# Dynamically modifying modules to avoid namespace collisions **URL:** \ **Category:** ruby-talk\ **Created:** [17 May 2006 00:19 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489 "2006-05-17T00:19:16Z")\ **Posts on this page:** 11\ **Page:** 1
**Author:** ![Wes\_Gamble](https://avatars.discourse-cdn.com/v4/letter/w/258eb7/32.png) [@Wes\_Gamble](https://rubytalk.org/u/Wes_Gamble)\ **Post date:** [17 May 2006 00:19 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/1 "2006-05-17T00:19:16Z")
All, I have a namespace collision problem between a Rubygem (htmltools 1.09) and a Rails framework class (in ActionPack 1.12.1). In order to use both, I will need to modify the namespace of one of them. However, I would really like to not have to modify the existing gems and I was thinking that maybe there might be some clever way to use the Module class to "wrap" the offending module redefine it's namespace. Has anyone ever done anything like this? Otherwise, I will be looking at "taking ownership" of the gem code and moving it to a locally - controlled location essentially freezing the version of that Gem for my application Thanks, Wes > **···** > > -- > Posted via [http://www.ruby-forum.com/](http://www.ruby-forum.com/). ---
**Author:** ![Wes\_Gamble](https://avatars.discourse-cdn.com/v4/letter/w/258eb7/32.png) [@Wes\_Gamble](https://rubytalk.org/u/Wes_Gamble)\ **Post date:** [17 May 2006 00:21 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/2 "2006-05-17T00:21:29Z")
Hey, I just thought of something. Is there a way to unload ("un-require") a module at runtime? Is there a way to load ("re-require") a module at runtime? Thanks, Wes Wes Gamble wrote: > **···** > > > All, > > > > I have a namespace collision problem between a Rubygem (htmltools 1.09) > > and a Rails framework class (in ActionPack 1.12.1). > > > > In order to use both, I will need to modify the namespace of one of > > them. > > > > However, I would really like to not have to modify the existing gems and > > I was thinking that maybe there might be some clever way to use the > > Module class to "wrap" the offending module redefine it's namespace. > > > > Has anyone ever done anything like this? > > > > Otherwise, I will be looking at "taking ownership" of the gem code and > > moving it to a locally - controlled location essentially freezing the > > version of that Gem for my application > > > > Thanks, > > Wes > > -- > Posted via [http://www.ruby-forum.com/\](http://www.ruby-forum.com/%5C). ---
**Author:** ![Logan\_Capaldo](https://avatars.discourse-cdn.com/v4/letter/l/7ea924/32.png) [@Logan\_Capaldo](https://rubytalk.org/u/Logan_Capaldo)\ **Post date:** [17 May 2006 00:34 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/3 "2006-05-17T00:34:34Z")
> Hey, I just thought of something. > > Is there a way to unload ("un-require") a module at runtime? Is there a > way to load ("re-require") a module at runtime? > > Thanks, > Wes I think you may be over-estimating the relationship between modules and files. One does not require modules, one requires files. One file may have many modules in it, or none. Regardless one can still 'require' that file. One way to hack something so it gets stuck in a module is: module Protect    eval(File.read("path/to/file.rb")) end Then you access all the classes defined in file.rb like: Protect::Node.new # etc. > **···** > > On May 16, 2006, at 8:21 PM, Wes Gamble wrote: > > > Wes Gamble wrote: > > > > > All, > > > > > > I have a namespace collision problem between a Rubygem (htmltools 1.09) > > > and a Rails framework class (in ActionPack 1.12.1). > > > > > > In order to use both, I will need to modify the namespace of one of > > > them. > > > > > > However, I would really like