Proposal: Concise instance initialisation

Brendan Eich brendan at mozilla.com
Tue May 24 07:49:17 PDT 2011


On May 24, 2011, at 7:44 AM, Brendan Eich wrote:

> You're right, this would be unambiguous because a block can't initialize a var. The only interpretation is an expression, which with Allen's proposal combined with object intiialiser shorthand would parse {!b}.

One more thing: {b} in expression context as shorthand for {b: b} is convenient. In destructuring context var {b} = c, e.g., is totally winning. But it breaks "alpha conversion" -- if you rename variables in a program with lexical scope, the result should be equivalent to the original program.

Ok, we can handle this by requiring long-hand expansion first:

... {b: b} ...

Then the {!b} form should expand to ... {!b: b} ... and that may be tolerable in the hoped-for block/object-literal disambiguation solution.

One thing that might be confusing: ... {!b} ... could be confused for ... {b: !b} .... People would learn, but the shorthand mixes uneasily with the retasking of ! and ~ from being unary expression operators to being property name prefixes.

/be


More information about the es-discuss mailing list