sorting 2d arrays using qsort

[yatindran]

hai
this is my 2d array.
int a[][6] = {
{5,2,20,1,30,10},
{23,15,7,9,11,3},
{40,50,34,24,14,4},
{9,10,11,12,13,14},
{31,4,18,8,27,17},
{44,32,13,19,41,19},
{1,2,3,4,5,6},
{80,37,47,18,21,9}
};
after
for(i=0;i<box;i++)
qsort(a,dim,sizeof(int),dim­_sort);
int dim_sort(const void *a,const void *b)
{
return ( *(int*)a - *(int*)b);


}


where dim=6, i get the foll o/p:
1 2 5 10 20 30
3 7 9 11 15 23
4 14 24 34 40 50
9 10 11 12 13 14
4 8 17 18 27 31
13 19 19 32 41 44
1 2 3 4 5 6
9 18 21 37 47 80
(works fine!)
now i want to sort each row of elements for above array,
like
1 2 3 4 5 6
1 2 5 10 20 30
....
etc.

what is the qsort routine and function, can anybody help me ?


cheers,
badri

 

[James Daughtry]

return ( *(int*)a - *(int*)b);
That's a dangerous way to compare two signed integers. You're risking
underflow, which is undefined.

now i want to sort each row of elements for above array

You can't do it with qsort because qsort tries to directly exchange two
items; an operation that arrays don't support. You can, however, create
an array of pointers to int, point them to each array in a, and then
sort that array by writing another comparison function that compares
the first mismatch in the two arrays being compared:

#include <stdio.h>
#include <stdlib.h>

static int arraya[][6] = {
{5,2,20,1,30,10},
{23,15,7,9,11,3},
{40,50,34,24,14,4},
{9,10,11,12,13,14},
{31,4,18,8,27,17},
{44,32,13,19,41,19},
{1,2,3,4,5,6},
{80,37,47,18,21,9}
};

static int *arrayb[8];

int cmp_int(const void *a,const void *b)
{
const int *ia = a;
const int *ib = b;

if (*ia < *ib) return -1;
else if (*ia > *ib) return +1;
else return 0;
}

int cmp_array(const void *a, const void *b)
{
const int **ia = a;
const int **ib = b;
int i, cmp;

for (i = 0; i < 6; i++) {
if ((cmp = cmp_int(&(*ia), &(*ib))) != 0)
return cmp;
}

return 0;
}

int main(void)
{
int i, j;

for (i = 0; i < 8; i++)
qsort(arraya, 6, sizeof(int), cmp_int);

for (i = 0; i < 8; i++) {
for (j = 0; j < 6; j++)
printf("%4d", arraya[j]);
printf("\n");
}

printf("\n");

for (i = 0; i < 8; i++)
arrayb = arraya;

qsort(arrayb, 8, sizeof(int*), cmp_array);

for (i = 0; i < 8; i++) {
for (j = 0; j < 6; j++)
printf("%4d", arrayb[j]);
printf("\n");
}

printf("\n");
}

 

[Lawrence Kirby]

hai
this is my 2d array.
int a[][6] = {
{5,2,20,1,30,10},
{23,15,7,9,11,3},
{40,50,34,24,14,4},
{9,10,11,12,13,14},
{31,4,18,8,27,17},
{44,32,13,19,41,19},
{1,2,3,4,5,6},
{80,37,47,18,21,9}
};
after
for(i=0;i<box;i++)
qsort(a,dim,sizeof(int),dim­_sort);

You could write this as

int dim_sort(const void *a,const void *b)
{
return ( *(int*)a - *(int*)b);


}

Be caseful, subtracting 2 ints can overflow so this isn't well defined
unless you can be sure that the difference between any 2 values in the
array is representable as an int. It would be safer not to make
assumptions like this:

int dim_sort(const void *va, const void *vb)
{
int a = *(const int *)va;
int b = *(const int *)vb;

where dim=6, i get the foll o/p:
1 2 5 10 20 30
3 7 9 11 15 23
4 14 24 34 40 50
9 10 11 12 13 14
4 8 17 18 27 31
13 19 19 32 41 44
1 2 3 4 5 6
9 18 21 37 47 80
(works fine!)

OK, you're sorting the contents each row individually which is fine.

now i want to sort each row of elements for above array,
like
1 2 3 4 5 6
1 2 5 10 20 30
...
etc.

what is the qsort routine and function, can anybody help me ?

