root@topaz.RUTGERS.EDU (Charles Hedrick) (07/30/86)
You are right that I have heard the complaint about interactive I/O a
number of times in the past. However it has appeared much less often
recently, since the various implementations seem to be converging on
the use of "lazy I/O". Most implementors seem to believe that the ISO
standard implies that this is the correct way to handle interactive
I/O. You complain that the following program will print an extra
prompt.
writeln ('enter an integer : ');
while not eof do
begin
readln (i);
writeln ('enter an integer : ');
end;
As I see it, you will have roughly the following dialog:
enter an integer: 123
enter an integer: ^D
at which point the program is finished. I don't see what is wrong
with that. Perhaps you want
enter an integer: 123^D
If you want to be able to recognize that, you will simply have to
use a separate call to readln. Presuming that you want to be able
to recognize ^D in either context, the program would look like
while true do
begin
writeln ('enter an integer : ');
if eof
then {exitloop} goto 666;
read (i);
if eof
then {exitloop} goto 666;
readln;
end; 666:
This tests for eof twice because I assume you want to allow it either
place. (Sorry about the goto's, but the logic is clearer that way than
with any alternative I could think of.)
As far as I know, the semantics of lazy I/O allow you to check for any
possible input, printing prompts at any point. This does not mean
that Pascal's input parsing is as powerful as scanf, or that its
output is as powerful as printf. Obviously they are not. Validating
input, to make sure that the user hasn't typed something bogus, can
require a bit more programming in Pascal than in C. For example, in
the above programs, if the user types an extra CR before the ^D the
program may think that there is one more number than there is. If you
want the program to be very careful about checking validity of its
input, you will need to write code to check that the user has typed
one and exactly one number on each line. But as far as I know, you
can always find a way to avoid having extra prompts and misplaced
input hangs.crs@lanl.ARPA (Charlie Sorsby) (08/01/86)
> As far as I know, the semantics of lazy I/O allow you to check for any > possible input, printing prompts at any point. This does not mean > that Pascal's input parsing is as powerful as scanf, or that its > output is as powerful as printf. Obviously they are not. Validating > input, to make sure that the user hasn't typed something bogus, can > require a bit more programming in Pascal than in C. For example, in Am I missing something? As I understand it, printf and scanf are not considered part of the C language but rather are part of the standard I/O package. Is this also true for writeln and readln in Pascal? If so, it appears that standard I/O packages are being compared rather than languages. If not, perhaps the comparisons are between apples and oranges. Is there some characteristic of the Pascal language that prevents construction of a Pascal standard I/O package that *would* be as powerful as that of C? Perhaps the existence of this characteristic is implicit in this discussion and that is what I am missing. If so, forgive my ignorance. (I realize that the lack of "separate compilation" in Pascal precludes use of precompiled libraries of functions.) Of course, there is also the fact that Pascal was intended as a teaching language to consider. -- Charlie Sorsby ...!{cmcl2,ihnp4,...}!lanl!crs crs@lanl.arpa
root@topaz.RUTGERS.EDU (Charles Hedrick) (08/02/86)
Because Pascal does strong type-checking, it would be hard to write anything like printf and scanf for it. Indeed if writeln and readln were not built into the compiler, you would not be able to define them yourself. A Pascal procedure must have a fixed number of arguments, each of a definite type. Thus the best you could do would be printint(int i); printreal(real r); etc. Unfortunately, I/O libraries are not something that can be tacked on later. I/O affects the definition of the language. This is one of the most serious problems with the languges in the Algol tradition. There is a tendency to think I/O is a detail unfit for computer scientists, which can be left to the "standard prolog".
gvcormack@watmum.UUCP (Gordon V. Cormack) (08/02/86)
> > Am I missing something? > > Is there some characteristic of the Pascal language that prevents > construction of a Pascal standard I/O package that *would* be as powerful > as that of C? Perhaps the existence of this characteristic is implicit in > this discussion and that is what I am missing. There are lots of reasons one can't write printf in Pascal, and they all contribute to the conclusion that Pascal is garbage. Just for starters, you can't (in Wirth's Pascal, anyway) even write a procedure that takes a quoted string as a parameter, unless its length is hard coded. Next, you can't write a procedure that takes a variable number of parameters. Finally, you can't write a *polymorphic* procedure that operates on many types, nor can you even overload the name "printf" to make it apply to many types. I wouldn't want anyone to take this article as support for C. For a long time I have been suppressing the urge to post an article to the effect: STOP!! You're both wrong. -- Gordon V. Cormack CS Department, University of Waterloo uucp: { allegra, decvax, ... }!watmath!gvcormack csnet: gvcormack%watmum@waterloo cdn: gvcormack@mum.waterloo.cdn
ali@navajo.STANFORD.EDU (Ali Ozer) (08/04/86)
In article <563@watmum.UUCP> gvcormack@watmum.UUCP (Gordon V. Cormack) writes: >...Just for starters, you can't (in Wirth's Pascal, anyway) even write >a procedure that takes a quoted string as a parameter, unless its >length is hard coded... True, but thanks to conformant arrays, you can do stuff like: procedure Foo (str: packed array [lo..hi: integer] of char); Then calls to Foo with packed arrays of any length are legal. And conformant arrays are part of the Pascal standard now, aren't they? Ali, ali@su-score.arpa, ali@navajo
jsdy@hadron.UUCP (Joseph S. D. Yao) (08/10/86)
In article <5918@lanl.ARPA> crs@lanl.ARPA (Charlie Sorsby) writes: [attribution omitted] >> As far as I know, the semantics of lazy I/O ... >As I understand it, printf and scanf are not considered part of the C >language but rather are part of the standard I/O package. Is this also >true for writeln and readln in Pascal? If so, it appears that standard >I/O packages are being compared rather than languages. ... Yes, you are missing something. Pascal I/O is part of the language, and not part of any library. It is a "built-in." -- Joe Yao hadron!jsdy@seismo.{CSS.GOV,ARPA,UUCP} jsdy@hadron.COM (not yet domainised)