Precedence of yield operator
Using generators for async is a clever hack, but it's just a hack. A proper solution will need a new keyword anyway (most languages seem to use "await" or something similar), which can get the better precedence.
On Fri, Jun 14, 2013 at 11:21 AM, Tab Atkins Jr. <jackalmage at gmail.com> wrote:
Using generators for async is a clever hack, but it's just a hack. A proper solution will need a new keyword anyway (most languages seem to use "await" or something similar), which can get the better precedence.
This is just not true -- the semantics of generators and of await as
proposed on this list are very very similar.
On Fri, Jun 14, 2013 at 5:33 PM, Sam Tobin-Hochstadt <samth at ccs.neu.edu> wrote:
On Fri, Jun 14, 2013 at 11:21 AM, Tab Atkins Jr. <jackalmage at gmail.com> wrote:
Using generators for async is a clever hack, but it's just a hack. A proper solution will need a new keyword anyway (most languages seem to use "await" or something similar), which can get the better precedence.
This is just not true -- the semantics of generators and of
awaitas proposed on this list are very very similar.
Yup, very similar, but not identical, which is why we have slight incompatibilities in the desired precedence, why you need to have a wrapper to handle the calling, etc.
Bruno Jouhier wrote:
While playing with my little async/await library, I noticed that I was often forced to parenthesize yield expressions as (yield exp) because of the low precedence of the yield operator. Typical patterns are:
var foo = (yield a()) + (yield b()) + (yield c());
That's actually a hard case, IMHO -- and "hard cases make bad law".
Many programmers would rather have the extra parens for uncertain cases (C's misplaced bitwise-logical and shift operators, vs. equality/relational; anything novel such as yield).
But the real reason for yield being low precedence is to avoid precedence inversion. Consider if yield could be high-precedence, say a unary operator like delete:
let x = yield a + b;
Oops, many people would want that to be equivalent to
let x = yield (a + b);
but if yield is at delete's precence level, it's rather:
let x = (yield a) + b;
Which is the rare ("hard", from law school) case.
For commutative operators such as +, over-parenthesizing is better again, because
let x = b + (yield a);
and
let x = (yield a) + b;
ignoring order of effects in shared mutable store, and ignoring floating point non-determinism, are equivalent.
The current precedence looks natural to me for a yield but less when yield is used as an await keyword. That's why I proposed to handle it with a different keyword rather than by changing yield's precedence. Await would have been a good candidate but it is not reserved.
Anyway, I don't want to make a big fuss about it. That's just something I noticed while writing examples of async code with generators. Adding the parentheses is not such a big deal.
2013/6/14 Brendan Eich <brendan at mozilla.com>
This is a familiar discussion from C#. I forwarded it to the mediator of that convresation and got a nice summary, pasted here:
---------- Forwarded message ---------- From: Mads Torgersen <Mads.Torgersen at microsoft.com>
Date: Fri, Jun 14, 2013 at 2:11 PM Subject: RE: Precedence of yield operator To: Dean Tribble <tribble at e-dean.com>
I’m not on the mailing list. Feel free to forward to it.****
In C# we have separate keywords too, and indeed the precedence differs as described below. For “yield return” (our yield) the lower precendence falls out naturally since it engenders a statement, not an expression.
** “await” is not a reserved keyword in C# either, but we managed to wedge it in all the same. Just adding await as an operator would lead to all kinds of ambiguity; e.g. “await (x)” could be a function call or an await expression, and the statement “await x;” could be a variable declaration or an await statement.
