After trying out various use cases string.each (synonym for .each_line) is very
weird. Here are some cases ....
# This one does not throw exception
s = "aaa"
s.each { s[0] = "b"}
=> "baa"
# This one does (it has a newline) (and matches the rubyspec test case)
s = "aaa\naaa"
s.each { s[0] = "b"}
=> Runtime Error: string modified
# Ok, if that is not weird enough (but maybe not completely whacked if you
consider each_line won't really iterate until there is a newline)
s = "aaa\naaa"
# modify s before iterating
s[0] = "a"
s.each { s[0] = "b"}
=> "baa\naaa"
This looks like MRI is relying on copy on write semantics to determine whether
a string was modified? Jim, says this is not the behavior for 1.9.
This seems like a good candidate for IR diverging on ...
rem
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Dave Remy
Sent: Monday, June 30, 2008 9:47 AM
To: [email protected]
Subject: Re: [Ironruby-core] MutableString.each
Translation: I need to check to assure that mri actually uses freeze to make
the string immutable while iterating. I should be able to test the string for
frozen. I just ran a test script, and nope, the string being iterated is not
frozen:
s = "aaa"
s.each { puts s.frozen? } => false
So I need to implement some other mechanism for making the iterated string
immutable during the iteration rather than freezing it. I wonder whether
other types of mutable objects that iterate have this same behavior ... I'll do
some more research.
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of John Lam
(IRONRUBY)
Sent: Monday, June 30, 2008 7:23 AM
To: [email protected]
Subject: Re: [Ironruby-core] MutableString.each
You should catch and rethrow the correct exception if you want to use the
frozen detection stuff. But you should check to see if it's actually frozen
while iterating - ie is this an observable side-effect of using each?
Thanks,
-John
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Curt Hagenlocher
Sent: Sunday, June 29, 2008 4:16 PM
To: [email protected]
Subject: Re: [Ironruby-core] MutableString.each
RuntimeException is defined in IronRuby.Libraries\Builtins\Exceptions.cs.
Is MRI really that inconsistent about which type of error is thrown when you
try to modify a frozen object? (Not that this would surprise me :(.) If so, it
might be cleaner if the call to FreezeObject could record or otherwise
influence the type of exception that we expect to throw.
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Dave Remy
Sent: Sunday, June 29, 2008 3:58 PM
To: [email protected]
Subject: [Ironruby-core] MutableString.each
I am working on one of the specs for string.each which is failing. This is
the spec that is failing.
it "raises a RuntimeError if the string is modified while substituting" do
str = "hello\nworld"
should_raise(RuntimeError) { str.send(cmd) { str[0] = 'x' } }
end
The idea is that the string that is being iterated over shouldn't be changed
during the iteration. It is easy enough to freeze the string in the first line
of MutableString.EachLine using:
RubyUtils.GetExecutionContext(context).FreezeObject(self); Which
appropriately throws an error when the string gets modified, however this
approach throws a "TypeError" and the spec wants a "RuntimeError" (message:
String#each raises a RuntimeError if the string is modified while substituting
FAILED, Expected RuntimeError, got TypeError (can't modify frozen object)). I
considered wrapping the code in EachLine that invokes the each block with a try
catch and then rethrowing a caught type error exception to a runtime exception
however I don't see a runtime exception type in RubyExceptions. Any
suggestions on how best to approach this?
Thx!
rem
_______________________________________________
Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core
