DFFML Weekly Sync, 2 Apr 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Weekly Sync 2021-04-02

Description

Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.fi5xc9hajd21

A

I'm not sure that let anybody else in.

B

It feels real slow today, oops, okay,.

A

So and now we're going to start writing friday.

A

A

The copy pasting becomes slightly more complicated.

B

All right april, 2nd great, so um where are we at shaw? Am I getting feedback on myself what's going on here.

A

uh That's weird.

B

I'm getting weird feedback and everybody's muted, so I don't know what's going on with that, all right, um what the hell! That's! Really weird: okay, anyways shaw: what do we got going on with you today.

C

uh I just wanted to give you a couple of updates. First, the data frame tr is ready for review.

C

The should I example is failing there as well. So again I don't know what's going on with that, yeah! Okay! Yes, that's fine!.

B

um Let's find out what the hell's going on here and then anything else going on.

C

uh Yeah, um I'm almost done with adding the exponential smoothing model as part of the forecasting models issue, so we should have a pr on that soon.

B

All right, fantastic all right is that, is that all from you today.

C

Yeah, uh I actually had a small question. um The jesus proposals are due tomorrow right and I have to send them to you or on the mailing list. You can.

B

You can send, I mean you, can send it to me, you can send it in. uh Let's see.

B

I should post the link shouldn't. I.

C

Today's meeting yeah my meeting link expired as well. So, oh really, that's weird.

B

Oh, you know what it is.

C

B

Know what it is is is I'm trying to reuse the same link from the other meeting, and- and so I don't know what we I don't it will. It makes everybody ask to join um so because it's the meeting from monday's meeting and it won't. Let me use the same meat thing. If you do, if you create the calendar entry, it won't. Let you it's annoying, um so I just put it in the description of the meeting and obviously this is maybe not the best idea.

B

So it's just it's rather unfortunate that we have to have a different link each time um but yeah who knows all right so and then hashem. I have the confidence I still haven't gotten around to that sorry, I'm pretty under pressure at work right now. um Let's see.

A

A

Yeah no problem.

B

And then, let's see so uh sahil, what do we have from you.

D

uh I have one more to discuss about that. One person thing and uh the ci lent tools to discuss.

B

Okay, which, which one was it yeah this one? Okay, this one great yeah. Thank you for working on this. um Let's see.

A

B

Oh yeah: hey, hey great great job, um nice, okay, so.

D

So it is actually a different issue, it is not the one person thing is not related.

B

D

Yeah, the one person in this thing is just like making it end and then uh removing the duplicate logs from the logins yeah, but and this this thing that uh uh you know like those bizarre numbers which are showing up those are.

B

D

B

Is related to this right.

D

Yes, it should be a different issue.

B

Okay, okay, okay, great um so yeah. Let's.

B

I think you're correct that it has to do um with the fact that there's no length provided, um let's see because negative numbers, would I mean what is the download code.

D

Because the download.

B

Oh, my gosh, that's the worst! I'm sorry.

B

Block size total size so total size. It must not. It must be returning like negative one for total size, because it doesn't know what it is. Yes,.

D

Probably that's the case with older ftp servers that I saw it in the.

B

Yeah minus one there yeah, okay, okay, so let's, let's maybe just let's see, let's put a comment on this and let's say: um okay, so let's merge this, um because this is this is making it hard to see. What's going on with, should I um let's see- and let's see thank you for doing this- um okay and change like no, we don't want to change.

B

This is rather misleading how it gives you a giant check mark here and then not all the tests are passing them, but can cause confusion.

D

B

um Yeah because this shouldn't affect any of that all right. Well, let's see here.

A

Let's merge this.

B

And then I yeah the lgtm is just I've tried to go update the damn file, um but it's got it's like all this. It wasn't failing and now, all of a sudden it's decided they must have updated something and decided that they're raising failure, but like we're following there's, I don't understand, what's going on the format of the file is wrong. All right. So, okay,.

A

Oops, I don't want.

A

A

B

Perfect great, thank you so let's say we've merged that.

D

The other to discuss is lending rules, so I came up with a very generic rule. uh I had proper documentation, but I don't have it right now. Oh.

