Sci Cloj Clojure visual-tools group, 23 Oct 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Clojure visual-tools meeting 15 - play.teod.eu, Kindly

Description

On Oct 23rd, 2022, the Clojure visual-tools group had its 15th meeting.
https://scicloj.github.io/docs/community/groups/visual-tools/

Teodor Heggelund demosntrated practices and ideas of knolwedge management at play.teod.eu.
https://play.teod.eu

Daniel Slutsky presented Kindly, a small library offering a common ground for standardizing data visualization and literate programming.
https://scicloj.github.io/kindly/

A

Hello, this is meeting 15 of the visual tools group, and here we are Lucas and Theodore and Mark and Danielle and we'll be discussing a few components of of the emerging ecosystem for note-taking, literacy, programming and Knowledge Management and such inclusion.

A

And what we have today is a tiny demo by Lucas of the Calvin notebooks, which is just a promo for a dedicated meeting really soon, hopefully, and then presentation by Kyoto about theodos practices of a certain knowledgement, Knowledge, Management, System enclosure and babashka and such and then I Daniel I will present this kindly project, which is a common ground. You know a proposal for Common Grounds to take notes in different tools and I will have a discussion and this visual tools group.

A

It is a group for Tool makers and Tool users in closure to collaborate and discuss things like that and um yeah. So uh I guess we we could begin uh by introducing ourselves so.

B

A

uh Mark, what would you like to tell me a.

B

Bit about yourself, even though.

A

You're so famous, of course,.

C

Okay, I'll give a shot. um Hi Mark champine um I've been a closureist for over a decade, uh I've been involved in lots of different ways. Attended. Many closure conjures I've uh still a lead organizer for the Boston closure group.

C

Although we haven't had a meeting in a while uh things went off the rails with covid, but um uh still trying to keep my hand in participate in conferences and uh I'm very interested in seeing the visual tools ecosystem Advance, you know, become more sophisticated, easier to use uh personally I'd like to be able to do rapid, prototyping of of uh graphing and charting.

C

You know for uh mostly for side projects as sort of a a way of doing exploratory data analysis that would eventually, uh you know, become uh hard-coded into uh you know some sophisticated uh ways of displaying and analyzing. uh Well, in my case financial data. That's! But that's um that's! Basically it in a nutshell, so glad to be here.

A

Thank you so much.

D

Yeah, uh so my name is Taylor uh I used to be a civil engineer uh when I started working more and more with software and now um doing more uh products. Product Discovery uh than before and I realized that I really care about knowledge and how we, how we model knowledge. So that's uh what I'm going to be focusing on today, um yeah uh Lucas.

E

Hi I'm Lucas um I got involved in the visual tools I've had by attending uh what was it called re-closure last year and I got in touch with Danielle um and I I'm, the main maintainer of omnitrace, a tracing library for like your codes and stuff, and one of the maintenance of calva, and today we're apparently going to talk about Calvin notebooks, which are a new thing and um which uh I built into Calvary in the weekend and which is pretty nice to be able to have Rich notebooks, like uh like the Jupiter notebooks and similar um just from your normal cover file a closure file.

E

However, it's it's on files apparently now um and one of the coolest things is that it's working pretty similar to like a clerk notebook that a lot of people have probably seen, um but you don't need like any special running environment. It's just like in your editor, like everything else,.

A

Wonderful I'm, Daniel I am a statistics person I, do statistics and I'm involved in review of the study groups and Dev groups like this one in closure, community and yeah, and today I'll show something I've been doing, which is this tiny little projects called kindly and Clay, which are about split search, programming and yeah. So I guess now, because if it makes sense, we could have like a short promo about notebooks.

E

uh Yeah I don't have anything like any like anything running right now. uh I just started this notebook up, but I found a short video uh by Chris. That already contains like the interesting little bits. So I can talk over that and we'll show like the the actual big demo next time together with Rick Chris, which I think makes sense because, like at least half of this is his work, um so yeah I guess I'll. Try sharing my screen. Let's see if I manage I've got too many screens these days.

E

uh Let's see yeah I think it's this one yeah and it looks correct, okay cool. uh So this is a video, that's posted in the portal Channel uh and Slick. uh If you want to like look at it yourself um and it's like a short demo of portal and cover notebooks interacting um and you can see, uh what's even needed to just start uh the Calvin notebooks, you just need a rebel connection.

E

um Oh well, that was not what I wanted. um You just need a rebel connection like you would normally have, and then all you do is right. Click a file go to open with select closure notebooks instead of the file, and then you'll get this a nice little notebook and an overlay now status, notebook where every top level form in that file is basically a cell that you can run.

E

um If you have a portal installed as well, the portrait extension in vs code, um each of those outputs get a little bit richer. We can see uh Chris running it now and you get basically Rich outputs in between all of those calls, and each of those cells is basically a running portal version and works like portal. Normally what you get like Rich contents and can click around in that get traces and whatever um we'll show a bit more of like the innards of.

B

E

The uh talk next time, probably um yeah, but the interesting bit around it- is that you don't need any special setup. You don't need any special files or anything. It's just use your normal closure files and does need don't need to be special files. One of the things that was always annoying to me about something like clerk is that you need code in between your defense or whatever that's actually executed, which is like.

E

Usually, we don't write our code like that right and we write our code by just having um just having Rich comments at the end, but like only usually having defense in between, because you don't want stuff executing when the namespace is loading right um and one of the cool things that um about these notebooks as well, is that uh these uh for statements are actually like in the file or just like inside a rich comment plot at the bottom, and this still gets split up into their own cells and can be run so like this could all be just defense, which don't really like actually have any interesting output, um and only the bottom whatevers would have actual output.

E

So you could like run your normal closure namespaces that you have for work without having to like put any special stuff into that file and which is really nice um yeah anyway. I think that's a nice trailer for next time or something um I'm happy to like talk about it more. uh If you want I, don't know if there's any questions.

A

Hello Kira by the way, thank you for joining. uh We were just you know, just beginning and having a really funny demonstration of kalvanov books by Lucas and promo for next time and um yeah, and now the other. You had the comments a few minutes ago when I stopped you about about the or comment on a question about Calvin notebooks. Should we come back to that or is it too late already.

