embedAst option to embed assertion's AST and tokens to make runtime side parser unnecessary

[twada]

This pull-request aims to introduce opt-in embedAstoption to embed assertion's AST and tokens to make runtime side parser unnecessary.

Context

The runtime side of power-assert (not a transpilation side) depends on ECMAScript parser (currently acorn) to parse canonical assertion expression at runtime. Browserify embeds the parser into built bundle for web browsers.

Problems

ECMAScript parser is huge. Therefore,

• it slows require time down
• power-assert built bundle for browsers is huge too

Bundled parser is an ECMA standard compliant one. Therefore,

• Cannot handle dialects such as JSX, Flow, etc...
• When JSX expression exists inside of assert(), parser causes parse error, power-assert runtime explodes immediately.

Solution

• Embed assertion's AST, tokens and its visitor keys to output code, means that they are passed to runtime side.
• Now runtime side doesn't have to parse canonical assertion expression since they already have AST, tokens and its visitor keys.

Advantages

• Reduce runtime side dependencies and its bundle size. Make runtime loading much faster.
• Can handle any dialects such as JSX, Flow, etc... anything we can parse could be visualized on the runtime side

When this pull-req lands, you can write JSX tags inside assertions.

assert(shallow(<ToDoItem item={item} />).find('.item-mark').text() === '✓');

Downside

• Slows down transpilation time
• Transpiled code gets bigger

TODO

• rough initial design
• JSX testcase
• spike implementation
• reorganize runtime side
• maint other tests
• benchmarking
• brutal refactoring
• a little optimization

This feature might be released with power-assert 2.0.0 (if there are some breaking changes) <- No. There is no breaking changes.

[twada]

Reorganization of runtime side has done. Time to move forward.