Node.js Node-API Team, 4 Nov 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Node.js N-API Team meeting - Nov 4 2019

Description

https://github.com/nodejs/bootstrap/issues/114

A

Distri I may not be able to stream it, because apparently you can only do that using the app and my app doesn't work. That's why I just had to rejoin using the browser. Okay, so it's gonna be sort of a shortcoming, but it looks like oh yeah.

B

We're just talking about next week's meeting I'll be away Gabriel's going to chair and everybody welcome to the N API team meeting for November 4th 2019.

B

We will look at the issues that we've got tagged for the milestone before we do that. Does anybody have any I'm announcements they'd like to make.

C

Landed a single progress work, so this is the next one implement to work. It has a slightly different with the previous one.

C

Much different.

B

Right I think I'm, not sure we would hold up for the release on on that. But I did wonder on that. One I haven't looked back to our past discussion. We looked at the different async worker variants and there was only a subset that we thought we needed to pull into into an API that that one, but I can't remember which them does anybody. Remember any of the more context on that discussion.

A

I remember something like we wanted the the progress worker because Dan has it, but otherwise they can stay in a library unless we really really need them or something like that and I believe correct me if I'm wrong, but I believe Nicola you were. You were maintaining like a library of like node API extras or something like that right or somewhere it pops up like we have like an old API extras at some point: yeah yeah.

D

E

I'm working on an extension of nappies, 3, yeah right.

A

Yeah, yes, that.

E

There is a very old issue where someone asked ad for this: I am.

E

Okay, the five five, the issue: five, five, another donate bi, so yeah I'm working on the is.

E

Cannot be extras now it's an experiment and.

B

Tory said: yeah yeah, five, five.

E

B

F

E

Very old, so yeah.

E

E

Yeah I I'm trying to to create these this extension to four-foot for the string and yeah. We can do the same thing with other stuff that we, maybe don't want, add in the core of immediately.

C

Then I think we didn't want to lend a unique variant of a stink rocket. Then the issue is to number.

C

82, it had to be had to be resolved, since we can't say over the homework complete of the text.

E

582 yeah I've.

B

E

C

This is now define a static, critical static method, so.

C

Possible to be overwritten.

B

Right now, where does async progress? Cue word for come from.

C

I have implemented in the committee cubic appear, so if we didn't intend with this focus cube, then I may be to pick these changes right.

B

C

I'm in sixties is there progress.

B

Cue worker in n, yeah.

B

G

B

We said we'd take the other one right so.

C

Like as a migration.

B

It can make sense.

B

Because this is the original.

B

D

Think in this issue they were requesting some sort of async promise a sink worker.

G

D

Sort of humus, variant and Nan does not have anything like that. So then he said: okay, if you were trying to create something like that, that would go in an external library, but now man does have an async progress worker, okay, progress, key worker! Excuse me right.

H

F

H

D

That's what he's trying to implement here right.

B

Okay, so that that's what I was trying to sort of understand. So if this is say, text yeah.

D

B

Know this just like the the biggest one is in then, and we want it makes sense to provide it as a migration path. Then that's. The kind of thing we have in the past said yeah. It makes sense the whole lid right.

D

So then I think that's what he's referring to and his implementation now of the async progress queue worker and then he he brought up those issues saying that those methods are static, private and could not be overridden Lucas.

A

I have a question so for for the is the difference between acing progress, a worker and acing progress queue worker simply that, with a seen, progress queue worker for for for every progress call, there is a call on the native thread, so there is the difference only that in the case of acing progress queue worker. There is no coagulation.

C

The difference might be the itch itch sent. Progress code would be killed to ensure each progress and progress already calling the JavaScript.

A

Thread, bread, yeah, okay, okay, so so then so then, basically is is, is is not the only different? The fact that if you, if you do a non-blocking call and and and it returns cute fool, then you have to handle that right in the case of acing progress, work or you don't have to handle that. But in the case of acing progress queue work or you have to handle that because you you have to keep trying until it doesn't say that BQ fool anymore right, I,.

C

Think there is no useful situation, the men they don't have. No.

A

No, they don't; no, they don't but- and we won't have one either in in the public. Api will not have Napa queue full, but internally acing progress worker is implemented using a threat, safe function right now. This frets a function can report nappy cue full. If you do a non blocking call. Yes,.

