quotable6

query

Promise .*? but

result-#moarvm.md

interesting, when running spectest with ASAN two Promise related test files has failures but ASAN does not complain. If I remove ASAN the failures disappear.
¦ MoarVM: break the Promise of the write, but instead cause the program to exit
the RT's on the other hand only say Promise broken, why that happened isn't mentioned but perhaps that is the way it should be?

result-#p6dev.md

Found nothing!

result-#perl6-dev.md

a summary of a paper about creating a control flow graph to detect bugs with Promises. it's for JavaScript, but might be relevant for Perl 6. https://blog.acolyer.org/2017/11/10/a-model-for-reasoning-about-javascript-promises/
One thread held the lock used to protect libssl against concurrent use and wanted to keep a Promise. The other was adding a then to the Promise, so holding the internal lock on the Promise, but GC ran during Promise construction, ran a DESTROY, which tried to do a lock acquisition of the lock used the protect libssl.
rakudo-moar 8d04be: OUTPUT«Tried to get the result of a broken Promise␤ in block <unit> at <tmp> line 1␤␤Original exception:␤ Type check failed in binding to value; expected Any but got Mu (Mu)␤ in block <unit> at <tmp> line 1␤␤»
rakudo-moar 8d04be: OUTPUT«Tried to get the result of a broken Promise␤ in block <unit> at <tmp> line 1␤␤Original exception:␤ Type check failed in binding to value; expected Any but got Mu (Mu)␤ in block at <tmp> line 1␤ in block <unit> at <tmp> line 1␤␤»
yeah, we could make that code change and put a little comment to where Promise has that attribute maybe? or the enum?

