GitLab Pair Programming - Frontend, 24 Aug 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Frontend Pairing - Jest speed reporter (Part 2)

Description

We continued working on: https://gitlab.com/gitlab-org/frontend/jest-speed-reporter

A

Hi, everyone hey what's.

B

A

I I will just uh I will just continue bootstrapping it here. Yeah, I'm.

C

A

I'm gonna kind of.

C

Do some stuff on the side.

A

Yeah yeah do that um it's should be pretty easy. So what do we have here? So I just copied my script. That's the script right, so we can do a yarn and what do we need? Axios? I think axios. Why do I have moment? I don't need moment. I don't need fail, that's good axios moment dot! Moment.We! Don't need right. I think. That's it yeah! That's it! We just need axios. For now um I just made it a private project.

A

I had already added the xml to json right, let's create a new folder, let's call it the git lib and then we can. You know new file. Api.Js.

A

And should be pretty easy, so what do we need to do? We have a gitlab ab ar this is all don't care about this stuff. What what is happening had some weird focus, probably um api.js, and let me just pop up enable coding assistance for this um const axios.

A

I don't think I need path here at all and we can just create the extra api and generally what I just do is like if process.n.

A

A

Console error: you need to define a.

C

Gitlab token right is: um is this something that says are these reports? Scene has to be authenticated to yes, okay,.

A

uh You often have to be it's really weird. We have some apis where you, um where you have to be like authenticated and it's annoying and yeah. Okay, uh we wanted to get lab api.

A

So this one is actually um jobs, nice yeah, so we we have a job iterator, but we don't really need that. We can that just do like a have a gitlab api iterator right, um because we basically just need a path, and in this case it doesn't oops in this case. It's just like uh api path and we can be api path.

A

Api path, api path, pipeline id can be gone and then it makes it a bit easier and we just iterate over the api path right um which allows us to be to do the following. um Const um get lab api.

A

Is it even named correctly like yeah, okay, okay,.

D

I'm not sure if the spelling's right, I'm terrible at spell.

A

Spelling and that's why I just copied it also: api iterator is a bit a bit weird, um okay, so.

C

Let's capitalize.

A

On git lab, so what what I would want to do is um oh yeah, the gitlab yeah yeah. No, I'm I'm not good. I like should be right right and the cool thing is now. We can just do a gitlab pipelines api.

A

Do we need that one or do we need the jobs uh api? So we we need the we nee. We need the latest pipeline for the project ah gotcha and then we're getting.

D

A

Yeah, so we we just need like uh pop-up. Where is it.

C

Can we can we do something smart too? Are we able to tell? Can we do a find pipeline, so we also guarantee it's a pipeline, that's successful and has um the jess job yeah. We can.

D

It's easy query.

C

Params, oh or is one of the job is one of the params include like job or something.

D

No, we can just make sure we get the latest pipeline. That's successful in past.

A

Let's, let's we make it watch the second paul you're. Getting ahead of yourself is a new gitlab api iterator, and we just have that url right um and what's the pipeline id of gitlab e, oh gosh, I don't.

D

A

D

Id uh the project id.

C

Oh gitlab work slash gitlab yeah, oh the actual id, though.

A

Yeah I mean we rene renamed the project once right. He did rename it once.

B

C

A

C

It did the id actually change, no.

A

The id that's the cool thing about an id, it doesn't change right. So, let's, let's do this? uh We want to have status.

A

Success right: we want to be master as well correct, yep, that's right, so we want ref to be master.

A

Obviously we need to teach the iterator about. Why doesn't jump there? We just like uh you know, like params equals, and then you know this params equals this, and then we ramps that we're passing it yeah you're right params, and then we can be like params.

A

This is a bit awkward and I think it could be reflected right, but the cool thing is like this params, because we obviously want the I mean. Maybe we want to overwrite per page, but the page at least should be there by the way. This iterator is actually pretty simple right, so it has an async. Next, as long is like. Maybe we should rename commits to like uh something more generalized identities or items is good.

A

I like items right so basically how this works is like uh unless you're on the last page right it uh it iterates through and as long as you has item have items it just you know, returns those items by the way. This also can be rewritten a little bit with a generator pattern, and then it's even more nicer yeah, because then you can just yield the values as long as you have them, and if you have no values anymore, you just jump back into your loop. I have that iterator somewhere lying around.

A

I wonder for which tool I've written that one, but anyway um should like be like good right and then we can be like four await uh and then pipeline off pipeline. It's a I mean it's an instance right, so it should be.

A

D

Yeah, that should work, but.

A

D

A

At some point right, yeah yeah, that's that's right, so we should have like a count or whatever right.

C

um If we run it right now, it's gonna just keep hitting it.

A

Yeah, uh so now we can all find where my token is right. We have this like uh tenth.

A

No, where is it like yeah there? It is yeah eval. So oh gitlab iterator is not defined. Gitlab api iterator. What is not defined yeah? We should properly export it as well right. So now we should get pipeline for pipeline. You see that oh it's yeah. Maybe we should uh and then in this loop I would simply you know, be like hey: let's create a jobs, iterator right.

A

New gitlab iterator and then we are just like project.

A

ah It's it's the same url, essentially just with.

A

Just with jobs in the end right and the pipeline id so yeah pipeline dot, id and jobs right, and then we do for a weight, const job of job. It's an italian one, apparently.

A

um And we can be like console um if, okay, let's be like const, just jobs, equals um empty array, an empty array. Exactly then we just drew like job name starts with, and we just do just space right. We could, because we have the jest integration, one which we don't want. um You know, but should be good for now. Then we just then we just do like a const data.

A

Gitlab api data equals gitlab api get and it will be.

A

Projects so essentially like this, but this time around with jobs.

A

And it will be job id right and now I need to cheat again.

C

We had built a um graphql query that uh there you get some of this um slash artifacts.

C

D

C

I like where this is heading.

A

So this is getting all artifacts right. uh Download get job, artifacts, okay, get a single artifact, so this is also really weird by the way like single artifact by job id.

A

What the heck like drop, shot by the artifacts and then.

C

We have the artifacts path that.

A

C

That's all all we need I'm fine if we just hardcode the assumption that we know yeah. What was the name, I that sound that rings bells.

A

Let me just check like real quickly because we changed it yeah. ah If you have it easier, you know what we're going to do. We're just going to push this wait a second and then we can be like hey if justjobs length, we just have break right and.

B

A

Console.Log jobs and the cool thing is, I think, the jobs, if I remember like output correctly.

A

Jess jobs is undefined. ha Yeah yeah give.

C

Let me move that to a um can. We move that to a function.

A

Yeah, that's a good idea. Let me let me let me just you know, uh yeah you're good.

A

For now we should be, but it ended somewhere right like ah and we can see here, and the good thing is that we actually have artifacts here as an array right. We should be able to. You know, that's a good. You know: const no async function get latest pipeline with uh or get latest gesture effects. Just just, I think it's get latest jazz jobs, yeah.

C

Yeah, I think that's good.

A

And I mean we can, you know, be like this is cool and we just you know, return.

A

And cons, just jobs equals weight, get latest chest jobs and then we can be like jess jobs, map x and then we do console vlog x.

C

Yep can you I.

A

Know it's 13.