D

uh I I think I remember what I said, but Lucas basically answered it during the presentation, so uh I'm happy to just leave discussion for later.

D

Should we move on to myself.

A

D

Okay, I'll share my screen.

D

It's kind of interesting that Kira joined because she got an informal attempt at explaining what it is I'm making, which I felt was really badly worded last time, so I'm getting a second chance now.

D

uh Are you able to see my screen? Yes,.

E

D

Yeah, so I'll just uh show you the thing first, so this is a website a normal website. It's just HTML, uh but I've spent quite a bit of time on content for it uh and I've kind of landed on a structure uh that I feel like it's working and I.

D

It really helps me think when I want to do something and just share it in public, so get feedback on drafts and and write, write out things uh and there's this go to random uh page button, which I kind of enjoy just just having to to go around yeah.

D

um So I'm just going to remove the emphasis on that and talk through some points.

D

uh So the reason why I'm sharing it here is that I'm interested in site, coach and Collective, Knowledge, Management and I think we have a job to do when it comes to Collective, Knowledge, Management uh and I'm quite excited about my current learning process uh when I want to learn something right now, I just make a page for it uh on my site and then I worked and I find that helps me think so. I'm going to share a bit today.

D

uh And I have this idea in my head, uh where we have sort of a cycloser knowledge archipelago uh where there are kind of islands, of knowledge that are Loosely connected and people can contribute uh and I think we're kind of we need I mean we, we don't have a central uh uh resource, uh uh Authority, so yeah, so if we can make it easy to create knowledge, uh collect and curate knowledge and easy to learn, I think uh we can have a bigger impact.

D

uh So a bit about my note-taking process, uh I've been journaling for a long time. uh I counted uh some journaling lines and uh I found out that I started out in 2017. uh I had 2 000 lines of journaling. uh Not here, then it increased a bit and then it's kind of become less, uh uh but it doesn't just become less because I'm writing less I'm I'm kind of writing more but I'm writing in different places.

D

uh So in 2020 was when I found uh Rome research which I use extensively um because then I can actually find my stuff later.

D

uh But then this year, I kind of got frustrated that I wasn't able to share easily uh what I made uh so I've tried to use the web for what it is, uh not products that pull you away from the web, but just use the internet.

D

Yeah so uh I guess I already said who I am but I used to do. Civil engineering uh writing reports, as a part of that uh which means I kind of have I, know I'm, not saying that I have different expectations to documentation other than from what I see uh in my team uh learn a bit of software and these days a lot of product, Discovery uh figuring out.

D

What is the right thing to build and I'm really dissatisfied with tools like notion because they pulled me away from plain texts: uh I'd rather use plain texting kits and get some rich linking so I'm I'm, very interested in how we can build knowledge together and lately, I've been inspired a lot uh by a book called uh the beginning of Infinity uh by davidoch uh and I. Guess we can just press enter on this thing. So this is my. uh This is my notes on that book and that's the Wikipedia pitch.

D

So what you're seeing right now is just a page in my system, uh but I'm going to go back and just continue from where I was.

D

Yeah, uh so uh on the right here, you kind of you have the sites, and uh here is the rendered version of the notes, uh but I figured we could kind of try to do something, a bit more interactive right now. uh This is C-Max, and this is the current folder index.org is the file that you're seeing which is text file that contains this?

D

uh uh This is just HTML. So if I search for story, I find that H2, which is the same as you're seeing there uh and if I go out, I found find a play, dot Eden file, which is a metadata. So that's the title. uh This is my default Readiness for incomplete contact contents, uh which is what happens uh when I create new stuff, and then it just comes into this pile of links.

D

Yeah so I asked you uh Danielle to kind of help me a bit uh with the navigation. So does this make sense? I could show how to create a page and deploy it. We could create a page for you.

A

That would be wonderful.

D

Okay, so we're going to start there, uh I have a create page function. uh This is a small emacs list wrapper for a Babushka CLI uh when I want to create a page I just press enter I, get asked for a page, ID I see that should have been page slug, uh see. If I can spell your name correctly,.

A

I, don't even know how to pronounce it and spell it. Okay,.

D

So now we got dropped into a new folder uh and there's an index.org page uh I guess we could.

D

uh Look you up on Twitter to find the links.

D

uh Yeah: okay, now I by habits, I pressed space, uh uh M.

D

And that's the make thing that runs make on top level and I'm going to look at the git status. So we say that the top level make file changed. There's now a daniels.ski index page, uh the top level index uh changed.

D

There was a new link uh and there were some changes to uh another index, a Eden index and a Json index and there's a new folder dial and there's a new file uh in index, slash by ued, uuid yeah, so I'm just going to uh add all those and write a really bad commit message and push.

D

uh So now, if we just go and GitHub uh um I find the thing, uh then we see here, that's browser is working, um and this is kind of uh this annoys me. uh I I don't have a build uh because I just check everything in so my get push takes a second and then I have to wait half a minute for Telfair to do its thing, um but we can just leave that running.

D

um And if I go back now uh into here now, I can insert a link and I can insert a link to Daniel and I can navigate to that link uh and in a sense now I'm kind of I'm cheating, uh because I'm leaning on something called orgrom, uh which is a knowledge management system in org mode. But I've kind of I I've been making an effort to abstract myself uh as far as possible. Away from that and I also see that yeah we're deployed.

D

So if I now go on, wait, try to find that out and there's Daniel.

D

Yeah and we can link to all the pages- that's what I did just there uh and there's a play.clj relations CLI. uh So we could have a look at that.

D

um I just have to move some Zoom menu away.

D

D

Yeah Wicked start that so this is a local preview and now I'm just getting a terminal in here and I'm pasting that commands and then that opens.

D

The same thing on localhost, uh so I could navigate to this page.

D

And we can go down to the demo.

D

And I thought I had uh live, updates, yeah, okay, it's working so I I. Think I'm just going to open uh for for questions right now. I could show you the kind of relations CLI, because I have some kind of kind of interesting metadata access. But uh what do you think Danielle.

A

Why not? Actually we have a lot of time.

D

Okay, are there any other questions to the things that I just told.

E

um What's like the advantage of your system compared to something like I, don't know, looks like or whatever.

D

I have uh direct uh control over the HTML, so I can do anything. I can do in HTML.