C

Yes, I think progress worker didn't bring the cue mechanism provided by a function. That's right.

A

So so that one didn't use it, but this one must use it. However, what what I'm interested in is like? Why do you need? Why do you need to change the the the the there's methods from private to public right, because you didn't need that for the progress worker and handling the return status of like of of that PQ full can? Is that why you need the to access? Those methods is because you're handling the status I.

C

Don't have any reason to just now using the kill mechanism provided by a tariff event, I just implemented them as what it was. No, oh.

A

I see so you're not using thread-safe function for this one yeah.

C

It's not using the Q macadam what yeah.

B

But if we used it in the progress worker, maybe the same approach should be used in the progress key worker as well.

C

Provided by charity function, it just used as a method to to check the actual progress, send.

C

C

D

Are you saying that you you just have like a the there is no queue size, so when you create the thread safe function, the maximum queue size is unlimited.

C

Maximum Keota is one and we ignored each. Let's say that you, for we don't care if it's a queue before we just we just clean each clean the queue on each function, execute on the transcript data, so.

C

Yeah, that's how we work in progress, work.

A

Right but but in this case, I don't understand in this case, this is like in the case of a acing progress queue work. Why are we not using the queue I did buy? It's a function.

C

Yes, the one I didn't tell the function. Is that and then provide don't limited me mm so.

A

C

A

See what you're saying, but you can achieve you- can achieve the the the the unlimited queue size. If you, if you, if you handle the the the queue full message right, because because basically, what what you're saying is you you will you will not you another right? Oh wait, wait a second. If it doesn't limit the queue size, then why don't you use a queue size of zero because it is.

C

A

Means unlimited queue.

C

C

A

Right, cool, okay,.

B

Okay, so this was the top is the first issue sounds like we've always already had a bit of a discussion on that one. It sounds like this it because it's in then are there others and then as well like how many more are in then that we don't have in an API.

D

I'm, seeing that there's another async queue worker, which is both executed, I think.

A

Yeah I think you were yeah. This is the one.

A

A

Yeah, that's what you're! What is that, so that that is like basically, threats a function.

C

A

Mm, it doesn't say much about it, it just says, run.

G

A

Worker, a sting has to be a little UV, both the execute and after work steps are taking care of you most of the logic, for this is embedded in an async worker. So, okay, what is what is the value? Add.

D

Unclear I definitely have to look for examples in the wild of how people.

B

D

B

Because that's the other question is: how often are these used? Yeah I mean.

A

I mean these millions, don't.

B

Want to fill any pie right, yeah.

A

Yeah I mean these the the the ones that were working on now they, basically they were requested right like where is it nan? Has it? How come you don't? Have it right right, I, don't remember this one coming up as as one that was requested, so I think I think maybe for your. If we're sort of passive about it, then you know we can implement it. If and when the need arises right.

B

So, do we get a request for the key worker.

C

The request of cookie I just come on and then to so I party.

A

Yeah, this is progress, queue of worker that you ported right, but acq worker I know.

B

I was but I was asking. Similarly, how did we get an a request? Oh they had people asking for the key worker. Well,.

A

Well, I mean yeah, I mean we did. What was it that got us started on these variants in the first place, right like we did right so.

B

We had this issue right here, which was for 73.

B

Streaming workers so a sync process, so this is like where we had somebody say: hey I'm, missing this right.

A

Right, oh yeah, that's right! Okay, so we just so for yeah, okay! So for progress worker we had. We had a request: okay yeah, so we haven't had one yet for progress. Cue worker looks like.

C

It's for streaming worker I think the progress queue work might be a more.

C

Suitable you couldn't see it, you know ensure each Co. The kind of progress will be ensured to go into a details. Good.

A

Right, so you can do that, for you can use it for streaming. That's true.

C

Anyway, I think the Ichiro 582 may have to be resolved to ensure we could have more variants in the world. Then we can't who can't override a.

C

Something before we can work complete better, so maybe I could split. I.

C

B

Think it was good it'd, be good to have it in to PRS just to understand. What's the implementation of async progress, cue worker, and which is something you think, we're fixing a sort of an extension problem.

C

B

It gonna break any existing code, no.

C

No, no, no! This has to be comparable. Okay,.

