Better string representation of errors

[macjohnny]

Resolves #2711

Before:

Chrome 59.0.3071 (Mac OS X 10.12.3) ERROR
 {
      "originalErr": {},
      "__zone_symbol__currentTask": {
        "type": "microTask",
        "state": "notScheduled",
        "source": "Promise.then",
        "zone": "<root>",
        "cancelFn": null,
        "runCount": 0
      }
  }
Chrome 59.0.3071 (Mac OS X 10.12.3): Executed 0 of 0 ERROR (1.348 secs / 0 secs)

After:

Chrome 59.0.3071 (Mac OS X 10.12.3) ERROR
 {
    "message": {
      "originalErr": {},
      "__zone_symbol__currentTask": {
        "type": "microTask",
        "state": "notScheduled",
        "source": "Promise.then",
        "zone": "<root>",
        "cancelFn": null,
        "runCount": 0
      }
    },
    "str": [
      "Error: (SystemJS) Object prototype may only be an Object or null: undefined",
      "    TypeError: Object prototype may only be an Object or null: undefined",
      "        at setPrototypeOf (<anonymous>)",
      "        at __extends$1 (dist/bundles/my-bundle.umd.js:2258:9)",
    ]
  }
Chrome 59.0.3071 (Mac OS X 10.12.3): Executed 0 of 0 ERROR (1.348 secs / 0 secs)

[googlebot]

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed, please reply here (e.g. I signed it!) and we'll verify. Thanks.

---

• If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Check your existing CLA data and verify that your email is set on your git commits.
• If you signed the CLA as a corporation, please let us know the company's name.

[macjohnny]

I signed it!

[googlebot]

CLAs look good, thanks!

[wesleycho]

The replace and split parts should probably be removed.

Also, this line changes the message variable's type from string to an object - this probably means that tests have to be updated.

[macjohnny]

the replace and split parts are for nicer output display in the console, since the object is converted to json.
With the split and replace, you get

"str": [
      "Error: (SystemJS) Object prototype may only be an Object or null: undefined",
      "    TypeError: Object prototype may only be an Object or null: undefined",
      "        at setPrototypeOf (<anonymous>)",
      "        at __extends$1 (dist/bundles/my-bundle.umd.js:2258:9)",
    ]

instead of

"str": "Error: (SystemJS) Object prototype may only be an Object or null: undefined\n\tTypeError: Object prototype may only be an Object or null: undefined\n\tat setPrototypeOf (<anonymous>)\n\tat __extends$1 (dist/bundles/my-bundle.umd.js:2258:9)",

[macjohnny]

the variable type is not changed, the message already is an object that will later on be converted to a string with JSON.stringify()

[macjohnny]

the message might be a string or an object

[macjohnny]

@wesleycho could you have a look at this again? this patch really helped me a lot when investigating preliminary import / systemjs issues that prevented the tests from being run and log a useful message.

[macjohnny]

@wesleycho could you please have a look at this again?

[larsduelfer]

I would also highly appreciate if this change is going to be integrated.

[EzraBrooks]

I agree with @wesleycho, I’d say the error should remain a string. Formatting of the message itself should be left to the client console or to formatting methods, instead of having the message be preformatted when attached to the object.

[macjohnny]

@EzraBrooks @wesleycho I can understand the formatting should not be done here. but since the message can be an object (which is the reason for this PR), it should be ok to transform it to a different object as I suggest in this PR. would you accept this PR if I removed the .replace(/\t/g, " ").split("\n") part?

[EzraBrooks]

Thanks, I’ll approve it once the CI checks complete.

[macjohnny]

@EzraBrooks it seems there is a problem with the travis ci build retrieving some commit message. however, all other tests pass. could you please have a look at it?

[EzraBrooks]

Please squash your commits into one that follows Karma’s naming conventions - Travis is failing on commit message validation.

[EzraBrooks]

Sorry, to clarify.. the commit messages all have a leading type on them. In this case, I would recommend feat: better string representation of errors.

[macjohnny]

@EzraBrooks done, waiting for the CI to complete

[EzraBrooks]

Looks like ESLint has a few style complaints.

[macjohnny]

@EzraBrooks I updated the PR, hope it is fine now

[EzraBrooks]

Looks good to me.