Language extension

The task is to extend the C or C++ language by writing a preprocessor. The preprocessor need only be a program that reads the extended source file, makes textual changes, and saves the resulting C or C++ source file.

Beginners
---------

Extend either C or C++ to use a Pascal style for loop. The grammer is:

Code:

---

for identifier := i (to|downto) n
  statement

---

For example, a loop to count from 1 to 10 would be written in the extended language as:

Code:

---

for i := 1 to 10 {
  printf("%d\n", i);
}

---

And to count down from 10 to 1:

Code:

---

for i := 10 downto 1 {
  printf("%d\n", i);
}

---

Intermediate
------------

Extend either C or C++ to perform run-time boundary checking on arrays. If a subscript would cause the array access to go out of bounds, the program should terminate with an appropriate error message. The check must be performed before the array is accessed, otherwise undefined behavior would make this extension useless.

Advanced
---------

This task is for C only. Write an extension that gives C a built in string type. The following operators are required for the C string type:

==, !=, <, >, <=, >=

C strings must be able to grow dynamically. The null character has no special meaning in a C string. Strings may use the subscript operator to obtain and modify a specific character. If the subscript is out of range then the string will grow to meet the demand. Negative subscripts will walk backward from the end of the string.

Entries will be judged on how complete and robust the extension is. Sign-up for the contest will last one month. When all entries that have signed-up are complete, the entries will be judged. Entries that have not signed-up will be accepted provided they are submitted before all signed-up entries.

So essentially, you would send a file through the preprocessor and it would do this?
Another great contest :D .
Could we do more than one part? IE Beginners, Intermediate, and Advanced?

>So essentially, you would send a file through the preprocessor and it would do this?
Yes.

>Could we do more than one part? IE Beginners, Intermediate, and Advanced?
Way ahead of you. :) The beginner extension only requires simple textual replacement. The intermediate extension pretty much requires a symbol table for arrays with size information. And the advanced extension is just plain tedious. ;)

dont u think even the one u classified as a beginner category is way ahead for a beginner. i am not sure though. can u further suggest what to do here. how to connect to the prepocessor. i mean how to make a check. like for example in case of tc we hit alt-f9 ctrl-f9 right. so when will the preprocessor take over and how to make it take over.
2. same to be told how to do in case of gcc.
clarify that pls.thanx.

Prelude means a seperate program. For example, you would create extcpp.exe, and run it. It would take in the name of the file and run through it, changing the extended code to C++ and leaving everything else alone. The program would then end. You could then send the pure C++ file through a true compiler.

>dont u think even the one u classified as a beginner category is way ahead for a beginner.
It's just simple textual substitution. I think anyone who can read from a file and parse a simple string should have little trouble with it.

>how to connect to the prepocessor.
The preprocessor is just a separate program that you run on the source code before really compiling it. For example, you have a .ec file that uses extended C:

Code:

---

#include <stdio.h>

int
main(void)
{
  printf("My extended C program\n");
  for i := 1 to 10 {
    printf("%d\n", i);
  }
  printf("Bye bye!\n");

  return 0;
}

---

Then you would run the preprocessor program independently of the compiler to get a .c file (assuming you call your preprocessor program ppec and it takes two arguments, the first being the extended C and the second being the resulting C source file):

Code:

---

$ ppec prog.ec prog.c
$ cat prog.c
#include <stdio.h>

int
main(void)
{
  printf("My extended C program\n");
  {
    int i;
    for (i = 1; i <= 10; i++) {
      printf("%d\n", i);
    }
  }
  printf("Bye bye!\n");

  return 0;
}

---

Then you run the compiler on that C source file:

Code:

---

$ gcc prog.c
$ ./a.out
My extended C program
1
2
3
4
5
6
7
8
9
10
Bye bye!