E

D

Of course, I I just draw an excalibur off quickly just before the meeting, so I'd like to share it. Okay,.

B

That sounds great yeah. Why don't you let's see? Well, let's see so one percent change pr. Okay, um just done with the exponential spinning model. There's nothing to talk about there. Is there anything to talk about here, shaw or is that just an update.

C

uh No just update.

B

Okay, great um so, and then the other thing from you sahil was you want to talk about the yeah okay, so.

D

The formatting limiting of commit master yeah commit.

D

B

Okay, great, thank you um and then hashem what else from you anything else.

F

uh Yeah, I actually pr uh a sample of uh the use cases.

B

Yeah I was looking at this.

F

B

They're very nice yeah. um Let's see.

F

um You can view it as a file.

B

Oh yeah, that's right, yeah what I had looked at it the other day. The only thing I'm I'm I'm wondering about. Is this the way that we're doing imports? You know we have a lot of examples that show imports like this, whereas we have a lot of you know we load, we can load the models via their entry point name by passing it to model.load, and I realize you know there's no, it's not document. So if you go to the plugins page.

F

Yeah, I know how to load those, but it won't allow you to. uh You know currently uh load multiple models. uh It's either all of them for just one.

B

F

About them, dynamically.

B

Yeah I'm just I'm saying though we can just do you can say, um let's see.

A

Let me view this file.

A

B

Yeah you know I I I I think what you've done is great. I'm just wondering the thing is that we don't have so we need to change something about our documentation, because the way that we do- um let's see, for example, model pi, torch, right and- and I think, there's there's more than just this, but we have let's see where is it?

B

I guess it's all sort of no there's no python example here yeah. The problem is, I think, the pro yeah well in the problem is we don't have a python example, but the thing is you don't know the import path. Looking at this, you just know that the entry point name right. So if I went to go um right, I know the entry point, but I don't know the import path. If I wanted to do you know from the fml model, pytorch import, one of these that's not going to work.

B

I don't think I can't remember, but I think we have a few of these like this um scratch. Where is that scratch one? No, this one does work yeah, okay, logistic, regression, yeah it's just there. We have an inconsistency in the way that this works right, so consistent behavior would be the this is the question: is you know what what? What should the? What should the behavior be um and consistency wise?

B

The thing is, you have yeah, so this gives you directly a class if you import the class right. If you do a model.load, then you'd be like logistic regression equals model.load and then the entry point name right um and but that just that adds. I guess it's you swap this line for that, and then you have the fact that it's documented here right. This is so it's more of a question of like how we either need to do it.

B

We need to do it. We need to have some more consistency or we need to have some more documentation in all of the models to make sure that we know what we can, how we can import them right and I'm kind of thinking at this point it would be you. You basically have two choices: right, audit audit, the models plugins page here to see you know to make sure that we have this example for everyone that somebody might want to use or switch it to logistic regression, equals model.load and then um scratch lgrs sag yeah.

B

So you see what I'm saying.

F

Yeah, um I actually tried to load two models at a time and it doesn't even allow that current.

B

Well, you did you try to pass two entry points at the same time,.

F

uh I I passed them separately.

B

Try to pass them separately and it didn't work.

F

B

Okay, let's check that out because that's that's a big problem.

A

Okay, okay and that's, this is doing a really poor job, taking notes.

A

A

All right so and let's see moving between models.

B

Okay, anything else from you hashem.

F

uh Yeah, I also uh wanted to ask if uh we really uh have to test the notebooks I mean uh I have created two examples. If you can see uh there are, you could say two workflows of uh the ways we can do this.

F

uh This one is a unit testable while the other one isn't the other notebook in this same.

B

Pr, okay, wait, one second! So oh fn! So what's the difference here, let's see so which one is testable and which one's not.

F

uh The one without fn is untestable.

B

Okay, because you didn't put it in a function, is that why yeah yeah, okay, yeah and wait. So why is this not testable.

F

uh I mean like uh the way uh textbook works, is it uh extracts functions and then you can.

A

B

F

B

Can you can it just extract.

B

F

um It didn't mention, but I'd have to check more.

