Jupyter Lab Team, 28 Sep 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: JupyterLab Team Meeting - September 28, 2022

Description

A meeting to share and discuss features, ideas, issues, and pull requests in JupyterLab and other Jupyter frontends. This meeting is open to anyone and everyone.

Join future calls via the Jupyter community calendar: https://docs.jupyter.org/en/latest/community/content-community.html#jupyter-community-meetings

Notes for upcoming meetings can be found on the agenda: https://hackmd.io/Y7fBMQPSQ1C08SDGI-fwtg

Past notes can be found on the JupyterLab team compass: https://github.com/jupyterlab/team-compass/issues?q=is%3Aissue+label%3A%22Dev+Meeting+Minutes%22

A

Okay, hello, hello, everyone welcome to the Jupiter lab weekly call, it is September 28, 2022 and I did have to look that up before the call, because I was not confident, but uh today we have 15 participants on the call at the moment. Hooray, that's awesome and agenda I'm, going to link that again. Real quick, please sign in add anything there. I see, we already have a few items.

A

If uh you think that your share is going to be about or longer than five minutes, please add it to the further discussion section, so we can get to it at the end and still make sure we make time for everyone earlier.

A

um Also because I'm, a nerd I'm going to remind you that this is a Jupiter Community event. So we do have a Jupiter code of conduct at jupiter.org conduct. That applies to any of these events at all times, and it includes me as well and yeah with that I think we're going to get started. Oh actually, before we get started. The one last thing I want to do is say I'm hosting the call today, because uh we're trying to rotate hosts. You can find more about that on that issue.

A

uh Please feel free to sign up if you want to co-host or anything I'm, just gonna bump that again now we're actually going to get started and on the agenda. I see Frederick here, first ready to go what.

B

Sure uh so, hello, everybody, uh sorry lots of bullet point. uh First one is an announcement for the Jupiter council members. You should have received one or maybe two emails depending on your spam policy. Sorry for that uh for voting for The Jupiter lab software steering committee representative, uh I'm, sorry I did that a bit late, so I moved the closing day for the election for October. 4Th I know that initially we try to Target October 3rd, but to let at least a week for people to vote.

B

So I decided to close the vote up on the fourth so next week we should be able to announce who is the representative for Jupiter lab. So that was the announcement as it has been discussed a couple of times now and in all Jupiter projects. uh Gita two-factor authentication are gonna, be enforced on Jupiter lab GitHub organization.

B

I propose to enforce it right after this meeting, except if somebody is a very good reason for me to delay that, but I think no, we have let everybody at the time for doing action about that and I think everybody, that's still active as activate that feature, so should be fine. So that was for the second bullet point.

B

um The third one is actually just a question, so there's been an issue on the extension cookie cutters that we discussed like I, think it was a month ago now about changing the license for the cookie cutters. The idea would be to switch it to cc0 I think it was Alex. The last time that uh was asking about that and the question I have is like: uh should we could we move forward or do we ever need to like get the unanswer for every authors that have contributed to that for changing the license?

B

Or can we just move forward? Presuming that people that haven't looked at that issue or having say that there are they don't they are not in agreement with the change of license are in agreement with it. Like that's, that's the question and then I I come back to the to the next point.

C

I don't understand the implication of this license versus BSD.

C

But I think what this is is it's more permissive and allows you to basically pick any lessons you like. Is that the reasoning behind it.

D

Yeah during our previous discussion, it's the equivalent of uh anybody can use it under any license. They want without any questions, it's a it's almost like it's unlicensed or uh it's the equivalent of like, uh like uh what's the term when music is free for everybody to use uh public domain.

D

Okay I see um yeah, it's a very intentional saying. Yes, anybody can use this and by putting it that way, rather than just not putting a license on it there. It removes the ambiguity, ambiguity of whether or not it's licensed.

D

That was my understanding of what this particular license means.

B

I see that yeah and the idea was also done to clarify the fact that you are fully free to change the license that come. That's the default output of the cookie cutter. If you, if you want and not start to explain, okay, you are free to change it because the license fires is generated with you. As the author of the extension that came out of the template, so it's just to simplify everything. Let's say.