You just have to remember that the array a is simply an array of elements
where each of those elements is itself an array, so you can write:

qsort(a, sizeof a/sizeof *a, sizeof *a, compare_rows);

Where sizeof a/sizeof *a is the number of rows in a, sizeof *a is the
size in bytes of a row. The comparison function would be something like

static int compare_rows(const void *va, const void *vb)
{
const int *pa = *(const int (*)[6])va;
const int *pb = *(const int (*)[6])vb;
int i;

for (i = 0; i < 6; i++) {
if (pa != pb)
return (pa > pb) ? 1 : -1;
}

return 0;
}

Lawrence

 

[Lawrence Kirby]

That's a dangerous way to compare two signed integers. You're risking
underflow, which is undefined.

You can't do it with qsort because qsort tries to directly exchange two
items; an operation that arrays don't support.

C doesn't directly supply an exchange operation for any type, arrays are
no different in this respect. In fact there is nothing in the definition
of qsort() that requires it to use exchanges as such, although many common
sorting algorithms do use them. The more fundamental operation is a copy.
qsort() doesn't know or care about the type of the elements of the array
it is sorting, in particular it doesn't care whether they are are
themselves arrays or not. In C you can copy (and therefore exchange a
pair of) ANY object by treating it as an array of sizeof(object) bytes
(specifically an array of unsigned char). An implementation of qsort() is
likely to use memcpy() or something equivalent for copying which works
fine on arrays. All it needs to do this is a pointer to the object and the
size of the object in bytes. All the necessary information is passed in
qsort()'s arguments.

You can, however, create
an array of pointers to int, point them to each array in a, and then
sort that array by writing another comparison function that compares
the first mismatch in the two arrays being compared:

You can do that, and there may be efficiency advantages to doing so (i.e.
copying a pointer is likely to be quicker than copying a row), but you
don't have to.

Lawrence

 

[James Daughtry]

An implementation of qsort() is likely to use memcpy() or something equivalent for

copying which works fine on arrays.

True, but can that be relied on to portably sort a two dimensional
array as you suggested?

 

[Lawrence Kirby]

True, but can that be relied on to portably sort a two dimensional
array as you suggested?

Yes. There's nothing that makes arrays different to any other type of
object for the purposes of qsort(). Remember again that the implementation
of qsort() knows nothing about the type of the data being sorted - C
doesn't have any form of runtime type information.

More fundamentally there is noting in the standard's specification of
qsort() that precludes it from sorting arrays whose elements are
themselves arrays.

Lawrence

 

[CBFalconer]

Yes. There's nothing that makes arrays different to any other type
of object for the purposes of qsort(). Remember again that the
implementation of qsort() knows nothing about the type of the data
being sorted - C doesn't have any form of runtime type information.

More fundamentally there is noting in the standard's specification
of qsort() that precludes it from sorting arrays whose elements
are themselves arrays.

Except you have to decide and codify what makes A1 > A2, or A1 <
A2, or A1 == A2.

 

[Lawrence Kirby]

Lawrence Kirby wrote:
....


Except you have to decide and codify what makes A1 > A2, or A1 <
A2, or A1 == A2.

As you do with any type of array element.

Lawrence

 

[pete]

hai
this is my 2d array.
int a[][6] = {
{5,2,20,1,30,10},
{23,15,7,9,11,3},
{40,50,34,24,14,4},
{9,10,11,12,13,14},
{31,4,18,8,27,17},
{44,32,13,19,41,19},
{1,2,3,4,5,6},
{80,37,47,18,21,9}
};
after
for(i=0;i<box;i++)
qsort(a,dim,sizeof(int),dim­_sort);
int dim_sort(const void *a,const void *b)
{
return ( *(int*)a - *(int*)b);

}

where dim=6, i get the foll o/p:
1 2 5 10 20 30
3 7 9 11 15 23
4 14 24 34 40 50
9 10 11 12 13 14
4 8 17 18 27 31
13 19 19 32 41 44
1 2 3 4 5 6
9 18 21 37 47 80
(works fine!)
now i want to sort each row of elements for above array,
like
1 2 3 4 5 6
1 2 5 10 20 30
...
etc.

what is the qsort routine and function, can anybody help me ?

