Performance benchmark about mnemonist

[maxiruani]

You should start the timer after creating the Trie

[Yomguithereal]

Hello @tomas-campodonico. Thanks for noticing this performance issue. I spotted an error in the Trie.get method and was able to get better performance (beating the ES5 array). Some things:

1. @maxiruani is right, you should put your timer after creating the Trie.
2. Try pulling the code & use the latest commit or wait for the npm release.
3. Try increasing your number to 100000 (one more zero) and you'll see a better benefit of using the Trie.

So, on the matter of benchmarking, you are right, I should start to make some in the repo to spot those kind of things etc. Usually I use matcha to do so. What do you think?

[Yomguithereal]

v0.10.2 has been released with the related commit.

[noopnik]

Same benchmark with 10000000 will failed on the Trie.from(words) line . v0.10.2, Node 7.4.0 .

[Yomguithereal]

Yes, but this is a memory issue. With 10M items node will run out of memory because the Trie needs to store pointers whereas a flat array is contiguous in memory.

[Yomguithereal]

This said, we can work together, if you want, to create a StaticTrie that could be stored in a flat array so it is more memory-efficient.