[Ruby 2.7] Add specs for pattern matching

[andrykonchin]

There is no official documentation yet so specs are based on:

• https://bugs.ruby-lang.org/issues/14912
• https://speakerdeck.com/k_tsj/pattern-matching-new-feature-in-ruby-2-dot-7?slide=23
• https://rubyreferences.github.io/rubyref/language/pattern-matching.html
• https://github.com/ruby/ruby/blob/master/test/ruby/test_pattern_matching.rb

[eregon]

Great, thank you for starting those!
We probably need to wrap them in eval <<RUBY to avoid syntax error on old versions.
I already ticked it in #745 to mark it's in progress.

[andrykonchin]

@eregon

We probably need to wrap them in eval <<RUBY to avoid syntax error on old versions.

Yeah, you are right.

TBH I'm concerned about readability because I use ScratchPad heavily e.g.

case {a: 1}
  in String[a: 1]
    ScratchPad << :foo
else
  ScratchPad << :else
end

ScratchPad.recorded.should == [:else]

but it could be done shorter with checking case expression result like

case {a: 1}
  in String[a: 1]
    true
else
  false
end.should == false

What would you prefer?

[eregon]

I would use the shorter second version, it could look like this with eval:

eval(<<~RUBY).should == false
case {a: 1}
in String[a: 1]
  true
else
  false
end
RUBY

(<<~RUBY syntax highlights correctly in Atom and RubyMine at least)

[andrykonchin]

@eregon

There is only one failing spec (on MacOS) not relevant to the PR:

CVE-2019-8325 is resisted by sanitising error message components for the 'loading command' message
Example took 15.006841999999978s, which is longer than the timeout of 15.0s

Actually there was another failing spec on Windows related to sockets. After re-running it passed but another spec, mentioned above, failed.

So I would like to merge the PR as is but waiting for your approval.

[eregon]

Do you know if this is intended behavior?

[andrykonchin]

@eregon

Looks like so - https://github.com/ruby/ruby/blob/master/hash.c#L4591-L4595

Keys passed as argument are just a hint to optimize performance. Struct#deconstruct_keys uses keys... Hash doesn't. At least for now. Hope it will be changed in the next release.

[eregon]

This is a great PR, thank you!

Sorry for the transient failure, I'd guess it was a particularly slow run, maybe we should just bump the timeout.