B

The other thing is like this: we could just extract the source from each one and run the whole thing. You know, because this is just I didn't realize that these are just like a nice little json file, so this is pretty sweet.

B

um This is very convenient um because yeah we can just extract the um source from each one of these write it out to a file and then run the whole file right and make sure the file works.

F

B

Sure. But at the same.

F

Time uh they are quite uh self-explanatory, like uh they at the spot. They give you the output of the cell, so I was uh trying to you know, understand the uh why we need to test these.

B

Well, we need to test everything. We can't not test something um because as soon as we don't test something, then it becomes out of like and we change something. Then it becomes out of date right, like or out of sync and.

A

B

It doesn't work and then we're showing users documentation that doesn't that doesn't work right um and so that it also means that you know if somebody makes a change that you know. If somebody makes a change, they don't know all of the pieces of documentation that need to be updated unless the documentation breaks. You know because then it'll flag it in the ci.

B

So it's really it's it's more for it's it's for users and it's it's a lot of it's for users and it's for contributors right, because that way you don't have to go, and you don't have to know ahead of time every place. You're gonna need to change. It just tells you when it breaks so, and this doesn't look too. This doesn't look too complicated to test. I think you basically just like you.

B

If you want to do it on your own, like if you can't use that thing, you could basically just grab, grab the source from each one of these write it all to a file, run the file and then make sure that the output is the same right.

F

All right, uh yeah yeah sure there are other ways to extract a python file out of the notebook as well. Okay,.

B

Yeah, whatever whatever you want to do here, um but yeah, we do need to make sure that they're tested. This looks great, though sure nice job.

B

um So let's yeah, I think, let's, let's do, let's switch it to model.load right and then let's switch it to model.lo load and then we'll come back, and we may revisit that. You know that decision later. If we get more consistency in the plugins pages, but for now um for now- and we know- I don't think we have enough consistency there and it might be confusing. So we more consistency better and I'm realizing. We need to try to change the rest of the docs too. I realized this the other day. um So, let's see so.

F

So, which one is your preference out of this? I.

B

Would prefer, I think this is more. This is this, is what people would probably write in a notebook right? They wouldn't write the phone because.

F

If the notebook supports top.

B

Level await, then, then, that's what people are gonna do right so, and this is less code. This is clear, you know, um so this looks great. um The only thing is, I guess, you're gonna, if you run into these like file blocks and stuff yeah. This is probably something where you want to use some kind of testing library right, because they're also going to start having to implement all these random things. um So.

F

Yeah right yeah.

B

Okay, great, this is great, um so we need to be able to test it um prefer.

B

Functionless and and what was that, the paper mill, what was it.

F

uh Yeah, it's that book. uh It's a module of paper mill, test book; okay,.

A

A

I'm just curious here.

A

How partners work.

A

B

Execute pass to notebook okay: this should be pretty easy to integrate into our existing testing infrastructure. So that's good, um because it's old python.

B

And you're saying, but are you saying that this doesn't work unless you put functions in there or.

F

No, not this one. uh I was talking about the test book. Okay, okay,.

B

This, because this looks I mean, will this raise.

F

B

I guess that's the question: does this raise errors? You know making sure that it does is key. So.

F

I'll have a go at this one.

D

All right great: yes, it does raise errors like in proprietary, stuff, okay, great.

A

Great um so, let's see.

B

All right um this one, I'm gonna, still have to review this one offline, because this one's too big, so I'm sorry just it just requires a lot of time.

A

Okay, so let's see.

F

I have one more thing yeah.

F

um I uh message you about it. uh Can I uh add, like demos, to demos, to my project uh like the ones who don't shoot it. I mean.

B

Yes, demos equals great yeah. Demos are great, more more more demos is, is more more better yeah. um That's that's.

E

B

um And when you say demos with like what do you mean just so we're on the same page.

F

uh Like uh I'm, not particularly uh highlighting a specific use case, but.

B

F

Building around a data set.

B

Okay and what and what's the data set.

F

um Just a second.

F

uh uh There were two demos with cli examples that didn't have code in the documentation: yeah.

