Rule: use a particular method of accessing range locations

[not-an-aardvark]

Nodes in an AST generally have duplicated range information: start and end properties from acorn, and range array properties. This rule would enforce the usage of one or the other.

// (this isn't a very good rule name, there should probably be a better one.)

/* eslint eslint-plugin/use-range-property: ["error", "always"] */

// Invalid:

foo.start;
foo.end;

// Valid:

foo.range[0];
foo.range[1];

/* eslint eslint-plugin/use-range-property: ["error", "never"] */

// Invalid:

foo.range[0];
foo.range[1];

// Valid:

foo.start;
foo.end;

[not-an-aardvark]

With use-range-property: always, the rule will need to somehow distinguish between node.loc.start and node.loc.end (which should always be allowed) and node.start and node.end (which should be disallowed).

[not-an-aardvark]

Closing because it probably isn't possible to disallow node.start while allowing node.loc.start in the general case.

[SimenB]

Closing because it probably isn't possible to disallow node.start while allowing node.loc.start in the general case.

"the general case" here refers to if people have destructured or something like that?

If so, could you have as a caveat that there might be false positives and only loc.start or loc.end is allowed (has to be directly accessed off of loc), or something like that?

Not perfect by any means, but still useful (e.g. would have prevented jest-community/eslint-plugin-jest#221 and by extension typescript-eslint/typescript-eslint#175)

[j-f1]

This could be implemented as a plugin that parses the code using a configured plugin and adds a getter that throws when accessing node.start or node.end to all the nodes. This could be used only when running tests for the plugin.