improve parsing speed

[ichiriac]

The lexer speed (reading a string from memory) is about 2,5 millions tokens / sec (depends on CPU).

It's about 3 times more slower than original PHP algorithm, but about 16 times more rapid than a jison based version (which is also more rapid than a regex version ...)

The lexer algorithm could be improved a bit, but there will not be much gain here.

The parser sometimes is 10x slower than the lexer, there is a lot of room to improve the speed, the most cost effective part is recursive parsing on structures like arrays, property chains or string concatenations.

[ichiriac]

Disk read speed (only I/O) : 12 Mb / sec

Lexer :

• Avoid costs with unput calls, instead use of appendToken in order to improve speeds
• Check if need recusivity/stacks on lexing states and avoid dynamic calls if possible (boolean states)
• Try to improve the unput function (columns problem)

Actual speed 5 Mb / sec
Target - improve 20% - 6 Mb / sec

Parser :

• Convert expression recursivity into a loop (avoid huge stacks)
• Add benchmarks & profiling

Actual speed : 2.5 Mb / sec
Target - improve by 50% - 3.7 Mb / sec

---

php-reflection is actually at 1.75 Mb / sec, the target is around 2.5 Mb / sec

[ichiriac]

• Replace array lookups with maps : https://jsperf.com/object-hasownproperty-vs-array-indexof/39
• Add output on tests suite in order to keep track of lexing and parsing speed