B

Okay, so anything more, we want to talk about in terms of async worker variants,.

B

If not, let's move on to 367, which is the workshop I.

B

Know Nicola Jim. If you want to take the.

D

Yeah regarding FETs a functions I began a sort of an introduction that goes through. I haven't posted it anywhere here, yet do so in the middle of the week, I guess with a brief introduction of how the thread-safe functions work with you know how, regarding the initial thread, count requiring releasing context and calls I want to how should link that up here, I've tried cleaning up a bit and Jim lad, because yeah I messaged.

G

D

Earlier to get that link from each end, well, I saw that there was nothing really up there. So I sort of took my liberty with what I have you know that to you all.

G

Right I really appreciate that Kevin. Thank you.

G

I am on the road right now, so I'm, not confident how good my internet connection is right now. Can you guys hear me yep.

A

G

Okay, okay, very good, okay, so Kevin! Thank you for that and I am continuing to work on the other sections as as I'm traveling Harry.

A

Okay, I have a quick question: I I added two slides about context, awareness, but I don't see them now did I add up to the wrong. No.

E

I I see I, see this right to strike.

A

So so, then is the link is the link at the top of issue. 367 that says slide. Is that the correct link then no.

E

A

Torrio's or list of candidates, no, no.

B

H

B

One it says 14 days ago, there's a link yeah.

E

And and you this is right in the right place, so oh yeah.

A

Okay, got it got it yeah yeah.

E

And 21 and I I created if you can take a look at 23, 24, 25, okay,.

G

E

Maybe I missing something and.

E

E

Everywhere, I have to change and yeah here, I.

E

Added what is in a na pi/4 and Apia v? That is the core, and the next slide is what we did in. No, they don't api.

E

And the last 25 is what we did about the tools so specific, rebuild and see make.

B

Candles definitely look good and.

E

Maybe I, if you agree about to the content, I can pass this ride to my friend. That is a designer that.

E

Say say me that won't just make a design more beautiful. If, if the content is good but yeah, we can.

B

F

B

Shop schedule Grayson a divider on if.

E

You want to remove something. It's it's good for me. Just.

B

The right, okay right context, awareness.

F

B

Brief history of any god, I, just wonder if this, what, if like 26, if.

B

Like 28, 27 and 26 should move up to be.

B

Like somewhere,.

E

Yeah because yeah in the at the beginning, I I think that.

B

Like maybe it's that they should just move above those new slides, because we talk about what is an API context, awareness and then like before. We talk, because when you get into like what happened, this no.

G

You know I talks.

B

About we should probably have V. What is that so I could move, though. So, let's just move these yeah.

B

Right to before that, and probably this one on man goes before that too right.

E

Yes, because and yeah, even if the code yeah, this is like the 20 career, because yeah yeah.

B

Yeah, those two yeah yeah, sorry, so that I meant to move those I just didn't yep like.

A

B

And then I might say.

B

Like wood, making moving the this down to just before the tutorials- oh okay, just to say, hey, that's kind of like now that you've done the whole thing. Add your badges right, so we kind of have the flow of like okay, who the people talking objective schedule. So what's not on any P I context. Okay, brief history, then add on know that on API, which is good a little bit a bit like what versus? How does that play versus man? Then?

B

Okay, what we do this year, a little bit of guidance on how you organize your project and then how to add, badges and then into the tutorials itself, right toriel's, which will up-to-date once we get the list and then supporting resources. Yeah I think that pretty good content, yeah.

E

E

I am on the slide: 38 I I.

E

E

Resources that we don't we don't have the last time, yeah yeah I, just see, make and yeah project near that now start the to support an API.

E

So you can write native a done using Russ and then a PR and a Jane, a P that yeah.

E

Is of projects that help to creator Rob.

E

E

Yeah, so it's good for you to come leave.

B

Look good to me.

E

B

Okay, so that certainly like yeah, if you want to get some help to make it look nicer, that sounds good, but I think the content is looking pretty good.

B

Any other comments discussion on this one before we move on.

E

Know about the slide but I'm working on adding other add-ons on the exam on the excel file.

E

Because because Vince and Weaver's.

E

Okay, you need the to go on the.

E

Eight seven on the Abby, stable, node, Michael, okay,.

B

Api stable so 387.

