This project has not been well-tested and should be considered experimental.
This project aims to allow full @babel/eslint-parser
configuration, but with
comment-parser
-based AST for JSDoc blocks and jsdoctypeparser
AST for
JSDoc types within those blocks.
There is also a @es-joy/jsdoc-eslint-parser/typescript.js
parser which
builds on @typescript-eslint/parser
.
It is hoped that this may spur development of official ESTree acceptance of such AST.
You can run our lint-test
script for an example.
- Using
no-restricted-syntax
to prohibit certain comment or comment type AST - Supplying comment or comment-type aware query selectors to
contexts
ineslint-plugin-jsdoc
In using the Babel ESLint parser, AST is produced which has the tokens
on
Program
expected by ESLint as well as the range
array expected on each
Node.
We add comment AST--or, rather, JSDocBlock
AST--by adding a jsdoc
property
to each item of the @babel/eslint-parser
-parsed code which does not have a
suitable ancestor to hold the comment block. We correspondingly
update the visitor keys to ensure that for every key's array of visitable
properties, jsdoc
is added, so it can be separately traversable. We also
ensure the comment-parser
and jsdoctypeparser
visitor keys are added to
our parser's visitor keys.
And we add a parent
for each Node so that our getJSDocComment
method can
find its way around (see below).
The tricky part with the jsdoc
property comes in with determining
which JSDoc should be associated with a given Node. In
jsdoccomment
,
we have a getJSDocComment
method, originally begun by ESLint itself to
determine where that should be expected.
This detection of the comment for a given structure is not a trivial matter. For example, with:
/* A */
const /* B */ aFunc = /* C */ function () {};
... for the function expression, we might look for the JSDoc Block at point C
or B first, but then if not present, look for it at point A. The
getJSDocComment
method uses such an algorithm, and we call this in our
parser currently on each node. If there is no JSDoc block at point B or
C, the node for the aFunc
VariableDeclaration
node should get the JSDoc
Block at point A added instead of the FunctionExpression
getting it.
Although abstractly, it may appear a better fit with FunctionExpression
,
the practical purposes of finding JSDoc blocks relates more to position and
code generation, which should be easiest to work with if in a position closest
to its actual location.
Note that the AST produced is described in the
jsdoccomment project. Additionally,
this parser adds to Program
nodes, a jsdocBlocks
property (corresponding
to comments
though for unattached comment blocks only), along with a
commentsIndex
integer proprerty to reflect the index to which this JSDoc
block corresponds with the comments
array.
npm i -D @es-joy/jsdoc-eslint-parser
The jsdoctypeparser
mode, jsdoc
, typescript
, closure
, or permissive
.
The latter is not generally recommended.
Determines how far up in terms of lines from a structure, we should allow the finding of comment blocks.
This is currently exposed but should probably not be changed and may be removed in the future. It indicates the indent to be used when parsing comments. (That may need to be auto-detected in the future.)
For indicating script
or module
Extra options to pass on to Babel (if not using the typescript
parser).
Any other options are passed onto the parser.
The changelog can be found on the CHANGES.md.
Brett Zamir and contributors.
MIT License, see the included LICENSE-MIT.md file.