What do you think about a C# 6 like nameof() expression for

Ron Buckton Ron.Buckton at microsoft.com
Sat Jun 15 22:03:09 UTC 2019


> At that point in the example code the identifer ```y``` does not exist.

That is not entirely incorrect. The identifier `y` exists, but its binding has not been initialized, otherwise you couldn’t refer to y in this case:

```
const f = () => y;
let y = 1;
```

> That fact can be utilized for an alternative approach to meet requirement of "Case 1. Function guard.".

I don’t disagree that there are “alternative approaches” to ‘nameof’ for many cases, but they all incur overhead.

At the end of the day, ‘nameof’ is a convenience feature that provides an incremental “quality of life” improvement, much like shorthand property assignments were a convenience feature (i.e. how hard is it to write `{ foo: foo }`). It’s not something that would be as much of a game changer to the language as async/await or yield were, but it would be a fairly easily spec’d and nice-to-have capability.

From: guest271314 <guest271314 at gmail.com>
Sent: Saturday, June 15, 2019 2:50 PM
To: Ron Buckton <Ron.Buckton at microsoft.com>
Cc: es-discuss at mozilla.org
Subject: Re: Re: What do you think about a C# 6 like nameof() expression for

> It doesn’t matter what the value of ‘y’ is, just what the lexical name of `y` is. `nameof` wouldn’t refer to `y` as an expression, its just pointing to the identifier.

Was not referring to the _value_ of ```y```. At that point in the example code the identifer ```y``` does not exist. That is, unless there was an expected presumption that ```y``` was defined globally prior to the ```const y = 1;``` declaration at the next line. Even then a variable declared using ```const``` cannot be referenced before "lexical declaration" or "initialization". That fact can be utilized for an alternative approach to meet requirement of "Case 1. Function guard.". Tested at Chromium 76 and Firefox 69. Have not used MS browser (or browserstack) in some time. Have not tried the code at Safari either, though the code is capable of adjustment for any JavaScript environment which throws a ```ReferenceError``` under the same case

```
function func1(param1, param2, param3, userName, param4, param5) {
  if (userName === undefined) {
    try {
      // throws ReferenceError
      // Firefox 69: ReferenceError: "can't access lexical declaration `userName' before initialization"
      // Chromium 76: ReferenceError: Cannot access 'userName' before initialization
      const userName = userName;
    } catch (e) {
      // match identifier at Firefox, Chromium ReferenceError message
      // adjust RegExp for IE/Edge, Safari, etc.
      const nameof = e.message.match(/(?!`|')[^\s]+(?=')/g).pop();
      throw new Error(`"Argument cannot be null: ${nameof}"`);
    }
  } else {
    // do stuff
    console.log(userName);
  }
}

try {
  func1(1, 2, 3); // throws error
} catch (e) {
  console.error(e); // get name of variable identifier
}

try {
  func1(1, 2, 3, 4); // logs userName at console
} catch (e) {
  console.error(e); // should not be reached
}
```

On Sat, Jun 15, 2019 at 9:37 PM Ron Buckton <Ron.Buckton at microsoft.com<mailto:Ron.Buckton at microsoft.com>> wrote:
It doesn’t matter what the value of ‘y’ is, just what the lexical name of `y` is. `nameof` wouldn’t refer to `y` as an expression, its just pointing to the identifier.

From: guest271314 <guest271314 at gmail.com<mailto:guest271314 at gmail.com>>
Sent: Friday, June 14, 2019 10:03 PM
To: Ron Buckton <Ron.Buckton at microsoft.com<mailto:Ron.Buckton at microsoft.com>>
Cc: es-discuss at mozilla.org<mailto:es-discuss at mozilla.org>
Subject: Re: Re: What do you think about a C# 6 like nameof() expression for

> I’m not sure I understand what you mean by “false-positive” in this instance.

