[WIP] Implemented split #66

[SamMousa]

The implementation takes care to not use string comparison functions to prevent needless copying of data.. This does make it a little bit harder to read, so might not be worth it.

Todo:

• [ ] splitStream
• [ ] splitCallable

Please check the implementation in #66 as well, it reads a lot easier but comes with the cost of having to do a substr call for every character in the data instead of only when emitting a value.

[nikic]

I'm not sure I understand the motivation for the manually implemented comparison, which is going to be very slow. Can't we use a strpos loop instead? Something along these lines...

$offset = 0;
while (false !== $offset = strpos($str, $deliminter, $offset)) {
    // ...
    $offset += strlen($delimiter);
}

[SamMousa]

I wanted to prevent substring copying; didn't think of using strpos. It is definitely a better solution that will also create more readable code.

Side note: why would the manual comparison be slow? Because it's implemented in PHP?

[nikic]

Side note: why would the manual comparison be slow? Because it's implemented in PHP?

For two reasons:

• Implementing it in PHP carries a lot of overhead here beyond the obvious, because the C implementation can use SIMD instructions. If the delimiter is a single character, then it is possible to check many (say 16) characters of the string for it at the same time.
• For long strings and long delimiters, a naive search has quadratic time complexity. The internal implementation will use a linear time matching algorithm if it is heuristically profitable.

[nikic]

This one looks good, so I went ahead and merged it. We can handle other variants in a followup...