melling@cs.psu.edu (Michael D Mellinger) (05/25/90)
I am learning X Window and have just found out that MIT borrowed the
Cardinal type from Pascal. Now in order to avoid all kinds of
warning messages, I have to declare my variables like argc to be
either of type Cardinal, unsigned int, or u_int(the way I used to
declare non-negative numbers). This might seem like a minor detail, but
three different ways to say something is a too much! Which style is
the more acceptable? Imagine teaching someone C or C++ and having to
explain to them why there are three names for numbers that are
non-negative. Maybe you could tell them that C (C++) is a language where
everyone does their own thing.
-Mike
------------------------- The problem illustrated -------------------------
main(u_int argc, char* argv[])
{
// XtAppInitialize() expects argc to be a Cardinal number
.
.
.
toplevel = XtAppInitialize(&app_con, "Xhw", NULL, ZERO, &argc, argv,
fallback_resources, NULL, ZERO);
.
.
.
}philip@Kermit.Stanford.EDU (Philip Machanick) (05/26/90)
In article <Fisw-a2@cs.psu.edu>, melling@cs.psu.edu (Michael D Mellinger) writes: > > I am learning X Window and have just found out that MIT borrowed the > Cardinal type from Pascal. ^^^^^^ Don't you mean Modula-2? Philip Machanick philip@pescadero.stanford.edu
melling@cs.psu.edu (Michael D Mellinger) (05/26/90)
In article <1990May25.193305.14920@Neon.Stanford.EDU> philip@Kermit.Stanford.EDU (Philip Machanick) writes: In article <Fisw-a2@cs.psu.edu>, melling@cs.psu.edu (Michael D Mellinger) writes: > > I am learning X Window and have just found out that MIT borrowed the > Cardinal type from Pascal. ^^^^^^ Don't you mean Modula-2? Philip Machanick philip@pescadero.stanford.edu Maybe it was Modula-2. I have never learned it though, so I can't be sure. I must have picked it up from somewhere because I have previously used it in my Pascal code. Anyway, doesn't it seem unsightly to see three names for exactly the same thing in C++ code. void foo(unsigned int a, u_int b, Cardinal c) { . . . } C++ is suppose to be a language for large projects. Reusability in the form of class libraries is suppose to be important(will this ever become a reality?). Although the three definitions are equivalent, mixing class libraries or code from different programmers makes the program look ugly (IMHO), if all three types declarations are used. I guess it's a question of style. I personally don't like unsigned int because it's two words and it's too long. Are there any generally accepted coding standards that are used by groups of people working on large projects(I guess MIT uses Cardinal)? -Mike
jgk@demo.COM (Joe Keane) (05/31/90)
In article <Ft*!k&2@cs.psu.edu> melling@cs.psu.edu (Michael D Mellinger) writes: >I guess it's a question of style. I personally don't like unsigned int >because it's two words and it's too long. This is funny. Do you avoid everything that's two words? The canonical name of the type is `unsigned int'. If you think that's too long, you can call it what you want. A common typedef is `u_int', used in the Unix kernel. Begin tirade... So what the heck is an `int'? Accoding to K&R, it's a bit pattern ``of the natural size for integer arithmetic on a given machine''. Similarly, a `float' is a ``single precision floating-point number''. These types are explicitly defined and named in terms of low-level machine resources. I think this is something Pascal and Modula-2 got wrong. Calling a machine word an `INTEGER' or `CARDINAL' and a floating-point number a `REAL' is a lie. I know what an integer is, and when you add two positive ones you better get another positive one. Similarly, real numbers obey associativity and similar laws which floating-point numbers do not. Furthermore, this naming paves over the distinction between abstract data types and their implementations. I think this is a bad thing to teach people who are learning to program. The world is full of things that screw up with more than 32767 (or 255 or 2147483647) of something. We need less of this, and Pascal confusing the issue isn't going to help any. At least C is honest about its type names. Demand `truth in advertising'.