C

That sounds reasonable to me. Are you asking on this call for objections, like? Is this the plan to do it? What what is the ask here.

B

Yeah is it like: can I move forward and open up your changing the the license on the route of the repository to clarify that and then propose cruising the issue I mean that will still let some window for people with the pr open to to come and bring objection. If there are some at last minutes but like just to I just want to to ask here before opening the pr, because if already somebody sees trouble, then I won't open a PR if it doesn't make sense.

D

My my only question is: do we know if there's any rules or precedent about whether or not the current contributors need to approve a license change from BSD to cc0, um because I know certain licenses, you can hop they're written in a way that you can hop between them without you like a unanimous approval, um because they're compatible I guess is usually the term, but is that true of going from dsd3 to cc0? Or do we need to get the sign off on every single code contributor?

D

It's not a long or a short list if I remember correctly, but I know from previous experience getting somebody who's, maybe only contributed one line to a project that they've never gone back to might be like pretty difficult.

C

So we're not changing the license to the cookie cutter project, we're changing the license that the cookie project, cookie cutter project template, generates right.

D

No we're changing license of the project so that people are aware that when it generates code, that code can be licensed any way they want I believe the plan was still to just generate a BSD license as an example like as a starting point, but the repo itself would be cc0. That was my understanding of our previous uh discussion.

B

Yeah I fully agree with what Alex has said.

E

B

And Jason commented just for um on the Note uh if things make sense, to open a PR and to get contributors to approve on that pure and I. Think I don't know of any other cases. But it's true that cc0 is much more permissive because, for instance, you are losing any sense of copyright so uh that definitely it's it would be better to ask for approval, but then I agree with Alex. We may end up with the the question of okay.

B

What do we do in two months times if there are still another some contributors that haven't approved, but maybe we can uh yeah see, what's happening, okay on the pr and come back with the question? If, if it raised.

A

I want to ask question but I'm worried derailing. Is it common to use Creative Commons licenses with code I feel like it's not even though I think it falls under them? Did I miss the reason that we decided to go with that because there are other permissive licenses. I feel like I see around more did or did I just miss a step trying to read through stuff.

B

The idea here is to say that if it's a template of something the best would be to say it's cc0, because it's not.

A

Actually, closing with it right.

B

And it's not actual code that you are executing is just like the text of it.

D

And if, from my experience in knowledge, cc0 is the only license that is written to be this like blanket permissible light, yeah public domain-esque license even super super other permissible. Ones Still have an actual license like a proper license attached that sticks with it.

B

D

But yeah uh yeah, like Steve, said in the comments. I, don't believe it is usually for code, because most people write code, one their name on it, even if it is free to share okay.

A

Just just checking um thank you.

F

To address a point that Darian brought up um because of large portions of the actual project are Incorporated in the generated code. I think that's why it's important to license the project. The the cookie cutter as a whole, uh because it's Incorporated in the in the generated output, William Stein in the chat, is mentioning wtfpl, which I think some people have adopted. My guess is the cc0 license has significant and more legal vetting than wtfpl.

F

A

B

I was saying, like I can probably move on for on to the next point, and thanks so I will open the pr and uh and ping the contributors for for asking for an approval, um and then the the next point I have is about releasing 3.5.0.

B

We have uh had some discussion with uh people at AWS, um so I'm a bit behind for doing the announcement, features and notification features then, because that AWS is really interested to be able to test Jupiter lab with Jupiter server V2, and that's already that engage it to release 3.5 without the notification and announcement features so that you can release it like starting almost today. If you want and and delayed the features about notification and announcement for another time.

B

So the question is: is it okay for everybody, or does somebody would prefer to delay 3.5 for including notification and announcement, for instance,.

A

I have no stake in it personally, but I'm fine with not delaying.

B

I will start the the 3.5 branches and and move forward dropping a notification and announcement features, uh and the last point is uh just a small reminder: there is the Jupiter lab performance meeting. Sorry I didn't write it fully uh just after this goal so feel free to join. If you are interested- and that was all for me thanks a lot.

