A proposal that enables .? and ?[]

[Mouvedia]

This is inspired by #77.

/*** USE CASES ***/

// optional static property access
catch a?.b;

// optional dynamic property access
var qux = 'foo-bar'

catch a?[qux];

// optional function/method call
catch mayBeNullOrUndefined?(...args);

/*** TRANSPILATION ***/

catch a?.b;
// equivalent of
a == null ? undefined : a.b;

PROS

• readability (declared intent)
• scannability (easily spotted)
• no conflict: as a keyword, catch cannot be used as an identifier (catch(a?.b) is possible)
• akin to proposal-optional-catch-binding (discards the error)
• short (subjective)

CONS

• cannot implicitly check every property of the chain without adding ? for each (cf comment-445671987)

[claudepache]

• akin to proposal-optional-catch-binding (discards the error)

As already said somewhere else, Optional chaining is not about discarding errors, and using the keyword ”catch” will maintain the confusion. For example, arguments?.callee will throw a TypeError if triggered inside a strict-mode function.

---

• short (subjective)

I disagree, catch x?[y] is significantly longer than x?.[y]

---

Also: Con: Using a keyword just for enabling syntax is a precedent (and imho not a good precedent). (In the past, "use strict" was intended to be mandatory for having new features, but later, it was decided that new features should be available also in sloppy mode.)

[Mouvedia]

For example, arguments?.callee will throw a TypeError if triggered inside a strict-mode function.

By discards the error I meant discards the error that you are expecting. Obviously any other error would throw normally.

I disagree, catch x?[y] is significantly longer than x?.[y]

Like I said it's better for clarity: it's explicit. And I consider ?.[] an atrocity. Also ?.[] is not consistent with ?. as explained numerous times in other issues.

[claudepache]

By discards the error I meant discards the error that you are expecting. Obviously any other error would throw normally.

Sure, you meant that. But what is obvious for you is not obvious for the lambda programmer. It is not clear why a catch keyword prefixing an expression discards only one specific error that would occur in one specific case when evaluating that expression.

[obedm503]

Maybe instead of using catch a different keyword could be used. catch already has meaning of catching all errors. maybe a keyword could help us get away from ?.[ and ?.()

[ljharb]

As well as the reasons already stated, it must be possible to do something like a.b.c.d where the .b and .d are optional but the .c is not.

a?.b.c?.d, for example, provides this - how would you do it with your suggestion?

[Mouvedia]

a?.b.c?.d, for example, provides this - how would you do it with your suggestion?

catch a?.b.c?.d;

[Mouvedia]

It is not clear why a catch keyword prefixing an expression discards only one specific error that would occur in one specific case when evaluating that expression.