E

um Wouldn't you have like the same thing if you're just like use something like I, don't know how ROM works, but like I've been using logs like for like a little bit now and they're, just building markdown files underneath right, so you could just take those markdown files and generate dhtml. You want out of them and like have all the other tooling for free, basically uh like your or Chrome extension, basically just like with all the other stuff and the markdown can create whichever HTML you want right.

D

So uh the the first part of my answer would be I, don't know. I'm using ROM I haven't really put that much effort into loxic and Rome has really poor support for plain text, because you're just sitting in in that outlining thing.

D

uh The other part.

D

I guess this is this: is all text based so there's no need to use any third-party products at all? uh I I think that's that says something about the reach, because it doesn't depend on having installed things and having an account and doing things. The the log sequence- okay, but uh I I'd, actually be interested in seeing uh people explore, building their knowledge base on loxic and and pushing the flexibility of what what they can do. Then.

E

Yeah I mean I, haven't used logs like a lot yet, um but at work we are doing uh like markdown base similar stuff just for regulatory texts and we actually like them, build an LSP on top of like that, markdown or whatever, and our tool to do extra stuff as vs code right and then you can say similar to your thing. You can ask questions about a text like what are the incoming links and whatever like I'm guessing.

E

You have basically the same tooling for your thing and um I'm, just wondering if there's like any advantages to having it in this system, compared to something like a markdown based system or whatever.

D

um Just a second there's, someone up the door, yeah, no problem.

D

Sorry about that.

D

My honest answer: I I, don't know because I don't have an overview over all the options. uh uh I know that I've I'm really liking my workflow right now. uh If there are other similarly good workflows, uh I wanted to just put them all next to each other and and see see what the differences are making.

A

D

A

Maybe I could mention one reason why I'm interested in this and that being small being something that is not the huge system as loxic. So possibly, this could be used as a rendering engine for other contexts like literate programming right like rendering our namespaces, and that is why I'm really really interested, possibly integrating with it.

D

Yeah, so that's that's a really good point, uh because I, uh let's have a look at that make file uh yeah. So it's huge. uh Let's find Daniel's page and I'm just going to copy this into another buffer uh so that we can split uh the lines a bit uh so.

D

There's a pan look commands that reads uh uh from org modes and takes an input, file and uh pipes that into a filter, resolve links and pipes that out into panic, so I don't really have any dependencies on uh input formats other than that they have to be convertible to by Panic.

D

So if we have a look at padlock, uh this small figure shows the supported input files so there's t2t and textile and tikiwiki and tsv yeah right now, I'm, just preferring or mod, uh but.

D

So there aren't Stopman independencies that kind of answer your comments. Daniel.

A

About the closure about the input and output of it and and what it does, yeah.

D

So uh uh this is just like a big folder, there's uh subfolders for each of the articles, uh lots of folders, um uh but if that's one interesting file, it's this file, so this is the CLI. So the whole thing is a babashka CLI and I was able to write the whole thing just in order to write content. So it's not like a lot of code. It's actually 500 lines now. uh uh But if we look at the commands.

D

uh Let's, rather go to the bottom uh and see that dispatch table so there's a command for creating Pages. There's the command for running filters, uh a filter as right now, there's only a filter to make sure the links resolved because I use uids and inputs and there's the command to create the index by u8 ID. There's a command to create the make file. There's the command to go to random page and there's the command uh to handle metadata.

D

um I was about to show that metadata command, so I can just do that now. So if I do play.clj, then I just see the available commands and if I do a release, it's kind of hard to type this inside of emacs, but I, don't want more stuff.

D

uh Then I see that I can do a from source to Target and sources can be from files or from lens so I'm going to do from files to lines. I think.

D

uh No that's wrong that doesn't work.

D

um There needs to be from files to Lance from files reads from the file system. So that's a lot. uh uh So, let's just look at the first uh line.

D

And now Zoom is in the way again, so this thing has a slug Draco, willink title Draco, willink, I read this and a uuid. uh So I could do this from files to lines. I could put this into a uh Alliance dot, EDM.

D

Then I could just edit that lines.vian I could do some kind of search, let's call them drug overlink, 2.

D

And if I now uh read that file s.in um pipe it into relations, but now I want to do from lines to what else.

D

Let's hope this works.

D

No, this doesn't work.

D

Kind of um a couple times- oh yeah, I I, didn't so I need to pipe.

D

And now it's written those things back into the file system. uh It also did a few other things uh yeah. It added some commas, because I pray to print the closure.

D

But here we go here's the Joker willing too, and if I make again now and preview uh up here, then the article is called Draco willing to at least in the metadata.

D

Sr uh I think that was the things I had planned to show. This thing is large, uh so I don't think it makes sense to read all of it in front of you, you guys, um so this was as far that we had gotten.

D

Are there other things? You are interested in seeing just specific questions.

A

C

uh Is is there illiterate programming aspect to this, where the output of running closure Expressions can be shown within the document.

D

S uh so I'm just using built-in org modes thanks uh we can have a look. So I wrote uh this article essentially about how to use pandlook from closure uh and, let's page, find, let's find the pandok article, uh and this thing is kind of large.

D

uh But here uh there's some literate stuff. So here I just include the jet BB uh dot sh uh uh and that's there, uh but that that's just pandok, so I haven't written any uh literate programming things, but it's kind of it feels like doing literate programming, because I can just change things as they go and yeah write programs to generate text files.

D

um I can show you. uh The index.clj uh I added I was really happy with this message. We interrupt this page with the methods from the visual tools meeting attendees. uh uh So I could just add this, because it's just a script, uh uh and here I, just uh uh I already just generate uh some HTML inside org modes, which is then converted to HTML again with pandok.

D

C

So just some comments, um I'm, a heavy emacs user and uh then or grown came along uh and I converted like a hundred of my uh org files into Rome format, you know I wanted all the back linking and uh so that's been um marginally successful. I I do find it a little bit cumbersome to try to maintain all the back.

C

Lengths and or Chrome has some limitations, um but this this seems really neat in the in its ability to uh facilitate collaboration, uh so uh yeah that it's it's pretty exciting that you know you have this Universal format based on on HTML, everybody can participate uh and using GitHub and so on. uh Standard, tooling, so I think it'd be pretty cool.

