Talk:Convex hull

I found some errors in the C code when I compare with c++ code : code C++

   // lower hull
   for (const auto& pt : p) {
       while (h.size() >= 2 && !ccw(h.at(h.size() - 2), h.at(h.size() - 1), pt)) {
           h.pop_back();
       }
       h.push_back(pt);
   }

code C :

  /* lower hull */
   for (i = 0; i < len; ++i) {
       while (hLen >= 2) {
           hptr = h;
           while (hptr->next->next != NULL) {
               hptr = hptr->next;
           }
           if (ccw(&hptr->data, &hptr->next->data, &p[i])) { <== mistake
               break;
           }

It should be

  /* lower hull */
   for (i = 0; i < len; ++i) {
       while (hLen >= 2) {
           hptr = h;
           while (hptr->next->next != NULL) {
               hptr = hptr->next;
           }
           if (ccw(&hptr->next->data, &hptr->data, &p[i])) {   <==== good code
               break;
           }

The same error is done in upper hull

Another error

C++ code

   // upper hull
   auto t = h.size() + 1;
   for (auto it = p.crbegin(); it != p.crend(); it = std::next(it)) {
       auto pt = *it;
       while (h.size() >= t && !ccw(h.at(h.size() - 2), h.at(h.size() - 1), pt)) {
           h.pop_back();
       }
       h.push_back(pt);
   }

C code

   /* upper hull */
                            <== t is fogotten
   for (i = len - 1; i >= 0; i--) {
       while (hLen >= 2) {
           hptr = h;
           while (hptr->next->next != NULL) {
               hptr = hptr->next;
           }
           if (ccw(&hptr->data, &hptr->next->data, &p[i])) {
               break;
           }

It should be

int t = hLen + 1; for (i = len - 1; i >= 0; i--) { while (hLen >= t) hptr = h; while (hptr->next->next != NULL){ hptr = hptr->next; } if (ccw(&hptr->next->data, &hptr->data, &p[i])) { break; }

You should test the code, and think about the results here. For example, "ccw" is a routine which determines whether the winding is clockwise or counterclockwise. But from which side? Anyways... since that's not a documented issue... ccw should work regardless of the winding direction -- as long as the winding direction is consistent. So... you have encountered a real issue here. But it's probably not a code correctness issue -- it's probably a lack of adequate documentation issue (which is a frequent problem for coders).

Similarly, on the second issue you brought up, just throwing code out -- without any documentation and without any test results -- does not adequately illustrate the issue.

That said, taking a close look at the implementation, like you have done here, is great. We all-too-often have had errors in code here on this site -- often for very understandable reasons. So double checking results is frequently a good thing. Thanks!

P.S. please sign your comments on the talk pages, using --~~~~. This helps readers comprehend talk pages. --Rdm (talk) 21:07, 22 December 2020 (UTC)

I'm sure of these errors because I tried this code for an exercice on Codingame (Encounter Surface) and I didn't get the good results. I compare with C++ and Java codes, found the mistakes, correct then as I explain and now it works ! May be the code is correct for the data given but it's wrong ! Trap D