C

Sorry on on line 13, the const.

A

A

I mean it shouldn't be that bad right like no well yeah, oh yeah, there we have it and you can see we have like junit xml, dot, gz right, artifact.

B

A

Works. This is by the way this is funny, so I think we actually need to change something in our pipeline by the way, because there is currently no api to retrieve this, but maybe let's just try it out. So we have you talking about retrieve the file type, so this has the file type junit and, as far as I know, right now for job, it's not possible to get the junit stuff right.

A

So if we have the jobs api, you can get the log file under trace right. So you can get like the that's the log right, the trace the job log, but you cannot create get the this. The thing is really sorry: we should be able to get the artifact right.

A

No, it's uh that's what I'm just going to try out right, so we have like the download single artifact from from by job id right. So, let's what? What do they have as a comment here? So this should be the file right?

A

Okay, now, let's do uh let's we can not. We don't need to do this to test this. We can be like const data is gitlab, api dot get and we just be like hey and we all don't need this, but let's maybe just take the latest job. So this is the job it.

C

Might be easier to actually map like from the job.

A

Yeah but I didn't want to go through it.

C

A

You're doing you're good yeah yeah, it's just more like hey, let's try it for for one and it should be junit xml, dot, gzip yeah. So let's try it. ah We also should do something with the data right.

B

A

Yeah- and this is the annoying part, because I think if I switch it out with the coverage front-end, one.

B

A

No okay! Well, that's good! He's scared! That's good! So maybe maybe let's go to the job I page go into browse.

A

This is the file name right, so maybe let's try instead of the gzip yeah artifacts. Let's just try like the junit jest dot, xml.

A

Undefined, why are we getting undefined undefined is rare right.

C

I think we're destructuring.

A

D

Maybe that's people without.

A

We need we need to await.

A

D

Okay, good, isn't it beautiful beautiful? Now we can convert this to json most likely and make it a little bit easier to build a front end.

A

Yeah- and I that's that's basically the stuff that I wanted to give you, because I know that I already have written this right. Yeah.

B

This is super helpful.

A

Yeah, so we just do this.

C

Oh cool, so let.

A

C

um Yeah and keep going ip, but so the way that you've done a handful of the handful of these pages projects.

C

You have, I correct me if I'm wrong a ci job that runs this script and creates like on a schedule and then creates like the static data that then your front end queries that static data got it. That's how you can circumvent the permissions stuff that exactly.

A

Exactly and it that makes it also easy, because it's just a static file, um it has other benefits so often, for example, here we are likely to have some json artifact right and what I'm also doing is I'm putting that json artifact into um into the artifact of that job, which is great because right now we just generated for the moment being. But if we have the json artifact and at some point are like hey, maybe we want to track some stuff over time.

A

We actually could go back and you know consume those json artifacts that we've built- and you know, you're, on independent from the gitlab pipelines and from red masters, and it's a bit more reliable, essentially right. No, I I love it and we.

C

Could even do an um ip and peyton what we were talking about earlier, like we can even do notifications in that scheduled job of like um yeah. That would that sounds really awesome. I'm excited about this yeah, I'm glad we went this route.

A

Yeah, okay, so what what do we need to do? We? We need to get like the the job, dot id right or maybe just do job here right and that's the data, so we have all of the data of this one, um so we can be like.

A

So I would I would go ahead and parse this xml first yeah yeah, let's, let's, let's do that? Okay, so const xml json equals require xml, json right xml to json.

A

And it's probably like parser.

A

Jason equals parser.

B

C

Think it's probably suitable.

A

Is it positive pass who do you think, son to json?

A

It's just two json.

C

You're right you're right, you're right, I'm sorry! Yes,.

A

C

Yeah, it's parser.json.

A

Is it positive to json so.

D

C

Can just destructure.

D

A

To jason right, yes,.

C

C

That, okay with you, I'm fine with that yeah, I'm gonna rewrite this in d now anyway. So here I'm I'm joking ip.

A

It's fine with me right, but.

C

What's cool is like this whole gitlab api iterator could be in its own file and like we could reuse it across projects. That kind of thing is really cool.

A

Yeah, I actually, you know, wanted to take the iterator and publish it, because I think it's it's really neat, because all of our rest apis worked like that and yeah. I think the only feature it would need in its current state that you have like hey. I want to have like a max thing right. I don't want to go through a thousand pages or something um by the way.

D

A

Also, how you can improve that's the reason why I didn't have the per page, so it's very unlikely that we go through. Let's say a hundred right. Often if you're like I want to have all commits of a project, you just say per page 100 to paginate faster right, but here, for example, it actually would make sense to like pay per page five, because it's very likely that one of the first pipelines actually has a successful jazz job right.

B

A

So you have smaller api results and whatnot.

B

A

So, um let's, let's, let's have a look. What happens here and we also probably should add some oh yeah. This looks good, you know what we're going to do is like hey.

A

I bought a board okay, so it says it it's jason. So maybe we do a const write file sync equals require fs right and you need.

C

To write anything here.

A

I I just want to write it because it's easier to go through a static json file than to some lock output, to see how.

C

Oh just so, we can find it.

A

Yeah, so we can just be like write file, sync and in the end we at least need to write a json file or an html file to look at right. I I prefer, for these kind of.

C

Like for these kind of scripts, to let bash do the piping if and like, let's just control, what's this output, and so I don't even have to worry about like receiving a certain input path or mess with anything like that. Yeah, that's that's actually smart, but this, uh but for just so we can see it makes sense. Yeah.

D

So we want to write the ci job or do you want me to do the front end or.

C

Yeah this sounds like a lot of fun um either way or are we doing it all now.

D

C

I would, I would hope.

D

You didn't have any.

C

Plans today peyton because well I think I think, once we get the data, um I have a few things I need to do today besides this, um but maybe, if you want to pair on all of it together, maybe monday peyton well, I can pick up some of it today.

D

I've got four, mrs in review, so I can just put something cool: do you.

A

See that guys that already looks like pretty good- let's maybe format it pretty soon. So I'd like.

C

Yeah, while it's formatted, because it's humongous um and one optimization we could make- is maybe instead of transferring it all to json and reading through it, maybe there is actually an xml reader that we could use because here's the end format, I would think, would be super helpful.

A

um You know you know to be honest, like this doesn't take long at all, um so I would just be like the only thing is like we have. This test suits global object, which we don't need, because it's basically the individual thing of those four I mean it re. Basically, what I think we are interested in is all the test suites, which is like this array right. So the only thing that I just wanted to know is like hey, give me the the test. Suites uh suites. Is this one right?

A

um You need to de-structure test suites um yeah, so uh tests test suites.

A

Oh so so it's test suites right and I basically want this all right, and so I can just be now like hey uh const test suites equals jess jobs, dot flat map, and then we have like job, and we do this thing where we do our await here. Oh async done.

A

Oh, no, that this doesn't work right.

C

Now, let's keep it like that? Well, I think if, if you pull into another function, the if we do like a fetch test suite for job that could make it like a really very readable mind, but it's I get what you're saying um um what I wanted to talk about while we're here, because I'm gonna have to hop off in a little bit. This is so freaking helpful. I really appreciate iep.