to not have to modify the existing gems and > > > I was thinking that maybe there might be some clever way to use the > > > Module class to "wrap" the offending module redefine it's namespace. > > > > > > Has anyone ever done anything like this? > > > > > > Otherwise, I will be looking at "taking ownership" of the gem code and > > > moving it to a locally - controlled location essentially freezing the > > > version of that Gem for my application > > > > > > Thanks, > > > Wes > > > > -- > > Posted via [http://www.ruby-forum.com/\](http://www.ruby-forum.com/%5C). ---
**Author:** ![Gregory\_Seidman](https://avatars.discourse-cdn.com/v4/letter/g/eb9ed0/32.png) [@Gregory\_Seidman](https://rubytalk.org/u/Gregory_Seidman)\ **Post date:** [17 May 2006 14:11 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/4 "2006-05-17T14:11:10Z")
[...] } One does not require modules, one requires files. One file may have } many modules in it, or none. Regardless one can still 'require' that } file. } } One way to hack something so it gets stuck in a module is: } } module Protect } eval(File.read("path/to/file.rb")) } end } } Then you access all the classes defined in file.rb like: } } Protect::Node.new # etc. This is a great idiom, but it needs to be able to function as a (nearly) drop-in replacement for require. This means that I need to be able to find the file in the search path that require would normally use. Consider the following: def find\_require\_file(filename)   # NEED IMPLEMENTATION HERE end def namespace\_get(str)   str.to\_s.split("::").inject(Object) { |ns,name| ns.const\_get(name) } end def require\_within(module\_name, filename)   file = find\_require\_file(file)   mod = module\_name   case module\_name     when Module,Class       #noop     else       mod = namespace\_get(mod)   end   mod.instance\_eval { eval File.read(file) } end This is untested, and I don't know how to implement that find\_require\_file, but the idea is that the following should work: module Foo   module Bar   end end require\_within Foo::Bar, "date" x = Foo::Bar::Date.today Anyone want to implement find\_require\_file and test the idea? --Greg > **···** > > On Wed, May 17, 2006 at 09:34:34AM +0900, Logan Capaldo wrote: ---
**Author:** ![Wes\_Gamble](https://avatars.discourse-cdn.com/v4/letter/w/258eb7/32.png) [@Wes\_Gamble](https://rubytalk.org/u/Wes_Gamble)\ **Post date:** [17 May 2006 15:23 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/5 "2006-05-17T15:23:42Z")
Greg, This is a good idea. Would make it easier to handle these sorts of issues. I may give it a shot. I'll let you know if I come up with anything. Wes > **···** > > -- > Posted via [http://www.ruby-forum.com/](http://www.ruby-forum.com/). ---
**Author:** ![Wes\_Gamble](https://avatars.discourse-cdn.com/v4/letter/w/258eb7/32.png) [@Wes\_Gamble](https://rubytalk.org/u/Wes_Gamble)\ **Post date:** [17 May 2006 17:20 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/6 "2006-05-17T17:20:17Z")
Greg, How does this str.to\_s.split("::").inject(Object) { |ns,name| ns.const\_get(name) } work? Wes Gregory Seidman wrote: > **···** > > > On Wed, May 17, 2006 at 09:34:34AM +0900, Logan Capaldo wrote: > > [...] > > } One does not require modules, one requires files. One file may have > > } many modules in it, or none. Regardless one can still 'require' that > > } file. > > } > > } One way to hack something so it gets stuck in a module is: > > } > > } module Protect > > } eval(File.read("path/to/file.rb")) > > } end > > } > > } Then you access all the classes defined in file.rb like: > > } > > } Protect::Node.new # etc. > > > > This is a great idiom, but it needs to be able to function as a (nearly) > > drop-in replacement for require. This means