result-#perl6.md

uses Channel much less than Promise and Supply, but that's probably a reflection of the problems he works on
mine is something like await Promise.anyof(Promise.in(10),start {...await $c...}) but that inner await seems to allow the outer to unblock...or someting
m: my $S = Supplier.new; my $s = $S.Supply; my $p = Promise.new; start react { whenever $s { "foo".say }; $p.keep; }; await $p; $S.emit("foo") for ^4; sleep 0.2 # Seems to work... but I'm unsure if there are tests/docs that promise it always will.
(I mean, Promise.in is fine, but the .then is less so)
hello. I am trying to get the Pelr6 advent Day 15 – A Simple Web Spider With Promises to run. It failed initially till i changed urls_seen to url_seen at line 35. but whenever i run it with a domain, the results are blank.
I'm pretty familiar with Promises in ES6, but I'm not sure what the differences are. I suspect that there are some.
hi, does someone know if there is a library function to get the first Promise ~~ Kept but try all of them in case they all break? I have something, but I think it is a bit elaborate https://pastebin.com/DUV1X7Vf
Timer::Breakable isn't actually cancelling anything. It's just wrapping the block passed to Promise.in() with a check to see if the code still needs run. So if you stop it the internal promise still runs but does nothing.... this is probably a really bad idea.
lizmat: I was thinking... the way that is sad to make a timeout, is to use Promise.anyof($promise-that-should-timeout, Promise.in($timeout))... but it doesn't really times out... the promise continue running...
the Promise.in(3) internally most likely *does* fire, but the code block is not executed because it's no longer inside the react block
somehow I read the block as being part of the Promise, but it isn't
rakudo-moar 2fb8c7: OUTPUT: «Tried to get the result of a broken Promise? in block <unit> at <tmp> line 1??Original exception:? Too few positionals passed; expected 2 arguments but got 1? in block <unit> at <tmp> line 1??»
A Proc::Async has one exit code (Promise) but may write to stdout/stderr many times (Supply)
rakudo-moar 0095cd: OUTPUT: «(&infix:<+|> WhateverCode SIGSEGV Slip &callframe Pair &keys SIGHUP utf8 &RETURN-LIST &infix:<∖> CurrentThreadScheduler PromiseStatus StringyEnumeration &infix:<ne> Distribution &infix:«<=» &slip &splice &trait_mod:<handles> &callsame &GATHER Backtra…»
rakudo-moar 9e8ecb: OUTPUT: «((Slip) (PromiseStatus) (StringyEnumeration) (Distribution) (int16) Nil (Setty) (FileChangeEvent) (UInt64) (Rational) (RatStr) (uint64) (Numeric) (Instant) (byte) (Bool) (Associative) (Iterator) (uint32) (ProtocolFamily) (Rat) (Enumeration) (Blob[uint16]…»
rakudo-moar 5592db: OUTPUT«((WhateverCode) (Slip) (Pair) (utf8) (CurrentThreadScheduler) (PromiseStatus) (StringyEnumeration) (Distribution) (Backtrace) (NFKC) (Duration) Nil (Any) (WrapDispatcher) (IO) (FileChangeEvent) (UInt64) (Order) (RatStr) (uint64) (Numeric) (Cursor) (Proxy) …»
rakudo-moar 5592db: OUTPUT«Doing Signal␤Doing WhateverCode␤Doing Sub␤Doing Pair␤Doing Signal␤Doing Sub␤Doing Sub␤Doing utf8␤Doing Sub+{<anon|71898608>}+{Precedence}␤Doing CurrentThreadScheduler␤Doing PromiseStatus␤Doing StringyEnumeration␤Doing Distribution…»
rakudo-moar d9c735: OUTPUT«(SIGSEGV WhateverCode Slip &callframe Pair SIGHUP &flat &RETURN-LIST utf8 &infix:<∖> CurrentThreadScheduler PromiseStatus StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace PF_INET6 &sinh &infix:<lt> &print &asech NF…»
rakudo-moar d9c735: OUTPUT«WhateverCode␤Slip␤Pair␤utf8␤CurrentThreadScheduler␤PromiseStatus␤StringyEnumeration␤Distribution␤Backtrace␤NFKC␤Duration␤Nil␤Any␤WrapDispatcher␤IO␤FileChangeEvent␤UInt64␤Order␤RatStr␤uint64␤Numeric␤Cursor␤Proxy…»
about failures from #perl6-dev: https://gist.github.com/anonymous/bfa6d799ccc890be63eafa5ea75009c2 + backlog copy; it's surprising Promise do not throw just like Failure. but nice :)
Plus, in Haskell you get things like Software Transactional Memory, which, together with constructs that work a bit like Promises and the like, give you a most-robust way to do something like variables, but safely, so that they _will not_ go wrong.
ZzZombo: it's a check for whether a Promise was Kept or Broken (but can be a simple bool)
Hello all, is there an object like Channel that may fail on send, what I mean: is there an object like Channel where send is a Promise? A bit like Socket::Async but without the Socket implemented
rakudo-moar 5f9103: OUTPUT«Invocant requires an instance of type Promise, but a type object was passed. Did you forget a .new?␤ in block <unit> at <tmp> line 1␤␤»
m: say Promise.^attributes
I was hoping I could do the equivalent of (&f1, &f2).map(&Promise.start), but it seems like a non-trivial thing to write
AlexDaniel, I'm not well read on Promise, but the docs say that start() returns a Promise that is kept by a Proc object.
rakudo-moar 02fdcf: OUTPUT«(SIGSEGV WhateverCode Slip &callframe Pair SIGHUP &flat &RETURN-LIST utf8 &infix:<∖> CurrentThreadScheduler PromiseStatus StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace PF_INET6 &sinh &infix:<lt> &print &asech NF…»
The second half of it is tossing `await` and instead shooting off a new Promise inside of which the callable calls will continue. Basically the equivalent of wrapping the whole thing in `start {}` but doing so only if at least one Callable returns a Promise.
This is my first attempt to use Promises. My search for a longest substring can work independently over the various string lengths, so I thought I might try to parallelize this. It's only going up to 20 instead of the full length because it's running wicked slow and I wanted to see if it would indeed process things not in order. It does, but am I doing something wrong to make it so slow?
rakudo-moar 04f0b2: OUTPUT«(SIGSEGV WhateverCode Slip &callframe Pair SIGHUP &flat &RETURN-LIST utf8 &infix:<∖> CurrentThreadScheduler PromiseStatus StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace PF_INET6 &sinh &infix:<lt> &print &asech NF…»
Hey all. jberger is trying to calculate pi using monte carlo method. I tried to write it with .race, but it was slower than single-thread way. Now, I rewrote it with Promises, but it *still* doesn't offer any speed improvement on my 4-core box. Any idea why? Is my code wrong: https://gist.github.com/zoffixznet/ae7bcf0efc99eba8eaadbd917238ff1a
jnthn: Is is by design that when an exception occurs inside a `Promise.start/then` block, the backtrace does not show the line-number and routine-name where it occurred, but only the line where the Promise ended up being awaited?