C

um I haven't realized it until now, but this has like been bothering me for a couple of months, and I just I just sit there worrying about it. Instead of not, instead of doing anything, I'm joking, um what I want to talk about is like, what's the ideal output of this data that we think could be helpful for me. What I would really like to see is um file uh the the test file yep.

C

ah I don't even.

A

C

All the jobs I want to see the test file and then and we've already.

A

Lost because there's no, we don't have the file in these outputs.

C

Okay, I knew that all right and that's fine. We can keep this. We can keep this name stuff. um This is why it's like, and if you ever see me.

D

On a review, we name all.

C

Our is this from the describe block. Yes, it is, if you see me in a review and I'm like oh yeah, it's really nice if these describes are all unique, and I.

D

C

I put the file name in the describe. This is the reason I do it, I'm about to go rename all of them.

C

Yeah it'd be really nice to this is this is opening.

A

Up a side see that this is really good, so they they actually have a config, so they have like class name template, and so we could just change the class name template to maybe contain the file or something oh use path for suite name. If you go up, there's use path.

C

For suite name, oh you set it to a boolean. I think that's 100, I think that's 100 what we want to do.

A

Yeah not deprecated, it's.

C

A

C

Name, template.

A

Yeah, so we have suit name, template and uh yeah.

D

A

Name, template equals file name and then the suits would all be right. Right, perfect. That is.

C

Perfect, so let's pretend that that's already there yeah, but I think.

A

We're all very.

C

Happy right and- and I p correct me if I'm wrong, because you've built more of these kind of dashboards than I have, but I'm I would just love to see, here's a file and then the slowest test of that file and then the time of that slowest test. So then, I know here's the files that we're going to need to refactor to fix these things.

A

So I I want to correct you a little bit so I think it there are probably multiple views right. So sometimes you have like a file. um Sometimes you have like a file, that's slow and you want to fix it, but sometimes you just have like humongous file. That's slow right! So that's fine right. If you essentially yeah, I think you're right so essentially what I probably would build. I would build one table.

A

um That is just like all the files and you know- and maybe you can expand it right, but that what I wanted to.

C

Emphasize is that time on line 14, that's to me personally, I'm not really worried about that.

A

You're basically worried about those right. You want the slowest. So why don't we? Why don't we build a table that you know and excuse the markdown? uh We need a readme anyway, guys right so peyton. This is something that I really hate. Writing, um so imagine imagine we have like uh like this one right.

A

So what what if like? Because this would be the file name right, uh so example- path, dot, spec, js, right and then we are just like here and then we just have time right. So we have like uh spec time.

A

And then we have like two seconds right whatever um and we have seconds, and we just sort that by uh we just sort that by longest running test, there's.

C

There's one there's one tweak to it that I think I would personally prefer, but I want to see if you guys think would be helpful or not because usually in these cases, when there's a file, that's slow, it's like all the specs in the files.

C

So it's gonna end up being really to me when I know like okay, what are my slow files because that's like how we create issues and the refactoring is like based on spec files, like it's really helpful for me to like, let's group it by file, and so what I'm really interested in is rather than the individual tests, because what I'm picturing would happen is I'll have like 20 all from one file. I see where you're going. I really.

D

C

Grouped by file, what's the slowest test of that file and then the time of that slowest, okay.

A

So you basically want to see example, okay, okay. So what what do you guys think about this uh time? I mean it would be interesting. So the whole thing took seven. You know seconds uh slowest test. It would be like.

D

A

Pop up when clicked right and then we have the slowest test time, which was like five seconds yeah right.

C

That sounds so beautiful, I'm getting I'm getting jittery just thinking about it.

D

Do we want to do the um display? This was like express and handlebars like a small node app, or what are we is that what you're thinking I.

A

Think so I think so no, no that doesn't make any sense. You know we have. It might be sorry for that, um because gitlab ci would allow us to just have it in pages I mean how large are these jsons uh ls.

C

Sorry go ahead. What we can say ip, I didn't mean interrupt.

A

um A lot of these a lot of these contain craft, and you know because, like here's, a lot of duplication and so on, so I think in in total. It's like four megabytes of data. I would just either be if you want to go crazy and go the next route, I would do a next route that just asynchronously gets like the combined json and chose it or you know, what's really fast, an html file yeah.

A

So I would just do it with part or whatever templating engine, and you know just render that table that we have outlined here. Yeah we'll do that.

C

I think I think, using a static site generator and just doing a no javascript solution makes makes the most sense and um because yeah it's it's. I don't think any there's going to be an any interaction unless you think we need to do like filtering. If someone wants to change the sort or something, but I.

D

Think for first iteration uh a static data table is going to be super helpful.

C

Yeah yeah um I've. I've only used hugo like for doing static site generating, um but I guess what you're saying oh, we could even use. We could use mustache, oh yeah, um because.

A

We could we could even use low dash templates.

C

Yeah but but those have a cli to it and all we got to do on our pipeline when we build the pages is just run: here's our data, here's, the here's, the mustache file and we output the index html and boom we're done um so yeah. It can even just be a mustache thing uh cool. That's I'm I'm giddy ip yeah. I want to push up a commit today. I have to awesome yeah. I love it. This.

D

A

No test suites. I cannot read test suites of undefined. Oh gosh,.

A

Why doesn't this work.

D

I is test suite an array I thought test suites and then sweets is like the root.

A

But that's it's just like it's. The root is an object right test, suite okay, const json. Let's, let's just do it like this right.

A

And then we can be like console.log json.

A

I mean that should be so. Is it ah it's actu? Is it actually doing json? No right it? It actually does does create json. So we need to do a json pause. Am I seeing that right? Oh no,.

C

Oh, you might be right, wait.

C

That doesn't seem like a very helpful npm package. Oh you can you can pass objects to to json and you can tell it to return an object instead of the string yep there. We have it we're good, but did you just do that? Oh jsonparse, you? Could you can't? You can't tell uh xml to json to do it by itself.

D

Oh okay, I was about to say they missed. The crucial part can do it themselves like.

C

There's an option, so you can pass um as a second argument. Yeah yeah to the.

D

A

C

Object true object. True.

B

Well, I guess you can just look for object.

A

uh Yeah default false yeah, okay, okay,.

D

A

Good, like yeah object, true yep.

C

And then we don't need the.

C

C

You can call concat on it since it's uh an array of things, yeah you're right. Well, you got to do the test. Suites equals test, suites concave.

A

A

ah Also, nice, that I wrote it into a js file.

D

It's valid javascript. I recently found out about dot take a few days ago with ruby that is pretty sweet.

D

C

Take it like it limits the the number.

D

Right like so, if you wanna, I had to put a max on something. So if you wanna, like only want the first 25 items out of an array yeah, I wish there was something like that and javascript isn't.

C

D

Javascript take isn't.

C

A

The thing now I've never used it.

D

Slice yeah, oh yeah, yeah, you got it, you got to do like zero and you know you got to pass in certain indexes that you want to do. Ruby is just syntactic sugar, so yeah.

A

So what essentially, what you're, saying, though paul is you want to this to be like a retrieve.

A

Retrieve uh j-unit right, artifact right.

C

Something like that or like or like um yeah fetch fetch. You know.

A

Speed report or something yeah so essentially, essentially what what you mean is like hey. We do this with a console log.

