Program runs too slowly with large input - C - c

Question

The goal for this program is for it to count the number of instances that two consecutive letters are identical and print this number for every test case. The input can be up to 1,000,000 characters long (thus the size of the char array to hold the input). The website which has the coding challenge on it, however, states that the program times out at a 2s run-time. My question is, how can this program be optimized to process the data faster? Does the issue stem from the large char array?
Also: I get a compiler warning "assignment makes integer from pointer without a cast" for the line str[1000000] = "" What does this mean and how should it be handled instead?
Input:
number of test cases
strings of capital A's and B's
Output:
Number of duplicate letters next to each other for each test case, each on a new line.
Code:
#include <stdio.h>
#include <string.h>
#include <math.h>
#include <stdlib.h>
int main() {
int n, c, a, results[10] = {};
char str[1000000];
scanf("%d", &n);
for (c = 0; c < n; c++) {
str[1000000] = "";
scanf("%s", str);
for (a = 0; a < (strlen(str)-1); a++) {
if (str[a] == str[a+1]) { results[c] += 1; }
}
}
for (c = 0; c < n; c++) {
printf("%d\n", results[c]);
}
return 0;
}

Answer 1

You don't need the line
str[1000000] = "";
scanf() adds a null terminator when it parses the input and writes it to str. This line is also writing beyond the end of the array, since the last element of the array is str[999999].
The reason you're getting the warning is because the type of str[10000000] is char, but the type of a string literal is char*.
To speed up the program, take the call to strlen() out of the loop.
size_t len = strlen(str)-1;
for (a = 0; a < len; a++) {
...
}

Answer 2

str[1000000] = "";
This does not do what you think it does and you're overflowing the buffer which results in undefined behaviour. An indexer's range is from 0 - sizeof(str) EXCLUSIVE. So you either add one to the
1000000 when initializing or use 999999 to access it instead. To get rid of the compiler warning and produce cleaner code use:
str[1000000] = '\0';
Or
str[999999] = '\0';
Depending on what you did to fix it.
As to optimizing, you should look at the assembly and go from there.

Answer 3

count the number of instances that two consecutive letters are identical and print this number for every test case
For efficiency, code needs a new approach as suggeted by #john bollinger & #molbdnilo
void ReportPairs(const char *str, size_t n) {
int previous = EOF;
unsigned long repeat = 0;
for (size_t i=0; i<n; i++) {
int ch = (unsigned char) str[i];
if (isalpha(ch) && ch == previous) {
repeat++;
}
previous = ch;
}
printf("Pair count %lu\n", repeat);
}
char *testcase1 = "test1122a33";
ReportPairs(testcase1, strlen(testcase1));
or directly from input and "each test case, each on a new line."
int ReportPairs2(FILE *inf) {
int previous = EOF;
unsigned long repeat = 0;
int ch;
for ((ch = fgetc(inf)) != '\n') {
if (ch == EOF) return ch;
if (isalpha(ch) && ch == previous) {
repeat++;
}
previous = ch;
}
printf("Pair count %lu\n", repeat);
return ch;
}
while (ReportPairs2(stdin) != EOF);
Unclear how OP wants to count "AAAA" as 2 or 3. This code counts it as 3.

Answer 4

One way to dramatically improve the run-time for your code is to limit the number of times you read from stdin. (basically process input in bigger chunks). You can do this a number of way, but probably one of the most efficient would be with fread. Even reading in 8-byte chunks can provide a big improvement over reading a character at a time. One example of such an implementation considering capital letters [A-Z] only would be:
#include <stdio.h>
#define RSIZE 8
int main (void) {
char qword[RSIZE] = {0};
char last = 0;
size_t i = 0;
size_t nchr = 0;
size_t dcount = 0;
/* read up to 8-bytes at a time */
while ((nchr = fread (qword, sizeof *qword, RSIZE, stdin)))
{ /* compare each byte to byte before */
for (i = 1; i < nchr && qword[i] && qword[i] != '\n'; i++)
{ /* if not [A-Z] continue, else compare */
if (qword[i-1] < 'A' || qword[i-1] > 'Z') continue;
if (i == 1 && last == qword[i-1]) dcount++;
if (qword[i-1] == qword[i]) dcount++;
}
last = qword[i-1]; /* save last for comparison w/next */
}
printf ("\n sequential duplicated characters [A-Z] : %zu\n\n",
dcount);
return 0;
}
Output/Time with 868789 chars
$ time ./bin/find_dup_digits <dat/d434839c-d-input-d4340a6.txt
sequential duplicated characters [A-Z] : 434893
real 0m0.024s
user 0m0.017s
sys 0m0.005s
Note: the string was actually a string of '0's and '1's run with a modified test of if (qword[i-1] < '0' || qword[i-1] > '9') continue; rather than the test for [A-Z]...continue, but your results with 'A's and 'B's should be virtually identical. 1000000 would still be significantly under .1 seconds. You can play with the RSIZE value to see if there is any benefit to reading a larger (suggested 'power of 2') size of characters. (note: this counts AAAA as 3) Hope this helps.