Using destructuring for function arguments
Seems like any identifiers in the arguments should always be defined in scope before ever considering what they will be assigned.
On Sat, May 31, 2014 at 11:59 AM, Nicholas C. Zakas < standards at nczconsulting.com> wrote:
I've been playing around with using destructuring as function arguments and have come across some odd behaviors that I'm not sure are intentional (or perhaps, not to spec). For context, consider the following function:
function setCookie(name, value, { secure, path, domain, expires }) { console.log(secure); // ... } // called as setCookie("type", "js", { secure: true });
What happens here:
secure === true
path === undefined
domain === undefined
expires === undefined
I'd say all of that behavior is as expected. However, if I omit the third argument completely, things get a bit strange:
setCookie("type", "js"); // throws error at console.log
What happens here is that none of
secure
,path
,domain
,expires
are defined. I can usetypeof
on them to protect against this, but then I end up with some lousy looking code:function setCookie(name, value, { secure, path, domain, expires }) { if (typeof secure !== "undefined") { // use it } if (typeof path !== "undefined") { // use it } if (typeof domain !== "undefined") { // use it } if (typeof expires !== "undefined") { // use it } // ... }
Strange, it seems should fail at trying to start destructuring process:
when ToObject coercion fails with the undefined
value set to the value of
the parameter.
My first thought was that this behavior made sense, since no destructuring can happen on undefined. However, the workaround for dealing with that behavior seems a bit heavy-handed.
I thought perhaps I could assign a default value, and that would solve the problem:
function setCookie(name, value, { secure, path, domain, expires } = {}) { console.log(secure); // ... }
Unfortunately, that resulted in a syntax error in Firefox. Traceur seems to have no problem with it.
So I really have two questions:
- Who is right about assigning a default value to a destructured parameter, Firefox or Traceur?
Well, the the FormalParameter
is the BindingElement
, and the later may
be the BindingPattern Initializer
, where the Initializer
is your
default value. From which chain Traceur is correct.
- Is the behavior of not having any bindings for destructured parameter properties correct? And if so, is it desirable?
Where do you test this? It seems it should fail at destructuring, and not set any binding pattern properties in this case.
I.e.
setCookie('type', 'js'); // throws right away, since no ToObject coercion
is made
setCookie('type', 'js', {}); // sets all binding props to undefined
With the default value of the pattern, the first call sets all binding
props to undefined
as well.
Dmitry
Nicholas C. Zakas wrote:
function setCookie(name, value, { secure, path, domain, expires } = {}) { console.log(secure); // ... }
Unfortunately, that resulted in a syntax error in Firefox.
Could you please file a bug against SpiderMonkey? Thanks,
Matthew Robb wrote:
Seems like any identifiers in the arguments should always be defined in scope before ever considering what they will be assigned.
Right, and they are in scope no matter what.
Seems to me that an implementation bug (can't have parameter default value for destructuring formal) is the only problem brought to light in this thread. Anyone see anything else amiss?
On May 31, 2014, at 8:59 PM, Nicholas C. Zakas <standards at nczconsulting.com> wrote:
I've been playing around with using destructuring as function arguments and have come across some odd behaviors that I'm not sure are intentional (or perhaps, not to spec).
Argument binding initialization takes place as part of people.mozilla.org/~jorendorff/es6-draft.html#sec-functiondeclarationinstantiation The actual initialization takes place in steps 24-25, essentially treating the entire parameter list as an array-like restructuring
Note that there has been significant tweaks to this section in each of the 3 latest spec. drafts. You favorite implementation may not match the current spec. text.
For context, consider the following function:
function setCookie(name, value, { secure, path, domain, expires }) { console.log(secure); // ... } // called as setCookie("type", "js", { secure: true });
What happens here:
secure === true
path === undefined
domain === undefined
expires === undefined
I'd say all of that behavior is as expected. However, if I omit the third argument completely, things get a bit strange:
setCookie("type", "js"); // throws error at console.log
correct, this should throw a TypeError exception because desctructuring patterns (in any context) require an object value to restructure. Since you did not pass a third argument there is no object to restructure into {secure, path, domain, expires}
What happens here is that none of
secure
,path
,domain
,expires
are defined. I can usetypeof
on them to protect against this, but then I end up with some lousy looking code:
not really, the thrown exception while processing the arguments should terminate the function and you should never start executing the function body
function setCookie(name, value, { secure, path, domain, expires }) { if (typeof secure !== "undefined") { // use it } if (typeof path !== "undefined") { // use it } if (typeof domain !== "undefined") { // use it } if (typeof expires !== "undefined") { // use it } // ... }
My first thought was that this behavior made sense, since no destructuring can happen on undefined. However, the workaround for dealing with that behavior seems a bit heavy-handed.