C

uh You know if, if you were to sort of um try to collaborate with a small group and maybe create a maybe a Wikipedia like article, you know with um you, know, extensive use of linking and um you know just try to work out. What are the?

C

What are the aspects of your your tooling that could be improved? What um what works, what doesn't work as well and uh especially try to create some uh cookbooks or uh procedures for people to be able to participate? You know so you know they, they would need to uh be able to use GitHub and probably um some some tooling, like maggot or similar, and especially you know how to create the the linking within documents to make a uh an interesting and usable collaborative doc. So um yeah.

C

If you could get just a handful of people together and say we're going to create an article about X and uh you know, guide everybody through contributing one uh one section or a few paragraphs.

C

um You'd have sort of a instance proof of of how your system would work in in a collaborative document. Development um process so looks.

D

Very promising, that's a really interesting idea. I I had never thought about that. I just thought that everyone would have their own sights and then we would just link between the sites, but there could be collaboration within the sites right there, Wicked yeah.

C

Like you say, build stuff together, exactly right, I mean um true, you could do it between sites or with insights. um Both would be uh I, think common use cases. uh Some you know, depends on the granularity right but um seems like it could also be something that would be useful in uh a cyclose community. There's lots of documentation efforts, um so yeah, really interesting stuff.

D

uh I'm kind of trying to be like as I'm, not I, don't want to push this down somebody's throat right now. It's it's very much tailored to my exact preference. So I don't expect uh other people to kind of love the same constraints, yeah um yeah, um so uh yeah. That actually leads us uh very neatly into a section that I had written. uh But all this is heavily adapted for for theater's preference.

B

D

So I I believe that a common toolkit should be extracted from personal preference, so play.tl.eu. It's kind of my knowledge, playground, backyard garden thing and I have this idea of extracting a toolkit uh for knowledge for babashka salads, it's kind of the working title, um but it doesn't really exist yet right now the only working system is the play.clj I've tried writing a few things, um but I wanted on, uh uh for instance, not have a designer constraint breed that you have to use emacs like we want to allow everyone to to contribute.

D

um So if I can kind of uh chop up the thing and uh serve some utility and avoid uh dependencies that we don't necessarily want to build in I expect that we don't want to require that everybody use emacs and our mod uh I figure that vs code, calva and markdown should be equally uh acceptable. I I I have no experience, writing cover plugins, but it's lucky that other people, people have um and and I understand that we should try to pull information from multiple sites.

D

uh That kind of follow up same interface, but perhaps interacting inside sites is equally valuable. I I, don't know yet um yeah. So this yeah.

C

You can try both things, uh so uh I didn't think of of the um pulling together aspect. So say you wanted to write a uh a tutorial or something, and it had three chapters. You could assign a chapter uh per contributor and then have some scripting that that assembles it into a final uh document, um so that that kind of thing I think, would be not a big leap from what you already have.

D

I can commentary on uh on the foreign reference torch because uh down here, there's this remote references thing and uh uh to pick something I'm curious about this is a book uh by David, Dutch and I kind of I create these things as entities, because then, when I want to do a page finds uh uh the beginning of infinity and look at the metadata, then this thing just has a uuid. Then I can link to that from anywhere uh and then the page is just a thing out.

D

So then I can update the references one in one place and I think we could facilitate like enter um in their site collaboration and not some way, but it's kind of hard to speculate without running experience.

C

Exactly yeah, the uuid is very helpful in in that aspect. You won't have any uh collisions in terms of the uh the artifacts.

D

uh And that's if we here uh uh I think I mentioned this yeah. So if we open this thing and do uh text mode, then we see that this is actually just a you ID and that's where the closure CLI comes in because I just put you IDs in the plain text and then I run the filter over the pan of Json. That swaps out any ID colon uh uuid uh with the actual URL.

C

Cool, that's! That's genius.

D

It's kind of funny, because that's that's really all you need to get that stable, linking I think.

D

Yeah so back into uh work mode.

A

It would be good to start converging to a conclusion yeah, for this part no rush, but that would be a good time.

D

I I agree uh it's nice to open up, but it's also nice to actually cover the agenda um yeah. So I would really like to see us create a knowledge archipelago for closure and I believe that uh with the community that we have uh it's possible, um there are current initiatives. Daniel and Kira are doing lots of important work, but perhaps we can activate more thinking phases.

D

And some current initiatives, uh Kira has the data Sciences code book, uh there's cloudupedia, which is kind of related and there's uh the uh libraries page that's Danielle uh has made, which is also.

B

D

D

Yeah uh can be skipped, we're skipping this one yeah and there's some contact in for for me, so I mean you guys are able to find me uh but feel free to talk to me about this. This is something I uh I find the collective Knowledge Management challenge to be a really interesting, Challenge and I want this to apply it successfully uh for for data science, enclosure.

D

There, uh uh thanks for the attention.

A

Thank you so much Theodore. It is really exciting because you know at least at the cyclos context we have been. We have had challenges and- and we had these hopes you were talking about- but some of that never happened, and then you actually were researching lots of options and and trying different tools and and practicing different ways for a while, since those early days of cyclos and probably much earlier and then seeing these fruits that we could actually start using.

A

It was really an exciting moment and then maybe after the kindly paths that we'll have now, we could chat a little bit about how this could connect to the idea of namespace as a notebook as we like to call it where we actually use our closure namespaces as the source, software and yeah um I stopped the recording for a moment, hello. We are back from a short break and now we'll talk about kindly and I will share my screen and we will have a tiny demo and then discuss it.

A

Yeah yeah so so uh currently is currently a version. Three of a tiny thing we have been using for a while, and the idea is to actually start using it widely very soon, and one of the reasons is coming. The science girls were talking about and, and many other needs to, write notes in a compatible way, and that is what we discuss in a moment. The idea is that kindly will be a certain way of expressing our nodes in closure namespaces and that it will hopefully be usable from different tools.

A

Currently, it is usable from this tool clay which I'm using now to render this page from a closure namespace that we have on the right and we are discussing it with a few of the tool makers, and it is also related to something which has been emerging in the last couple of weeks by casting Bering.