A

Thank you for all those updates. um Yeah. Thank you that was you you're working on a lot of different things and, what's that's cool to see great, can we move on to durian.

A

C

Cool um so let's see.

C

As Isabella mentioned at the top, please uh sign up to host this call if you feel like it, there's a team Compass pinned issue and if you cannot modify the top level comment to put your name in just write, a comment and we'll we'll do it for you um and if you have any questions, uh I am happy to help uh any of the previous hosts are probably going to be really nice as well and help um yeah.

C

Please do it and the other thing is I linked a lumino PR that has been open for a couple weeks now and it has to do with um processing message, Loop uh propagation of messages to the widget tree uh in a background tab. So when you first load Jupiter lab the um application class uh instances of shell, which is a lumino widget and the events propagate downward from that shell to all of the children nodes.

C

um It's a bit like if you're familiar with react, the react, root, level component and all of its children, but with luminal. The messaging that happens is really lightweight and automatic. So you're not passing things like props. It's just your widget. By virtue of being a widget is a message Handler and it receives things like.

C

Oh, there was a resize, so you can resize your re-render, your stuff, or there was a focus or there was an update, request Etc, and this was not happening when you were loading, Jupiter lab in a background, so suppose you type in the URL for a hosted Jupiter lab instance you hit go, and then you switch tabs to read something because you know your environment's slow or something.

C

It turns out that the UI rendering wasn't happening because the message Loop was predicated on at base level, using the request, animation frame function in the browser and in the um non-browser context it was using the set immediate function. What this PR does is two things. One is that it allows the pull class, which also is something that occurs at a regular interval to know when your document is hidden and any pull subclasses can access this dot hidden and have behavior that might be different than the superclass and it adds a concept of lingering.

C

So a poll will tick one additional time in the background, even if it's back off, even if its strategy is to not tick when hidden just in case one time it'll do it, you can change linger to zero. You can change Linker to 19, it doesn't matter, but by default, it'll linger one more time and it also changes the way. The message Loop um defers propagation of messages to not use, request, animation, frame or set timeout.

C

It instead relies on the async nature of promises, so it caches a single resolved promise and then every single function call happens in the dot, then of that cached promise and it allows you to cancel because the function that runs lets you intervene by flipping a flag. So when you schedule you get back a function that if you execute it, it'll actually unschedule your thing, and this is not a public API change. This is internal to message. Loop, all the test pass.

C

In fact, the tests are a bit more aggressive now to make sure nothing got changed and the performance profile seems pretty much identical to the request animation frame in some contexts. It seemed like it was faster. Sometimes it seemed marginally slower, I tried it in Firefox and Chrome I didn't try it on Safari, but I.

C

Think I would I would be surprised if there's any difference in Behavior there, and uh hopefully what this means is nobody notices anything except for people who were loading stuff in the background they should notice it loading faster and that's it for me. Unless somebody has questions.

C

A

You're requesting a review with this right or one more review.

C

Well, I specifically asked Vidar to review it because Vidar made some insightful comments that led me down a rabbit, hole to re-implement the whole thing, um but I'm not asking anyone to reveal unless they feel like it.

A

Awesome, thank you. I'll be quiet for a few seconds for any surprise comments.

A

Okay, keep your secrets: everyone good work, um Jeremy you're! Next on the agenda, if you're ready.

G

Yeah, just a quick announcement that um there's a new uh for the old priorities uh out today and there are a couple of changes so beats test. um There's the new extension manager um using by pi as the back end and also a bunch of performance improvements and this horror RTC refactor. So all of this using um and you can try it on binder if you want and that that's it.

A

I mean that's, it sounds like a wonderful, that's it, but thank you so much for the updates thoughts on that more silence.

A

Awesome, a good meditative time, um I am checking the uh we don't have anything else on the agenda, uh but we do have time if anybody wants to throw something in. We also don't have anything on the additional discussion yet um yeah. Let me get you that link, but I would like to throw in we've had a little bit more discussion in the chat about licenses.

