Sorry for the confusion, this is a genuine error - but with the wrong message.

In order to use rules which require type information, you have to provide a reference to your tsconfig.json.

I have opened #132 to address this. I'm afraid with the huge migration of everything, we haven't got around to improving the docs too much yet.

Thanks @JamesHenry.

Just an FYI, re: performance...

In a project containing ~200 x *.ts files, I recorded some timings for a full linting run before/after enabling this rule and adding parserOptions.project: <path to tsconfig.json>:

Rule off

"rules": {
  "restrict-plus-operands": "off"
}

real  0m10.094s
user  0m16.304s
sys	  0m0.720s

Rule off + parserOptions.project set

"parserOptions": {
  "project": "./tsconfig.json"
},
"rules": {
  "restrict-plus-operands": "off"
}

real  0m54.436s
user  1m15.434s
sys   0m8.106s

Rule on

"parserOptions": {
  "project": "./tsconfig.json"
},
"rules": {
  "restrict-plus-operands": "error"
}

real  1m17.327s
user  1m42.875s
sys   0m9.295s

Simply setting parserOptions.project to point to my tsconfig.json caused the linting time to jump from ~16s to ~1m 15s (presumably setting this option causes tsc to run during parsing?).

Then with the restrict-plus-operands rule, the total run time goes to ~1m 42s.

This 6x slowdown may make it impractical for some large projects to enable this rule (or any future rules that rely on tsconfig.json reference being set).

Out of curiosity, how does it compare to TSLint for any rules that require type checking? c.f. #55

Tagging @uniqueiniquity

Hi @scottohara, thanks for sharing the perf numbers you're getting.
@aboyton, I'm in the process of writing a handful of semantic rules with the intent of doing a larger-scale perf comparison. In my experience so far, the two tools have been pretty comparable, but I haven't fully tested yet. Moving forward, I'm planning on iterating on the support for generating semantic information so we can get that perf hit down as much as we can. The current support is definitely a first cut.