Handwritten parser

[motet-a]

This subjective and opinionated patch fixes many issues but has
several drawbacks.

The parser should be usable in strict mode now (see #9). It is not
longer generated by jison. parse.js contains a simple handwritten
recursive decent parser. It should be easier to debug and to hack, but
there is definitely more code and one can prefer parser generators.
Note that jison seems to have big maintainance issues.

Note that Yargs is not usable in strict mode because it depends on
this package.

I also simplified the lexer (IMHO).

I wrote more Mocha-based tests in test/index.js. I really like the
embedded tests in README.md, but I was afraid to bloat it with
boring spec-compliance tests.

This patch does not introduce a breaking change and does not change
dependencies (however, it changes devDependencies).

[kemitchell]

@motet-a: Thanks very much for this!

Two quick points:

1. Do you license this work under The MIT License, as well?

2. Would you object to squashing the last fix commit into the original rewrite, and opening a separate pull request for the relaxed-mode implementation? I can also do this for you, if you prefer.

[motet-a]

1. Of course! Should I add any other licensing information?
2. I’ll do it right now.

[kemitchell]

It's late in Lyon! ;-P

[kemitchell]

No, no more license hassles. We should find an appropriate way to credit you, however.

Thank you for splitting the PR!

[motet-a]

I have updated the commit message. The tests should pass on older Node.js versions now.

[kemitchell]

@motet-a, I've invited you to the jslicense GitHub org, which in turn gives you write access to this repo. I will make PRs for all non-trivial changes going forward. Please do the same. We can look after each other ... not to mention the code! 😄

As a practical matter, this is now your package as much as it belongs to anyone. Congratulations! It's an important piece of community infrastructure.

That is not my way of saying that I am about to disappear on you. Whether it's dealing with SPDX or npm or other projects, I think there will be plenty more work to do. 😛

I wish I knew how to write this part in French:

This is one of the best rewrites I've seen in open source. It's obvious you took the time to read and understand the spec, and also the existing code. The code is good, and you submitted it in the most respectful and positive way possible. You don't need to hear it from me, but you're good at this.

Technically, I share the opinion that a handwritten parser is best for the task. It was on my list to rewrite in this way for months, but I never found the time. I think you've done it cleaner than I would have.

There are a couple small things to mention:

1. I've already gone through on my own branch and reverted to ES5 syntax in a few places. On another project, that would be a matter of taste, and I would defer to you. In this package, which goes out with npm and runs on quite a few legacy Node systems, we need to stick to ES5, for maximum compatibility. I think it was just {value} syntax, anyway.

2. There was some more work to do in package.json to make it ready, which again, I've already done. I think we can do without a separate util module, too, to keep things as small as possible.

If there is anything else, I will bring it up via PRs.

[motet-a]

Wow, what a long comment! I am really happy to join jslicense. I will try to contribute more to open source projects.

I will make PRs, I’m used to it.

Regarding your questions:

1. Of course, transpiling with Babel has drawbacks and it’s easier to stick with ES5 in this little project. Since it is used by npm and Yargs, it's important.

2. Of course, it is nicer without util.js. I wrote it because I thought Array.prototype.map was unavailable on old Node.js versions, and oneOf() was just ridiculous compared to your ||-based replacement 😅

[kemitchell]

My comment was too long. In short: Welcome!