that I need to be able to > > find > > the file in the search path that require would normally use. Consider > > the > > following: > > > > def find\_require\_file(filename) > >   # NEED IMPLEMENTATION HERE > > end > > > > def namespace\_get(str) > >   str.to\_s.split("::").inject(Object) { |ns,name| ns.const\_get(name) } > > end > > > > def require\_within(module\_name, filename) > >   file = find\_require\_file(file) > >   mod = module\_name > >   case module\_name > >     when Module,Class > >       #noop > >     else > >       mod = namespace\_get(mod) > >   end > >   mod.instance\_eval { eval File.read(file) } > > end > > > > This is untested, and I don't know how to implement that > > find\_require\_file, > > but the idea is that the following should work: > > > > module Foo > >   module Bar > >   end > > end > > > > require\_within Foo::Bar, "date" > > > > x = Foo::Bar::Date.today > > > > Anyone want to implement find\_require\_file and test the idea? > > > > --Greg > > -- > Posted via [http://www.ruby-forum.com/\](http://www.ruby-forum.com/%5C). ---
**Author:** ![Logan\_Capaldo](https://avatars.discourse-cdn.com/v4/letter/l/7ea924/32.png) [@Logan\_Capaldo](https://rubytalk.org/u/Logan_Capaldo)\ **Post date:** [17 May 2006 20:03 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/7 "2006-05-17T20:03:24Z")
def find\_require\_file(filename)    $LOAD\_PATH.each do |directory|      [".rb", ".so", ".dll"].each do |ext|        if File.exists?(f = File.join(directory, filename + ext))          return f        end      end    end    # otherwise we didn't find it    raise 'Could not find #{filename} in $LOAD\_PATH.' end > **···** > > On May 17, 2006, at 10:11 AM, Gregory Seidman wrote: > > > def find\_require\_file(filename) > >   # NEED IMPLEMENTATION HERE > > end ---
**Author:** ![Daniel\_Schierbeck](https://avatars.discourse-cdn.com/v4/letter/d/dfb087/32.png) [@Daniel\_Schierbeck](https://rubytalk.org/u/Daniel_Schierbeck)\ **Post date:** [17 May 2006 18:06 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/8 "2006-05-17T18:06:03Z")
Wes Gamble wrote: > How does this > >   str.to\_s.split("::").inject(Object) { |ns,name| ns.const\_get(name) } > > work? It takes a string on the form "A:🅱:C" and splits it into an array; ["A", "B", "C"]. When calling #inject on that array, you iterate over each element, yielding them to the block (the second parameter). The first parameter is the result of the last call to the block, or, if it's the first element, the default value, here Object.    %w(A B C).inject(Object) {|parent, child| parent.const\_get(child) } is the same as    Object.const\_get(:A).const\_get(:B).const\_get(:C) Hope that made it clearer Daniel ---
**Author:** ![Logan\_Capaldo](https://avatars.discourse-cdn.com/v4/letter/l/7ea924/32.png) [@Logan\_Capaldo](https://rubytalk.org/u/Logan_Capaldo)\ **Post date:** [17 May 2006 20:14 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/9 "2006-05-17T20:14:51Z")
Actually in retrospect, I should have used double quotes in the raise and if you know you're always gonna be eval-ing these files, you probably want to get rid of the .dll and .so extensions. > **···** > > On May 17, 2006, at 4:03 PM, Logan Capaldo wrote: > > > On May 17, 2006, at 10:11 AM, Gregory Seidman wrote: > > > > > def find\_require\_file(filename) > > >   # NEED IMPLEMENTATION HERE > > > end > > > > def find\_require\_file(filename) > >   $LOAD\_PATH.each do |directory| > >     [".rb", ".so", ".dll"].each do |ext| > >       if File.exists?(f = File.join(directory, filename + ext)) > >         return f > >       end > >     end > >   end > >   # otherwise we didn't find it > >   raise 'Could not find #{filename} in $LOAD\_PATH.' > > end ---