I thought perhaps I could assign a default value, and that would solve the problem:
function setCookie(name, value, { secure, path, domain, expires } = {}) { console.log(secure); // …
yes the above is exactly what you are expected to do. Or you could include some defaults in the default object:
{secure, path, domain, expires } = {secure: false}
}
Unfortunately, that resulted in a syntax error in Firefox. Traceur seems to have no problem with it.
that sounds like a Firefox bug
So I really have two questions:
- Who is right about assigning a default value to a destructured parameter, Firefox or Traceur?
Traceur
- Is the behavior of not having any bindings for destructured parameter properties correct? And if so, is it desirable?
no, it is supposed to throw as discussed above.
On Sat, May 31, 2014 at 1:59 PM, Nicholas C. Zakas <standards at nczconsulting.com> wrote:
- Who is right about assigning a default value to a destructured parameter, Firefox or Traceur?
Traceur is right.
- Is the behavior of not having any bindings for destructured parameter properties correct? And if so, is it desirable?
I think you're mistaken about this. What happens is that
function setCookie(name, value, { secure, path, domain, expires }) {
console.log(secure);
// ...
}
behaves like
function setCookie(name, value, options) {
var { secure, path, domain, expires } = options;
console.log(secure);
// ...
}
and the error is in trying to unpack options, which is undefined, not in later trying to use the variables.
In addition to what Allen said, you could also do something like this:
function setCookie(name, value, { secure, path, domain, expires } = cookieDefaults) { }
where cookieDefaults
is defined elsewhere.
Or you could do something like:
function setCookie(name, value, { secure = false, path = ".", domain = "", expires = now() } = {}) { }
where you are specifying default initializers within the destructuring pattern.
Or, for more readable code:
function setCookie(name, value, options = {}) {
let {
secure = false,
path = "",
domain = "",
expires = whenever()
} = options;
// Do stuff
}
Kevin
On Sat, May 31, 2014 at 1:41 PM, Brendan Eich <brendan at mozilla.org> wrote:
Matthew Robb wrote:
Seems like any identifiers in the arguments should always be defined in scope before ever considering what they will be assigned.
Right, and they are in scope no matter what.
Seems to me that an implementation bug (can't have parameter default value for destructuring formal) is the only problem brought to light in this thread. Anyone see anything else amiss?
No, seems it's the only problem. Also, just to add: esprima parser also has this bug yet (can't have param default value with patterns).
Dmitry
Dmitry Soshnikov wrote:
No, seems it's the only problem. Also, just to add: esprima parser also has this bug yet (can't have param default value with patterns).
Nicholas kindly filed
bugzilla.mozilla.org/show_bug.cgi?id=1018628
to track the SpiderMonkey bug. If you file an Esprima one, could you cite it to the list? Thanks,
On 5/31/2014 1:54 PM, Allen Wirfs-Brock wrote:
What happens here is that none of
secure
,path
,domain
,expires
are defined. I can usetypeof
on them to protect against this, but then I end up with some lousy looking code:not really, the thrown exception while processing the arguments should terminate the function and you should never start executing the function body
This sounds like there's a second error in the Firefox implementation. If an error should be thrown when the destructured argument is omitted, then Firefox is incorrectly continuing to execute the function body.
Am I reading this correctly?
I've been playing around with using destructuring as function arguments and have come across some odd behaviors that I'm not sure are intentional (or perhaps, not to spec). For context, consider the following function:
What happens here:
secure === true
path === undefined
domain === undefined
expires === undefined
I'd say all of that behavior is as expected. However, if I omit the third argument completely, things get a bit strange:
What happens here is that none of
secure
,path
,domain
,expires
are defined. I can usetypeof
on them to protect against this, but then I end up with some lousy looking code:My first thought was that this behavior made sense, since no destructuring can happen on undefined. However, the workaround for dealing with that behavior seems a bit heavy-handed.
I thought perhaps I could assign a default value, and that would solve the problem:
Unfortunately, that resulted in a syntax error in Firefox. Traceur seems to have no problem with it.
So I really have two questions:
Thanks.