Today I came across a very strange scenario that raises a syntax error
in pretty much all Ruby implementations. I have absolutely no
explanation for the following behaviour:
x :y
#=> bar.rb:1:in <main>': undefined methodx’ for main:Object
(NoMethodError)
x = nil
x “y”
#=> bar.rb:2:in <main>': undefined methodx’ for main:Object
(NoMethodError)
x = nil
x :y
#=> bar.rb:2: syntax error, unexpected ‘:’, expecting $end
Why does that latter case raise a syntax error during parsing? Why would
this be invalid Ruby? Why is it only invalid when supplying a symbol
argument, and why only when the local variable with the same name is
already defined? I would expect all three examples above to be run and
abort with a NoMethodError.
The behaviour is consistent in Rubinius, JRuby, Ruby 1.9 and Ruby 1.8.
What’s going on, does anyone know?
means that x is a method and anything that follows it is a parameter.
Since you don’t have a method called x you get “undefined method `x’”
just like it says in the error message.
When x is a reference to a variable by itself it just returns its
value, but when you put a parameter after it it looks for a method
called x, because you are passing it a parameter “y” and the only
things you pass parameters to are methods so you get the “undefined
method `x’” message again.
In the final case again you have no method called x and therefore a
parameter, in this case the :y, is unexpected. In that you don’t pass
parameters to a variable only to methods.
It’s good that you checked it out in all the various implementations
of Ruby but the error message was quite clear, “undefined method `x’”
means that Ruby thinks x should be a method but you don’t have a
method x defined.
On Mon, 17 Jan 2011 20:02:58 +0900, Peter H. wrote:
called x, because you are passing it a parameter “y” and the only
things you pass parameters to are methods so you get the “undefined
method `x’” message again.
In the final case again you have no method called x and therefore a
parameter, in this case the :y, is unexpected. In that you don’t pass
parameters to a variable only to methods.
I think Rolf’s point is that in both of those last two cases there is
no method called x - the only difference being that in the first case
he
tries to pass a string to the non-existent method in the second a
symbol. So why is the first (correctly) throwing a NoMethodError and
the
second gives a syntax error?
Well, and I have to admit I am guessing here, the source is parsed in
the context of the code that precedes it. So ‘x :y’ is parsed when
nothing (no method or variable) called x exists so the parser is
thinking that x should be a method, because it has not been told
otherwise and the :y makes sense as a parameter to a method.
In the case of ‘x = nil; x :y’ the ‘x = nil’ means that we now have
something defined for x (a variable) which the parser can then take
into account when it gets to ‘x :y’. Whereas before with ‘x :y’ it
assumes that x is a method because x has not been defined. But with ‘x
= nil; x :y’ we know that x is a variable and the parser no longer
assumes that x is a method and tries to make sense of ‘a variable’
followed by ‘a symbol’. Which make no sense, thus the error.
I don’t know if this is the case as I do not routinely browse the Ruby
parser code but from what I know this makes sense - even if I may not
have explained it clearly.
called x, because you are passing it a parameter “y” and the only
method x defined.
What I don’t understand is the difference between x “y” and x :y
irb(main):001:0> x = nil
=> nil
irb(main):002:0> x “y”
NoMethodError: undefined method `x’ for main:Object
from (irb):2
from :0
irb(main):003:0> x :y
SyntaxError: compile error
(irb):3: syntax error, unexpected ‘:’, expecting $end
from (irb):3
from :0
I think Rolf’s point is that in both of those last two cases there is no
method called x - the only difference being that in the first case he tries to
pass a string to the non-existent method in the second a symbol. So why is the
first (correctly) throwing a NoMethodError and the second gives a syntax error?
Exactly. And to be entirely clear, the fact that the method does not
exist doesn’t even matter. If the method does exist, the same syntax
error is raised if the first argument is a symbol. And when using
parentheses, no syntax error occurs…
I think the explanation is simple: you (most likely) found a bug in the parser.
You should report this to ruby-core@
In my humble opinion I think it’s even simpler than that.
Sorry if I’m wrong, but…
If you have an object (expression returning an object reference) what
are the syntactic constructs allowed after that?
When the variable is not defined the parser can think that it is a
method name (an undefined method error) and the “Symbol” is an
argument. But, when it is defined, it returns an object, and the colon
is not allowed syntactically in this context.
I just couldn’t figure out why using constants renders different
behavior.
Look
nil is an object.
Try this:
nil:y
nil :y
nil: y
nil : y
nil : y
class A; end
a = A.new
a:y
a :y
a: y
a : y
A:y (No Method)
A :y (No Method)
A: y
A : y
Only “A:y” and “A :y” raises a Method error. The others, all of them
raises Syntax error.
(tested against ruby 1.9).
I think the explanation is simple: you (most likely) found a bug in the parser.
You should report this to ruby-core@
In my humble opinion I think it’s even simpler than that.
…
If you have an object (expression returning an object reference) what
are the syntactic constructs allowed after that?
A symbol should be allowed anywhere any other literal is allowed.
I don’t have an explanation for why using the symbol gives a
SyntaxError rather than a NoMethodError, but the above three of your
new examples seem to me to be the key examples, together with maybe
the following on Ruby1.9.1. Note that using “eval” with x not being
defined as a variable also raises a SyntaxError.
two types of error message:
NoMethodErr === #<NoMethodError: undefined method `x’ for
Results in:
ruby 1.9.3dev (2010-12-28 trunk 30412) [x86_64-darwin10.5.0]
/Users/benoitdaloze/Desktop/test.rb:210: warning: `:’ after local
variable is interpreted as binary operator
/Users/benoitdaloze/Desktop/test.rb:210: warning: even though it seems
like symbol literal
/Users/benoitdaloze/Desktop/test.rb:210: warning: possibly useless use
of a variable in void context
/Users/benoitdaloze/Desktop/test.rb:210: syntax error, unexpected ‘:’,
expecting $end
p :sym
^
If I remember correctly, it worked with Nobu’s patch. It must have
been reverted since.
Please post your ideas/suggestions about this bug, it is not so easy
with the operators.
So, for the record, the Symbol is not accepted because it is believed
as part of the ternary operator ? :
There are also a number of bugs fixed. Special thanks to Jozef Aerts, John
Sillers, Chris Tornau and a number of other users for input and suggestions
It might help if you say what it is in the announcements.
The core API of VTD-XML has been expanded. Users can now perform
cut/paste/insert on an empty element.
This release also adds the support of deeper location cache support
for
parsing and indexing. This feature is useful for application performance
tuning for
processing various XML documents.
The java version also added support for processing zip and gzip files.
Direct processing of httpURL based XML is enhanced.
Extended Java version now support Iso-8859-10~16 encoding.
A full featured C++ port is released.
C version of VTD-XML now make use of thread local storage to achieve
thread safety for multi-threaded application.
There are also a number of bugs fixed. Special thanks to Jozef Aerts,
John
Sillers, Chris Tornau and a number of other users for input and
suggestions