Re: Bug #47643 - array_diff slowdown

Can we get some opinions on this. I'm willing to make the changes, but I want to get some sort of consensus on this. Basically I believe the problem is that patch 42838 should be reverted and the documentation should be updated. Someone on IRC had disagreed with me so I want more opinions before I do anything. On Fri, Aug 27, 2010 at 4:01 AM, Lonny K <lonnyk@gmail.com> wrote: > I wanted to discuss this bug to try to get it resolved. I briefly > discussed this over ICQ and had differing opinions then the person I was > talking to, so naturally I wanted a bigger audience. > > Quick summary of what's going on: > * 47643 is about array_diff being slow. > * This was caused by the patch for 42838 > * The problem with 42838 is basically that 0 != '0' > * Here is the diff: http://www.lonnylot.com/42838.diff > The lines that causes the slowdown are: > > - while (*ptrs[i] && (0 < (c = diff_data_compare_func(ptrs[0], ptrs[i] > TSRMLS_CC)))) { > - ptrs[i]++; > + while (*ptr && (0 < (c = diff_data_compare_func(ptrs[0], ptr TSRMLS_CC)))) { > + ptr++; > } > > > I feel the fix should be reverting 42838 because I feel 42838 wasn't a code > issue, but a documentation issue. It is noted in some places that comparing > a string to an integer results in the string being changed to 0. That is the > issue in 42838 and b/c this is excepted in other places it should be > excepted here. The documentation should be updated to say to typecast your > comparison to (string) if it is going to be important to your code. >