**Author:** ![Gregory\_Seidman](https://avatars.discourse-cdn.com/v4/letter/g/eb9ed0/32.png) [@Gregory\_Seidman](https://rubytalk.org/u/Gregory_Seidman)\ **Post date:** [17 May 2006 20:48 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/10 "2006-05-17T20:48:06Z")
} Greg, } } How does this } } str.to\_s.split("::").inject(Object) { |ns,name| ns.const\_get(name) } } } work? Magic 🙂 So suppose you have a string "Foo::Bar" and you want to retrieve the module (or class, or other constant) Bar from the module (or class) Foo. (For simplicity we will assume that both are modules.) First, we make sure it's a string and split it on "::" which results in the array ["Foo", "Bar"]. Now we use inject with its optional starting object. For the first iteration of inject, ns = Object and name = "Foo" so the value of ns.const\_get(name) is Foo (the module). On the next iteration, ns = Foo and name = "Bar" so the value of ns.const\_get(name) is Bar (the module). That's the end of the iteration, so the value of the entire line is Bar, which is the value of the method call, which is what you wanted in the first place. } Wes --Greg > **···** > > On Thu, May 18, 2006 at 02:20:17AM +0900, Wes Gamble wrote: ---
**Author:** ![Logan\_Capaldo](https://avatars.discourse-cdn.com/v4/letter/l/7ea924/32.png) [@Logan\_Capaldo](https://rubytalk.org/u/Logan_Capaldo)\ **Post date:** [18 May 2006 01:01 UTC](https://rubytalk.org/t/dynamically-modifying-modules-to-avoid-namespace-collisions/27489/11 "2006-05-18T01:01:45Z")
One more thing, this won't work with gems unless you use require\_gem. e.g. % irb irb(main):001:0\> irb(main):002:0\* def find\_require\_file(filename) irb(main):003:1\> $LOAD\_PATH.each do |directory| irb(main):004:2\* [".rb", ".so", ".dll"].each do |ext| irb(main):005:3\* if File.exists?(f = File.join(directory, filename + ext))irb(main):006:4\> return f irb(main):007:4\> end irb(main):008:3\> end irb(main):009:2\> end irb(main):010:1\> # otherwise we didn't find it irb(main):011:1\* raise 'Could not find #{filename} in $LOAD\_PATH.' irb(main):012:1\> end =\> nil irb(main):013:0\> find\_require\_file('parse\_tree') RuntimeError: Could not find #{filename} in $LOAD\_PATH.          from (irb):11:in `find\_require\_file'          from (irb):13 irb(main):014:0\> require 'rubygems' =\> true irb(main):015:0\> find\_require\_file('parse\_tree') RuntimeError: Could not find #{filename} in $LOAD\_PATH.          from (irb):11:in `find\_require\_file'          from (irb):15 irb(main):016:0\> require\_gem 'ParseTree' =\> true irb(main):017:0\> find\_require\_file('parse\_tree') =\> "/usr/local/ruby/lib/ruby/gems/1.8/gems/ParseTree-1.4.1/lib/parse\_tree.rb" irb(main):018:0\> Since of how rubygems alter the load path. If you are using a gem, as long as it doesn't use the deprecated (and evil) autorequire attribute, you should still be able to safely stick it in its own namespace. > **···** > > On May 17, 2006, at 4:14 PM, Logan Capaldo wrote: > > > On May 17, 2006, at 4:03 PM, Logan Capaldo wrote: > > > > > On May 17, 2006, at 10:11 AM, Gregory Seidman wrote: > > > > > > > def find\_require\_file(filename) > > > >   # NEED IMPLEMENTATION HERE > > > > end > > > > > > def find\_require\_file(filename) > > >   $LOAD\_PATH.each do |directory| > > >     [".rb", ".so", ".dll"].each do |ext| > > >       if File.exists?(f = File.join(directory, filename + ext)) > > >         return f > > >       end > > >     end > > >   end > > >   # otherwise we didn't find it > > >   raise 'Could not find #{filename} in $LOAD\_PATH.' > > > end > > > > Actually in retrospect, I should have used double quotes in the raise and if you know you're always gonna be eval-ing these files, you probably want to get rid of the .dll and .so extensions.