rakudo-moar 1d1256: OUTPUT«(SIGSEGV WhateverCode Slip &callframe Pair SIGHUP &flat &RETURN-LIST utf8 &infix:<∖> CurrentThreadScheduler PromiseStatus StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace PF_INET6 &sinh &infix:<lt> &print &asech NF…»
i.e. osx has some problem with `await Promise.allof(@promised)`, but `for @promised { await $_; }` works, with the difference being the first one doesnt seem to sink all the promises so they never get out of Planned state
azawawi: "phantomjs returned Proc::Async promise: Promise.new(scheduler => ThreadPoolScheduler.new(initial_threads => 0, max_threads => 16, uncaught_handler => Callable), status => PromiseStatus::Planned)" but it's using the wrong port
rakudo-moar 775271: OUTPUT«(SIGSEGV WhateverCode Slip &callframe Pair SIGHUP &flat PromiseStatus CurrentThreadScheduler &infix:<∖> &RETURN-LIST utf8 StringyEnumeration &splice Distribution PF_INET6 Backtrace &GATHER &callsame &postcircumfix:<{ }> &print &infix:<lt> &sinh &GENERATE…»
rakudo-moar 42a583: OUTPUT«(SIGSEGV WhateverCode &FILETEST-X Slip &callframe Pair SIGHUP &flat PromiseStatus CurrentThreadScheduler &RETURN-LIST utf8 &MAKE-DIR-LIST &infix:<∖> StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace PF_INET6 &sinh &…»
llfourn: I can't Promise, bad pun, right now as I need to wrap my head around this and use it. but I will eventually (after my move to another town and the build of the ground of my summer house, not too long)
a quick style/taste question if I have a method that will return a Promise but it can detect that it just won't work up front is it best to create a new Promise and break it with an exception or do the Promise.start anyway and throw the exception in there?
await Promise.allof($p1, $p2) # should do it, but seems allof throws if one of the promises fails
rakudo-moar e6f360: OUTPUT«SIGSEGV WhateverCode bit &FILETEST-X &callframe Pair SIGHUP &flat utf8 &TRANSPOSE &MAKE-DIR-LIST &infix:<∖> CurrentThreadScheduler PromiseStatus StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace &sinh &first-index &…»
rakudo-moar a6bb0b: OUTPUT«(SIGSEGV WhateverCode bit &FILETEST-X Slip &callframe Pair SIGHUP &flat PromiseStatus &RETURN-LIST utf8 &TRANSPOSE &MAKE-DIR-LIST &infix:<∖> CurrentThreadScheduler StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace &…»
Suppose I want to make a Proxy that backs a Promise. I can implement STORE to replace what it will return with, but I can't really implement FETCH (unless I either die, or freeze, or something). But, it would still make sense for me to do something like += 1 on it, but I can't do that if I can only define STORE/FETCH. So maybe MODITY/FETCH makes more sense, and storing is just a special case of MODIFYing
rakudo-moar a6582d: OUTPUT«SIGSEGV WhateverCode bit &FILETEST-X &callframe Pair SIGHUP &flat utf8 &TRANSPOSE &MAKE-DIR-LIST &infix:<∖> CurrentThreadScheduler PromiseStatus StringyEnumeration Distribution &splice &postcircumfix:<{ }> &callsame &GATHER Backtrace &sinh &infix:<lt> &f…»
not sure if it's exactly a Promise, or just a block that'll get called the first time you access it, but yeah
any really - you can get plain cue({}) and cue({}, in => 1) from Promise but every and times not
i'm not sure what kind of async library one would look for, with Promises, Supplies and start/await in core, but that might just be because i'm not particularly knowledgable
all accesses *should* go through the .vow method in Promise, but they don't have to
I'm reading up on Promises, and I wonder: How do I break a Promise? I tried "$promise.vow.break;" but I'm getting: Access denied to keep/break this Promise; already vowed
moritz: When I run "my $breakable = Promise.new; my $p = Promise.anyof(start { sleep 5; 42 }, $breakable);", I'm getting: Too many positionals passed; expected 1 argument but got 2
yer welcome :) wrt alarm(), I'm pretty sure we don't have it in lieu of higher-level timing and async stuff like Promise.in()...could get at it w/NativeCall if you really wanted but then you have to make it work on both posix and windows again
raydiak: Actually, I was asking about alarm() simply because I wanted to use a Promise, but reading from STDIN in a Promise is a bug in Windows Raduko.
It makes a Promise object, but shouldn't "for lines" never return?
mickcy_ca: as for the consumer, i'm starting a Promise that contains the while which emits - i think .list or .from-list could work as mentioned with gather/take, but i didn't figure out how that'd work
Promise tries to type-check its arguments, but doesn't success
Does it break the Promise, or does it keep the Promise but the result is a Failure?
suspects he's observing Promise.allof spinlocking or some such. 100% CPU usage, but no input is coming in :-/
I'm very, very confused. If I do «await Promise.anyof(@threads, Promise.in($short))», and in one of those threads call sleep($long) or Promise.in($long) (or anything long-running?), the anyof doesn't return for the Promise.in($short) but blocks on the entire thing...
In S17 the documentation for a Promise refers to the excuse method, but Promise.pm defines cause.
method Promise() { my $l = Lock.new; my $p = Promise.new; my $v = $p.vow; my $t = self.tap(-> \val { $l.protect({ if $p.status == Planned { $v.keep(val); $t.close() } }); }, done => $v.break("No value received"), quit => -> \ex { ...similar to more code but .break }); $p }
or do we need to change Promises into Channels, and use the .closed attribute on the Channel
the only thing i could imagine is Promise.care giving a handler that you can call .forget on and when the last .care has been .forget'd it would redeem ti, but that seems fragile and racy
lizmat: You need to await that Promise.allof, but maybe somebody managed to break allof too...
Promises are a bit of both. You can poll for the result, but .then is reactive.
TimToady: I originally had it as Promise.sleep, but lizmat wasn't keen on that...I can kinda see why-ish...
japhb__: Promise is getting a fairly big refactor but it's largely "on the inside" so I'm hopefully it'll not regress anything you're doing.
There will also be a way to get a "keeper" object from a Promise that lets you encapsulate the right to keep/break it, but it's optional
Thinking about it... The "Promise" analogy works well for read operations, but gets slightly weird for write operations...
jnthn: so Promise.sleep(1) makes a promise that doesn't do anything, but takes a second to do it?
Ulti: It's a Promise, but when you try to use the result it throws the exception.