Happy Holidays! This week, we discuss tackling some queue serialization issues, debating migrations versus seeders, and discussing some new ideas for Laravel Core PRs. We also discuss PHP 8.4’s latest features, why down methods might be risky, and how custom stubs could make development smoother.
Creators & Guests
What is The Midwest Artisan?
The Midwest Artisan Podcast brings you insights and discussions on all things Laravel, with a healthy dose of general web development and tech talk. Join Andy Hinkle and Dalton McCleery, two passionate Laravel developers from the Midwest, as they share their expertise, explore the latest trends, and dive into the world of coding. Whether you're a seasoned developer or just getting started, this podcast offers something for everyone who loves tech and Laravel.
Connect with us:
Andy Hinkle - X/@andyhnk
Dalton McCleery - X/@DaltonMcCleery
Cover Art by Robert Wade (robertcreates.com)
Andy Hinkle (00:02)
Hey everyone, welcome to another episode of The Midwest Artisan. I'm your host, Ane Hinkle.
Dalton McCleery (00:07)
And I'm your cohost Dalton McCleary.
Andy Hinkle (00:11)
How you doing, Dalton?
Dalton McCleery (00:12)
Doing good yourself, surviving.
Andy Hinkle (00:13)
Good man, doing good. Yeah, I've been through the ringer with a queue serialization. You've pair with me a little bit on this. And so you kind of know some of the gist. Like, did you know this is a thing? So I'll set the scene. I've been investigating. Yeah, I've been investigating an issue with queue serialization. I have a job.
Dalton McCleery (00:27)
Mm-hmm. Mm-hmm.
Yes, give me the context.
Andy Hinkle (00:41)
it goes back and pulls a nightly report. So it goes back to the previous day, pulls the report storage, etc. So, I have it in the console of the scheduler. do like schedule, quote unquote job, and then the job class and put it daily. Right. So I want this to run every day, pull yesterday's report, save it, do the whole thing. So in my job class.
I have a constructor, so it's like public function construct. This is an extra job function class. Then inside public function construct inside the, inside the parameters, like open construct inside there inside the parentheses, I have public question mark carbon, like illuminate support carbon. like so to set it as an all, so either a carbon object or an all variable date equals an all. And I apologize just talking code on a, on an audio podcast or isn't
Dalton McCleery (01:33)
audio podcast.
Andy Hinkle (01:34)
But just imagine you have a constructor inside your constructor, have public carbon variable date equals null. Date or carbon has a question mark in front of it, just saying, you know, it can be null or carbon object. So, we put tests behind this, like one of the tests even is like if it's null to ensure that the date is getting set to yesterday. So when you do pass null,
it will inside the constructor do something like this date. if the data set use that, if it's not set, go today, sub day. that way it's getting yesterday's report, not today's report. So what I was finding is that, it would create a new carbon instance of today, no matter what, like if it's no, unless it's, unless you pass a carbon object to it, but if it's no.
Dalton McCleery (02:12)
Mm-hmm. Mm-hmm.
Andy Hinkle (02:28)
It will automatically just create a carbon instance, even though you still are really set it as no, just so strange. and the reason that we have a variable like of Null instead of just setting it properly is that we have in other places where you can go like, and fulfill like dates in the past. So you can do.
like date ranges in the past and it just kind of loops through a start date and end date creates a, know, carbon object for each individual date into this job. You can pass it the date, et cetera. But in this case, I'm like, Hey, if I don't pass it anything, if it's just null, give it yesterday. And so that way it's clear separations of concerns are good. However, so when you schedule it in your console, this is where it just gets really tricky. So date, when you do a date,
Dalton McCleery (03:12)
Mm-hmm.
Andy Hinkle (03:16)
It's a new carbon instance of today, not yesterday. So it's so strange. I have never run into this. And so if, but the biggest frustration is if I fired this job manually, like through the driver of like sync or like a non-console thing, like a non-schedule works fine. Like you get today's update like you get yesterday.
But so did a lot of digging. It's like something in the job serialization process. So like when when the schedule is running or setting up, it's something like the deconstruction of the serialization or something where it takes that null and creates just a new carbon object of today. It's so weird. So I moved it to the handle to the handle method, like inside the handle method it can do. But that's so weird, too, because inside the handle method.
Dalton McCleery (04:06)
Mm-hmm.
Andy Hinkle (04:10)
Like it's so hard to, to handle because it's always a carbon object. Even if you deliberately set it as null, it's the strangest thing.
Dalton McCleery (04:17)
Mm-hmm.
Andy Hinkle (04:19)
So yeah, I put it in handle method and what I did inside the handle method, I use a method inside there to just, Hey, call this up and return the right carbon instance. Just very strange and kind of a quirk and like the queue serialization, but little along with that when did there have, have you, I mean, we kind of during halfway through this is like, I need a pair of somebody cause I'm going to lose my mind. So I talked to you about it you're like, you had a lot of great like debugging things. You just say like, Hey,
because we figured out we couldn't, we couldn't replicate it at all locally. We tried and tried. so with, one point we thought it was a vapor issue that then, once I figured it out, I tried it over and forge and it's an issue. Like it happens over there too. And then I was able to finally replicate it locally by just running the whole like schedule, you know, on your, like schedule command on your CLI. Finally got to replicate locally. yeah. It's a, it's just weird that like.