However, in C# “await” is only allowed inside methods marked “async”, and since there weren’t any of those around before the feature was introduced, it is not a breaking change. Inside non-async methods, therefore, “await” continues to be just an identifier.****
I don’t know if a similar thing is possible in EcmaScript. But I believe that a low-precedence yield as a substitute for a high-precedence await is problematic: you never want “yield a + yield b” to mean “yield (a + (yield b))”: the things you await – Task, Promises, Futures, whatever you call them – just don’t have operators defined on them, and it would be silly to parse them as if they might and then give errors (at runtime in EcmaScript, at compile time in e.g. TypeScript).****
Mads
I don’t know if a similar thing is possible in EcmaScript. But I believe that a low-precedence yield as a substitute for a high-precedence await is problematic: you never want “yield a + yield b” to mean “yield (a + (yield b))”: the things you await – Task, Promises, Futures, whatever you call them – just don’t have operators defined on them, and it would be silly to parse them as if they might and then give errors (at runtime in EcmaScript, at compile time in e.g. TypeScript).****
This is exactly the problem I ran into: yield does the job of await but the syntax rule feels a bit silly.
This is not too dramatic because the errors are either detected at compile time (yield a + yield b) or at runtime by my little library (yield a + b).
This is all on the agenda for ES7. It cleanly missed ES6 in May 2011(!).
I'm maybe biased, but I would love to consider "yield" as a function. Indeed, it calls another code, which can return you a value. That looks very similar to a function call to me. If we do this, the predecence becomes intuitive again:
var x = yield(a) + yield(b);
yield(a+b);
I think there was a proposal to allow parenthesis-free function call at some point at the root of a statement. When it lands, you'll be able to do things like
yield a+b;
as a normal statement again. In the mean time we can just use parentheses, that's not a huge issue and it helps clarity.
But maybe this is too late to change ES6, it's just an idea I had while reading this thread, not a strong call for change.
-----Message d'origine---
If this is on the agenda for ES7 maybe ES6 should at least reserve "await" as a keyword in function* scopes.
Le 15 juin 2013 à 10:18, François REMY <francois.remy.dev at outlook.com> a écrit :
I'm maybe biased, but I would love to consider "yield" as a function. Indeed, it calls another code, which can return you a value. That looks very similar to a function call to me. If we do this, the predecence becomes intuitive again:
var x = yield(a) + yield(b); yield(a+b);
I think there was a proposal to allow parenthesis-free function call at some point at the root of a statement. When it lands, you'll be able to do things like
yield a+b;
as a normal statement again. In the mean time we can just use parentheses, that's not a huge issue and it helps clarity.
But maybe this is too late to change ES6, it's just an idea I had while reading this thread, not a strong call for change.
There is an issue when the argument of an intended parenthesis-free function call happens to begin with a parenthesis:
yield a*(b + c) <==> yield(a*(b + c))
yield (b + c)*a <==> (yield(b + c))*a instead of yield((b + c) *a)There is an issue when the argument of an intended parenthesis-free function call happens to begin with a parenthesis.
Just like there are issues if you want to return an object literal from a lambda in JS, you've to use parentheses around it. Not a big deal.
Anyway, I'm not saying we should support paren-free function calls, I'm just saying that if we ever do, that would also apply to the "yield pseudo-function" of my proposal, too.
If we do provide such special syntax, then I suggest (a variant on some else's suggestion -- I forget who) that "yield" is to "function* as "await" is to "function!", amplifying the association of "!" with asynchrony. Rather than say "async function" or whatever, we'd just say "function!". Since "function!" would be new in ES7, there's no reason to reserve anything in ES6 to prepare for this.
On Saturday, June 15, 2013, Mark S. Miller wrote:
If we do provide such special syntax, then I suggest (a variant on some else's suggestion -- I forget who) that "yield" is to "function* as "await" is to "function!", amplifying the association of "!" with asynchrony. Rather than say "async function" or whatever, we'd just say "function!". Since "function!" would be new in ES7, there's no reason to reserve anything in ES6 to prepare for this.
Mark, is there a particular rationale for the reuse of "!" as the eventual send operator and here for similar semantics? I recall this also appeared in you Distributed Electronic Rights paper. Putting its strong existing meaning aside (negation), when I see "!" I think of "urgent" things—quite the opposite of the behavior described in the paper and above.