E

Yes, three, eight, seven and I think that Jim I need to be promoted again. How.

B

Let me do that, participants right. Oh, we got something else too: cool, okay and.

H

G

B

E

Here in the last government, Vincent.

E

Reported a good list of a native, a dog that does not that don't don't have any P I porting and the filtering these these lists, because yeah some of some of these has a branch with an API but or maybe already have apparatus so yeah I'm I'm, going on this list to try to get other other naked on to propose at at workshop. Okay.

B

Sounds good okay,.

B

Anything else on that before I move on, it's like we're, making good progress on the workshop sure.

B

Okay, the next one is yes to a light port Jim anything new on that front. I.

G

Don't have anything new on that.

B

The next thing is the burn downs of list of issues, so any specific issues. We can take a look at what's going to the nude the next release as well, but any issues that people think we should call out.

H

H

E

On loaded on API, the it's, the the last issue.

E

Reported and yeah this guy said that there is a difference between.

E

The file on the module publishes on NPM and what is in the the github repo I H I, did check I check it out. The branch and yeah I tried to to do a deef, but I.

E

I didn't find the the the the any difference between what is in non models. If you install the the the know, the the package or and the the branch.

D

E

D

Did the same thing when I saw this issue by checking out the branch packing the branch and checking how it compares an NPM install and the only differences were in like the package.json yeah.

G

D

Because it was formatting of the package.json, there was no change in yeah.

G

D

Or anything yeah yeah I, don't know, I was afraid that I.

E

Did a mistake so yeah thank you Carrie, so I have the same result. So I think that this is not a problems. Okay,.

B

I will see what if he comes up with anything specific but otherwise, because those die I assume that those like the changes to the package Jason was done as part of the publishing tool. Yes,.

E

B

That formatting, when you publish it so yeah.

E

Because when you get the the you installed, the model the packages on is is different, so it's yeah. If.

D

He's just looking at like the integrity hash that is different because the package.json formatting is different. You know yeah I. The actual contents of the code is the same. Should we but book or I I, don't know exactly how the packaging is done in order like what why it's different, when you do NPM pack forces what's packaged on the I'm, not sure what.

B

If you did it, if between the package.json, what's different between the two of them, it's.

D

Just like, like the contributors section, it's like one line per contributor versus being it's like the the depth of the JSON right yeah. That's, basically, the difference.

B

I'm just curious whether we should make that change in our so that it does not yeah.

E

I I tried to check rapido stage nappy, you know no tej, a nappy, the deprecated oh and index Dodge is, and it's.

E

There are no difference so alright, no.

B

But I'm just I'm just wondering is it: does it make sense for us to take the formatting that they convert it to and blind that in the repo just so that next time there isn't any difference.

D

Yeah I could see. I could see that being okay, but then again it's really not a problem. You know yeah.

B

Unless there's some expectation like if somebody says hey I, want to check that what's in the repo matches, what's in the release, yeah like it's, not a reasonable expectation that people look and do people look at that regularly if it it's from my perspective, it's like, if we can somehow make it less likely that people look at it and say: hey, that's different and it's not very hard. Maybe it's worthwhile.

E

E

Only on the package we can, we can shape the package JSON like NPR yeah,.

B

Yeah, unless, unless there's a reason that doesn't make sense or whatever I I say why not yeah, we can.

D

Probably do that for the new, oh release, exactly.

B

Like I wouldn't worry about it for the existing one but yeah we.

D

Can say the new one.

B

Will do that so yeah? If that's what it turns out to be I, think that would be the only action if there's something else that we can see.

B

F

Other thoughts or.

B

Any other issues we should.

B

Many other the new issues or the.

D

New issue for the the gate, context, threats, a function, gate context, failure, I'm issue; five, eight one, a force.

A

For ninja list yeah.

D

Yep this is reported and there's a PR out.

D

For this as well, but it was basically a simple, simple fix. The only thing was that, since the gate context call can be done in any thread, it does not have a a nappy environment nappy end. So none of the air. You can't like create a javascript error to be caught or anything like that. So the only way that we could properly do an error that I saw was doing a fatal error which would you know and I think Gabe. He saw that and he said that was okay, yeah.