A

uh If it's okay, I'm going to read some comments, so they're captured in the recording, because I think this is a good discussion, um we're okay, what am I seeing here. So we have. uh There are several public domain equivalent licenses. We have a Wikipedia link that lists those uh a very good argument or that's what Jason grout sent the list of things William Stein said. A very good argument for CCO over wtfpl is that the wtfpl is explicitly banned for use at Google.

A

um Jason found that CCO over the unlicensed is encouraged, so we're seeing kind of some discussions. It's continuing. Does anyone want to speak more on those or is it good just happening in the chat.

F

uh The one interesting thing, one of the interesting things that I saw from the Fedora project was it says it's also noteworthy that some MIT variant licenses MIT, contains explicitly the right to sub license.

F

uh Some ID variant licenses which contain the right to sub license are closer to True public domain declaration than the one in the unlicensed text and in fact the in particular the MIT zero license removes the attribution requirement, so it becomes effectively a public domain license uh according to the Wikipedia article. So that's another possibility, but I think it MIT license, does not have a patent clause in it. So it's got the same problem as cc0 as far as Fedora says.

A

Okay, yeah, that's actually what I was wondering um why cc0 over MIT in that case, but that makes sense thanks for hunting this down any further licensing insight.

E

This is licensing for, what's a missed that part, oh.

A

For um cookie cutter, the extension cookie cutter I can link again from that.

H

A

Yeah, diving into that very fun.

A

Okay, any other! Oh you! Okay! We have more on the agenda but go on yes, an.

C

Issue you open Sylvan.

E

That wasn't another life.

A

E

A

I

Performance meeting in like an hour, I saw that on my calendar.

A

B

In 40 minutes, yeah.

I

Or actually it says at 10 A.M in the calendar? Okay, so in 30 minutes, yeah.

B

Depending on your time zone, but.

A

I'm, seeing one more thing on the agenda: actually, we can back up uh empty notebooks salon.

E

Oh yes, I want you to bring this up um so um The. Notebook format does not require for a notebook to have at least one cell.

E

It's totally fine to have an empty notebook or the the current implementations in the front end always add an empty cell. Whenever you find yourself with an empty notebook, this actually is problematic um in especially in the case of real-time collaboration.

E

So there are several Corner cases where you could find yourself with. Let's say, a network with two cells and two different users delete a different cell and the auto merge would result in an empty notebook and everyone will start adding cells and, as many tears as you may have, and so you find yourself with many cells that are empty, because not everyone has the same cell, they create a new one.

E

um And moreover, this constraints of having at least one cell uh whenever like the cell is deleted, is so the logic seems to be spread across the crowd they the code base in in.

E

So it's a bit untidy, and so what we would like to do is to allow for empty notebooks in the UI, so new notebooks would still have a a new mtcr when you create them or will deleting that cell would be totally okay in terms of ux. We are thinking of two things. First, we should say something in the notebook widget about uh rather than showing a blank page, so that the person doesn't doesn't think that the notebook widget is broken.

E

There is actually something it's just empty um and it is also Loosely related to some discussions that we've had in the past to add buttons at the bottom of the notebook to add new cells, just like in vs code. So in the case, where should we add such buttons?

E

In the case of an empty Netbook, you would still have a markdown code and raw buttons for creating such a new cell manually, so yeah I. Just we just wanted to um bring this up the first time, because there is already a traffic request opened by Carlos.

E

um Incidentally, this pull request also um fixes um issues with race conditions, about creating that you know new cell, which, in the case of RTC, can result very easily in notebooks having two empty cells when you open the open them so yeah, that's it. um If people have ideas of comments, please feel free to share them.

A

A quick question before, because you're going to answer me faster than binder can load. Does this PR include the ux UI from the discussion? Another issue.

E

No I think I. We would like to maybe have something many more in there, but try to decoup all the two discussions, because we feel that the ux UI question will probably have more stakeholders and people giving opinions and making mock-ups and and would take more time to get merged.

