Json Lexer Benchmark upgrade.

[bd82]

• Extracted to a separate file.
• Read input relative to module, not to working directory.
• Test versus expected number of tokens including whitespace when possible.
• Added Chevrotain 0.25.0 to JSON benchmark (derived from Moo's definition).

[tjvr]

I'm not keen on moving the JSON benchmark to a separate file.

[bd82]

I will try keeping it in the same file. It is just a bit of a mess right now...

[bd82]

Is Chevrotain tracking line numbers here?

Chevrotain always tracks line numbers, but in different ways. There are Three token types, some of which are class based and some are basic Object structures. The full position information (Line + column + offset -> start and end for all variants). Is either saved on the Token object, or computed on demand depending on the Token Type.

The computation is very efficient and actually saves previous computation results to avoid unneeded work on future position request (even on different tokens of the same input text!).