I hadn't thought of the "urgent" implication -- interesting. I agree that this has the contrary connotations which is unfortunate.
E and AmbientTalk use "<-" which I always thought read well. In E, the "then" concept is named "when" and is also supported as a syntactic form using the "-> syntax. These work well together, as both indicate
asynchrony, with the direction of the arrow indicating what is eventually being delivered to what.
In JS, "a <- b" already parses as "a < -b" so introducing a "<-" token is a non-starter. Process calculi languages like Pict (based on Pi Calculus), and languages inspired by process calculi notions like Erlang and Scala use an infix "!" for message sending, with the receiver on the left and the payload on the right. This started with Tony Hoare's CSP (Communicating Sequential Processes) and was picked up by Occam, where it was synchronous. Technically it is synchronous in the Pi Calculus too. But in Pict, Erlang, and Scala it is asynchronous. (Pict is based on the asynchronous subset of the Pi Calculus.)
More importantly for our audience, for JS (and similarly for Scala), there's also a nice mnemonic rule: Wherever you can write an infix "." you can write an infix "!" instead, changing the meaning of the construct from synchronous to asynchronous. The "!" symbol has a "." at the bottom of it, so you can think of it as "an enhanced dot" or even "a dot with a trail" if you'd like.
In any case, once the eye gets used to it the need for these mnemonics rapidly drops away. Having lived with both "<-" and "!", I've come to prefer the "!". I'm not really sure why.
Hmmm, my initial thought was against this (and for this specific case i'm not sure i've been sold on it), but I do kind of like the concept. Alas this is me we're talking about so I do have a few issues with it
- I'm a little concerned that this could make for easily misread code, even for other characters (! obviously disappears into line noise in english)
- Adding a new syntax for a single (though admittedly major) use case is worrying to me, i don't really want to see ES disappear into line noise territory where code changes meaning based on a single character
- It also seems a little inflexible
After reading Mark's email I did ponder what it would mean if we allowed "operators" to be described, e.g.
function .???(...) { throw "Because he hates us all" }
"This is terrible why is oliver suggesting this".???
I'm really not enamored with this, but I thought I'd throw it out here
Why not use the ! marker both for function declarations and function calls then:
function! foo(a, b) { }
var v = foo!(a, b);
Of, for better symmetry:
function foo!(a, b) { }
var v = foo!(a, b);
When you have a lot of code sitting on top of async APIs, you end up with a high proportion of async calls. So a discreet marker is better than a heavy keyword like await.
Hi Bruno, yes the idea is to use the infix ! for async calls, but not in quite the way you mean. Have you seen strawman:concurrency and research.google.com/pubs/pub40673.html?
The basic idea is that without syntactic support, today in Q to send an asynchronous message, you'd say
Q(p).send('foo', a,b)
In what has been codified of DOM promises, you'd say
Promises.accept(p).then(o => o.foo(a,b))
which wouldn't even work distributed, since there is no local o on which to do a local ".". Both of these are so much heavier than
p.foo(a,b)
as to make asynchronous and distributed programming remain second class compared to local synchronous programming. So the full proposal is that
p ! foo(a,b)
would mean to eventually send the message on the right to the receiver designated on the left. Likewise for
p ! (a,b) // asynchronous function call
p ! foo // asynchronous get
p ! foo = x // asynchronous put
delete p ! foo // asynchronous delete
and the computed forms of all these
p ! [name] (a,b)
p ! [name]
p ! [name] = x
delete p [name]
Over a RESTful transport, these would turn into POST, GET, PUT, and DELETE. Fortunately or unfortunately, because of CORS and UMP (something they agreed on), I expect RESTful PUT and DELETE to wither away and die in browser-to-server traffic. Our recent electronic rights paper uses only POST and GET, even though it uses only server-to-server traffic. If it can't practically be used browser-to-server, I expect it won't be used much server-to-server. Nevertheless, we should keep these forms of "!" in the language.