/* BEGIN new.c */

#include <stdio.h>
#include <stdlib.h>

int int_comp(const void *a, const void *b);
int a6int_comp(const void *a, const void *b);

int main(void)
{
int a[][6] = {
{5,2,20,1,30,10},
{23,15,7,9,11,3},
{40,50,34,24,14,4},
{9,10,11,12,13,14},
{31,4,18,8,27,17},
{44,32,13,19,41,19},
{1,2,3,4,5,6},
{80,37,47,18,21,9}
};
const size_t dim = sizeof *a / sizeof **a;
const size_t box = sizeof a / sizeof *a;
size_t i, j;

puts("\n/* BEGIN new.c output */\n");
for (i = 0; i != box; ++i) {
for (j = 0; j != dim; ++j) {
printf("%d ", a[j]);
}
putchar('\n');
}
putchar('\n');
for (i = 0; i != box; ++i) {
qsort(a, dim, sizeof *a, int_comp);
}
for (i = 0; i != box; ++i) {
for (j = 0; j != dim; ++j) {
printf("%d ", a[j]);
}
putchar('\n');
}
putchar('\n');
qsort(a, box, sizeof *a, a6int_comp);
for (i = 0; i != box; ++i) {
for (j = 0; j != dim; ++j) {
printf("%d ", a[j]);
}
putchar('\n');
}
puts("\n/* END new.c output */");
return 0;
}

int int_comp(const void *a, const void *b)
{
return *(int *)b > *(int *)a ? -1 : *(int *)b != *(int *)a;
}

int a6int_comp(const void *a, const void *b)
{
size_t n;
const int *a_ptr = a;
const int *b_ptr = b;

for (n = 0; n != 6; ++n) {
if (b_ptr[n] > a_ptr[n]) {
return -1;
}
if (a_ptr[n] > b_ptr[n]) {
return 1;
}
}
return 0;
}

/* END new.c */

 

[pete]

In C you can copy (and therefore exchange a
pair of) ANY object by treating it as an array of sizeof(object) bytes
(specifically an array of unsigned char).

/* BEGIN new.c */

#include <stdio.h>

void swap(void *s1, void *s2, size_t n);

int main(void)
{
char odd[] = "1 3 5 7 9";
char even[] = "0 2 4 6 8";
double three = 3.0;
double four = 4.0;

swap(&odd, &even, sizeof odd);
printf("odd %s\n", odd);
printf("even %s\n", even);
swap(&three, &four, sizeof three);
printf("three %f\n", three);
printf("four %f\n", four);
return 0;
}

void swap(void *s1, void *s2, size_t n)
{
unsigned char *p1, *p2, *end, temp;

p1 = s1;
p2 = s2;
end = p2 + n;
do {
temp = *p1;
*p1++ = *p2;
*p2++ = temp;
} while (p2 != end);
}

/* END new.c */

 

[CBFalconer]

As you do with any type of array element.

Of course. But most users are used to simply comparing basic
types, or at worst passing things to strcmp. Here they might have
to think about the meaning of things.

 

[CBFalconer]

.... snip ...

int a6int_comp(const void *a, const void *b)
{
size_t n;
const int *a_ptr = a;
const int *b_ptr = b;

for (n = 0; n != 6; ++n) {
if (b_ptr[n] > a_ptr[n]) {
return -1;
}
if (a_ptr[n] > b_ptr[n]) {
return 1;
}
}
return 0;
}

In other words you have come to the arbitrary decision that the
highest indexed component of the array is the most significant.
Fine, but you should document it in the comparison routine.
Whether or not this is what the OP wants is unknown.

 

[pete]

... snip ...

int a6int_comp(const void *a, const void *b)
{
size_t n;
const int *a_ptr = a;
const int *b_ptr = b;

for (n = 0; n != 6; ++n) {
if (b_ptr[n] > a_ptr[n]) {
return -1;
}
if (a_ptr[n] > b_ptr[n]) {
return 1;
}
}
return 0;
}

In other words you have come to the arbitrary decision that the
highest indexed component of the array is the most significant.

I don't understand what you mean.

Fine, but you should document it in the comparison routine.
Whether or not this is what the OP wants is unknown.

OP gave an example of what he wanted.
You snipped it:

/* BEGIN new.c output */

5 2 20 1 30 10
23 15 7 9 11 3
40 50 34 24 14 4
9 10 11 12 13 14
31 4 18 8 27 17
44 32 13 19 41 19
1 2 3 4 5 6
80 37 47 18 21 9

1 2 5 10 20 30
3 7 9 11 15 23
4 14 24 34 40 50
9 10 11 12 13 14
4 8 17 18 27 31
13 19 19 32 41 44
1 2 3 4 5 6
9 18 21 37 47 80

1 2 3 4 5 6
1 2 5 10 20 30
3 7 9 11 15 23
4 8 17 18 27 31
4 14 24 34 40 50
9 10 11 12 13 14
9 18 21 37 47 80
13 19 19 32 41 44

/* END new.c output */

 

[CBFalconer]

... snip ...

int a6int_comp(const void *a, const void *b)
{
size_t n;
const int *a_ptr = a;
const int *b_ptr = b;

for (n = 0; n != 6; ++n) {
if (b_ptr[n] > a_ptr[n]) {
return -1;
}
if (a_ptr[n] > b_ptr[n]) {
return 1;
}
}
return 0;
}

In other words you have come to the arbitrary decision that the
highest indexed component of the array is the most significant.

