This code can definitely improve. It can invoke undefined behaviour on systems with a signed char (on x86_64 for example) if string contains "negative" characters (say the input contains some UTF-8 codepoints). This is because tolower is actually supposed take an unsigned char cast to int, or alternatively EOF, with all other values invoking UB. Plus, you probably don't want to use tolower anyways as it is locale-dependant. Edit: And also, inconsistent formatting, please move all your pointer asterisks * to one direction! Remember that the optimal type for sizes, counting, etc in C is size_t from <stddef.h>.

tolower() is defined in ctype.h header file, shouldn't you include it?

I agree with this.
But wouldn't it be nicer to get rid of tolower? One can use a switch and "hardcode" both ascii values for x and o respectively. Also we could omit the char *p since we can iterate on str, but that's very likely optimized away by the compiler anyway.

use a single var to keep the count and then its perfect

Please include more information if this is still an issue after 1 year.

You have typo in your code and it did not work, It should have been '\0' instead of '\n'.

So the following fix works.

string to_postfix(const string& infix) {
  int pos = find(infix.begin(), infix.end(), '\0') - infix.begin();
  return to_postfix_NEW(infix.substr(0, pos)) + infix.substr(pos, infix.size() - pos);
}

simpleAndNestedLoops(SolutionTests)
 Your interpreter should jump to the command straight *after* the matching "[" on the iteration where it hits the "]" and *not* the matching "[" itself expected:<10000
01000
00[1]00
00000
00000
00...> but was:<10000
01000
00[0]00
00000
00000
00...>

I think the spec for ']' differs from the one of this test case. My solution is in Java.

Line if (token.length()>1) mPlus.remove(mPlus.size()-1);
is redundant and can be removed.

Got the same problem when submitting python code.

How do I handle "Unknown Error"? Help please!

Add Rust too, please!