Was referring to


const x = nameof y; // "y"

const y = 1;

Where ```y``` is ```undefined``` an error is not expected to be thrown? Is ```y``` declared globally, using ```const``` or ```let```, or not at all? The use case  described by OP


function func1(param1, param2, param3, userName, param4, param5) {

   if (userName == undefined) {

       throw new ArgumentNullError(nameof userName); // `ArgumentNullError`

is a custom error, derived from `Error`, composes error message like

"Argument cannot be null: userName".

   }
checks if ```userName``` was ```undefined``` before using ```nameof```.  If error checking is a substantial portion of the proposal, why should an error (```y``` being ```undefined``` though referenced) be ignored when referencing an undefined identifier though concentrate on coercing a name from a different potentially undefined property?

Consider this case:
```
const someObject = { value: 1 };
function setValue(value /*1*/) {
  if (typeof value /*2*/ !== "number") throw new TypeError(`Number expected: ${nameof value /*3*/}`);
  someObject["value" /*4*/] = value /*5*/;
}
```
If you rename the parameter `value` of the function `setValue` in an editor with a rename refactoring, you want to rename the symbols at 1, 2, 3, and 5, but not the string at 4.

Not gathering the purpose or value of ```nameof``` usage in that case. If the value is not a "number" then why does the value or name matter?

Since the primary use case appears to be an editor environment, why cannot the editor be programmed to recognize the custom JavaScript ```nameof```` function or operator? Then it would not matter if this board concurred with the ```nameof``` functionality or not. Both CLI and GUI editors (and JavaScript) are generally extensible. FWIW, some time ago incorporated features into gedit for HTML templates; should be a similar process to create custom scripts for the various editor environments where users rely on such programs for code composition; now simply write the code by hand and test in different environments, without taking the time to customize or rely on an editor - take the time to test the code where the code will actually be run where errors, if any, can be evaluated in the context in which a specific output is expected. To each their own. What needs to be implemented outside of what the users which advocate for ```nameof``` cannot implement themselves?

As mentioned earlier do not rely on "an editor with name refactoring" to compose code. The code has to be tested (outside of the editor environments) anyway. Test the code itself, here, not the editor.

On Fri, Jun 14, 2019 at 9:49 PM Ron Buckton <Ron.Buckton at microsoft.com<mailto:Ron.Buckton at microsoft.com>> wrote:
I’m not sure I understand what you mean by “false-positive” in this instance.

Consider this case:

```
const someObject = { value: 1 };
function setValue(value /*1*/) {
  if (typeof value /*2*/ !== "number") throw new TypeError(`Number expected: ${nameof value /*3*/}`);
  someObject["value" /*4*/] = value /*5*/;
}
```

If you rename the parameter `value` of the function `setValue` in an editor with a rename refactoring, you want to rename the symbols at 1, 2, 3, and 5, but not the string at 4.

Ron

From: guest271314 <guest271314 at gmail.com<mailto:guest271314 at gmail.com>>
Sent: Friday, June 14, 2019 2:43 PM
To: Ron Buckton <Ron.Buckton at microsoft.com<mailto:Ron.Buckton at microsoft.com>>
Cc: es-discuss at mozilla.org<mailto:es-discuss at mozilla.org>
Subject: Re: Re: What do you think about a C# 6 like nameof() expression for

How is that behaviour related to the use cases presented by OP? Would such behaviour not lead to false-positive relevant to the 2 use cases?

On Fri, Jun 14, 2019 at 9:36 PM Ron Buckton <Ron.Buckton at microsoft.com<mailto:Ron.Buckton at microsoft.com>> wrote:
> `nameof whatever` → `Object.keys({ whatever })[0]`, but I'm a bit confused why it'd be better to type `nameof foo` in code, rather than `'foo'` - if you change `foo` to `bar`, you have to change both of them anyways.