A

Yeah yeah yeah, there is no other alternative and and I looked at the implementation, forget context, and- and it doesn't really it pretty much returns- nappy. Okay, unless you pass some unless you pass a null for one of the parameters and since we control the parameters, because we don't directly pass anything through, you know we pass internal objects into the call. Then it should never be null. So we should be okay right.

D

Okay, so then, this here this is this: implementation I have to update it with the new master. That's been merged with, since the TS FN copy constructor was merged, so not to update that right. Okay, then there and the existing PR from last week, though, is the blocking call and non blocking call addition for I'm calling the TS FN directly. Oh yeah,.

B

That's 580, yes,.

D

580 I'm pretty sure the gate context will have conflicts with this one. Regarding yeah, regarding like the tests, you know because the adding the same lines right around there, but on so if this one could be reviewed and merged in eventually, because I guess this one we had listed as a nice to have for the the OH release.

A

Kevin for for for for for the regular threats a function, are we using the context in any way or are we exposing it? We okay.

D

Context via the data callbacks that you send yeah they it's part of the parameters for your callbacks, okay,.

A

Okay, but so okay, when you expose it, we don't daisy chain it, though right like. If you give me, if you give me a pointer as your context, then it is that pointer that I give you back right, not not I store. Some I store, your pointer in one of my my pointers and then I store that pointer and then, when you get context, I retrieve my pointer and I retrieve your pointer from my pointer.

A

Do we do this hip hop or or do we store whatever pointer we get from the user directly in the context.

D

If I think there is a wrapper around it, because it's this convertible context struct that, yes, has this operator overload to be able to be cast to any type. Okay.

G

So imagine that yeah.

D

We have to create a new object for that.

A

Okay, but but that's that that would mean that a threats a function created using using an existing object. You know its context will not be.

G

A

This type, so when you call get context you cannot, you cannot assume that it's an object of that type, so so the the API is starting to diverge at this point right, because in some cases you can assume you have to remember that this was created directly and therefore don't try to cast it to this convertible thing right, just pass it straight through so so there's an additional special case that needs to be made in get context. For, for, for this direct call right.

D

Okay, let me verify that that we are creating that structure. I believe we are I can't remember off the top of my head, but yeah. That's a good point. I'll take.

B

This, oh, but yeah, I think Gabriel. If you could take a look at 580, yeah.

D

I have 580 the underlying call, yes, mm-hmm.

D

Yeah, that's true. Okay, now take a look.

B

And the other one, do you want to bring up four I.

D

Have nothing else? Okay,.

B

Unless anybody else has something else, let us look up the next release. One.

E

Yeah, maybe we add another task above Sparky Jason.

B

Better, let me just.

D

Okay, I just looked at the thread-safe API, you know the the car I was incorrect. The context pointer is passed directly, as is to the.

A

D

A

D

A

D

It's the get context, C++ method just returns that convertible context struct. Oh.

A

Okay and okay, so, let's see okay voice, start context: context: okay, okay, all right; okay! So so we sort of wrap it when we return it, but we store the actual pointer we get from the user. Okay, that's good right! It's perfect! That is perfect!.

B

Okay, so I think like on our list here, we've added the one about updating the package Jason, the only other one on the list that we haven't completed is 580, which is the one we just were talking about. I think yes, so I think Gabriel. If you could take a look at that and give it.

D

Your thumbs up.

B

We can get that landed, okay,.

A

E

Have a curiosity about the recommendation in the core, for example, if I go on navigate today to value I cannot see the nappy birthstone.

F

E

F

The last the lost okay.

F

So which so, if.

E

You go yeah, maybe you search for nappy got date back you.

E

That is okay. Here are dead in academia and we don't have the nappy version 5 and it's it's it's for all the the version 5. So oh yeah.

B

Right: okay! Well guess we just missed that yeah! We.

E

Said the key new yeah I'll.

B

Go and add that that's I can do that pretty easily yeah.

E

Yeah because he died, I was searching about what is in okay.

E

E

B

E

Yeah he's another strange thing about the wind Asia.

E

Okay in the arrow rendering we have that the struck to not extend the juror info he's added in an API version, for is it right, because on top we have that these destructor, his other, did not be Russian. One yeah.

B

That doesn't look right either I'll double-check that but I think cuz that extended oh yeah, yeah.

A

That was all very beginning: yeah.

B

Okay, yeah I can fix that one as well.