---

I am signing up for the first one, I want to sign up for the second one but I am not sure what you mean. I understand run-time checking of an array for out of bounds limits, but what do you mean by subscript

The subscript operator is [].
Basically, if the number within said operator is too high / low, then an error should occur.

here is what I get from it

Code:

---

int main(void){
    int array[10];
    int i;
    for(i=1;i<=10;i++)
          array[i]=i;
    return 0;
}

---

this or is there more to it. This is a bound overrun. But it is precompiled are you talking about a user inputting something. Then see if it is too big in the code itself?

One small correction. On the pascal for loop the actual syntax is:

Code:

---

for identifier := i (to|downto) n do
  statement

---

On the C string do we have to provide normal 0 base counting for the accessing of a character? Does it have to support any length or is their a limit?

>On the pascal for loop the actual syntax is:
I know, but I was going for a decent merge rather than an exact duplicate of the Pascal syntax. A duplicate would be more difficult.

>On the C string do we have to provide normal 0 base counting for the accessing of a character?
You aren't required to, but it would certainly make sense for most C users.

>Does it have to support any length or is their a limit?
The limit is when you run out of memory. :)

linuxdude: What I mean is this, based on the code you posted. The loop execution will go as follows:

Code:

---

array[1] = 1;
array[2] = 2;
array[3] = 3;
array[4] = 4;
array[5] = 5;
array[6] = 6;
array[7] = 7;
array[8] = 8;
array[9] = 9;
Error.

---

Because 10 is out of bounds for the array, a run-time error is given and the program terminates without accessing the out of bounds index. The equivalent explicit check would be:

Code:

---

int main(void){
    int array[10];
    int i;
    for(i=1;i<=10;i++) {
        assert(i >= 0 && i < 10);
        array[i]=i;
    }
    return 0;
}

---

what about different formatting. I am checking to see if it is a normal for loop, but someone might do for\t( or for\n( or for ( or for( do you know what I mean. Do I have to keep looking until i find a ( or a variable to see what type it is?

>Do I have to keep looking until i find a ( or a variable to see what type it is?
Pretty much, yes. :)

I guess I will go for the advanced, but when is it due?

Quote:

---

Originally Posted by Prelude

Sign-up for the contest will last one month. When all entries that have signed-up are complete, the entries will be judged. Entries that have not signed-up will be accepted provided they are submitted before all signed-up entries.

---

Looks like about a month.

on the first assignment what about syntax errors. I am looking for a ( or a word Can I assume they won't have

Code:

---

for )

---

or something like that.

Um linuxdude you are going from Pascal to C so you shouldn't be looking for

Code:

---

for (

---

should be more along the lines of

Code:

---

for i := 0

---

I figured we could have both types of for loops in the code

Ah, I would guess if there is a syntax error then you ignore it :)

One last question. Will all the for loops conditional part at least be on one line? Also can we assume that variable name isn't a keyword??

Do not assume anything :)

>but when is it due?
As long as it takes everyone to finish. When everyone who signed up in the first month submits and entry, the contest will be judged and the results posted. So feel free to take your time and do a good job. :)

>I figured we could have both types of for loops in the code
Yes. The extension shouldn't effect use of the rest of the language at all. :)

>I would guess if there is a syntax error then you ignore it
If it's a syntax error with the extension then you should handle it appropriately. I don't expect the preprocessor to completely understand anything more than the extension that it implements, but valid extension syntax should usually compile into valid C or C++ code.

>Will all the for loops conditional part at least be on one line?
Not necessarily. C and C++ are free-form languages. ;)

>Also can we assume that variable name isn't a keyword??
Adding a check to ensure that the variable is a valid identifier and not a keyword would be a very good idea. The variable name does count as the syntax of the extension. Besides, such a check is relatively easy anyway. ;)

what about a loop like this. How is it in PASCAL?

Code:

---

for(i=0;i<250;x+=2);
or
for(i=0;i<100;x^=20);

---

Do we have to incorporate those. And if so what is the syntax?

Pascal for loops only increment or decrement by 1.

hmm, i'd enter this contest, but i'm too busy working on my actual language, lol. I'll just submit an OST->C++ converter program

>is that what you mean?
That would be a good start, but don't forget that the string has to be able to grow and shrink on its own. So using a statically sized array in the preprocessor doesn't meet the requirements.