E

While a an initial technical removal of that constraint, we'll solve some of the issues we have in RTC, we can have a placeholder UI and then you know work together on the improved version of the ux um yeah, although yeah I think actually Gabrielle I may have posted some things on the issue, just so that yeah Darian.

C

So is the idea that we unblock development, but for 4.0 there is a sort of sign off on the ux.

E

No I think uh adjustment in two different periods: yeah; no, no. We we need to have a nice TR yeah for a nice ux. For for that, oh definitely.

C

That makes sense.

F

Fyi I think there's probably assumptions through the code that assume there's yourself so, for example, I think active cell is always assumed to point to a cell in the notebook um when you do the above or below keyboard shortcuts Etc. So uh this this may have some subtle ramifications throughout the code base and interface. Changing uh changes, no.

C

Cell I think is uh it'll. Always you have to always check. It can be no.

F

J

Is uh is right, so the Arctic cell can be new, but all over the place, we assume is not known. We, it can be known only because of the initialization, because at the beginning we're not in the document. We don't have a model, yet it added later, but uh yeah everywhere in the code will assume is normal. I did change that already on my peer, it could be on their cases that are invention, good.

F

I'm glad it's on your awareness, but the odd Point here is well okay. So if it's just assumption that's great, if it's actually changing the interface in a in an incompatible way, then we want to make sure it's on a 4.0 boundary major major version, boundary.

A

Do we um no ways we could test those things like? Are there ux tests to make sure that that's not getting overlooked somewhere that that require I I? Ask that really weird. Do you know what I mean, though,.

J

Yeah, these uh there's actually a lot of tests. The reason it just did because there is a lot of tests that check specifically for uh not having an empty notebook uh so yeah. That's why I want to add at least about uh on this pair, so I can change the test to check that there is a button instead of uh as at least one set and yeah.

A

Go ahead, William.

I

um I I, don't like this PR for what it's worth um for two reasons one is, it seems like your motivation, is that you I mean why don't you just have this all get created on the back end and I guess? The answer is because you still haven't done some stuff. That needs to be done in the back end, but that's not a really good motivation to change the whole user interface.

I

It's just a motivation to fix the back end and then the other thing is there's all kinds of really nice users, user interface stuff, like creating a new cell by clicking on a toolbar with an existing cell, and if you don't have any cells, then that goes away so, like a user will be like the way I know to make a new cell is to click here and then they accidentally delete all their cells. Now suddenly, they're like ah I, have no idea how to make a new cell.

I

Obviously there's, like a you, know, a menu item at the top, but it's still you know for a new user. It can be a frustrating experience if they have to know two different ways to do something so.

E

Specifically for the ux we we, uh we mentioned the idea of having something at the bottom of the notebook which will be at the top when you have an empty notebook with a plus button for a new cell. So you would. This would be right in the middle differently.

I

Now, there's three ways to make a new cell instead of I mean I, don't.

E

Know the other issue is that the NP format does not require to have a sales, and so now you I mean just independently of RTC.

E

You find yourself with an editor that always mutates a document when you open it, you can't just open it and look at it and not you know, change it because it doesn't know how to have no cells and to me that's already something and then overall, like these sorts of constraints, actually don't play well with The Oddities, for the reason that I explained it that two changes that are valid can be can result through the automatic merging into a state.

E

That's invalid, and so in order for the constraints to be available to be represented by see what it is. uh I think it's better to actually not have this mechanism to read the server after the fact.

E

But that's true. We could do it in the back end in Wi-Fi, like remove interior the logic and keep Carlos's PR, and maybe only keep the logic in Jupiter white docks so that there is only one crdt peer that is responsible for creating the cell.

E

uh Michelle resistant.

H

To be a special kind of a request when the last cell is deleted, when the functions requests to delete the last cell that it it doesn't get technically deleted ever, but it's just cleaned.

H

So that's my first question and my other notes is that if Jupiter lab 4 is going to be a base for Notebook 7, that would be a substantial change for the for some of the users, and would there be a way to restore the previous behavior of having a cell always present and while I understand that it might not be uh problem from the majority of users?