A

Then we can do also be like hey. um We can. Maybe here, do a console one and like uh got data for pipeline, and then we template string makes do something like what's the thing that we want to do. It's like pipeline dot id.

A

Right: uh okay,.

A

Success found some just jobs right: four, okay, and then we can be like here console one uh get jest.

A

Just j unit artifact for yeah.

C

Job id man I wish we had this level of you- know: logging in our application.

A

So you're like node, retrieve junit, and then we can just pipe it into junit, combined dot, json right, yep.

C

Oh, it's so pretty.

C

A

Okay, so we have oh huh yeah. Obviously we need to console log jason uh stringify test suites too right, yeah yeah. Thank you. uh No, that's not what I want. I want this.

C

Yeah zsh keeps you from overwriting file, yep.

A

That's cool, so we have like this should be the combined json right, yeah, oh yeah. This looks right, yeah.

C

A

And the thing that you're saying is like you want to have.

A

Pug yeah, I want. I want those.

A

You want those pugs, yes does. Puck have.

C

A cli- I don't know I I know mustache does um and I'm not I'm more familiar with using mustache templating than uh pug, but yeah. Then we just run. It doesn't even have to be a party here. That'll just be its own script. That.

A

Yeah, I know I I just wonder if it makes sense to just you know: you're already instantiated the thing now: you're dumping it out of memory in order to read it into another thing, and it's like hey, essentially create our own pipeline and we can just you know, be like async get test. Suites.

C

Right because no you you bring up a good point. Mustache can also read from a stream as well, so because we're not.

C

When, when we do console log, it does let us be, I think, make those optimizations, uh because um cli can read from a file lots of different cli things can read from a file or or a string or a string, and so yeah mustache. You can also do a stream input which would work as well, so we don't have to dump to disk or anything. If we wanted to do.

C

That, but it's not it's not a huge deal, and that's that's just my um that's how I tend to approach these things.

A

So one of the things that's yeah- I like that too, if you can plug in things together right. So but it's like uh hey, let's, let's maybe just do name right. Yes,.

C

We need we need.

A

We need the time right.

C

Yep but then I think we'd want to do the slowest test calculation here. So you know.

A

Yeah, exactly that's what I mean right yeah, so um we can be like, and so this is this one right and we can be like return uh and then it's name. It's time uh is pause in time. 10 we have. Oh, it's not positive. It's pass float anyway, because we just use number yeah um tests is just the count of tests.

A

So is this like at least a number? No, obviously it isn't so you you would just be like number tests.

C

If test is a number, that's a that's that's it's! This number number is a function that converts strings to numbers or any numbers yeah.

A

Okay and then we say slowest, and it's just like test case, um you want to create a function, called find slowest test.

C

A

Okay, function find slowest test case right.

A

And we just do this and we can be like return, uh oh test case here. We can at least go plural right. Yes, let's please do that uh cases, and then we can do a map and.

C

I guess we we could you can't do it in one reduce um or we could split it up.

A

You need to loop over it anyway,.

C

Well, we need to get the slo, we need to first calculate the slowest value and then find the one that did the slowest. If you do a reduce, you can just keep track of okay. Do I replace my accumulator?

C

You know what I mean yeah yeah, so you could do the redos and the two two params for and you don't need to pass the second arc to it either it'll it'll already start kicking for you um pass the accumulator hey, I usually call the first one yeah and that that works too yep and then yep yeah that works. That's that sounds really helpful.

A

um Okay, so we just do that if I don't provide an argument, does it start with the second one and puts the first one in the you.

C

Know what I mean yes, it is, it does that um well,.

D

B

C

Be slow, I think, will actually be undefined. Well, let me double check uh yes, you're right. I think you do want to pass null to it. That makes a lot of sense, because these are different types. You can't use like a reduce to do summing um and it will like include the values, but we don't need to do that um yeah. I think that's the best way to do it. So if we don't have a slowest we're going to return the current yeah. This is a good approach.

A

So, what what do we? uh We have time and name right. So we again have time name uh equals current right and we can be.

A

If slowest we return time number time and name yep right, so this is already bs, so maybe just do cons name. We just do it for name right. I think it can be cons. Time equals number current time. That would be cool destructuring with a map function, yeah.

C

What one that one pattern I've been doing lately is: um if, if I know I'm going to like convert something, I'm not really going to use it I'll call it like name time, but then I I rename it the time arg yeah like I, I throw like arg or pram at the end. If I know like I'm, not really going to use this, I need to convert it first or something like that, but um yeah. This makes sense.

C

A

Time is smaller.

A

No, if slowest time is larger than time, we return slowest right. Otherwise, we return.

A

C

Right, this little time is larger than time. Yep yep yeah.

A

Okay, so we can also be like hey yeah, yeah yeah yeah. If time.

A

Slowest time right that works and then we can just return slowest. I love it yeah. So if we now do that, we have to test suites.

A

Should didn't someone right wanted to write the front end.

C

I think I think peyton was was talking about doing that.

A

Touch template dot, pug.

A

Okay, so we get. What does it complain about? Const require pug right and we can be hey, compile the file to us and we have test suites and that's the template right. Yeah.

C

So that uh peyton can pick up the the template part. Do you mind if we just console log that to test suites for right now.

A

What do you? What do you guys? What do you think if we just do the following uh ah wait? We also have the problem that is like he can do the template, I'm just going to be like hey html, uh huh see what we need a right file sing.

C

Yeah you should, I would suggest, use a mustache. The cli does all this for you, um you don't have to code it.

C

It can be pretty nice.

A

Yeah, I don't I don't know the thing I don't like. We would need to serialize it to jason, for example, and jason is really. If there are any problems, you know you don't get the benefit, because the json has to be passed completely anyway.

A

Well, uh the performance benefit, for example, right and uh yeah.

C

A

C

That's true, I don't think uh because we're aggregating by test name, the json is gonna, be significantly smaller since they're activating by test file, rather than doing each test name.

C

But I don't I don't. I don't feel super strongly about it, but I would just say: hey, let's, let's. I just really want to see the data right now and that's uh yeah and then we'll figure out what we do later.

A

Oh, by the way, we probably also one thing that we want to do with the test: suites right.

A

By the way, haha so that's actually a good call out lucas slowest time uh slowest test time um and it's like that and uh is like slowest test name all right.

C

Well, wait we already we're already prefixing this thing with slowest. No, we are not well I'm saying yeah. We could do that.

A

I like that, because then you know you don't have to look into that object.

C

A

C

C

You're good now.

A

Yeah, I think it's fine either way um iteration. So this is like okay pug each. What? What do you want now? Do we just do a paragraph, each value in what is it test tweets.

A

Oh actually, I didn't want to do that. I I just wanted to be like pre and then be like equals jason stringify test suites null 2.

A

Let's see what happens.

C

Do we need to give um do we need to give our like json object, a root object now, because I think it's just an array.

A

C

A

You can, oh time is not defined. Why is time not defined? Oh, it's added so assignment yeah.

A

Get got data for pipeline. I see a new pipeline just finished. It's crazy.

A

Cases reduced is not a function there. You go. Why not is test case.

A

Looking at the combined data, it's test case right: do we maybe have skip bonds where it's not defined.

C

That would be so scary.

D

A

I mean reduce for an empty array should be pretty efficient. What why is test k is not.

