Bug #20043: `defined?` checks for method existence but only sometimes

• Status: Open
• Assignee: -
• Target version: -
• ruby -v: ruby 3.3.0dev (2023-12-05T21:25:34Z master 56eccb350b) [arm64-darwin23]
• Backport: 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN

[tenderlovemaking (Aaron Patterson)]

When an expression is passed to defined?, it will sometimes check if a method in a sub-expression is defined and sometimes it won't.

For example:

$ ./miniruby -e'p defined?(a)'
nil
$ ./miniruby -e'p defined?([a])'
nil

In the above case, Ruby will check whether or not the method a is defined, and it returns nil. However, if you use a splat, it will not check:

$ ./miniruby -e'p defined?([*a])'
"expression"

The same thing seems to happen with method parameters:

$ ./miniruby -e'p defined?(itself)'
"method"
$ ./miniruby -e'p defined?(itself(a))'
nil
$ ./miniruby -e'p defined?(itself(*a))'
"method"

Oddly, defined? will check contents of arrays, but won't check contents of hashes:

$ ./miniruby -e'p defined?([[[[a]]]])'
nil
$ ./miniruby -e'p defined?({ a => a })'
"expression"

I think all of the cases that refer to a should check whether or not a is defined regardless of splats or hashes.

[Eregon (Benoit Daloze)]

I wonder if defined? should only be defined for constant paths and method call without arguments, e.g. defined?(RubyVM.keep_script_lines).
Is there any use case for defined? besides those?

IOW, I think the simpler defined? is the better (less complexity in Ruby implementations, I think little value in practice).
That could mean don't recurse into nested nodes or so, except for constant paths.

[byroot (Jean Boussier)]

Is there any use case for defined? besides those?

Instance variables for sure: defined?(@ivar).

Also checking for super: defined?(super).

Some people use it for a sequence of calls: defined?(foo.bar.baz).

[ko1 (Koichi Sasada)]

• File clipboard-202312191405-owvpy.png clipboard-202312191405-owvpy.png added

It seems a bug from Ruby 2.2.

[Dan0042 (Daniel DeLorme)]

I didn't know about this way of using defined? and it's interesting to me that we can check if multiple things are defined at once. Although the 3rd case below is another example that seems odd. But both that and defined?([*a]) return "expression" since ruby 1.9, not 2.2

x = 1
defined?(x) and defined?(y) and defined?(z) #=> nil
defined?([x, y, z])                         #=> nil (shorter!)
defined?(x && y && z)                       #=> "expression" (???)
defined?(x & y & z)                         #=> nil

[tenderlovemaking (Aaron Patterson)]

• File general - ruby-lang - 4 new items - Slack 2023-12-19 16-10-39.png general - ruby-lang - 4 new items - Slack 2023-12-19 16-10-39.png added

ko1 (Koichi Sasada) wrote in #note-3:

It seems a bug from Ruby 2.2.

I think itself was added in Ruby 2.2. Seems this bug is perhaps from Ruby 1.9?

Dan0042 (Daniel DeLorme) wrote in #note-4:

I didn't know about this way of using defined? and it's interesting to me that we can check if multiple things are defined at once.

Yes, I thought so too. I can understand the utility of defined?([x, y, z]), but I can't tell if that was intended? Especially considering the x && y && z case.

Also defined?([x, y, z]) will return "expression" if x, y, and z are defined. But [x, y, z] seems like an expression regardless of whether or not those are defined (the expression just might raise though).

[jeremyevans0 (Jeremy Evans)]

I submitted a pull request to fix this: https://github.com/ruby/ruby/pull/9500

[ko1 (Koichi Sasada)]

In future, should we deprecate defined? usecase which return expression?