I don't understand what you mean.

Fine, but you should document it in the comparison routine.
Whether or not this is what the OP wants is unknown.

OP gave an example of what he wanted.
You snipped it:

So he did. However, a decision has been made as to the relative
significance of the components. Notice that the opposite decision,
based on the highest order component of the array, would have given
the same result as far as his example actually went. So would an
array sort based on the 3rd (index 2) item of each array.

Which is not a criticism of your comparison function, but simply
pointing out the arbitrariness of the decisions taken. This sort
of thing will always come up when comparing arrays or structs,
while there is instinctive agreement on the meaning of comparison
between numbers or strings. For strings, the standard spells it
out anyway.

 

[pete]

pete wrote:

... snip ...

int a6int_comp(const void *a, const void *b)
{
size_t n;
const int *a_ptr = a;
const int *b_ptr = b;

for (n = 0; n != 6; ++n) {
if (b_ptr[n] > a_ptr[n]) {
return -1;
}
if (a_ptr[n] > b_ptr[n]) {
return 1;
}
}
return 0;
}

In other words you have come to the arbitrary decision that the
highest indexed component of the array is the most significant.

I don't understand what you mean.

Fine, but you should document it in the comparison routine.
Whether or not this is what the OP wants is unknown.

OP gave an example of what he wanted.
You snipped it:

now i want to sort each row of elements for above array,
like
1 2 3 4 5 6
1 2 5 10 20 30
...
etc.

So he did. However, a decision has been made as to the relative
significance of the components. Notice that the opposite decision,
based on the highest order component of the array, would have given
the same result as far as his example actually went.

I still don't understand.

int a6int_comp(const void *a, const void *b)
{
size_t n;
const int *a_ptr = a;
const int *b_ptr = b;

n = 6;
while (n-- != 0) {
if (b_ptr[n] > a_ptr[n]) {
return 1;
}
if (a_ptr[n] > b_ptr[n]) {
return -1;
}
}
return 0;
}

gives:
9 18 21 37 47 80
4 14 24 34 40 50
13 19 19 32 41 44
4 8 17 18 27 31
1 2 5 10 20 30
3 7 9 11 15 23
9 10 11 12 13 14
1 2 3 4 5 6

int a6int_comp(const void *a, const void *b)
{
size_t n;
const int *a_ptr = a;
const int *b_ptr = b;

n = 6;
while (n-- != 0) {
if (b_ptr[n] > a_ptr[n]) {
return -1;
}
if (a_ptr[n] > b_ptr[n]) {
return 1;
}
}
return 0;
}

gives :
1 2 3 4 5 6
9 10 11 12 13 14
3 7 9 11 15 23
1 2 5 10 20 30
4 8 17 18 27 31
13 19 19 32 41 44
4 14 24 34 40 50
9 18 21 37 47 80

So would an
array sort based on the 3rd (index 2) item of each array.

I think I understand that, but that would be silly.

Which is not a criticism of your comparison function, but simply
pointing out the arbitrariness of the decisions taken. This sort
of thing will always come up when comparing arrays or structs,
while there is instinctive agreement on the meaning of comparison
between numbers or strings. For strings, the standard spells it
out anyway.

What do you mean about strings?

 

[CBFalconer]

CBFalconer wrote:
.... snip ...


I think I understand that, but that would be silly.

Why? It depends on the needs. For an example we have some device
out there reporting its position through a gray code encoder. For
each position it reports a magnitude. We collect a herd of these,
and the cycle of positions is one of the arrays in your 2d array.
Its sequence goes 000, 001, 011, 111, 110, 100 (ok, so its not a
proper gray, but only one bit changes at a time avoiding gross
errors). We use these as index values into the array.

Now we want so treat those arrays as attaching greater importance
to the earlier entry. Write the comparison function.

We can complicate it further by saying the magnitudes stored in the
arrays are also gray encoded. This is a quite reasonable thing to
do, in order to avoid major errors due to race conditions between
bits.

 

[pete]

Why?

I don't that think OP's specifications are as cryptic
as you seem to think that they are.

 

[pete]

I don't
think

that think OP's specifications are as cryptic
as you seem to think that they are.

 

[pete]

There were about one or two typos.

 

[CBFalconer]

I don't that think OP's specifications are as cryptic
as you seem to think that they are.

Probably not, but I could write a lot of software that would meed
the specs and be totally useless. This sort of thing leads to
lawsuits and other things that enhance the lawyers standard of
living. And all I really said in the first place was "document
what you wrote".