GitLab

Closes #10

close: #9

It will make parser output a lot more human readable.

It includes a simple implementation of a no-undeclared-var rule.

We add a tsconfig.json with sane defaults at the repo root and extend it
in each packages through a tsconfig.build.json.
The tsconfig at the root will be used for lintin

The `parseCommaSeparatedList` allows to parse lists of variables in
selection and list of triplets in restrictions

It will allow to handle comma delimited lists in parser.

- configure compiler to put built files in `build` directory (and
  make sure these files are not linted or compiled)
- ensure the test files are not compiled on build
- add `build`, `watch` and `lint` scripts to parser's package.json
- add a README file
- add a .npmignore files to make sure only relevant files are included
  in the distributed package
- add .tgz files to .hgignore

This will be exposed in the public API of the library.

The lexer was emitting an Unknown token when encountering a space
character. Now it just ignroe it, allowing to read successive tokens
separated by spaces.

This is common in parsers to use infinite loops (with `true` as
condition) and use a break to leave the loop. Make the linter accept
that.

Let's start by making it able to read a single identifier or keyword.

The `startLine` and `startColumn` can be used when the document is
embedded in a broader document. This is the case for example if the rql
document is embedded in a typescript file.