On Sun, Jun 16, 2013 at 7:25 AM, Mark S. Miller <erights at google.com> wrote:
Hi Bruno, yes the idea is to use the infix ! for async calls, but not in quite the way you mean. Have you seen < strawman:concurrency> and < research.google.com/pubs/pub40673.html> ?
The basic idea is that without syntactic support, today in Q to send an asynchronous message, you'd say
Q(p).send('foo', a,b)In what has been codified of DOM promises, you'd say
Promises.accept(p).then(o => o.foo(a,b))which wouldn't even work distributed, since there is no local o on which to do a local ".". Both of these are so much heavier than
p.foo(a,b)as to make asynchronous and distributed programming remain second class compared to local synchronous programming. So the full proposal is that
p ! foo(a,b)would mean to eventually send the message on the right to the receiver designated on the left. Likewise for
p ! (a,b) // asynchronous function call p ! foo // asynchronous get p ! foo = x // asynchronous put delete p ! foo // asynchronous deleteand the computed forms of all these
p ! [name] (a,b) p ! [name] p ! [name] = x delete p [name]
Should be
delete p ! [name]Thanks a lot for the pointers and for taking the time to debrief me on the proposal. I had stumbled on the concurrency paper a while ago but never took the time to read it thoroughly. I surely will!
It is good to see that the problem is being attacked from a generic angle with an eye on a concise and efficient syntax. I like a lot the idea of putting async on an equal footing with sync.
On Sat, Jun 15, 2013 at 7:46 PM, Mark S. Miller <erights at google.com> wrote:
I hadn't thought of the "urgent" implication -- interesting. I agree that this has the contrary connotations which is unfortunate.
E and AmbientTalk use "<-" which I always thought read well. In E, the "then" concept is named "when" and is also supported as a syntactic form using the "-> syntax. These work well together, as both indicate asynchrony, with the direction of the arrow indicating what is eventually being delivered to what.
In JS, "a <- b" already parses as "a < -b" so introducing a "<-" token is a non-starter. Process calculi languages like Pict (based on Pi Calculus), and languages inspired by process calculi notions like Erlang and Scala use an infix "!" for message sending, with the receiver on the left and the payload on the right. This started with Tony Hoare's CSP (Communicating Sequential Processes) and was picked up by Occam, where it was synchronous. Technically it is synchronous in the Pi Calculus too. But in Pict, Erlang, and Scala it is asynchronous. (Pict is based on the asynchronous subset of the Pi Calculus.)
More importantly for our audience, for JS (and similarly for Scala), there's also a nice mnemonic rule: Wherever you can write an infix "." you can write an infix "!" instead, changing the meaning of the construct from synchronous to asynchronous. The "!" symbol has a "." at the bottom of it, so you can think of it as "an enhanced dot" or even "a dot with a trail" if you'd like.
In any case, once the eye gets used to it the need for these mnemonics rapidly drops away. Having lived with both "<-" and "!", I've come to prefer the "!". I'm not really sure why.
www.youtube.com/watch?v=OmSbdvzbOzY 10 minutes. Directed by Chuck Jones of Bugs Bunny fame. 1965 Oscar winner. All worth it for the punch line.
While playing with my little async/await library, I noticed that I was often forced to parenthesize yield expressions as (yield exp) because of the low precedence of the yield operator. Typical patterns are:
var foo = (yield a()) + (yield b()) + (yield c()); if ((yield a()) && cond2 ...) ...Looks more LISPish than JSish to me.
The low precedence plays well with yield statements, as it lets you write
yield a + b;very much like you would write
return a + b;But it does not play as well with yield expressions.
The only solutions I can think of would be to have a different keyword for yield expressions (await?) with the same precedence as other unary operators.
Anyway, this is only a small annoyance and I know it's getting late into the game. But I thought I'd raise the issue anyway.