Last result local '_'

[veelenga]

Inspired by https://github.com/pry/pry/wiki/Special-Locals#Last_result

Super simple (5 LOCs) but very useful 🔨

icr(0.23.1) > a = 42
 => 42
icr(0.23.1) > _
 => 42
icr(0.23.1) > _ + 1
 => 43
icr(0.23.1) > 1 + _
 => 44
icr(0.23.1) > _.to_s
 => "44"
icr(0.23.1) > _.to_i
 => 44
icr(0.23.1) > inc = -> { _ + 1 }
 => #<Proc(Int32):0x1002ea350>
icr(0.23.1) > inc.call
 => 45
...

[jwoertink]

Since we can use an underscore for ignoring block values, I think this would still match something like x.each { |_| nil } which would be

"x.each { |_| nil }".to_s.gsub(/\b_\b) { "42"} #=> x.each { |42| nil }

Maybe if the whole command was just _ so command.to_s.gsub(/^\b_\b$/) { last_value.to_s.strip }?

[veelenga]

@jwoertink yes, also _ can be used in such way:

a, _ = [1, 2]
a #=> 1

Making it just a "whole command" does not make sense for me, because the general purpose of _ is the following:

icr(0.23.1) > User.last
...
icr(0.23.1) > user = _
icr(0.23.1) > user.firstname

Will try to fix those issues. Thanks for a comment

[jwoertink]

So I had an epiphany while I was sleeping. You're right that it can't be a whole command, so my regex won't work. However, I think a regex is the wrong way to go anyway. I feel like it should just be a macro! That's basically what it is anyway. That way if _ is used in a block, or method argument it's not caught.

[veelenga]

@jwoertink did you mean Crystal's macro? Unfortunately, a compiler does not allow methods/macros named _.

[jwoertink]

aww lame. That would have been the perfect use. Oh well.

[greyblake]

as last -> as the last

[greyblake]

Despite, the Crystal itself does not support _ as far as I can see, the changes look OK to me 👍

[veelenga]

Well, seems like it requires much more effort than this naive implementation. Currently, it could bring more problems than benefits. If someone has ideas or wants to try to implement it, i am ready to help ;)

[veelenga]

Valid chunks of Crystal with _ that will produce an unexpected result with current implementation:

a, _ = [1, 2]

def foo(x : _)
end

def foo(x : _, _ -> Int32)
end

def some_proc(&block : Int32 -> _)
  block
end

[1,2,3].each { |_| puts "yo!" }
{"k" => "v"}.each { |k, _| puts k }

"super _ b"

Maybe we can use some other symbol(s) rather than _. For example, !! (just like a similar bash command)...

[jwoertink]

I like that !! lines up with bash, but maybe still an issue because of doing a = 1; !!a. Though, doing a regex for !! would be a lot easier since it's probably only used in a string or at the beginning of some value.

I do agree; however, that something other than _ could be used. It's nice to work like IRB, but I don't think we need to be 1 <-> 1 match with it. Especially since there's that paste command that was added. What about __? 2 underscores. I can't think of any place that would be used.

[veelenga]

I think __ is a good idea... refs #63