Re: output of enumeration types

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On Wed, 20 Apr 2005 02:17:26 +0200, Staszek Goldstein wrote:

> Uzytkownik "Stephen Leake" <stephen_leake@acm.org> napisal w wiadomosci 
> news:mailman.54.1113954069.24457.comp.lang.ada@ada-france.org...

>> As far as I can see, it is "just as easy". There is a package, that
>> someone wrote, that does text output for type String. Since you are
>> writing a new type, you need to write the output package for that as
>> well. Once that is done, both are "just as easy" for anyone else to
>> use.
>>
>> Note that Character has a special place in this discussion, since by
>> definition it is what gets written to a "text" file.
>>
>> So to do "output" with Roman_Number, you must specify some way to
>> convert its components to Character.
> 
> I do not feel convinced by the argument. If you want to write a number, you 
> can qualify it
> to be of some specific type and output it using, say, some instantiation of 
> the
> ada.text_io.integer_io package. Although the situation is very similar 
> here - I just want to be
> able to output character literals or arrays of such literals, I have no 
> generic package at
> my disposal and no possibility of converting them directly to the 
> "character" or "string" type.

What you refer are remnants of Ada 83 [flawed] approach which hopefully
someday will be replaced in a more OO way, where 'Image would become
object's attribute, dispatching and thus requiring no type specification in
any form. Note that generic instantiation (of Integer_IO etc) is a sort of
type specification.

>> Here's another tidbit that points out why "universal character" wont
>> solve this problem:
>>
>>   type Roman_Digit is (Foo,'V','X','L','C','D','M');
>>
>> is a perfectly legal type declaration!
> 
> I am not sure if this is really a problem. If 'V' is of a "universal 
> character type", then it is directly
> convertible to any character type which contains it as an enumeration 
> literal. Also, writing
> character(rm), where rm is a roman digit, should do no harm - this should 
> raise constraint_error
> or program_error if rm is not a character literal. More or less the same is 
> true of
> integer(long_int)... By the way, I have not experienced many applications of 
> the "mixed" types
> up to now.

Universal character as a common ancestor for all character type is a good
idea, but as Stephen has pointed out, it would not solve this particular
problem. You need a way to convey that Roman_Digit is a character type
derived from universal character. Presently Ada does not have character
class. Neither you can create a subtype of Character non-contiguous ranges
of characters (BTW, character maps might be what you are looking for.)
Further you'll need universal strings and the string class, also useful,
but non-existent in Ada. Too much work.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de