G

F

Handwritten digits and the flower species classification and other that I also have taken out some data sets like heart disease. uh Uci data set.

B

Okay, I'm not familiar with that.

F

Yeah, it's supposed to you, know you're supposed to predict. If someone could have heart disease, heart diseases or something okay, yeah.

B

um Okay, so we want to add code um to non side. Note so side note uh if we see docs that are only cli examples, so this is just a side note to this discussion. If we see docs, if you guys notice that docs only have cli examples, we want to add python examples to those as well.

B

F

And, and could I add, uh notebooks to those as well.

B

Yeah all right, okay, the yeah. So the thing is okay, so the notebooks we're going to need to figure out. How do we display the notebooks? I mean we can just reference people to the github, but it would be nice if we could figure out how to display the notebooks in. Let's see swingsidepython notebooks thanks, I python notebook.

B

Somebody must have done this right yay.

B

This seems like something someone would do: yeah 2014 now: okay, that's not it's not optimistic! um Convert to rsv.

A

Oh well, that's great um how to.

B

Non-Trivial to use with read the.

B

B

Yeah, it would be nice if we can figure out how to make.

B

um Oh god it would be nice if we can figure out how to make the ipython notebooks display on the website um just so that all the docs are in the same place.

D

I have seen such notebooks being displayed in read the docs websites, but.

G

D

They didn't have the theme we are using. They have that pandas kind of theme. You know oh yeah,.

B

Yeah and it probably tied in well, we can switch the theme, it's not a big deal. um You can switch this name all day, okay, great! So let's change the extension of the parts source parser, oh yeah- and this should just be thank you.

B

Oh yeah. This shouldn't be too hard to do. Anyways. I was just looking at the way that the markdown files get converted and basically the whole thing has this: they have this like intermediary representation that doc utils does so. Basically you you parse the files you build. The syntax tree like you, build the you, build the you, build these node objects and then so, basically the to the output writer. It doesn't matter if the input was rst or markdown you, because they built these node objects uh in memory, and then it writes them out.

B

So you this looks like somebody probably uh when it looks it didn't work for me: um red notebook, okay, yeah, so we'll we'll figure this out.

B

Yeah look into it yeah. Let's just see, this is still around.

A

Okay, great yeah, shanks extension uh yeah. Let's just make sure they still exist february, 28th, okay, great, hey, sweet.

B

All right, nice, very nice, looking good, oh yeah, hey darts, oh yeah, yeah yeah, oh yeah! This is beautiful, yeah, oh yeah! This looks great. Oh we'll have to steal whatever theme they're. Using what theme are they using.

F

um Yeah, this looks good.

B

Oh, my gosh, all right. Okay, this looks fantastic yeah, great okay, um I'm glad we figured all that out. Okay, yeah that'll that'll be nice; um okay, so.

A

Thanks again, all right, okay uh touch the paper we want to be able to.

B

To display them on the website on the docs site,.

B

B

So then last thing so 20 miles. No bucks uh model.load wasn't working for multiple models and I wonder if this is ipython specific, so we're about to find.

F

F

uh Could you use the same two models just to be sure.

B

Yeah, let's do it good idea.

A

Where'd you go okay,.

A

A

A

A

Okay, so slr and linear.

A

A

uh Psychic all right.

A

Yeah, oh python, 2.

B

Get out of here.

A

No one'll name data classes; okay, what version? Oh god? What machine is this? What.

B

No mod lie. That's a lie: python 3.6! Why are you using python 3.6? Where did just come from what is going on? Okay, all right.

A

A

All right, we'll just create a virtual.

A

B

Ridiculous ridiculous, I tell you, I don't understand how I don't do anything and my python environment gets messed up. It's just.

B

Well, it looks like they're loading for me um so yeah, I don't know it's loading for me. It might be an ipython notebook. Specific thing might be an issue, that's the specific type python notebook. um So, let's see.

F

I'll check again and uh follow up on this okay, great.

A

It works with uh in a regular python. File might be notebook.

B

Specific okay um and then nitesh. What did you want to talk about.

H

I want to discuss the issue that is 8 31, where we are going to save the config dictionary of models.