C

Let's see if we can, I would, I would suggest, console logging at find slows test case. You can even exit after you get the first one and just double check what it.

C

A

A

Oh yes, obviously, um this xml parser the problem with xml. If you have like the tag multiple times, it will convert it to an array, but it will. You know.

C

Okay, well, that's! Well, that's actually an easy thing to do. um We could do on line thirty. Three.

C

A

um At radius, I would.

C

Yeah, I would do if array is not array and then just do. I would just do a recursive call of wrapping it within a ring.

A

ah Yes, uh yeah yeah. I know what you mean now like like why a recursive call well.

C

I mean you could do it that way. We could. We could change the param to be what we wanted to be, or we could just do if we're not an array, then return find slows test case wrapping in an array. uh But, however, you want to do it. It's fine.

A

ah So you're like if array is array cases you would just do like return, find slow. Oh that's, really. Smart yeah! I've never done it. That way.

C

Oh yeah, it's kind of like the you know, having those like top level if guards like it's kind of like some.

A

Yeah I know I I usually.

C

Do it the other way.

A

uh I I always do it just iteratively and then it's then f is not defined.

A

D

A

I mean we got further right, like.

C

Oh man wow, oh we're, not sorting it yet! Oh, I can't wait for us to sort it. Can we sort it? Please.

A

Yeah, but do you want to solve it by total test time or by slowest test slowest test time.

C

um And we would do the sort on the at on the out right after the outer map.

A

Right yeah: that's you're right, um yeah, okay, so this is a b and then a dot slowest dot, time minus b, dot slowest dot time and then because it's obviously wrong.

B

The other way around yeah.

C

Yeah, how does that? How does that work? So, if the bigger one, it's always going to do it, the ascending of the sort value, so this is going to. I think this is gonna, be ascending. I think we do want it the other way, but let's just double check yeah, I'm probably wrong too.

A

Oh jeez, why is that.

C

A

How's your day going man mine, it's good. My sister was uh anna just finished her bachelor thesis, oh nice, and put it into the post. Today I looked it's the slowest one cool jesus. Why isn't it like ah sorry for the bless? For me I know you're a religious guy thanks. I just meant some. I just meant some random guy in america.

C

A

Jesus, I'm sure, jesus and.

C

Jesus will forgive you.

C

A

Would be so cool if.

C

A

Had a jesus on the team like.

C

That would be pretty awesome.

A

No I'm going to you. ah It's recorded right ah dang. Now everyone.

C

Well, you already.

A

What'd you say everyone will see that I'm using vs code now for the final. This is not good. People are like okay, so it's just you know what I often just then do.

B

A

Yeah go for it, I mean you know, what's also, what we also could do, we could just do plus a uh no. My, my mine, minus and.

C

A

B, oh now, that's.

C

An interesting way to think about it. Okay, I I've never been able to like get a really mental model. That really makes sense here.

A

Oh well, okay, uh touch, dot, get ignore right and we can be like hey what what what do we need like? We would definitely want to ignore node modules yep to get ignore.

C

And I guess also the any generated file which would be the um yeah.

A

So the thing that I probably would do is like mkdr public.

C

Yeah, that's a good idea and that's what you know. Here's another benefit too, with you know letting the terminal do the glue code. uh It's just so then you have one place that all of these you know, directories and stuff good reference and the way paths are what what do you mean? Oh, like.

C

C

Because we ignore the public directory, we want to write end up. Writing there. It's a little like outside. You know the responsibility of the script, to like know the path that it's putting stuff to and anyways.

C

um I'm probably gonna make it. I'm probably gonna pipe things with the cli uh later.

A

Do we have I, I just wonder you know, because huh so you're saying mustache okay come on, you know I did all the things my way I'm like mustache npm is that is that where it is yeah.

C

I think you use npm yeah.

A

C

There's a ruby uh thing too, like I think you, I think you already have it like on your cli. um How would I have it in my cli if it's like in your.

A

Brew- uh it's not in my brew. Why should it be in my blood not.

C

A

A

Is like command line? Does it have oh yeah? It has like okay, so we we we just do that like is that yeah yeah, so we do where's my puck thing index no move pug. We we just call it like template dot, moustache, yeah and oh yeah.

D

A

So this is like how we do it.

A

Okay, so we do mustache output, html or, in this case, public dot index html yeah right. I think it's.

C

Called template.

A

C

A

What we can do here is like public uh uh output, dot, json, oh cool yeah, and then we have template dot. Mustache.

C

A

Okay and what do we do for index.js, so we just do console log.

A

The basin stringify.

C

What's what's what's his name test suites right yeah, but I think we do want to wrap it in a object like like you did on line 78 through, because I think we we do want to have like a root object. We reference. Otherwise, I think you could use dot and mustache.

A

I'm not entirely sure so this is so and how is a do we have like iteration? There should be an example right like ah just like that: no yeah, what just like this yeah, that's it, and so we we we just go in here and be like hey and then we we are like hey. uh I don't know lee yep lee and you know ul oh yeah ul, and then we have like what's his name template. So this was test suites right yep and we are like test suites and then.

D

Here we are in there.

A

Yep and it's like a name and then I don't know, and then we have like a slowest dot name, and then we have slowest time yeah. I'm.

C

I'm 75 sorry, that's gonna work.

A

Mustache not found.

A

No, no, no worries, so we actually can do like yawn and then we just do no mod on run or no. You know what I'm yeah, the the diabetes, the dot bin uh and we are like yan, remove, pug and.

A

So this is something I already don't like. Our xml library is shitty because it has like some thing it compiles. Oh, it has a native.

C

A

Yeah, oh wow, yawn, add mustache.

C

um Yeah, let me double check the um the cli.

A

uh Oh could not pause view as jason ah yeah. That makes sense, because we just we still want pug.

A

Okay, let's see what happens: oh.

C

Yeah, it comes with a.

A

Thing: whoa publix, oh yeah, see sweet. Okay, you convinced me mustache is the go-to. From now on.

C

Nice, ah I am so freaking excited this is so helpful. So what is the slowest I want to see? What's the slowest one.

A

um Like us, wait wait a second. We need to build it first. What do you mean yeah? It builds right now.

B

A

Changed the output.

B

A little bit like.

C

B

C

Cool um you know, I I see it being pretty. Oh wow, okay, the slowest one awards handlers for sure, the slowest one that shows up all the time, but.

A

But it's so these timeouts that we are seeing.

A

Is it um these timeouts that we are seeing? Is it like the whole suite times out or just the test times out? It's an individual test.

C

um Gotcha yeah, it's an individual test times out, but you know usually what that means is that we have to refactor the whole suite or we have to refactor the whole file, uh yeah, so yeah um cool and then we'll. Let uh peyton get crazy with the um styling of this uh making it really pretty yeah you're right. I think.

A

I think that's good, like that's that's good right, so the thing that we already could do, which is nice uh just in order to like, uh like it's like thing here like console and I'm leaving this one for peyton.

A

Five, like five slowest test, suites.

C

Yeah I like that.

A

So if you run it now, we have leaves at the output right.

C

Is there a way to get like node to cache? Some of these requests we're making.

A

Yeah I mean to be honest: if you run it like once an hour, it should be fine, because we should have new pipelines anyway, and that's it right by.