I think the approach and method is for you to figure out.

>that means that i will heavily use realloc()?
Well, you don't have to. You could use malloc to allocate an array of some insanely huge size so that realloc or anything similar is unnecessary. ;) As for the actual implementation, that's up to you to decide. You get to weigh different options and choose the one that fits the task best. :)

Quote:

---

Originally Posted by C+++C_forever

ok, one more question and i am done: what's better:
a.realloc() ,or
b. free() and then malloc() ?
thanks

---

I don't like realloc. Any benefits you could get are destroyed by unpredictable behavior. You're effectively tricked into thinking that you're getting something when you're simply introducing subtle bugs. So rather than this:

Code:

---

T *temp = realloc(p, new_size);
if (!temp) {
  error("Memory exhausted");
}
else {
  p = temp;
}

---

I prefer to do this:

Code:

---

T *temp = malloc(new_size);
if (!temp) {
  error("Memory exhausted");
}
else {
  memcpy(temp, p, old_size);
  free(p);
  p = temp;
}

---

Now you can be sure of what's happening. Instead of hoping that any pointers into the memory block will not be invalidated, you ensure that they are so that they can be reseated accordingly. You also avoid sticky issues where realloc actually frees the memory instead of resizing it.

>but isn't somehow like that realloc() is implemented?
Yes, but realloc may or may not change the address of the original pointer which may or may not invalidate any pointers into the original memory. I'd rather be sure that it did so rather than hope it didn't and potentially get burned later when accessing a pointer outside of my address space.

>here you meanthat it first frees the memory, then tries to alllocate a new block and it fails, right?
No, I means that if the size is 0, realloc is the equivalent of free. No resizing is done, the memory is lost. The part that's sticky is that if the size were somehow set to 0 by accident, you'll end up accessing freed memory later.

>what do you mean here?

Code:

---

int *a = malloc(5 * sizeof *a);
int *p = &a[2];

/* Assuming success */
a = realloc(a, 10 * sizeof *a);
printf("%d\n", *p);

---

The pointer p may not point to valid memory after the call to realloc, so the printf call is unpredictable. It may work as expected or invoke undefined behavior. The only way to ensure that this doesn't happen is to reassign p before calling printf:

Code:

---

int *a = malloc(5 * sizeof *a);
int *p = &a[2];

/* Assuming success */
a = realloc(a, 10 * sizeof *a);
p = &a[2];
printf("%d\n", *p);

---

>are you the one who will judje our solutions? Do you have that much time? OR are you going to get paid for that?
I wouldn't mind getting paid. ;) But I'll find the time to judge.

>but the same problem would appear with the other method of malloc+free...
Yes, that's true. But I prefer having a guarantee such as "this method will invalidate any pointers into the memory" instead of something potentially problematic like "this method might invalidate any pointers into the memory". The two approaches are similar in size, but the former gives a definite guarantee as to what will happen while the latter is wimpy and error prone. ;)

It's a stylistic choice though, so if you prefer realloc then more power to you.

Finally got the beginner done. Yeah bug free and everything. I just need to make a few more changes. First make all the code the same. Sometimes it is ! and then sometimes it is ==NULL. Almost there

How do I turn in my code

Whoops, sorry about that. Email it to [email protected]

When is the closing date for the competition?

Quote:

---

When is the closing date for the competition?

---

Quote:

---

Sign-up for the contest will last one month.

---

Looks like today would be the last day for sign-ups.

>When is the closing date for the competition?
When everyone that signed up before today submits. But you don't have to be signed up to submit an entry.

Quote:

---

Originally Posted by Prelude

>dont u think even the one u classified as a beginner category is way ahead for a beginner.
It's just simple textual substitution. I think anyone who can read from a file and parse a simple string should have little trouble with it.

---

You think so? Really?

Code:

---

int main( void )
{
    int x[BUFSIZ];

    printf("%[10]",(3+1-7*4/11+200)[x]));
    return 0;
}

---

You mean to say it's simple string parsing for it to work with the results of something on the IOCCC which conformed to the standard? The above was just a quick example, but I'm sure there's way worse that would break it.

