs:
http://www.mrneighborly.com/
http://www.rubyinpractice.com/

 

[Gregory Brown]

NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO! NO!

There you go being an asshole again Eric. (please stop)

if Kernel.respond_to? :gem then
gem blah
else
require_gem blah
end

DO NOT go aliasing methods on top of each other. #require and
#require_gem do different things, much like #sub and #sub! do
different things, respect that.

I'm aware of this. I actually need to just lock version and do a
normal require.

I'll do it that way.

 

[Eric Hodel]

alias is really a somewhat scary function. I can't think of a good
example right now, but i've seen a couple different surprising things
with alias, wheras alias_method is simple and has normal behaviour.
Perhaps someone on the list can help me out with this.

You're seeing phantoms. There's no difference between alias_method
and alias on the inside.

case NODE_ALIAS:
if (NIL_P(ruby_class)) {
rb_raise(rb_eTypeError, "no class to make alias");
}
rb_alias(ruby_class, rb_to_id(rb_eval(self, node->u1.node)),
rb_to_id(rb_eval(self, node->u2.node)));

static VALUE
rb_mod_alias_method(mod, newname, oldname)
VALUE mod, newname, oldname;
{
rb_alias(mod, rb_to_id(newname), rb_to_id(oldname));

 

[Gregory Brown]

You're seeing phantoms. There's no difference between alias_method
and alias on the inside.

Hah. That'd explain why I am scared.

I forget what the cases were, but Sebastian Delmont did a talk on
things he found surprising in Ruby, and the stuff he pulled up for
alias vs. alias_method was pretty convincing.

 

[Eric Hodel]

Hah. That'd explain why I am scared.

I forget what the cases were, but Sebastian Delmont did a talk on
things he found surprising in Ruby, and the stuff he pulled up for
alias vs. alias_method was pretty convincing.

any slides online?

 

[Gregory Brown]

any slides online?

I didn't see them on the nycruby list. :-/
I'll email him and see if they're posted anywhere

 

[Gregory Brown]

Why? (Curious, not argumentative.)

I guess it really is just the method vs. keyword thing.
alias is available everywhere, which means you might be aliasing in
the wrong place

Also, alias can work on global variables and regex backreferences.

it's also scary to me to see: alias new_method old_method
where alias_method :new_method, ld_method seems more natural to me.

I think I may have been overly fearful about the use of alias, but I
guess there is no good reason not to use it as long as you know what
you're doing.

regards,
-greg

 

[James Edward Gray II]

You're seeing phantoms. There's no difference between alias_method
and alias on the inside.

Well, you can override alias_method() if needed, but not alias.

James Edward Gray II

 

[Ryan Davis]

Well, you can override alias_method() if needed, but not alias.

And that changes or invalidates what Eric said... how? Not only did
he qualify with "on the inside" but he also backed it up by showing
said inside.