Style/NumericPredicate broke $CHILD_STATUS == 0

[ShockwaveNN]

Steps to reproduce the problem

Create temp.rb file with

require "English"
pid = spawn('exit 0')
Process.wait pid
puts $CHILD_STATUS == 0

Run ruby temp.rb #true
Run rubocop temp.rb

It show warning

temp.rb:3:6: C: Style/NumericPredicate: Use $CHILD_STATUS.zero? instead of $CHILD_STATUS == 0

If you fix it to

require "English"
pid = spawn('exit 0')
Process.wait pid
puts $CHILD_STATUS.zero?

It broke

ruby temp.rb
temp.rb:4:in `<main>': undefined method `zero?' for #<Process::Status: pid 13846 exit 0> (NoMethodError)

RuboCop version

$ rubocop -V
0.46.0 (using Parser 2.3.3.1, running on ruby 2.1.10 x86_64-linux)

[Drenmi]

Yes. This cop has been known to cause some false positives. The reason being we can't know what type the receiver will be at runtime (since we're doing static code analysis. 😀) We make the assumption that if a receiver is compared with 0, that receiver is a duck type of Numeric.

You can disable this for the line affected only using:

if not_a_numeric == 0 # rubocop:disable Style/NumericPredicate

or, if this is causing enough false positives that it's not worth it, you can disable it entirely in your .rubocop.yml file:

Style/NumericPredicate:
  Enabled: false

😊

[mikegee]

For the particular example you posted, Process::Status#success? is an improvement over either sort of numeric comparison.

Process::Status allows numeric comparison to be compatible with code written a very long time ago. $CHILD_STATUS has been an instance of Process::Status since at least Ruby 1.8.6.

[bbatsov]

Process::Status allows numeric comparison to be compatible with code written a very long time ago. $CHILD_STATUS has been an instance of Process::Status since at least Ruby 1.8.6.
👍 2

We should definitely add a cop for this - it'd be trivial to implement.