Fix raise error when the expectation target is an error instance. #236
Conversation
|
Related #232 |
|
Also this gave some errors related to differs when I ran locally, I want to see if it's a local problem or if travis picks it up. |
|
I also get these errors on master, any ideas @myronmarston ? |
| def expected_error_class | ||
| (@expected_error.is_a? Class) ? @expected_error : @expected_error.class | ||
| end | ||
|
|
There was a problem hiding this comment.
This is interesting -- I didn't realize 1.9.3- supported rescuring an exception instance but 2.0.0 doesn't. Good find! I noticed something when playing around in irb:
2.0.0p0 :001 > $exc = StandardError.new("boom")
=> #<StandardError: boom>
2.0.0p0 :002 > def try
2.0.0p0 :003?> raise $exc
2.0.0p0 :004?> rescue $exc => e
2.0.0p0 :005?> puts "Rescued #{e}"
2.0.0p0 :006?> end
=> nil
2.0.0p0 :007 > try
TypeError: class or module required for rescue clause
from (irb):4:in `rescue in try'
from (irb):3:in `try'
from (irb):7
from /Users/myron/.rvm/rubies/ruby-2.0.0-p0/bin/irb:16:in `<main>'
2.0.0p0 :008 >
As the error indicates, rescue SomeModule is valid, too...and your is_a?(Class) conditional here will skip that case.
On top of that, I was just realizing that this logic will wrongly allow it to pass when you do this:
error = StandardError.new("foo")
expect {
raise StandardError.new("bar")
}.to raise_error(error)When a user passes an exception instance to raise_error, I think the expectation is that it will only pass if the raised error is == to the passed exception instance. The change here will allow it to pass if the exception class is the same.
I'm thinking that we may want to do a bit of a larger refactoring of this matcher....rather than having a rescue expected_class and another rescue Exception, maybe we should just have the rescue Exception part, and then the actual and expected exceptions can be compared using == and ===.
Regardless of the approach taken, it would be great to have additional spec coverage of the exception instance case to handle the example I put above.
(BTW, here's the 2.0.0 change: http://bugs.ruby-lang.org/issues/4438 )
There was a problem hiding this comment.
Hi @myronmarston. I've updated this with some new specs to actually assert around this behaviour. So what I felt the best way to do this is compare whether actual == expected, or actual.is_a? expected. Thoughts?
Signed-off-by: Sam Phippen <samphippen@googlemail.com>
| # This clause should be empty, but rcov will not report it as covered | ||
| # unless something (anything) is executed within the clause | ||
| "http://eigenclass.org/hiki.rb?rcov-0.8.0" | ||
| if @actual_error == @expected_error || @actual_error.is_a?(@expected_error) |
There was a problem hiding this comment.
I believe rescue actually uses === for matching (at least, that's what Exceptional Ruby said), so I think we should use that here:
if @actual_error == @expected_error || @expected_error === @actual_error
|
@myronmarston how's this looking now? |
Closer! But see my comment on |
This closes rspec#232 Signed-off-by: Sam Phippen <samphippen@googlemail.com>
Signed-off-by: Sam Phippen <samphippen@googlemail.com>
Signed-off-by: Sam Phippen <samphippen@googlemail.com>
Signed-off-by: Sam Phippen <samphippen@googlemail.com>
Signed-off-by: Sam Phippen <samphippen@googlemail.com>
|
pew pew pew. Also rebase against master. |
|
Looks good. Feel free to merge once travis is green. |
|
@myronmarston should I make my expectations more lax, or is there a way that I can get 1.8.7, 1.9.2 and ree to give me the strings instead of pointers on that travis backtrace? |
|
Can you interpolate the exception instances into the expected failure string to match what the matcher does? |
Signed-off-by: Sam Phippen <samphippen@googlemail.com>
|
@myronmarston so I think this is now ready for merging. Anything I've missed? |
Fix raise error when the expectation target is an error instance.
|
nope :). Thanks as always! |
|
literally awesome. 🐴 |
Not sure if this is the best way to do it, but it certainly seems to work. Thoughts?