Serialize or Implode

Question

I need to store lots of two-dimensional arrays inside database and was not sure what to use: serialize or implode. So I did a few tests, to find out which one is working faster and came to the conclusion it was serialize:

Execution times: 1'000'000
Serialize: 1.4974119663239 seconds
Implode: 2.5333571434021 seconds
Explode: 4.0185871124268 seconds
Unserialize: 1.6835169792175 seconds 

So the question: Why is implode+explode so much slower then serialize+unserialize?

PS: I found this question already, but it is not exactly what I am asking.

Constantine

Answer 1

My idea is that explode/implode operate on strings that's why, while serialize/unserialize output/input a string at the very ending/beginning. Probably the more strings you implode and the longer string you explode, the slower it is, have you tried?

Which to use I don't know, it depends if you later wish to handle output string, I think serialized string is more difficult to parse. But to store it in database or file I'd use serialize.

Ellenaellender

That sounds somehow reasonable. And yes, the larger the array, the longer implode/explode works. I already tested that too. – Constantine

@DainisAbols I meant that it is not linear dependency. Is implodeing four strings exactly 2 times slower than imploding two strings? I guess it's more; and imploding eight strings would take more time than imploding two strings four times (all the strings should be the same length of course). – Ellenaellender

Implode 2 elements: 1.8937180042267 seconds, Implode 4 elements: 2.4380650520325 seconds, Implode 8 elements: 3.6058769226074 seconds Not exactly two times, but close to it, while serialize changes with the same progression type, but slower rate. – Constantine

So I was wrong, imploding more elements is somehow faster. Then concatenating strings must be the weakest link. – Ellenaellender

@DainisAbols You could compare the implementation of the implode function with the serialize function – Considerable