Talk:Comparison of ALGOL 68 and C++

Latest comment: 12 years ago by Rp in topic ":="

Isn't this discussion missing about a zillion C++ features..? —Preceding unsigned comment added by 140.211.173.16 (talk) 22:11, 13 May 2010 (UTC)Reply

":="

edit

The phrase ":= - assignment operation symbol (to avoid confusion with equal sign)" sounds a bit confusing: Does it mean specifically that c++ hasn't that symbol, or does it mean that c++ has no distinction between comparison- ("==") and assignment-operator ("=")? Phresnel (talk) 14:32, 12 January 2009 (UTC)Reply

There are several different scenarios that come into play:

Intent ALGOL 68 C++
Assign a value 888 to a variable x x:=888; x=888;
Compare two value IF x = 888 THEN ... FI if(x==888){ ... }
Define a constant INT x=888; const int x=888;
*Initialise a variable INT x:=888; int x=888;
*Allocate a variable from the HEAP REF INT x = HEAP INT;
or simply:
HEAP INT x;
int *x = (int*)malloc(sizeof(int));
Compare address of two pointers REF INT x, y;
IF x :=: y THEN ... FI
int *x, *y;
if(x==y){ ... }
Compare value referenced by two pointers REF INT x, y;
IF x = y THEN ... FI
int *x, *y;
if(*x==*y){ ... }
*Name a new type MODE LONGREAL = LONG REAL; typedef double longreal;
*Name a new record type MODE CUST = STRUCT(STRING name, address); struct cust {char *name, *address};
*Name a procedure PROC f = (REAL x)REAL: ( code; result ); float f(float x){ code; return result }
*Name a new operator OP ↑ = (REAL x,y)REAL: x**y; n/a
*Set priority on a new operator PRIO ↑ = 9; n/a

C and C++ both have the irony that the "=" operator uses the "equals" character, but the usage is neither symmetric nor commutative as it is actually a right to left assignment. And where an actual equality comparison is required (in C++ and C) one is required to use the "==" operator.

C descended from ALGOL 68 and BCPL, neither had "==". And both used ":=" for assignment, so I am not sure why C broke the equality.

To quote Dennis Ritchie (April 1993). "The Development of the C Language"

"Other fiddles in the transition from BCPL to B were introduced as a matter of taste, and some remain controversial, for example the decision to use the single character = for assignment instead of :=. Similarly, B uses /* */ to enclose comments, where BCPL uses //, to ignore text up to the end of the line. The legacy of PL/I is evident here."

Also note that initially "proto C" used PL/I type declarations, but when Bourne arrived at AT&T this was changed to ALGOL 68 type declarations including the void type.

NevilleDNZ (talk)

Yes, I know. What I am confused about is that the phrase sounds very much like "C (and relatives) doesn't use different symbols for the assignment- and equality-operation", but which is clearly not the case. And "One is required to ..." sounds like personal opinion. I personally, as a Programmer of C and C++ (and PHP, C#, ...) think it is more clumsy to use := as the assignment operator (especially in an imperative language where assignment is one of the most common operations), because both on an english as well as on a german keyboard it is very slow to type that combination (compared to other operators).

I have had exactly that experience of clumsy- and slowness when I was a Delphi programmer for some years a while back. Hence, I personally think one is more "required" to type := than a single =.

Phresnel (talk) 06:48, 13 January 2009 (UTC)Reply

re: "clumsy- and slowness" of ":=", here is an "anonymous quote" from "COMPILER CONSTRUCTION" - W. M. McKeeman:

"If PL/I is the Fatal Disease,
then perhaps Algol-68 is
Capital Punishment".
   - An Anonymous Compiler Writer

Consider ALGOL 68's:

FOR i FROM j BY k TO l WHILE m < n DO INT o:=p; print(q) OD

Versus C's:

for(int i=j; i<=l; i+=k){ if(!m<n)break; int o=p; printf("%d",q); }

What is not obvious that in ALGOL 68 the programmer had to go into or out of SHIFT-LOCK a grand total of 14 times. Effectively making this example line 14 key presses longer to type. If your keyboard has a CAPS-LOCK (eg a Selectrics keyboard, or an IBM 3270) then this drops to only 7 extra key presses.

Alternatively - as in the actual ALGOL 68 spec - the reserved word's can be in a bold typeface, while this isn't CAPITAL punishment as such, it still makes for extra typing. eg:

for i from j by k to l while m < n do int o:=p; print(q) od

BTW: you can download the ALGOL 68G Interpretor from http://sourceforge.net/project/showfiles.php?group_id=114223 And if you are feeling a bit retro, then add a code sample to http://rosettacode.org/wiki/Category:ALGOL_68 - there are already 100+ samples there. This is a novel way to compare ALGOL 68 with C++. NevilleDNZ (talk)

Heh, that anonymous quote is funny, I didn't know about it, but am I also interested in Compiler Construction in my spare time.

for(int i=j; i<=l; i+=k){ if(!m<n)break; int o=p; printf("%d",q); }

This code sounds rather artficial: m and n don't change (I assume they are not volatile), so why move them inside the for loop? Plus i is a loop-local variable, which means no code after the loop can depend on it. As we are in C and not in C++, the following is semantically equivalent.

     if(m<n) for(int i=j; i<=l; i+=k){ int o=p; printf("%d",q); }
v.s. FOR i FROM j BY k TO l WHILE m < n DO INT o:=p; print(q) OD

Still bad code. What is o=p for? Nevermind for the rest, let me minimize it all:

     if(m<n)for(int i=j;1>i;i+=k){int o=p;printf("%d",q);}
v.s. FOR i FROM j BY k TO l WHILE m < n DO INT o:=p; print(q) OD

I don't know ALGOL 68, but I try to minimize it (correct me if I am wrong):

     if(m<n)for(int i=j;1>i;i+=k){int o=p;printf("%d",q);}
v.s. FOR i FROM j BY k TO l WHILE m<n DO INT o:=p;print(q)OD
White space is mostly ignored by ALGOL 68, so you could do this:
v.s. FORiFROMjBYkTOlWHILEm<nDO INTo:=p;print(q)OD
At least the compiler would know what to do... NevilleDNZ (talk)

Don't want to make this a flame :D. The reason why I find := clumsy is the reachability without breaking fingers ":": lowest row on keyboard, "=": uppermost row, then, both are nearly at the same column.

I will have a look at those examples, thanks!

- - Phresnel (talk) 15:25, 14 January 2009 (UTC)Reply

Oddly enough, Unicode has some related characters:

≔ 0x2254 8788 COLON EQUALS
≕ 0x2255 8789 EQUALS COLON

I guess if I puzzled long enough I'd figure out where ≔ and ≕ came from. I seem to recall that ALGOL W punch cards had a ≔ character. A human would have to look long and hard to find a keyboard that supported this character.

BTW: the "artficial" code:

for(int i=j; i<=l; i+=k){ if(!m<n)break; int o=p; printf("%d",q); }

Was cut and paste directly from the Unix SysVr5 kernel. (not ☺)

NevilleDNZ (talk) 11:13, 4 February 2009 (UTC)Reply

Prof. Koster told us that the := is meant to be an approximation of a left arrow for keyboards without one. Rp (talk) 14:35, 11 May 2012 (UTC)Reply