C

A

This is the only data that um that's the only data that um echo success fully run ran our build. um That's the only there that that I would. um You know that I would also like to have on the page like which commit are we talking about which pipeline? Are we talking about right? So this is also.

D

A

Yeah, that's that's a problem where we moved it to that function, right that you probably need to change it around a little bit like hey um as soon as you see one of the jazz jobs here right. You probably want to return the pipeline, then so that we have it in the output as well.

C

Yeah, that's a that's a good point. I'm going to create a an issue for that and our are you going to create an.

A

Issue before we had the first.

C

uh Pipeline you said pipeline.

C

Traceability uh yeah.

D

C

What pipeline did this run on what commit um and then is that that's about it right, oh and when probably when did it run? That's a good idea, um yeah. So.

A

Our metadata yeah, so we have to get ignore.

A

So what else do we need? It's like readme,.

A

uh What else do we need? It's like index.js, like hey, readme,.

A

Run bash build.sh.

A

And let's look at the public.

C

Folder, do they need to run yarn at all? First, all right.

A

See that oh it's in there cool yeah, that's generally how I do it because then at least here in the gitlab ci armor, I.

D

A

Need to add, like one one more thing and just one thing right yeah, um so let me just delete this file. We don't need that one. We don't need that one. The rest already looks important. Now, let's just like prettier, there should be. Like ah I mean we basically just have one file like right.

C

A

Is it index.js yeah.

C

That sounds right.

C

Oh it's pretty.

A

Yeah, it's also the new prettier. You know the one that you don't want to use yet there's one important, there's also warframes. I think it's just like uh it always wraps. uh If you just have one argument and then an anonymous function, it just raps them.

C

I I do that by habit and prettier changes it for me, so maybe maybe I do want to use it. I.

A

Don't know so you know what now we're going to do some cheating uh yeah. So, let's, let's, let's have a look into the ci yama default image, looks good default, gitlab.org test. I don't think so. This is actually a good thing that we can put here like we want to have a frozen log file.

A

um We don't have any further tests here right. So we can, you know node, ah I mean it looks okay, build. We don't need to.

C

Separate those we might want to run the test um on an mr, uh like just just run the build script on if we have a merge request, so we don't end up like breaking master.

A

Or maybe yeah yeah yeah you're right, so we have public here uh here we do actually a gcip. So it's fine, so we only want to do pages on main, um but we don't need to test script because we don't have anything to test yet right uh right, we'll just test the build that makes sense. Okay, yeah so and oh, this is a bit weird. Why is it? Is it like that, like this should be like that? This should be like that.

A

This should also be like that, and this should be like that- and you know we have this crazy tool. It's called prettier and we just can.

D

A

Use it to write a yaml and it looks pretty again nice, so here I called it, you know build sh. Do we need anything apart from curl and bash and get I don't know, even though I don't think we even need those right. Let's see if it breaks.

C

I mean shouldn't already have bash: does alpine not use bash.

A

ah You're right we definitely no, the the alpine one are really tiny, so we don't have.

C

Do we need, can we just use regular shell, like it's not doing anything, super fancy right.

A

I think you're right. um The only thing that I really like and what I wanted to do is like I had this cat right, webpack memory, metrics and I want to be like uh I really like this unofficial shell mode so in case something.

D

Fails right now.

A

That's a good call, let's do it yeah, and I think that should be it right.

C

uh Yeah, I would guess: well, let's, um if we go back to the, if you go back to the cia yeah, if you go back to ciamo stages, do we have.

A

ah Yeah you're right, we need build yeah, we need build and there's another thing that we should do. I mean I don't think it's a big problem here, but it's like ci jamo reference, um it's a thing that we don't do yet at gitlab, but that I really like and it's uh default as default default, yeah global defaults and what I really like to do is uh I don't know why I did put that. Oh yeah see we even have like a nice yarn cache here in case we ever update the version.

A

I think we probably.

D

Also can do like.

A

12 18 already, we probably even could do 14, um but the thing that I wanted to do is with artifacts and in artifacts it would be cool yeah. So we have expire in yeah, so yeah. Let's get idea so like it. It would be like artifacts expire in uh that's.

B

In 30 days,.

A

Yeah I like that, that's it okay git commit minus a. How do we call.

B

A

Push to product.

B

Push to production.

A

Now we have a properly named default branch as well all right, so it should deploy pages right like just speed reporter we already should. Oh yes, pipeline is failing, that's good! Oh no, unknown key files.

A

What jobs build cash can save.

B

A

This is not indented, probably so. This is by the way, one of the most annoying things in order to.

A

Yeah now it's valid um git check out minus b x c. I uh maybe let's run prettier again: uh git commit minus a minus minus amend and oh no, no! No, we definitely don't want to do that. I don't want to amend like we just want to say, like hey, properly indent cash key.

C

A

Push get them all.

A

C

A

I'm just going to go ahead and be like. uh Oh, I.

C

Want to see if the I want to see if the pipeline works so okay, it looks like it's working.

A

Yeah, that should be like hey. We have pages, oh man.

C

I'm so excited, are you excited as well? This is gonna, be super helpful, um so.

A

Like what would.

C

A

This is the problem because we use that uh we use that xml puzzle. So what was the other one fast? Xml parser, no.

C

We have to change our parser, oh because we're using it because it doesn't have the native dependencies that it needs for building its thing. That's frustrating.

A

So, what's array mode.

A

You know what index uh const parser right, frustrating you know on the plus side is like yarn, add fast xml parser yarn remove, what's the other one xml to json yep and we can just be like. Oh what does.

B

A

C

Get traversal object like yeah, I guess get and that's really cool get traversal object, and so this is going to. Let us actually do it in a more optimal way. Let's just convert it to json for right now, but I think it traversal object is going to let us it's not going to copy all the data over. It just lets us now actually traverse it. So we can do um our parsing, like like we were uh that'll significantly save.

A

Space so so here we probably have to say array mode. What's every mode, I don't know so I I.

C

A

Don't like that they do. They have any dogs like what you have. These attributes right for json.js object. The xml.

C

um Just a sec, I'm looking at the.

A

Where does it, you know tell me which attributes it has like? Oh yeah, so we have every mode.

A

C

A

I assume it's the thing doing the thing that we said before right.

C

So I mean you know they had.

A

C

An example on the on their read me just there that was doing what we wanted to do. Can we check it out with us which one that second, the second example.

D

C

No, I meant the last one. Sorry um if you go down, oh no, this is json json, never mind. I read that wrong.

A

Yeah xml to json, you know and it's like hey array mode. Hopefully we are like. Oh wait. No look.

C

What they have to do, though, so, okay, that this is interesting, so you see they first have to get this traversal object on the last two lines.

C

Okay, that's one way of doing it. I guess another way of doing it is on the line above those two lines. You can just do parts.

A

B

A

Okay, so where's the two json. We don't need that right. It's just pass or pass and options.

C

I was right, it is parser parse.

A

C

Library bash built see what happens: uh oh lots of pipelines, so those are all successful pipelines.

A

That don't have the j-unit artifact wow.

C

That's a lot that don't have the j unit one or they don't have the chest thing.

