[patch][cla] dojo.exists behavior is a little confusing

[ckbcowboy]

I'm not a dojo user, but I noticed this today:

dojo.exists acts like it should be called dojo.truthy or dojo.methodExists.. it seems to be geared towards testing properties that will never be set to 0, false or empty string.

for example, in firebug console, on http://dojotoolkit.org/:

var foo = { a: 1, b: 0, c: false, d: null, e: '' };

console.log([
  dojo.exists('a', foo),
  dojo.exists('b', foo),
  dojo.exists('foo.c'),
  dojo.exists('foo.d'),
  dojo.exists('foo.e'),
  dojo.exists('foo.f')
]);

// logs:
// [true, false, false, false, false, false]

// but should probably log:
// [true, true, true, true, true, false]

A fix would be to change !! to !== undefined.

• Cowboy

[lipik]

I agree. Got bitten by this just last week. I vote that this be prioritized - fixed is simple, the !!obj["val"] check in exists() should be replaced with typeof obj["val"] != "undefined", or even val in obj.

I am not sure that the latter is supported in IE6, though.

[James Burke]

It looks like dojo.exists is just an artifact on top of dojo.getObject() and it is optimized for finding objects and methods on objects. I will update the docs to clarify its purpose.

Note that for the getting the behavior you want mentioned above, it is much easier (and faster) just to use the "in" operator in Javascript:

("d" in foo)

will return true.

[James Burke]

(In [17325]) Fixes #9001: clarify docs on the purpose of dojo.exists.

[dante]

-6 bytes, +11 bytes. teeny patch. unit tests pass, expanded unit tests.

[James Burke]

I'm fine if the patch uses false instead of the /* falsey */0 thing, if the concern was number of bytes. I would rather have the code clearer. Other than that feel free to commit it.

[dante]

(In [22878]) fixes #9001 - thanks cowbow for the initial filing. going with full and negligible byte increase approved. of course you're \!strict