A

This case Commerce, is a fascinating project to create a certain standardized way to express notes, and it is similar and different in many ways and I think I'm very inspired by what cast and has been doing, researching and I think we should discuss how these things could be compatible, and this current version of currently is really a result of discussing um it. With a few of you and really very much inspired by conversation we've had in the past in this group. Let us just briefly mention the problem.

A

We have many tools for writing, notes, writing, visualizing things they're doing this practice called literature programming, and we have many existing nodes written in different ways and each time somebody is writing some notes. They actually need a certain way to express how things should be displayed. So, for example, let us look into this color tutorial by Thomas, which is written in clerk right.

A

So here we can look inside and uh sorry for my slow internet, so I'm sure, oh yeah, so this beautiful color tutorial is written in club and it needs a certain way to express that something is a color. And so in this case it was defined by this function that Thomas written, which is adapted to clerk, so that Claire would know how to show Colors. So the notation here is eventually clerk specific and the name space which is generating. This is clerk specific and maybe to look into another example, maybe uh um yeah.

A

Let us look into the beautiful tablecloth, API docs, also by tomash right this amazing page. This was written in markdown in actually in our markdown, and it has its way of saying how things should be displayed, and so each of these ways of writing notes has those little differences and things are not copy, based friendly as cast and called it, and maybe another example.

A

This vcrj tutorial by uh yeah never mind, maybe because of the slow internet I skip a little bit and in the past, in the psychological Community, we have been using node space to write, lots of notes and the reason we stopped it and actually, in a way to kind of stopped.

A

The progress with this node space tool is that it didn't have a clear path for being copy-based friendly and being able to use notes-based nodes in the art tools which are emerging like kalva, notebooks and portal and clear canal, and the goal is to have a way to make things copy paste friendly. So, ideally, what we want is that somebody writing a tutorial would be able to write it in a way that will have its own semantics of saying how what things should be displayed like like this thing is a hiccup note.

A

This is a Vega plot, but it would just work with existing tools and also with Future tools which do not exist at the moment of writing. The tutorial you know here is like a proposal of how we could do it so conceptually we want to have a certain notion called a kind which is a keyword saying.

A

This thing is of kind hiccup, for example, so for a given context like a little piece of code that should be rendered and the value resulting from evaluating this code for a given context, we could have some inference inferring, what kind it is and then the kind is the information that different tools could use to decide how to display it. So, for example, all the tools we are talking about, they have a certain way of displaying hiccup as HTML, so tools which are interested in using the advice of kindly could see.

A

Oh, that is hiccup, then I will show it as hiccup, and the inference of kind could be done in a way which is specific to the user specific to a given tutorial, because maybe in some tutorial it would be convenient to write things in a certain way, but since kind inference is decoupled from actually implementing the current, which is displaying pickup writing a given tutorial with.

B

A

Of kind inference should just work with different tools if they are interested in using kindly's advice. So that's that's the ideology and of course there is a challenge here and and we'll see in a moment that some details are not obvious, but let us see how we could use it. So the idea is that we.

A

Can have we can the user can Define certain functions which are called advice so in these functions they have the logic of taking a given context and deciding what's the relevant kind is so, for example, some user may wish to invent a special kind, or maybe let us use this imaginary kind, that some some tools could could possibly wish to use.

A

So possibly let us imagine, we have a certain kind called ABCD that some sewers know how to render. So. This user wants all the numbers in the notes to be rendered as this kind of thing ABCD, and maybe they want all the all the strings to be rendered at this time, a efgh, so they could Define. This device functions that take a given context and if that context has a number value for example, then it adds to the context. The relevant kind information and this advice function would be used by all the relevant tools.

A

If the user configures the given namespace to use that advice and then let us try, these advice doesn't see they're working. So we can ask for kindly advice for this context, where the value is 3. With these advices we have defined, and then yes, it does recognize that it has kind, ABC and and so on, and the the idea is that what the user would do in the namespace is something like that they would set the devices to a given set of advices they're interested in and then all tools.

A

When the tools ask kindly for a given context, then that would just work and all the tools we learned. Oh, this thing is of kind ABCD and in a moment we'll see a few namespaces like that, where the user defined a given set of advices and then all tools would enjoy this kind inference.

A

Here in this namespace, we were using actually the default current inference.

B

A

Is something that you know behaves sensitively so, for example, here some map was written, so we could just see the map, so it didn't have any special kind, and but let us go to another namespace. So um here is like one example namespace where you know that could be a tutorial, somebody's writing.

A

So at the top of their namespace they decided to have you know only this advice where all values, if the value was 3, they would get this special kind, ABCD and otherwise no no kind information is added to the context and then, if we render you know, if we look into this night space, then then we see yeah. When we have this context of a value of three, then uh the kind information will be added. So this call to kindly advice is not something the user would do.

A

It is something the tool makers would do so that their tool would know how to render a given namespace and in a moment we'll discuss it further. But maybe uh let us go to the more detailed namespace where we see the default behavior of kindly so here we already say default setup. So we get the defaults, a behavior of kind inference and then let's see what happens so so first, you know when we take like a given context. Value is some math.

A

Then we ask for kindly's advice.

A

Then you see, it doesn't add any kind information, because if this simple map doesn't have any any kinds to be inferred about it, it is just a map, but we could, by by this default behavior that we could specify the kind of things in certain ways. So, for example, in this situation we could say yeah. Let us take this value but consider it as something of kind ABCD. And then you see it does get the kind information and there are the different ways to write it and and I'm skipping a little bit.

A

But the idea is that by this default Behavior it is possible to express the additional kind information for a given value or a given a form to be evaluated. So it would behave accordingly and- and maybe let us see how things look this way so, for example, I created some hiccup data structure like saying the word hello in purple color.

A

So if I just write my cut my hiccup, then there is no special kind information and it is just rendered in the data structure, fine and but if I consider it as kind of Hiccup, then I could see it actually rendered as a purple hello, and there are different ways to write it by. You know calling this special kind function and also by attaching metadata to the code, and so by this default behavior of country.

A

All these ways would work- and we see here the clay tool, rendering these Asians, showing the purple hello and hopefully any other tool that would like to use kindly is advice would just need to pass the relevant information kindly and then realize. Oh, that should be considered as Hiccup and then display a purple, hello and um yeah, and some kinds uh are inferred by these words. So, for example, if we have a VAR, then it would get the kind bar.