A

I think you know array mode. True is probably the thing that we want, but to be honest like let's be like hey, um maybe we just move it. You know pause xml right, uh oh no, pause, uh uh test dodge. No, it's fine right, pause and and extract uh pause, junit and extract, and we just have like the xml here right.

A

uh Past junit and extract xml um function, uh pause up pop, so const object equals this right and then we can be like console log object and yeah, and then let's see that we get our array.

A

C

C

So sometime when you're feeling froggy um yeah, you should you should check out uh the way the way that dino like suggests handling its dependencies, um because your dependencies, you import by a url and there's no like package json or anything if you're working like how do I then, like maintain that I have like you know, I'm using the same package project.

C

Oh, what's this cannot parse the usb yeah.

A

Obviously, because we're piping right, so I need to uh call the note the script directly uh ah test suite chest suite, so it actually looks like you know,.

A

It to me it looks like the thing that we want yeah, it sure does.

A

Oh, I have a feeling, that's interesting, do they did they parse it.

C

A

Do you see like no, they they they.

A

ah This is undefined, so they definitely it's like this one. But theoretically the test suite already should have attributes and it doesn't have any attributes. So my assumption would be you know it's just empty empty strings, because it's not pausing the attributes right. This is so. This is bogus ad pass attribute value.

A

I don't wanna, I oh no.

B

C

Maybe we should just use a different docker image for um for updates.

A

Yeah you're almost there believe.

C

Me, oh man, this seems very granular and not having the defaults. I would expect it to have oh shoot.

A

What why doesn't it, you know ignore: ah ignore attributes false.

A

This is like, why would you ignore attributes.

C

huh What is this? Oh, okay, but we don't want it. This attribute prefix thing right, so it's interesting. I can appreciate a little bit what they're doing here, but it's also not helpful.

A

So so test cases, oh test suite right, so I I mean I'm I'm fine with that. It's like we don't want. The note value right. So we can just be like attribute. Prefix is just empty. Yeah yeah, yeah, yeah.

C

That makes a lot of sense.

A

And then I'm like hey give me the first test suite, maybe also you know, give.

D

A

Like the first test case,.

C

A

Just in order to have some.

A

This looks good right, I think so yeah. So what do we need to return? We need to return like object, test, sweets, zero. I think dot test suite yeah and now we should get like this humongous jason again.

C

And there's like.

A

This looks good, we don't want to log and we definitely want to run prettier now.

A

Well, hit commit minus a uh out xml parser.

C

Since you've ran it a couple of times, do you want to add a job um or add a package.json script for print the prettier command you're running.

A

Yeah, but that means I would need to add uh pretty.

D

Oh, we don't have to be right installed.

A

You're right you're right there we're good okay, so this should now be done right. In the meantime, oh.

C

The um that job failed, it failed so easy to find gitlab tokens.

A

Yes, that's good, that's good, right, right, um so settings, so the questions like which, which, which token do we define here- and this is definitely where I'm going to pause the share for a second so yeah here we.

A

Go I'm wondering like uh we. We definitely oh shoot. What what am I doing like what uh yeah I'm going to find? Like a token, that's that's, okay, uh repository yeah. Do we use the the robot?

A

uh No, I I'm. I know we have a token for the webpack memory hooks right, so it has a gitlab token. So I'm just going to use that token as well, um so yeah. Okay, that makes sense get lab token. Is that protected so.

C

A

Yeah, it's a token, I don't actually know who this token belongs to could be mine.

A

You know it's like, maybe maybe we should create a non-privileged account that we use for like.

D

Scripts, like these.

A

Just in order for rate limits and whatnot that then cannot do shenanigans right yep. uh I think.

D

A

Yeah so resume share. That's I. I retried the job.

C

Okay, do you um do? Is there any traceability like? Is that token and one password or something.

A

That token, I basically copied it from another playground project and I just wanted to go to my favorite website.

A

So I think we want to run it at the 27th minute right generally um so, and every hour right.

C

The 27th minute.

A

Why not it's uh random yeah, oh cool it built it successfully, so we can run it in the 27th minute utc and, to be honest, like hey, this is probably enough if we run it from one till seven.

B

A

C

A

C

A

One two: five.

C

Oh yeah, that makes sense. That's a good call.

A

uh Every hour during the work week, less.

B

Every hour in the work week.

A

Okay uh next run in 52 minutes, so that's actually smart because it doesn't run uh before the merge request is done right, um but the quick thing that we can do is we can check check the artifact of the build job.

A

A

Oh wow, do you see that ah that's the complete thing, so the complete thing took 17 seconds right.

C

But you you see like like there's some, so what would then be really cool is have something you know, another smart layer on top of this to know like here's, here's our heavy offenders, but then, if there's a new offender, that kind of approaches and like like getting some notification of that or like, if there's not an issue existing for this kind of that, that would be another really cool. Yeah.

A

You know what the cool thing is we could do now, um because we also saved the jason away. We actually could get the five last runs or the five last times we built that thing or we could extend it to you know, get the json artifacts of the five last runs and actually show us. uh You know the average on the last five runs right yeah. That should be really easy to do.

C

Right, I think I think probably um an even smaller iteration would be um we build. We. We build the public thing from this, but then we could also um use the json and make sure that we have issues created for each one of those files um with a certain label.

A

Yeah, but that's also why I'm like hey: let's: let's uh you mean that you that you have a script, that you can run that you get, that gets the json and creates issues for the ones that are over a second or something.

C

C

Yeah that would be, um and then we we don't create the issue. If that issue already exists like for that file, we could we can parse that somehow, like that'd, be really cool. The next step here, though, is definitely to change that junit config. um That's one thing.

A

C

Have the file path, the file path.

A

Yeah yeah so yeah, so that's also funny because the slowest test you know the test name already contains like thing so yeah. I I agree with the just config: it's it's nice too, that we can use the file path. um Oh the old note spec huh what it took a second and it didn't do anything.

C

Old notes, close the notes, class.

A

Are we getting the slowest test? Yes, we definitely are. Are we.

C

I mean we should be.

A

Anyway, it doesn't really matter that much.

C

The um the individual tests won't necessarily add up to the total, um because the the total test sometimes has like justice and bootstrapping stuff that it doesn't use to.

A

Yeah, so this it means, like this old note, spec, if it doesn't do anything and takes a second, it's like probably the whole file is skipped or something right, like all of that,.

C

Test setup, script um and importing, and we do lots of stuff on import, sometimes um that that'll all affect things before we've even started tracking an individual spec uh yeah, but yeah. This is so freaking helpful, really appreciate it. So.

A

You know what's annoying, though, and what we probably should do. um I I I like the format that you're saying, but I think we should log every spec that's over a threshold right. So, for example, here in the awards handler you you're, like hey, you know it took 21 seconds in total, but it could be that there are five more tests that took three and a half seconds. You know what I'm saying: it's like yeah.

C

And that makes sense, I I'm uh and I'm less concerned about the total time, I'm more concerned about that 3.7 seconds time. uh Yeah. That's.

A

What I mean, but but.

C

If there are five more that are 3.5 seconds, that we've done.

A

C

So my goal, my my biggest concern is like okay, but if it's like, if I just see a whole bunch from awards handler, I'm gonna lose some of the like other ones, yeah and but I I agree, it'd be nice to see.

A

But by the way.