jejones@mcrware.UUCP (James Jones) (06/01/90)
In article <2875@demo.COM> jgk@osc.COM (Joe Keane) writes: >I think this is something Pascal and Modula-2 got wrong. Calling a machine >word an `INTEGER' or `CARDINAL' and a floating-point number a `REAL' is a lie. >I know what an integer is, and when you add two positive ones you better get >another positive one. Similarly, real numbers obey associativity and similar >laws which floating-point numbers do not.... > >At least C is honest about its type names. Demand `truth in advertising'. Ah, but...I know what an integer is, too, and the whole reason for defining integers is to get negative numbers. Doesn't this make "unsigned int" an oxymoron? It would've been nice to have "nat" (for "natural number") instead; not only is it truth in advertising, but it fits in with the brevity-uber- alles nature of C (at least pre-ANSI, otherwise we'd have "vol" instead of "volatile" :-). :-) :-) :-) :-), of course! James Jones (Opinions? Can a company be said to have an opinion? Assuming it can, I know of no company whose opinions have any correlation with mine.)
multics@clotho.acm.rpi.edu (Richard Shetron) (06/03/90)
How about the declaractions in PL/1. declare a fixed bin (7), b fixed bin(15), c fixed bin(31); which gives you an 8 bit binary integer for a, a 16 bit binary integer for b, and a 32 bit binary integer for c. The sign bit is not included in the bit count. YOu can also declare items as unsigned and full PL/1 compilers support unalingned data types as well. You also have: declare x float bin(24), y float bin(56); /* IBM 360/370 forms */ These specify the equivalent of a c float and double. The (precision) is optional and I'm not sure of the defaults on all machines, but on IBM 360/370 and Multics fixed bin defaults to half word (15 IBM, 17 Multics), and float bin to full word. instead of bin you could say decimal if you needed decimal storage and arithmetic instead of binary. If something is fixed you can optionally specify the location of the 'decimal' point. FOr those interested the keyword real is assumed. any of the above could also be declared complex if you need complex numbers and arithmetic. The compiler is always free to supply more precision then the programmer asked for, but will give warnings/error messages if it can't meet the requirement. It was nice haveing float dec(59) on Multics for some statistical work I did. It might have been slower then float bin, but I didn't need to develope software multi variable precision routines to get my answers. A good bureaucracy is the best tool of oppression ever invented. Richard Shetron USERFXLG@rpi.mts.edu multics@clotho.acm.rpi.edu
multics@clotho.acm.rpi.edu (Richard Shetron) (06/04/90)
How about the declaractions in PL/1. declare a fixed bin (7), b fixed bin(15), c fixed bin(31); which gives you an 8 bit binary integer for a, a 16 bit binary integer for b, and a 32 bit binary integer for c. The sign bit is not included in the bit count. YOu can also declare items as unsigned and full PL/1 compilers support unalingned data types as well. You also have: declare x float bin(24), y float bin(56); /* IBM 360/370 forms */ These specify the equivalent of a c float and double. The (precision) is optional and I'm not sure of the defaults on all machines, but on IBM 360/370 and Multics fixed bin defaults to half word (15 IBM, 17 Multics), and float bin to full word. instead of bin you could say decimal if you needed decimal storage and arithmetic instead of binary. If something is fixed you can optionally specify the location of the 'decimal' point. FOr those interested the keyword real is assumed. any of the above could also be declared complex if you need complex numbers and arithmetic. The compiler is always free to supply more precision then the programmer asked for, but will give warnings/error messages if it can't meet the requirement. It was nice haveing float dec(59) on Multics for some statistical work I did. It might have been slower then float bin, but I didn't need to develope software multi variable precision routines to get my answers. A good bureaucracy is the best tool of oppression ever invented. Richard Shetron USERFXLG@rpi.mts.edu multics@clotho.acm.rpi.edu