A

Then that is maybe interesting because Theodore you had some ideas about how a closure VAR could be rendered in a beautiful way with including documentation and such. And if you have like an image for example, then it does get the kind of a password image. Then here you see, the clay tool also views this image already, as as the image, instead of in any other way, because that because the clay tool asks for kindness advice- and it is oh, that value is an image, so I would show it as an image.

A

So that is a little taste of how things look in the clay tool and the idea is to to allow all tools to actually behave this way. So, ideally, this namespace could be viewed in clerk, for example by writing a small adapter where we've used in the case of cloud. We use those predicates that clerk as for viewers and those predicates, we just call kindly and and pass the value and get the resultant result in any comments. So far is it? Is it actually confusing what I'm saying maybe.

B

It is confusing.

A

So if maybe I'll just comment you see here, we see how a typical tutorial looks. So we are not having explicit calls to advice on the user side. The call for advice is something the tool we would do so.

B

A

B

A

Was calling the kindly advice to realize that this thing was of kind hiccup, right and yeah just to clarify sorry.

D

I have a question but I I'm kind of guessing that it's a big one, so you can choose either to answer or just wait. uh It's kindly a protocol.

A

No, but there is it, it is extensible by protocols as one of the ways to extend it. I.

D

I mean protocol not in the closure protocol sense, but in the like, uh IP and stuff, because it kind of sounds like we're trying to Define sort of a narrow waste between providers and showers. Oh yeah yeah, actually.

A

Then yes, it is then yes, it is. It is a stateful protocol, because you see we can do some some mutation of the, how it is configured to behave. But after that, after we, for example, set up the defaults, then it is a certain function that could be formalized. Their tools could use to so that the.

B

A

Of kindly is could be considered a broad one. Yes, so that is something a tool would call to realize. Oh this value, it actually doesn't have any kind information, so I would just render it my different ways.

E

But the only function that a tool would actually call would be the kindly advice function right if we, the setting uh in between that you have like on most of those things where it's like well, please consider this a hiccup wouldn't be something that the tool would do. That would be something in your namespace right. Only it's. Basically, this kindly advice would be the only function. That's ever called pile well, I mean maybe there's in the future some but like right now.

E

The only function that would be used by the tools would be the kind the advancing and all the rest would be like stuff that you use to have different kind outcomes. Basically right.

A

Yeah, the the the behavior adding advice actually specifying that a certain value as a certain kind. All that is happening on the user side. On the tour side, we only call kindly advice, yeah.

E

Yeah, that makes sense otherwise, we'd have problems with the stateful stuff. If we transfer in between different jvms or different runtimes right, if you're displaying stuff runs in cl.js or whatever you could, you wouldn't be able to set uh certain things. The way that you set the defaults right now right.

A

A

Talk about two problems, if, if that is making sense yeah, so one problem is that not always the value is enough for the context, because sometimes- and that is controversial. Actually we are not sure about this detail, and there are. There is a long conversation with it with cast and about it and and in the last few days, so some users were asking for this kind of notation, saying yeah. This thing I wish to attach this little piece of metadata to the code.

A

To say this thing should be rendered as hiccup right and- and that is something that clerk does as well, and it is not obvious because to make that work, the tools would also need to pass additional information to the advice function. They would need to ask not only the the resulting value, but also the form which was which was um evaluated. So we see it here, for example, so that is how kindly behaves at the moment.

A

If we pass a given form like this one plus two with additional metadata, then kindly would be able to realize oh yeah, that is of kind ABCD. So you see, the metadata was part of the code, but it is not metadata or the resulting value, and these equations they they are kind of necessary, I. Think because some users like this habit and also it has this advantage that when we attach metadata this way, then the resulting value is not touched by metadata.

A

In some situations, it is desirable that it is only used as a certain annotation for how things should be rendered, not changed, actually the resulting values, and so that is one controversial detail. Should we actually support that and would actually the tools or the tool makers like the idea of having to pass also the form and not only the value.

A

Another trouble is the jvm, so some situations need the actual jvm to infer the kind. So, for example, here we recognize this Java type Java class and realize oh, that should be viewed as an image. So, for example, if I understand correctly, Calvin notebooks pass the value to the JavaScript.

D

A

And then it is decided how to display things. So, even though kindly.

B

A

Cljc, that would be too late, because the information of the Java class is currently not passed to the JavaScript side in in Calvin outbox. So this dependency on the jvm is something to consider how we could support that and concretely, you know if we have a buffered image, which is something that we sometimes have.

A

How do we like to infer that it should be include as an image, and so you see we have those dilemmas, and that is something that you know: I'm really hoping to start with you, Lucas and but for now that is the concept and, um and the hope is to convert to something that kind of works, do all the necessary compromises and be able to start using it and actually write notes which are copy based friendly in the beginning.

A

We'll only use this clay tool, which is just a tiny tool to support that and gradually we will hope to make adapters for different tools. So clerk should be easy because it has this predicate uh way of defining of inferring viewers, but it would not support the form, bypassing the form metadata at the moment. So there is a trouble there and other tools may have other challenges, but that is the current situation: yeah, it's making sense, actually yeah yeah. Thank you for that.

A

Thank you for understanding situation and- and you see we are in this moment that, where compromises are needed and possibly we should make things seem simpler and more stupid till they actually work. Yeah.

B

That's what I was going to ask this is really cool. um I was just going to ask about I guess what you just mentioned was like how what you think it'll take to integrate this with um the different like viewers or notebooks or whatever. So like clerk, the Calvin notebooks, maybe portal um so yeah, it sounds like I can see what you're saying it makes sense that you could implement this for clerk by basically just writing like a set of custom viewers, and then it should be like pretty straightforward.

B

That's pretty cool, but then I guess the hiccup with no pun intended. uh The problem with like uh calva would be the jvm, because right is: is there any way to like what does it? Is there any way for a cowboy plug or a vs code plug-in to get access to a jvm, runtime? Well,.

E

Call myself does have it right because you're connected to the rebel um but I think on that side, the easiest way would be to actually run the thing with the kindly around it uh with the get advice thing around it, because that would be run in the rebel right and then we would get the whole expression with its kind back from the rebel, and that would be basically thrown right into a portal and then you're done, um which is pretty nice because then we basically don't have to do anything on Karma's side anymore um and one of the fun things around covers.