C

A

One by the way, one concern that we actually have with the total time is right. Now we don't use something like knapsack, so what knapsack is actually doing on the ruby side?

A

It's grouping tests by length in order to have jobs that are that have a similar length in the end and right now, because we are just you know, it could be that, like a watson, love which takes like 21 seconds- and you know all the other ones end up in one of the gest jobs and so that one jess job actually runs considerably longer wow.

A

That's really interesting. Maybe.

C

A

For example, if you look at this right, it's like hey alert management table um yeah, and so this is the reason how they consume like stuff right and so with this file. Maybe if we would split it into, we could write our logic. Hey, let's try to get that json file where we have just you know the length of all the tests.

A

If you don't get, it fall back to the default. You know hey just you know one to four, but but we could basically group them into specs that long run the same time or something but yeah. That's that's.

C

A really I like that, I, like that, a lot um yeah. I think I think it's a lot of stuff that we can improve upon. um That's that's interesting about grouping tests that take a similar time, um because that can help your fast ones go really fast.

A

ah So actually, I thought about the other way around right, so group them like hey, it's just going to take not that long right, yeah, so that all you know all the tests are like you know, equal in the end, like all the jobs are.

C

Have the equal length- ah oh that's interesting. Okay, I was trying to thinking of. I was thinking of like this example. You just highlight it's like why in the world is that taking 14 seconds? But maybe, but you know these are here's what's scary too. These are unit tests and this unit test takes point.

C

Seven seconds and unit test is where I'm getting all my value coverage from like you could easily have 20 unit tests for a single subject and if you're at the 0.5 second per test, you know you're looking at 10 seconds for the whole suite without any kind of setup time, and we have so many singletons in our app that I know like a lot of the and that's one. This is a huge downside too, with I think all the singleton stuff that we do um just every single test file has a whole new.

C

um You know required environment, so we end up bootstrapping, all the singletons and stuff, some of which we don't even need in some tests.

A

So you know you know what what that brings me back to it's actually like the thing that you're, also interested in that we currently don't show, is uh total time right. This is the total time, uh so the total time you're also interested in test count, which is.

B

Just yeah and average.

C

Is that in there oh, I guess we can count it. That's a good point.

A

I love it, and this is just like: hey average is like ah so you. You would be like time uh like time time, arg and tests like tests arc, and then you would be like hey const time equals.

A

Const tests equals like.

C

This, but this time oregon tests arc um yeah.

C

That is highly a stylistic preference but uh yeah that that average, though, isn't going to be completely accurate, because that includes all of the like startup stuff. um So you could end up in a really weird situation where the average is greater than the slowest.

A

ah Yes, no, I know so you would basically would be like hey average, and then we just it's a rate over the test cases. Yeah we'll we'll need to do.

C

um We'll need to sum find find some of test case uh with the tab. Yeah.

A

Something like that and then you're like average length, it's like it's like this, like function, yep and then cases, and then it's just like we just copy this bit, because it's smart right and here reduce it's just like some cur and then was it. What is it like a current and then we just do sum plus current number current time right and then we do zero and.

D

A

We do through cases, length.

C

C

um But we really want to we. What we'd want to do um if we're being just extra finicky is add a guard for if cases is an empty array and just returning zero, because we don't want to divide by zero and blow up for some because we actually have where we do this. We pass an empty array actually to here.

A

Like this yep, it's beautiful, it's super beautiful, like good check out.

A

This is gonna, be really nice. Oh unexpected token, oh yeah.

C

You gotta have.

A

It's like uh add, average and test count to the output. It's also funny. You know that we are like hey, let's, let's, uh oh no! No! I don't want that. I want that um that we are like hey. Let's pause, the test argument. Instead of just taking the test case length.

C

Yeah, hopefully those are the same.

C

I don't know one one thing too, and this this you know it kind of runs.

A

C

How is the, how is the yama invalid? Oh, was your master out of date? Oh yes, you're right, good re or the main animate.

A

Yeah, so, oh god, git rebase minus I and then it's uh original main.

A

Oh, what too many origins how?

A

Oh no, we actually have a conflict.

C

Oh, oh yeah, that's that's yeah. Let's just.

B

C

At the same spot,.

C

uh So one once extreme side note like, since we were talking about it, because I was selling it to you, um yeah with dino you import based on a url, and so, if I want to have like okay, I have multiple files. Now in my project, how do I make sure I'm importing from the same third party?

C

They don't have like a package json. What they suggest doing is having a dependencies module that you kind of it's. It's basically forces you to do this adapter pattern for every every dependency. Where you do this import export thing, which is what's really cool, is oh I've imported. I bumped a version, but now that I've, you know kind of been forced to do this adapter. I can actually just fix that api difference and my dependencies module as opposed to like having to you, know unbox throughout my entire code base.

C

However, I've uh was coupled to the previous api, um which kind of came up when we were doing when we were converting xml to json to the fast xml parser. um We could have ended up not touching any kind of our uh main application code, just changing. How do I.

A

But you know my argument to that is you can also do that with node.

C

A

B

Definitely do the same thing.

C

What's interesting is how dino, like kind of forces you to and it's it's just a really interesting paradigm, um I would suggest checking it out.

A

Yeah, you know what I'm now interested.

A

In oh, yes, the average like, oh, why it's like? Why can't we even do icons.

C

Oh they're, off a little bit.

A

Is this like a gl button? Did someone replace that with a gl button, yeah looks like it, and so this you know broke everything again, just because.

C

Since we were move, you know moving from fun awesome to svgs.

A

Yeah at some point we should be. Oh, the average is interesting.

A

So where's the one that was really long like 14 seconds or something like was it a varts handler? No, which one was it like see? That's also funny like awards handler yeah, and now you have one that's slower, and now it just took like 1.9 seconds earlier. It took like 3.6 seconds so.

D

A

Variance seems to be very high, but so in this one it's really interesting that the average is only 0.588 while the longest one took 1.926 right. So.

D

A

Also be interesting: how much does your slowest test in that file? You know deviate, um because that tells you a lot right um like hey, because if, if you like have an average of 0.4 but the longest one took 0.45, it actually means that hey. This is probably like fine right, because you just have a bunch of unit tests or whatever you know yeah well like this. One is also like 54 tests. Average uh 90 milliseconds is that 90 minutes.

B

C

0.1 yeah yeah.

A

Yeah so yeah, no, it's like yeah 90 milliseconds, no, nine.

C

That's what that's interesting, because you know you could end up, because when unit tests are fast like they're really fast like when unit tests are fast, it's like 0.01 milliseconds is very possible, but when they're slow like so you get a huge spread really an average average is gonna, be tough because you will get a huge spread. um I I think another another interesting thing to capture here would be like: let's just see how many, how many test cases are above, like a certain threshold.

C

um So maybe the average is like 0.01, but we still have 20 cases that are above you know, 0.5 or one second or something um yeah.

A

I I think we have a good setup to build, improve this page right, yeah.

D

A

Yeah, that's that's! That's all. We need um yeah going to stop to share. Oh gosh. We were here for one hours and 45 minutes, hey.

C

It was, it was worth it, it was fun. I.

A

You know, let me.

C

Let me stop you.

A

Yeah I like coding, even though I shouldn't do that that often.