P3
We got this feedback from ###@###.###. We need to
fix the spec accordingly.
> Minor issues:
> * p.10 and p.33 use ".length" differently; is it "array.length"
> or "array.length()"? If it's "array.length", is it an lvalue?
"length" is an operator that can be applied to arrays. Strings however
are instances of class string and define a method called "length". This
means that "string".length() is valid when applied to a string.
By the way, I can't math the page numbers with the copy of the spec
that I have. What date does it have on it?
> * p.16 shows two methods in class "A" without definitions. Is
> this a legal syntax? If so, how does a developer define these
> methods (i.e., I assume it would be via the "::" syntax from
> C++; this should be documented somewhere).
This is an error in the spec. The methods should have a definition.
We don't allow methods to be defined outside the class definition.
I find this very confusing.
> * p.20 uses the phrase "not visible" as synonymous to "not
> accessible". C++ uses access rather than visibility, having
> to do with a nasty little rule called the Hiding Rule. Since
> you don't allow operator overloading (in particular, you don't
> allow overloading of the assignment operator), you can probably
> get away without the hiding rule. If this is the case, you
> may be able to define "visibility" and "accessibility" in a
> nearly synonymous way. In any event, paragraph 4.8 needs to
> carefully state whether access==visibility or access!=visibility.
Right. acces!=visibility. This should be made clearer.
> * 4.10.1: "while one thread is using it" needs to be defined.
> I.e., what exactly constitutes a "use" of a variable? Does it
> mean "while any method is executing on the indicated object",
> or "during the middle of a statement within one of these methods",
> or "between sequence points of a statement within one of these
> methods"?
That is a good point. The implied definition is "while a method
is using it". I'll make sure this will be made clearer.
> * 4.10.2: Please describe "specially" in the sentence,
> "Variables marked transient are treated specially when instances
> of the class are written out as persistent objects".
Making a variable transient sets a bit in the variable slot. If this
transient bit is NOT set the variable's value is saved to a file when
writing out the object. This feature is rather old and is currently
not implemented.
> * 4.10.4: What is the C signature of a native method? For example,
> what's the C signature of a method "barney(int,String)" of class
> "Fred"?
This is defined in a seperate document called "Implementing Native Methods".
The mapping is very straight forward. I can send you a copy of the
document if you are interested. It is not part of the spec because each
native language requires different bindings.
> * 4.10.5: Regarding "A method that overrides a superclass method
> cannot be abstract", this feature would be useful if (1) the
> superclass method is abstract, and (2) the "override" narrows
> the return type.
True. We will review this restriction.
> * 4.10.5: Regarding "It is an compile-time error to ... attempt
> to call an abstract method directly", how would one call a
> method "directly"? Does Java allow a "::" syntax in a method
> call a la C++?
You can call a method directy using super (ie: super.foo() calls
foo in the direct super class). It is a minor point but avoid finding
out at runtime. By the way, you can always qualify classes using a fully
qualified class name (use . instead of :: though).
> * 6.2: In the last example, I believe "FooBar" should be
> "Employee_List".
Yes.
> * page 35: I believe the code: I r = (I)new("I");
> should be replaced with: I r = new I();
The new("I") expression is actually legal. It allows you to instanciate
classes by name.
> * page 39: The title of this section is "MIndex". Is this a
> typo?
That is a typo. Thanks for pointing that out.
> * page 40: "synchronize" should be "synchronized".
Again, you are right!
fix the spec accordingly.
> Minor issues:
> * p.10 and p.33 use ".length" differently; is it "array.length"
> or "array.length()"? If it's "array.length", is it an lvalue?
"length" is an operator that can be applied to arrays. Strings however
are instances of class string and define a method called "length". This
means that "string".length() is valid when applied to a string.
By the way, I can't math the page numbers with the copy of the spec
that I have. What date does it have on it?
> * p.16 shows two methods in class "A" without definitions. Is
> this a legal syntax? If so, how does a developer define these
> methods (i.e., I assume it would be via the "::" syntax from
> C++; this should be documented somewhere).
This is an error in the spec. The methods should have a definition.
We don't allow methods to be defined outside the class definition.
I find this very confusing.
> * p.20 uses the phrase "not visible" as synonymous to "not
> accessible". C++ uses access rather than visibility, having
> to do with a nasty little rule called the Hiding Rule. Since
> you don't allow operator overloading (in particular, you don't
> allow overloading of the assignment operator), you can probably
> get away without the hiding rule. If this is the case, you
> may be able to define "visibility" and "accessibility" in a
> nearly synonymous way. In any event, paragraph 4.8 needs to
> carefully state whether access==visibility or access!=visibility.
Right. acces!=visibility. This should be made clearer.
> * 4.10.1: "while one thread is using it" needs to be defined.
> I.e., what exactly constitutes a "use" of a variable? Does it
> mean "while any method is executing on the indicated object",
> or "during the middle of a statement within one of these methods",
> or "between sequence points of a statement within one of these
> methods"?
That is a good point. The implied definition is "while a method
is using it". I'll make sure this will be made clearer.
> * 4.10.2: Please describe "specially" in the sentence,
> "Variables marked transient are treated specially when instances
> of the class are written out as persistent objects".
Making a variable transient sets a bit in the variable slot. If this
transient bit is NOT set the variable's value is saved to a file when
writing out the object. This feature is rather old and is currently
not implemented.
> * 4.10.4: What is the C signature of a native method? For example,
> what's the C signature of a method "barney(int,String)" of class
> "Fred"?
This is defined in a seperate document called "Implementing Native Methods".
The mapping is very straight forward. I can send you a copy of the
document if you are interested. It is not part of the spec because each
native language requires different bindings.
> * 4.10.5: Regarding "A method that overrides a superclass method
> cannot be abstract", this feature would be useful if (1) the
> superclass method is abstract, and (2) the "override" narrows
> the return type.
True. We will review this restriction.
> * 4.10.5: Regarding "It is an compile-time error to ... attempt
> to call an abstract method directly", how would one call a
> method "directly"? Does Java allow a "::" syntax in a method
> call a la C++?
You can call a method directy using super (ie: super.foo() calls
foo in the direct super class). It is a minor point but avoid finding
out at runtime. By the way, you can always qualify classes using a fully
qualified class name (use . instead of :: though).
> * 6.2: In the last example, I believe "FooBar" should be
> "Employee_List".
Yes.
> * page 35: I believe the code: I r = (I)new("I");
> should be replaced with: I r = new I();
The new("I") expression is actually legal. It allows you to instanciate
classes by name.
> * page 39: The title of this section is "MIndex". Is this a
> typo?
That is a typo. Thanks for pointing that out.
> * page 40: "synchronize" should be "synchronized".
Again, you are right!