Ruby Forum Ruby-core > [Feature #440] Better introspection for methods (declaring class, arity)

Feature #440: Better introspection for methods (declaring class, arity)
http://redmine.ruby-lang.org/issues/show/440

Author: Lasse Koskela
Status: Open, Priority: Normal


Tools relying on introspection would benefit from having access to 
information such as:
1) where was a given method declared, and
2) how many arguments does a method accept.

Currently, for 1) one needs to parse the output of 
Object.new.method(:name).to_s to resolve where a method was inherited 
from.
For 2), one can only get the number of required arguments with 
Object.new.method(:name).arity, which isn't sufficient if you want to 
use the optional arguments, too. With the current implementation of 
arity, one needs to just "assume" that one can throw an arbitrary number 
of arguments at the method and see if it fails.

The use case for these enhancements is coming from the open source Robot 
Framework (http://www.robotframework.org). It's a testing harness where 
you plug in implementations of test cases in the form of "keyword 
libraries" implemented in Python, Java, or (hopefully soon) Ruby. In 
order to report available "keywords" (actions available for the test 
author, implemented as public methods) to the user, the framework needs 
to scan available keyword libraries for suitable methods and determine 
how many arguments each keyword accepts (ideally, with the names of 
those arguments). Improving Ruby's introspection capabilities would be a 
boost for Ruby's viability as a Robot Framework keyword library 
implementation language.

1) would seem like a simple thing to do, given that the information is 
already available for Method#to_s. I'm not proficient with C, but I 
might be able to implement that myself as a patch if someone could point 
me to the right direction regarding the location in the source files.

Issue #440 has been updated by Koichi Sasada.


Current trunk seems to support your requirements.  Could you confirm 
about this issue?
----------------------------------------
http://redmine.ruby-lang.org/issues/show/440

Issue #440 has been updated by Lasse Koskela.


The first aspect of the issue has been improved via the 
"source_location" and "receiver" methods being available on the Method 
object.

It seems to me that the arity stuff, however, is the same as with the 
1.8.x branch? That is, aspect 2) of this issue is still current.
----------------------------------------
http://redmine.ruby-lang.org/issues/show/440

Issue #440 has been updated by Roger Pack.


if you don't mind a hack then
RubyVM::InstructionSequence.disasm
might help.
Looks like it would be nice to add something to do this, though, since 
it seems difficult to tell from the #arity call to see how many optional 
and non optional arguments there are.
Cheers.
-=R
----------------------------------------
http://redmine.ruby-lang.org/issues/show/440