A

B

All right: okay, we're gonna oops. Where did I go? Okay there we go all right, let's wrap up, let's see so confidence pro move between models, notebook done confidence, pr offline. Still, sorry too long um did from pr that one's probably going to be offline too. Unless there's any question, let's see this should be a quick thing. So, let's just make sure because oh yay see I passed.

B

Okay, so there's still a lot of changes in.

B

C

Yeah, it's still showing me uh my uh the changes from my previous pull request, even though we closed that- and I don't know where it's doing that so, okay, I'm not exactly.

B

But this is the branch right.

C

B

Okay, this new github- they finally rewrote it and go, and now it works.

B

Source panels, data frame- okay, I think a merge commit happened here. That was not what we wanted, so I think we need to go through and source trading white.

B

Space all right, so what happened.

A

B

Search data frame move yeah. I think that merge sent us into a bad spot. um So, let's see, I think what I would recommend is, if, um like the question is really like: where did you diverge from master? um I think I'd recommend just pulling out the files create a new branch, pull out the files and put them in a new branch because something's a miss. You know.

C

All right I'll, create a new branch, and so uh okay I'll create a new branch and submit a new pull request. Okay, great yeah, uh something.

B

Went wrong with git um create.

A

It is annoying new and just put changes related to this pr.

B

In it um and things like so, if you did like it check out, that should be new and then you did get check out. Oh wait.

B

Why is tab complete, stop working everything stopped working, ah I rebooted, okay, um so yeah. So this is your new branch right, it's on whatever, so you can do git check out and then what is it.

B

What everything sucks now great.

A

uh What was it called? What was that branch I lost.

B

It now um oh well, you you can do git checkout, you can say like get check out, um yeah, yeah and then the file and then you'll be good. You you, you got it um yeah great um okay, so I have a hard stop here at 10, which is why I'm sort of frantically going through things all right so and then we have sudanshi.

B

Okay, sudanshi: what are you? What do you have uh today.

E

uh I had to ask some questions regarding proposal. Okay, yes, so there are two questions, so are we taking it right now.

B

E

Okay, so my first question is like about the sql scorer method. Is there, so what I actually figured out is that every uh every model itself also has a score method.

E

So do I also like need to implement that as well in the proposal.

B

um Well, I mean, I think that well so every model has its own score method right, just kind of.

G

E

Kind of similarly to how.

B

We were doing originally right.

E

Yes, originally.

B

So you probably can just make one score that calls the score method of that model right. Does the model still get passed to the score? I can't remember how this works now.

E

Yes, more context gets passed.

B

So that's just one score, then so that should that should be pretty low maintenance to implement right.

E

B

Okay, okay and that would basically yeah that would effectively replace the existing. What we had like you know. That would effectively be the same thing as the test that existed in scikit. Before this. um I think um I don't know well yeah, I would just say, implement implement it as, like.

B

You know a path, the thing that just says you know: okay, let's see so um each scikit model has its own score function um and you, you know just make a score for this um name it something name it something that implies it's going to call the the score function of the model. You know, yeah like like scikit score or something like sky scikit model score or something or something that uh that currently describes what it does.

B

Okay, yeah, I think you know, I don't think that should be too much work right.

E

Yeah, that would be it. Okay,.

B

E

More question is: uh what are the expectations in the in the in the gsoc project? Pre-Made data flows for the data cleanup.

B

What are the expectations.

E

Yes, so actually, I was actually like. I figured out like what are the cleanup operations that we can perform and so like there are like, uh like I figured out like 10, 10, 10 or 12 of them, but uh originally like, and I'm planning to implement, like seven of them. Okay,.

B

E

For whatever, like, I have like, decided what should be like the expectation for it, okay,.

B

E

B

E

The proposal yeah.

B

What should be in the proposal? Well, so, uh okay, so.

B

This is a quick one, so so, basically I would. I would say that um you do focus focus on example. Data sets make sure that you have.

B

You know three to four example: data sets that you're going to clean up uh right like write tutorials, where you use the operations you've decided to make to clean up the data.

B

Does that make sense.

E

Like oh, okay,.