E

uh Like calvas, notebooks and portal listed it's the same thing. Kind of um the basic display is just Eden or like markdown, I, think and like a two or three more but I didn't want to spend any extra time on creating visualizations of use or whatever, because it's so much easier to just display it in Portal. Right and now. If we get kindly integration in the portal, we magically have it in uh Caldwell notebooks as well. That's.

B

Really cool okay, nice.

E

And yeah, as is that the difficulty is kind of getting the kindly function to run in the same jvm.

D

E

I think that can happen if you just like put the kindly call with it as to be executed in the jvm world like the rest of the call is executed anyway, um we can probably have like a special rapper that, like does that for you and you don't have to write the company thing around it right and then I don't know, put some like metadata on the NS or whatever and say well. This NS always gets around every call. This kindly call which hopefully doesn't expose usually um and you're done.

B

That's really cool and then yeah would this be like? Are you picturing trying to work with all of the tool maintainers to integrate, kindly support, but then it would kind of be like an optional like um not an improvement. What am I trying to say an optional like other way of using the tool, so basically like a tool, could you could support kindly for with clerk, for example, and like not use it at all?

B

And clerk would still just work like normal, but then like is that the goal to have it eventually merged into like the main tool, so that, like you, could say, yeah you can copy paste this uh notebook text from your from calva to clerk to portal and it'll all it'll render in all of them, or you can just write the standard clerk way and that will also work like. Is that and I it's up like the vision or whatever.

A

B

A

Yeah, thank you for clarifying that yeah. So, for example, with clerk, let us say we have a kindly compatible notebook, yeah yeah. Then somebody might be interested in using that with clerk, so they may need to call one function to uh adapt click to kindly because maybe clev doesn't support it by default, so that one call of function would just uh Define those viewer predicates that that needs, and then it would just look and so just work for all values that have kind information right.

A

Other values that don't have kind information who just have the usual clear, Behavior.

B

So so I see what you're saying now I think I get it better. Now like yeah kindly like you could include kindly in your project and then call this like setup.

B

You know set up kindly advices or whatever and then yeah and then for the rest of the notebook. It would just know how to render stuff.

B

That's really cool, because then I'm, just thinking of like that's just easier, because I'm like yeah, that can be implemented. That can be like a community effort and there's no need to actually get like. Obviously it would be ideal to get buy-in from Martin or whatever, but if it's, it can be like a totally optional add-on, it doesn't have to be merged into the main actual thing same for and same for portal or whatever it sounds like portal.

B

Nbs code are totally on board, so yeah, that's cool I, really like this I think this is super cool.

A

Thank you yeah and then total has this submit function where you can pass a value and they say typically, you bind that to the closer tab so that all everything which is tapped goes to submit right. So it could just add a submit function that figures out the device before actually submitting. So it could be a you know: A variation of the total submit function and so for different tools.

A

It might look just a bit different how things should be done, and but for that we have time what we need I think is to to convert something that does make the necessary compromises and then just start writing notes already and then the other tools will eventually hopefully converge to being able to view those notes. But you know that is also needed for for motivating the need for adapters actually having notes which could be usable.

A

So that's the plan to spend some time asking for for feedback from Tool makers and then convert to the necessary compromises and focus on writing notes. Actually right.

B

And then actually start doing, work and notes and whatever, instead of talking about talking about talking about nodes, yeah but yeah, but this is the software developer way. You know it's my whole life just writing tools and never do any actual work.

E

um Don't question Danielle um like does the kindly advice function? Can I just throw anything in there and it would basically run the normal thing because so I'm thinking about like the Calvin name spaces, if I just wrapped, every single cell magically with the kindly advice, even like defense and deaths and whatever, because like I, don't I mean I could make distinctions. But I would be wrong at some point right, because somebody would have their own deafen McCall written and that would explode so could I, just like put it around everything without having any problems.

A

Yes, yes, you pass a map to the advice and you get a map where the the map you get would possibly have and additional kind field, and possibly it could change the value. So sometimes it is actually useful to get one value as inputs and then kindly with the kind the advice would say. Oh actually, you want to display that other value, which is a better representation, a better visual representation. So possibly, oh actually, is that problematic right? Should we avoid that, possibly right? We should think right.

E

A

E

Have it in uh maybe have it in a different uh field than the map right having like value, always be the value that you actually passed, so it stays the same and having like display value or whatever, as the extra thing uh that might be understood by the tool or might not be, uh but at least your thing doesn't change anymore. Just the biggest thing that can happen to you is that you now have like extra keys in your map, which usually doesn't destroy anybody in the closure name: closure Community red.

E

Usually we don't care about extra stuff yeah. Yes,.

A

E

What happens if I don't give it a map, like? Is that, as I said, if I'd say like kindly advice and then put deaf something in there, yeah.

E

Because that would be like the important thing for it to just then well, I mean I, don't care what it gives me back, because I mean there's no real interesting thing that a Dev can display anyway. I mean it would be just be the bar right, but virus usually don't have any interesting thing to display, but the important thing is that, like I can wrap every single cell magically in that, and it's still at least that's the thing that should happen.

A

Interesting, so that is now how it works at the moment. At the moment, it is expecting a map where there is a at least a value member in the field in the map, and that the value would be the value right and but yeah I'm not sure, should.

E

We change it, yeah I mean we could also like have a second function right that is safe to just pass anything in right, and that could do the thing and underneath it calling like if it gets in a map, it's gonna, kindly advise. Otherwise, it's just executing the thing and doing nothing kind of uh like my point is just I need something that I can always do right, because calva itself doesn't want to like do interpretation on like your forms and whatever, because it's gonna get wrong. Sometimes.

A

Yeah that makes sense. Thank you. Yeah and uh I see the clock. Maybe some people may need to leave and Lucas I'm tempted to ask you more about kind of notebooks. So, but maybe we could stop for a moment and for those who may need to say goodbye, say goodbye and if somebody wishes to say some concluding words, it is a good time and then those who could stay could chat a little more.

A

C

I need to go so.

B

Yeah continue so much fun. See you later.

C

All right take care, yeah.