H

Some users might have this current experience as part of the workflow or some extensions might assume that there is a cell and that wouldn't be shown up in this PR. So I'm just saying that there is a quite of a quite a lot of things which might go wrong with this change and we need to be really sure that it's worth it and right now, I'm, just not convinced.

H

So sorry, just last one point that adding I I understand that this removes some of the codes special cases that are in the code, but it adds a new UI which we need to test and ensure accessibility for ETC, and maybe there needs to be three battles to add a cell code and mark them cell Etc. So just let's make sure that it's worth it yeah.

E

So before um so, no this purpose does not add a UI. It removes liquid strands to have an empty cell to have one, and so there is no means to know whether a certain removal removes the last cell, because you could have messages Crossing from different spheres, actually editing together the same notebook.

E

We are both editing, the same notebook, I, delete cell 1 unit cell two messages cross, and we both are, you know, end up with an empty notebook uh and so and it's actually something that does happen so uh in mostly around initialization and and whatnot. So basically, we don't want to have to have all of these Logics everywhere real in sales and checking whether you were removing the last or whatnot. So that's actually removing a lot of special casing.

E

The uh the new UI would be is something that has been proposed independently of this question, which was to add buttons at the bottom to create new cells, and it does exist in vs code. So, for example, um I'm going to share my screen.

E

uh No, that's not the one.

E

So something like this, where you can, we would basically have these three buttons at the bottom code, markdown raw, that's the easiest sort of uh experience you could have and in the signal markup you can actually um uh just hit enter and if the focus isn't good, you get taken and decode itself. So no click is required.

A

um To kind of recap, what I think I'm hearing? So if we are splitting these discussions in two, it seems like, though there is still a discussion that needs to happen with the with the fix on right, even separate from the UI.

A

um Okay, any other other thoughts there, I'm assuming this is moving to the pr or I. Don't want to cut this off. So we have time.

I

I guess my one quick other thought is what text editors do like. If you open a file, that's completely blank, it makes it look like there's um a single blank line um like the open edit, something in Vim or code mirror, but there actually isn't a line there and when you start typing it, it puts the line there. So you could still have it where um it looks like there's a cell that isn't actually nothing actually happens. Unless the user starts editing like there, there isn't really a cell there unless the user starts editing.

E

Yeah we have a mocha for that, like a cell widget, but no cell in the.

I

E

In the model and when you start editing, it creates the cell.

I

E

I

A positive existing text, editors.

J

That's the only solution that I could think of by keeping the they said. Basically, uh a standalone set that we are up to the UI, but is not on in the model, and once we receive a Source centered signal, then we add it to another model: uh the Water another suggestion from mic, I, think cleaning the last cell uh I. Don't think it's our idea because of what Silvan just said. Maybe two users get it. uh There is only two sets on the notebook it uses that it's one set.

J

We will end up with two uh empty sets. You know, but also I, don't think it's a good idea, because I believe we could end up in a situation where one user cleans the cell that eats all the cells, and so we clean the last one. Instead of removing it and another user is editing that cell, for example, changing it from a culture to a markdown.

J

We could end up, maybe in a situation where we have a code cell with the metallic data of our marginal I, need to double check that but I think it's possible. So I don't think it's a good idea to clean the the last set so yeah.

E

Okay, so that was it I, like the zenf python, remove special cases.

E

um But um uh yeah thanks for suggestions, we will be I think there will be more discussions about this for sure.

A

Yeah, it sounds like it. This is interesting. Definitely got me thinking about stuff.

A

I am checking the agenda one more time. I don't see anything. There. That's been added to the agenda or additional discussion, so I'm going to be quiet for a few seconds and let you all decide but then I'll stop recording and we can have a non-recorded discussion if needed.

A

Going once going twice, okay, thank you all that was a lot of good updates and interesting discussion that I'm going to be thinking a lot about. So uh thank you and for anyone watching in the future have a good rest of your day.