B

Look at data like finds, find some data sets that require cleanup right and then your project focuses on implementing the operations and then um and then writing tutorials that show. You know why, like that, that show the value of that right through through real data, set examples.

E

B

All right, um okay, anything else, sir.

E

Oh, no, that's it! Thank you all.

B

Right, great um so 8, 31 man, I keep trying to type this issue and I get it wrong all right. So, oh yeah, this is rather nondescript. Isn't it um let's see so this is basically I believe that what was happening here was, um you know, models right now they have their config dictionary and they like the way that okay. So let's look at this.

B

um Yeah, where is that a dot.

B

B

All right there, so here's the config dictionaries right.

B

A

A

Does this have an export now.

A

All right, bullsh, oh.

B

Okay, there we go.

B

So each of these has a config dictionary right, um and so, if I wanted to load so if I wanted to save this model right, the model gets saved to its directory right, but also okay,.

A

uh I wish let's.

A

H

Okay, yeah, as as we are saving the models right so yeah, I didn't get what what's the use of this saving the config as well. So.

B

Yeah, that's what I'm demonstrating all right, yeah one! Second, um it will become clear in a second, so basically.

B

uh So so um loadable.

B

B

Okay, so when we instantiate so when we leave the model to to save it,.

A

Effect type exact.

B

Value trace back okay, so when we leave the model here so this is the you know, example example loadable um and so we're going to add this class method.

B

um That's like you know that, basically we're going to add a method that um we're going to add a method that allows us to load from the directory or file. You know within the directory right, and so we can basically just say um uh what is I'm just going to name it a because I apparently can't think um so. We when we exit, we say self.config.directory, you know slash a dot json, because I can't think um and then we write out um the contents of the config dictionary.

B

And then that way we can go in in this a method here. So this is model one, and so we saved the config of model one and then in a method. We basically say you know instantiate the class, um and you know if I'm given a directory load, the a file read text, json loads.

B

Star star json.loads, so basically, when you, when you leave the model, save the save the file or save the config to the file, and then you implement this method that loads the model data from the file right, but you probably want to override the directory here to be. You know whatever the directory is now just in case like the directory got moved right. So um let's see how do we do that.

A

Or yeah directory.

B

B

The other thing is that, like you know, this file name is sort of susceptible to okay, fine, be like that. This file name now is sort of susceptible to the fact that, like what, if something overwrites it.

B

Oh, I put that in the wrong spot. Didn't I.

B

Do some loads yeah, so this file name is now susceptible to the fact that, like something something could overwrite the file name, you know and then let's see, where'd it go. Curtain accuracy was never awaited, oh because we're doing from dfml.no async import train.

A

B

Okay, so now we can basically take um you know, and this would be we do this on on model itself. Actually we can just do it on model, um and then we can take this and say. Example: loadable, you know dot a.

B

I guess this is yeah whatever you know, you're implementing this on the model based class right, and so then you can just say um model one dot, config director and now it should give you oh yeah, it's basically.

B

Basically this type of thing: oh yeah. This is directory.

B

Come on json object, really sure blah blah blah. Oh yeah.

B

You get the picture, um do you get the picture.

B

Does this make sense.

H

B

Okay, so this is also sort of a part of that archive storage, one. So, basically.

H

Let's say file or directory, so at the high level we are saving the config so that we can manipulate uh yeah.

B

H

B

Yeah we're saving the config. Oh that's why um we're saving the config, because.

B

We, if we, if we.

B

If we save the config on x, like on exit from the model, then we can.

B

If we save the config on exit from the model, then we can use that config file to like reload the model right, because the model has state in the directory and it has state in the config and the config should always be serializable right, um and so, if I wanted to load the same model like, if you guys you guys are pricing, joblib and stuff.

B

If I wanted to load the same model right, the thing is, I would be pointing at a specific directory, so I need some sort of like file, and this is if this is why it feeds into that other project. um Where we're doing um uh you know the the directory becomes location and stuff. So this is part of that too. This is like a pre-workout. This is like this needs to get done as a part of that.