If you are using an editor that supports rename refactoring, its generally easier to rename the symbol `foo` and have all references (including `nameof foo`) be updated. You cannot safely automatically rename `'foo'` to `'bar'` since an editor or language service cannot guarantee that by the string `'foo'` you meant “the text of the identifier `foo`”.

From: es-discuss <es-discuss-bounces at mozilla.org<mailto:es-discuss-bounces at mozilla.org>> On Behalf Of Jordan Harband
Sent: Friday, June 14, 2019 2:29 PM
To: guest271314 <guest271314 at gmail.com<mailto:guest271314 at gmail.com>>
Cc: es-discuss <es-discuss at mozilla.org<mailto:es-discuss at mozilla.org>>
Subject: Re: Re: What do you think about a C# 6 like nameof() expression for

`nameof whatever` → `Object.keys({ whatever })[0]`, but I'm a bit confused why it'd be better to type `nameof foo` in code, rather than `'foo'` - if you change `foo` to `bar`, you have to change both of them anyways.

On Fri, Jun 14, 2019 at 1:31 PM guest271314 <guest271314 at gmail.com<mailto:guest271314 at gmail.com>> wrote:
Am neither for nor against the proposal. Do not entertain "like"s or "dislike"s in any field of endeavor. Am certainly not in a position to prohibit anything relevant JavaScript. Do what thou wilt shall be the whole of the Law.

Have yet to view a case where code will be "broken" by ```nameof``` not being a JavaScript feature. "robustness", as already mentioned, is a subjective adjective that is not capable of being objectively evaluated as to code itself. That description is based on preference or choice.

In lieu of the proposal being specificed, use the posted code example of ```Object.keys()``` that "works".

```
function func1({userName = void 0} = {}) {
  console.assert(userName !== undefined, [{userName}, 'property needs to be defined'])
}
```

provides a direct indication that the property value is required to be defined. Note that the example code posted thus far does not first check if ```options``` is passed at all, for which ```nameof``` will not provide any asssitance.

Usually try to meet requirement by means already available in FOSS browsers. Have no interest in TypeScript or using an IDE.

FWIW, have no objection to the proposal.

On Fri, Jun 14, 2019 at 7:53 PM Stas Berkov <stas.berkov at gmail.com<mailto:stas.berkov at gmail.com>> wrote:
guest271314, what is you point against `nameof` feature?

If you don't like it - don't use it. Why prohibit this feature for
those who find it beneficial?

I see `nameof` beneficial in following cases

Case 1. Function guard.
```
function func1(options) {
...
   if (options.userName == undefined) {
       throw new ParamNullError(nameof options.userName); //
`ParamNullError` is a custom error, derived from `Error`, composes
error message like "Parameter cannot be null: userName".
 // `Object.keys({options.userName})[0]` will not work here
   }
}
```

Case 2. Accessing property extended info
Those ES functions that accept field name as string.
e.g.
```
const descriptor1 = Object.getOwnPropertyDescriptor(object1, 'property1');
```
vs
```
const descriptor1 = Object.getOwnPropertyDescriptor(object1, nameof
object1.property1);
 // `Object.keys({options1.property1})[0]` will not work here
```
2nd variant (proposed) has more chances not to be broken during
refactoring (robustness).

It would make devs who use IDE more productive and make their life
easier. Why not give them such possiblity and make them happy?
_______________________________________________
es-discuss mailing list
es-discuss at mozilla.org<mailto:es-discuss at mozilla.org>
https://mail.mozilla.org/listinfo/es-discuss<https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmail.mozilla.org%2Flistinfo%2Fes-discuss&data=02%7C01%7CRon.Buckton%40microsoft.com%7C4036bd0c96744cab8c4208d6f1db7291%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636962322162641169&sdata=U%2B%2Fq939hNKeRQlHnXNvoEPvcoi01IR3T56Jv7eOJSlg%3D&reserved=0>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.mozilla.org/pipermail/es-discuss/attachments/20190615/af8325ba/attachment-0001.html>


More information about the es-discuss mailing list