rspec

Comments and changes to this ticket

• 

  Aslak Hellesøy August 23rd, 2008 @ 01:49 PM

  This could work, but I think there is a risk that some people will do things like:

  
    check body.weight == 77
  

  and get a lot of false positives. Or am I being paranoid?

• 

  bahuvrihi August 23rd, 2008 @ 05:19 PM

  That's a good point. One option is to validate the check line includes a should/should_not statement.

  You can get at this information by using SCRIPT_LINES__ (which, if set to a hash, will store all lines parsed into the interpreter) and caller:

  
    # A method serving as the dumping ground for 'should ==' statements
    # that otherwise cause a 'useless use of <method> in void context'
    # warning. Useful if you're running tests with -w.
    #
    #   check 1.should == 1
    #
    # Check will validate that the line calling check contains a
    # should/should_not statement; the check fails if it does not.
    def check(return_value)
      caller[0] =~ /^(([A-z]:)?[^:]+):(\d+)/
  
      check_file = SCRIPT_LINES__[$1]
      violated("could not validate check: #{$1} (#{$3})") unless check_file
  
      # note the line number in caller
      # starts at 1, not 0
      line = check_file[$3.to_i - 1]
      violated("check used without should/should_not statement: #{line} (#{caller[0]})") unless line =~ /\.should(_not)?[^\w]/
    end
  

  That's the method... the attachment illustrates should/should_not statements executing without warning and a false positive causing a violation. Run '% rake spec' to get it going.

  The only downside to this, as I see it, is that you have to use SCRIPT_LINES, which will require a bit more memory and probably be minutely slower... but neither is a big deal, I imagine. Maybe there is a way to determine if you're running with -w? There must be... then you only set SCRIPT_LINES when running with -w, and have 'check' raise an error/warn otherwise. That way it's kind of an optional, use-it-if-you-want-it system.

  What do you think?

  • check.zip 3.9 KB