B

um So basically, if you write out, so you write out the context of the the contents of the config somewhere in that directory right, and that way you can basically just use that config file to reload. All of the parameters that were in oh well, this needs to be like that too. um You can use this config file to reload all the parameters like that, all of the properties of the models config um and so right. So we we basically just um does this. Does this?

B

Does it make sense now or because, basically, you save all the config properties, you re, you load all the config properties from the file, and then you override the directory property to be what it is now just in case somebody moved the directory because they dumped out the directory with the rest of it right, and that way you can load.

A

B

Okay, so this is for some. I don't understand why it is not working slr model model, a dot json, I don't understand it- doesn't make sense um because it should have it should have trained. One should have should have done that, but that's okay, um whatever.

B

Public uh uh issue 831, finally got it right. um If, oh, I shoot 831. um so need to make it so bad.

A

Config properties.

B

Are saved, slash loaded from file or can be.

A

A

A

All right there we go so we'll put this up there.

A

We will see.

B

Okay, so let's talk about that linting of commit messages, real quick! Do you want to go ahead and present sahil?

B

D

So can you see my screen great.

B

D

So so this is the this is the generalized comment structure where we have our comment message at the end and various subparts separated by colons, uh so it could be like three or four that we don't have depth beyond four. I guess so, I'm just trying to uh generic one and then what we would do is like take this parts and add them to the root uh and make a path out of it. So it should be a directory or a file. uh Extension of the file is not necessary.

D

If it's not our directory, then we will check for if that's, there is a file with that name. Yep, and this would need to just some things would need to change like if we are putting something inside dffml, we would need to start it with some specific words so that I can replace that specific word and I would say, like we have something like core: okay.

B

D

B

Is we're going we're about to move basically everything? That's not core out of it, so we can safely assume that you know if you see slash dffml in the first part, then you know you should you should remove it right.

D

uh No, no I'm saying that just if, if a commercial message starts with core that means uh that file or that path, this part, if, if it is code, it would be replaced with tffml, that's it. Okay,.

B

Okay, so, and I'm saying, though, we have an existing, you know I'm. I think this is, I think in theory, this is good, but we're I we don't want to change the way that it works right now, because the way that it works right now implies that you know you're doing things in in the in the core package right like so, if we all of a sudden start, prefixing everything with core everything's going to be prefixed with core right, you see what I'm saying if we move.

A

All the second party plugins.

B

Out then, every we're just going to have all our commit messages are going to be prefixed with core and then core doesn't really mean anything right.

D

Okay, so we can like, uh if we don't find it in the root, we will look for it in the ffml.

D

We can do that yeah exactly yep, okay, so I'll change it and, uh for example, if you need to address some things like if we go up in the meeting minutes- and we see like we have some things, starting with the maintenance or plug-ins and some that sort of stuff, so we can have a prefix suffix symbol to say that, like this, this thing is not a part of the path and we can ignore that safely.

D

Okay, so these are this is this is the basic thing I had on my head that.

B

Sounds good yeah? No, this sounds great yeah. I want to hear more about four um because I think you know that's. That's. That's. That's things like so this is stuff. Like you know. We talked about cleanup and.

D

Yeah stuff, like that, okay yeah, let's.

B

D

Enumerate those yeah, so we can like put a dash or something like that, uh and that would just skip that part like if it is in the start, and it has a certain symbol of prefix or suffix. We can decide on that whatever it is okay, and we can just ignore that safely, like that, it is about plugins, but it does not have to be uh inside a plugins directory like that.

B

Okay- okay, great all right, so, let's and and I think I think, you're probably good. To start writing. You can probably start writing some code right now and then you know you'll you'll! You have some variables that you can throw in. You know for four, um but you can probably start writing some code at this point right.

D

Yes, yes, all right and another thing that would be I'm.

B

Sorry, I have to run. I have to run right now, um but um can you send me this question on gitter.

D

uh I have already asked it on.

B

Okay, you've already asked, and I github issue: can you link it link me to it, because there's just so much going on it's hard to see hard to get everything great all right, thanks everyone. um If, if we didn't get your thing, please um you know please please raise raise on on gitter here and and ping me um ping me directly too, if, if you're, if, if you haven't gotten a response, all right thanks have a good one.

B