Path: utzoo!mnetor!uunet!husc6!bloom-beacon!think!ames!sdcsvax!ucsdhub!hp-sdd!hplabs!hpda!hpesoc1!hpcllla!hpclisp!hpclscu!shankar
From: shankar@hpclscu.HP.COM (Shankar Unni)
Newsgroups: comp.lang.pascal
Subject: TP problem
Message-ID: <950003@hpclscu.HP.COM>
Date: 19 Dec 87 00:31:49 GMT
References: <159@conexch.UUCP>
Organization: HP ITG/ISO Computer Language Lab
Lines: 41

>
>In article <1808@rayssdb.RAY.COM> Richard A. Brooks writes:
>>In article <936@sask.UUCP>, Kevin Lowey writes:
>>> In article <16249@topaz.rutgers.edu>, Jonathan Joshua writes:
>>> >
>>> > [ the preceeding discussion about enumerated type I/O ]
>>> [...]
>> Try this:
>>program TEST;
>>
>>type
>>  ON_OFF    = (ON,OFF);
>>var
>>  v         = ON_OFF;
>>begin
>>    readln(v);
>>    writeln('Switch is now ',v);
>>end.
>>
>>    Upon compiling the above, you will recieve an I/O not
>>    allowed error at the "readln(v)" and the "writeln(v)",
>>    if the readln is removed.
>

Eric Pedersen replies:

>......                        To accomplish this, the compiler would
>have to keep the actual character names of every enumerated type in the
>object program, wasting lotsa space, and complicating the TEXT library
>I/O routines.

Well, a few compilers (decent ones, btw) written for larger machines DO 
support this feature exactly as you describe (a table is dumped somewhere
for each enumerated type for which I/O is requested). And yes, the code
generated for the I/O *is* a little gnarly. But boy! is it worth it!! Input
may not be too important (and may be counter-productive, because user typos
will not be handled too gracefully), but output is certainly useful, if for
nothing else but for debugging purposes.

Shankar Unni.
...!hplabs!hpda!shankar