Actually, I'd really like to see it done:

Code:

---

#include <stdio.h>
#define W 1000
int main( void )
{
    int x = 0, y = , z[W];
    for(x=0,y=W;x<W;x++,y--)
        printf("%d",(z[z[w-x+y]])[z]);

    return 0;
}

---

Now with pure text subsitution, show me that all of those are in bounds at all times.

Quzah.

>You mean to say it's simple string parsing for it to work with the results of something on the IOCCC which conformed to the standard?
You're referring to the intermediate task. But technically it's still textual substitution, even if the process of setting it up is more difficult. And for the sake of the contest I don't require the commutativity of subscripting to be taken into account, though if a solution does it then it would likely be good enough to win. ;)

>Now with pure text subsitution, show me that all of those are in bounds at all times.
Well, that example would fail to compile, but assuming it was written in haste, the principle is the same. It's far from simple, but still text replacement.

Quote:

---

Originally Posted by Prelude

>Now with pure text subsitution, show me that all of those are in bounds at all times.
Well, that example would fail to compile, but assuming it was written in haste, the principle is the same. It's far from simple, but still text replacement.

---

You can also have macro, string names, function calls in combination with [ ].

Quzah.

>You can also have macro, string names, function calls in combination with [ ].
So you're saying that it's impossible? ;)

Yeah. Pretty much. Unless you have a program with no user input at all, it'd be impossible to know for sure that functions which return a value which is used to index an array would return a value in range. You could always FORCE a fit, by wrapping a modulus of the array size around it some how, but I still think you couldn't do it due to the complexity of potential array name, brackets, value ordering.

[edit]You'd have to actually have a program with no input whatsoever, and it couldn't use any random values, and all poniters would of course have to be initialized, as well as variables before using them, otherwise you could pull that "unknown" value... ;)[/edit]

[edit2]
Actually...

You might be able to do it if you wrote a program that would replace all [ ] chunks of code which would replace all operatins on arrays with function calls, and it would instead of being an actual array, be a dynamicly allocated chunk of memory which the only operations done on it were by said "operator overloading" functions.

But if you're doing that, you may as well just write your own damn language...
[/edit2]

Quzah.

I disagree. This isn't a syntactic checker and it certainly isn't expected to do it at compile-time. The idea is to replace the subscript with a suitable assertion that will halt the program with a debug statement before there's any undefined behavior. Using a table of existing arrays and sizes, the parser should be able to tell what is an array subscript and what isn't (a simulated array with pointers isn't, btw). Because the index would be used more than once in the replacement text, a global variable for saving it can be defined by the preprocessor.

In light of all that, it's not as impossible as you suggest, especially since I don't require a one pass algorithm. ;) The hardest part is the parsing, and as I already said, I don't expect entries to be complete in that they handle every legal construct, just the most common/reasonable ones.

Quote:

---

I don't expect entries to be complete in that they handle every legal construct, just the most common/reasonable ones.

---

Then they'd be incomplete, and not a full working extension. If they can't parse the example I've given, and anything like it, then they can't be considered a full working extension.

Code:

---

char array[SIZE];
0[array+functioncall()]=functioncall()[array];

---

In otherwords, I bet I can break any "working" end result, using full ANSI compliant, no-warnings, no-errors, -pedantic code to do so.

Quzah.

>Then they'd be incomplete, and not a full working extension.
If you disapprove so much, you aren't forced to participlate. ;)

It's your contest. I'm just surprised you'd consider a non-working example working. Have fun.

Quzah.

>I'm just surprised you'd consider a non-working example working.
Well, you have to consider that I spent all of 15 seconds devising all three of them. The result is usually a task that's much harder than it seems originally because I think of things that I would find interesting and I have to scale it back to make it doable for the level of people who enter. You'll notice that the talented programmers like you and Salem don't regularly participate. ;)

How about an implementation that simply inserts:

Code:

---

(()%size)

---

around all array indexing. ;)

Quzah.