B

A

So you don't record it I think it makes sense for those who could stay and then yeah Lucas. What do you think about this idea of passing not only the value but also the form which was evaluated right? That is one of the controversies here and, and the reason is that yeah some tools may not find it comfortable to pass this additional information and it is kind of a user need and yeah I.

E

Mean in my case, it would be easier right because, like I would only pass the forms.

B

E

Because I would just take the thing, wrap it and kindly whatever I don't care. You can like provide me. An extra function and I can wrap it with it, and I will throw that into the rebel and expect the inner thing to run like normal and get some kind of value back that you won't then displayed and I wouldn't do this always because, like not, everybody is going to use currently and whatever. But we can't just have one more metadata thing on the namespace right and then calva.

E

That's easy to read like I'm, hoping I'm, not gonna, make a mistake when reading like a little metadata thing on the namespace and if the namespace is like a kindly namespace or something I'm, gonna wrap every single thing in this kindly call before executing, and if it's not well, then like I, don't need to do it and we're gonna have like some weirdness around some tooling right uh like something like tracing.

E

It's gonna expect like this gonna, see the thing and tell you well you're traced into this thing and you're gonna see the cell and it's not gonna have to kindly thing, because we don't want to show that it's also gonna get cold. So there's gonna be some weirdness, but I don't think it's gonna be like major, and probably it's gonna be easy to explain why more happened than as I was actually written there right. So yeah, no I, think for us. It makes total sense to throw in the whole form.

E

That's like you said. uh The the whole form would be because of metadata right and we're already doing the metadata craziness anyway, because portal is using metadata to say which kind of viewer you want to use so we're already in that minefield.

B

A

Yeah yeah, then, because I I will ask you more about it in the coming days. I'm hoping and thank you for your patience. I know you have been kind of open to talk about it for a while and I I was very slow with kindness and.

E

It's all good, like I've, been busy as well, so.

A

Yeah I imagine yeah as always yeah um yeah. Maybe in a moment we should stop the recording unless there's more and yeah I.

D

Just have a question to other kindly API interface, uh when you register advice is that kind of global immutable.

A

Yeah there is an atom which is the list of advices, and the user can change these atom to make to be any list of things that should happen.

D

Because when reading your notebook, I would kind of expect that I would I should be able to see that in a notebook and if we make that mutable per namespace rather than mutable globally, then the user would have to put it in their own namespace and they could just like they fit in somewhere else and then reuse it if they want to. But I guess that would require you storing that data differently.

A

That is clever, so possibly we could have both and then encourage the practice of of making it namespace specific, but at least have the default Behavior.

D

Otherwise, globally, right yeah, it could all make sense, I mean either you could kind of provide multiple viewers, and then you just uh uh read the viewers nearest the user first and pick one of those and then fall back to generals or yeah.

E

Yeah, it also makes a difference what you want to use the whole thing for right because for a tutorial, I want to like Define all the kindly stuff at the beginning. Otherwise it's just not gonna work, but if I've got like a huge project that, like I'm, using only for myself or where I'm Distributing the whole project and not just the namespace I, can have like one namespace that gets loaded into Zelda, kindly craziness in it, and everything else does need to know about it.

B

A

Now here is the beautiful thing we could keep evolving that part. We don't need to decide now, because the two only asks for kindly advice without specific details of what advice the users can keep evolving, how they specify the devices any given. Notebook will have its own version of kindly oh. No. Actually, there is a travel of versioning because the tool might depend on yeah. It actually might break yeah. So we should be careful but yeah.

A

We should be careful but positively, possibly if we figure out the versioning problem, maybe dividing it into two libraries or something we could keep all versions of kindly working like in all defaults or an old way to specify advisor and and allow it to keep evolving. But that does request separating to two different libraries right and.

C

A

The idea we want an old notebook to keep working, even if it used an old version of kindly's defaults that is actually needed right.

E

A

B

E

Mean like the newest version, could always just translate from like old kind, please to new kind, please. But if it like your old thing, produced kindly old, hiccup or whatever and you've got the new thing running right now, the new thing could still understand what all Dash pickup is like produced a new one right or just produce, both of them or whatever, and think we should be careful because, like at some point, you're going to overwrite every possible useful keyword, uh but.

B

Are you talking about like the kindly Library itself, changing the way it annotates certain types of viewers? Are you talking about like changes in the underlying.

E

No I'm yeah, you can't really remember itself different kinds of Kinds yeah.

E

I mean I, don't know if those are actually useful to change, but uh anytime, you think. Well, this is not going to change it's kind.

B

Of yeah that's a good way to think it always will change.

E

We're gonna Define numbers being something, and then next week we're gonna be like no numbers are actually something else. Yeah.

B

Yeah yeah yeah, but that's right, yeah like if it's internal at least like if it's all done under the hood, really then yeah. You could kind of obscure those implementation details from the the end user, probably assuming like it's still compatible with the I, think one thing that was like how to keep it to keep kindly compatible with, like all past versions of the notebooks I think will be another challenge or maybe maybe similar but I. Think it's again, it's possible.

B

If, like I, don't know, I, don't know if every made every combination of kindly in every notebook could work, but that would be cool.

E

Itself wouldn't be too problematic. The more problematic thing would be like different versions of the tools that show it right.

B

E

They all of a sudden say well, my hiccup now is different and it was like two years ago um and it's just gonna have like a different display right and you're expecting notebooks, usually to stay the same over versions like exactly the same, not my.

A

E

The same and that would kind of break right, but I, don't think, there's any way to fix it anyway, unless you how to code the output.

B

A

Yeah yeah, so here is where we should be clever and we'll keep thinking about it, and maybe yeah and I'll keep asking you about it and any.

E

Comments before making the first few versions, I mean like that's why the first few versions are beta thread or Alphas or whatever we want to call them. It's actually a good thing to break the first few versions, I feel like usually because then you notice that you actually learned something new and like have better behavior for the future.

A

Yeah, that is important. You.

E

I mean that's: why you're on P3 now of currently right breaking.

A

The first two right right right, right, yeah, yeah and it should buy just by this principle. It should remain Alpha. While we start writing notes for a while and and yeah yeah yeah. um Should we maybe stop the recording in a moment and those who can stay can stay further yeah. So anybody any concluding comment about you.

C

Know what you were.

A

Presenting um so goodbye to our listeners and see you on the next time.