E

Okay, this is that and.

B

Even this, an added envy 10.6 doesn't make sense. Yeah.

A

That's way too way to resent.

B

Like it would have needed to be like it should be, probably nappy version 1.0.0 right, yeah.

E

Yeah yeah yeah.

B

I'll double-check that that's the right thing, but then I'll update it to that cuz. That's pretty much what it should be: okay,.

B

Any other things: oh, no, no, no, okay, okay! So for the next release, yeah. If we got that one thing and I think we're, then pretty much ready to go.

B

Well, I think we, you know, we got a good set. We probably shouldn't wait to add anything else in we can, whatever whatever we do, can be in the next version.

B

Okay, any other things, people think we should talk about this week.

D

I think there was a request to see if we could move the meeting time. Oh yeah.

A

D

B

What times do people have in.

B

What times do like are we is the day. No good is at the time.

A

B

A

Pretty late for you right so.

C

A

We move it. I I would be okay with moving it up one or even two hours like I'm, okay with 8:00 in the morning. I could.

G

Move it up two hours, also, which would be eight o'clock for me. Yeah.

B

Okay, so that would be at 11:00 o'clock Eastern. Yes, yes, okay! Let me just check my calendar, though.

E

It's two hour: yeah.

A

Twice earlier yeah, it's probably.

E

A

You as well right, yeah,.

E

Yeah yeah right.

B

Yeah yeah, as long as it's not too early for the Pacific people.

E

Cabin you have the same time right I'm in Berlin, so.

D

B

Actually, and surprisingly, that's works for me as well, so, okay, let's do don't move it two hours, so I will change it as we speak.

D

And then Nicola, if you have the time in the slides, you may need to update the slides.

E

Sorry, Kevin I I didn't.

D

Hear you, if you have the meeting time in the slides that you have you'll, have to update this your slide for.

A

The workshop the.

D

E

B

A

B

Okay, so now wait a second.

F

B

Change it in the right calendar, so we want to change it for the 11th right and we will make it so it can fall. Wait a second conflicts with the security working group meeting.

B

Wait because we can't have two streaming at the same time right now. That meeting is only every so often, let's just see.

B

Group meeting was this week, but.

B

Yeah, so it's like one: it's once a month, yeah.

A

So, on those days I guess we can do. We can do like an hour only an hour earlier and then every other time. We can do two hours earlier than today, because I I don't see any meetings for between 9:00 a.m. like I. Don't see any meetings immediately following the the Security working group meeting or we could do it three out of four weeks: oh yeah, okay, yeah and then just skip skip skip the week of the Security working group meeting yeah, that's probably. Okay. All.

B

Right I guess yeah like we can, whatever people prefer I.

D

Think it would be better to have it all at this, always at the same time, I. Don't.

B

Kind of what I'm thinking.

D

Too so I would rather have it be one hour later earlier, instead of two, so that is always at the same time or do the two hours and then skip the third every other once a month, I think skipping once a month is fine, because then you but I.

A

Just know my my mic, my alarm clocks, gonna, wake me up anyway,.

A

B

Like I'm, looking at the I think because one one hour earlier, I'm probably have a conflict, I think there's every other week. I have a conflict. Now the.

G

The other thing is, if, if we back it directly up to that other meeting and we have to get off the streaming or vice versa, you know like if we go over a little a it's, not an issue now.

D

G

Be the three and skip one.

B

Yeah I think that I mean I think that would probably be fine. We can always like if there's something we really think we need to discuss. We could do a meeting without the stream, but yeah yeah, okay,.

G

So I'll schedule.

B

That for or if I can easily do that, as so that is going to be 11. My time.

B

Repeat every week, I think actually I'm just gonna have to make it every week and we'll just cancel if it's a conflict, because I don't know that I can easily schedule all, except for that one.

D

Right you'd have to do three different.

B

Meetings and even not I'm, not sure, we'll work out where, like where you have five weeks or.

D

Something right.

B

It might but let's yeah, why don't we just schedule for that time and then because it's only once a month we'll just handle the once a month ad hoc, yeah, okay, okay, okay, that's good any other things! We should talk about.

D

Nothing here, oh I'm,.

B

Good okay, great thanks for everybody's time and we'll talk to y'all later later, bye, bye,.