From: Eric Sosman on
Wojtek wrote:
> Lew wrote :
>> Lew wrote :
>>>>> Map <TimeZone, String> map =
>>>>> new TreeMap <TimeZone, String> ( new Comparator <TimeZone> ()
>>>>> { .... } );
>>>
>>
>> Wojtek wrote:
>>>> No arguments that this does work.
>>>
>>>> However I was replying to your statement "'TimeZone' can easily be a
>>>> map
>>>> key, yes, even for a 'TreeMap'. " which appears to say that you do not
>>>> need a custom Comparator:
>>>
>>
>> Even though I had explicitly mentioned a custom Comparator in the post
>> to which you were replying? Come on, now. Your explanation is
>> disingenuous at best.
>
> To me a 'key' is the part which goes into the 'K' part of Map<K,V>.

Yes, that's right.

> For a Comparator, the signature is Comparator<T> where the 'T' refers to
> a 'Type': http://www.j2ee.me/javase/6/docs/api/java/util/Comparator.html

... which means only that the K in Map<K,V> is the same as
the T in Comparator<T>, or is a subclass/subinterface of T. You
shouldn't read too much into the particular letters chosen to name
type parameters; the names are as arbitrary as those of method
parameters, and carry no deeper meaning.

Does Map<String,String> bother you, because String == String
but K != V? If not, K != T shouldn't bother you, either, even if
you have a Map<Book,Date> and a Comparator<Book> with Book == Book.

--
Eric.Sosman(a)sun.com
From: Wojtek on
Eric Sosman wrote :
> Wojtek wrote:
>> Lew wrote :
>>> Lew wrote :
>>>>>> Map <TimeZone, String> map =
>>>>>> new TreeMap <TimeZone, String> ( new Comparator <TimeZone> ()
>>>>>> { .... } );
>>>>
>>>
>>> Wojtek wrote:
>>>>> No arguments that this does work.
>>>>
>>>>> However I was replying to your statement "'TimeZone' can easily be a map
>>>>> key, yes, even for a 'TreeMap'. " which appears to say that you do not
>>>>> need a custom Comparator:
>>>>
>>>
>>> Even though I had explicitly mentioned a custom Comparator in the post
>>> to which you were replying? Come on, now. Your explanation is
>>> disingenuous at best.
>>
>> To me a 'key' is the part which goes into the 'K' part of Map<K,V>.
>
> Yes, that's right.
>
>> For a Comparator, the signature is Comparator<T> where the 'T' refers to a
>> 'Type': http://www.j2ee.me/javase/6/docs/api/java/util/Comparator.html
>
> ... which means only that the K in Map<K,V> is the same as
> the T in Comparator<T>, or is a subclass/subinterface of T. You
> shouldn't read too much into the particular letters chosen to name
> type parameters; the names are as arbitrary as those of method
> parameters, and carry no deeper meaning.
>
> Does Map<String,String> bother you, because String == String
> but K != V? If not, K != T shouldn't bother you, either, even if
> you have a Map<Book,Date> and a Comparator<Book> with Book == Book.

Semantics!

We work in a precise field. An extra semi-colon can make a world of
difference (brought down a phone system a few years ago).

So yes, the labelling DOES make a difference. Whereas a Book is a Book,
where it is used does change its meaning. So in Map<Book,Date>, Book is
a key and in Comparator<Book>, Book is a Type.

Otherwise the Javadoc author would not have made that distinction.

It is used differently. As a key, it must provide its own comparison
methodology. As a type, the Comparator makes the comparison, possibly
using an external conversion such as I18N.

--
Wojtek :-)


From: Wojtek on
Wojtek wrote :
> We work in a precise field. An extra semi-colon can make a world of
> difference (brought down a phone system a few years ago).

Ah sorry, it was a break statement:
http://users.csc.calpoly.edu/~jdalbey/SWE/Papers/att_collapse.html

--
Wojtek :-)


From: Lew on
Wojtek wrote:
> Semantics!

The word "semantics" means "meaning", so it isn't trivial.

> We work in a precise field. An extra semi-colon can make a world of
> difference (brought down a phone system a few years ago).
>
> So yes, the labelling DOES make a difference. Whereas a Book is a Book,
> where it is used does change its meaning. So in Map<Book,Date>, Book is
> a key and in Comparator<Book>, Book is a Type [sic].

Now I see where your difficulty lies and I'm much more sympathetic to your cause.

In both contexts the type parameter signifies a type. In 'Map <K, V>' there
are two type parameters, indicating the types to which the Map is bound. The
first is the type of the key, and the second is the type of the value.

There is no semantic difference between expressing that as 'Map <K, V>',
'Map <T, U>' or even 'Map <foo, bar>'.

> Otherwise the Javadoc author would not have made that distinction.

There is no distinction. The choice of letters in the Javadocs is arbitrary
and designed to help us understand the usage of the two types managed by the
'Map', but regardless a type parameter is a type parameter is a type parameter.

> It is used differently. As a key, it must provide its own comparison
> methodology. As a type, the Comparator makes the comparison, possibly
> using an external conversion such as I18N.

That is not true. Take a look again at the Javadocs for 'TreeMap' to which I
pointed you earlier:

<http://java.sun.com/javase/6/docs/api/java/util/TreeMap.html#TreeMap(java.util.Comparator)>

Note that they describe the comparator type parameter as '<? super K>' - that
is the exact same 'K' as in the 'TreeMap <K, V>' type parameter.

For further understanding, check out the rules for type parameters in the JLS.

--
Lew
From: Lew on
Wojtek wrote:
>> It is used differently. As a key, it must provide its own comparison
>> methodology. As a type, the Comparator makes the comparison, possibly
>> using an external conversion such as I18N.

Lew wrote:
> That is not true. Take a look again at the Javadocs for 'TreeMap' to
> which I pointed you earlier:
>
> <http://java.sun.com/javase/6/docs/api/java/util/TreeMap.html#TreeMap(java.util.Comparator)>
>
> Note that they describe the comparator type parameter as '<? super K>' -
> that is the exact same 'K' as in the 'TreeMap <K, V>' type parameter.

I was unclear, in part because your antecedents are hazy. It is true that a
'TreeMap' that uses "natural" order does not need a 'Comparator' and that one
that needs a 'Comparator' does need a 'Comparator'. It is not true that the
type parameter 'K' in 'TreeMap <K, V>' is different from the type parameter
'K' in the 'TreeMap' constructor argument 'Comparator <? super K>'.

--
Lew