Dalton McCleery (04:57)
Mm-hmm.
Andy Hinkle (05:20)
So TLDR and all of that, a little long winded. If you have a new job class and you have a construct in there, if it's like you put public carbon date equals null, will set the carbon, it'll create a new carbon object, which I'm curious if it does that for users and models and all of that, I think. That's kind of like how that works with serialization, but I don't know. Have you seen this? Have you seen anything like this? Yeah.
Dalton McCleery (05:20)
Yeah.
No. And I was going to, I was going to pull my hair out when we were, when we pair programming on it. Cause I couldn't figure it out either. I was like, it's probably just something really dumb. Something's something easy. no, it's actually, it's actually quite hard. no, this is actually really, really hard. Why, why is it doing this? And yeah, we, tested all kinds of thought was like, I think at the end of that call, we were like, it's probably just a vapor thing. The way that vapor has, you know, the queue set on a different version of the application and sort of that, but.
Andy Hinkle (06:07)
Mm-hmm.
Dalton McCleery (06:15)
I mean, did, how, much did you, did you test on it? Cause I was going to ask you if you tested it on forge. Cause the last that we talked about it, that's, that was what I knew. I didn't know if you'd fixed it yet. I was going to ask if it worked on forge. I was going to ask if it worked with, with date time instead of carbon as well. I wonder if it's just like a carbon thing or if it's an actual date time issue. Cause it's definitely something with, with that type hint is like, you can't have a null version of this specific type.
Andy Hinkle (06:30)
Mmm. Mmm. I to try that. Mm-hmm.
Dalton McCleery (06:43)
like a carbon, it just doesn't happen. See, that's super, super odd and super very specific. I I guess not specific enough because I've put jobs that have dates as their constructors. I've put classes that have dates as their constructors. And it seems so odd to me that it would always default to today. So yeah, I'm curious on how you did fix it. Because if it does default to today, how do you know that it was passed null?
Andy Hinkle (06:46)
Yeah.
Yeah.
Dalton McCleery (07:13)
in order to figure out that it was yesterday. You know what mean? Like how did you actually fix it? You said you put it in the handle method and you have some other stuff. I wanna dive just one level deeper.
Andy Hinkle (07:17)
Yeah.
Sure. Yeah. So you brought up vapor and yeah, cause we've been, we've been bit by vapor before without going to a side tangent, but like, for example, you can't set jobs past like a retry after 15 minutes. So if you have a flaky API, you have to set it to retry within those 15 minutes. Like you can't try three hours later. You just have to continuously retry, which is annoying, but just AWS, it's not a vapor thing directly. It's AWS. It's upstream. But yeah, so we thought we kind of like, we're starting to wonder because of
Dalton McCleery (07:45)
which is dumb.
Andy Hinkle (07:53)
we've dealt with Q issues in the past in vapor. Maybe it's some like interesting AWS vapor thing. But yeah, I ended up, it's really hacky and I don't like it. I'm probably gonna go figure out how to refactor it. what it is, is I check if the time is midnight on the date requirement. And if it is, I do today, I'm sorry, if it's not midnight.
I do today, sub day, basically because I know at that point it's, it's giving a new carbon object, but it's kind of like, I hate this logic because I have to know as a developer, anytime I pass to it, I have to make sure it's midnight. So I wrote a stupid comment in there, but I'm going to refactor it cause I don't like it. It obviously smells funny. So because anywhere else you're passing the dates into there, it's always going to be midnight. And since it's going to create a new carbon object, will very likely not be midnight to the millisecond.
And so because of that, I know that that it's, it's doing some weird stuff in the constructor. It's setting that carbon object for some reason. And I just have to reset that and just do today's update. So that's kind of my current implementation. I hate it. It smells because I, have to know, like if you pass the random today, carbon object, sub day, it's not going to work. It's going to like, it's going to take that and change it. So, man, it smells terrible, but it works for now.
I was able to deploy on a Friday with that change. Hated it, but hey.
Dalton McCleery (09:22)
Friday scares me also it's just while you were talking I had the weird thought of if someone who's not a developer or Good with Laravel if you've made it this far in this podcast and you're like, wow Why are they talking that their code smells? Just just know that That just means hacky, right? It just it smells funky, but it's very clever. That's very clever that you basically trying to see if it's a
Andy Hinkle (09:44)
Yeah, good smell.
Dalton McCleery (09:51)
system generated date versus like if you're passing it a date. It's very clever.
Andy Hinkle (09:55)
Mm-hmm, so I'm doing now, yeah. Yeah, because if you just do new carbon, if you just write date equals new carbon, it's gonna give you now. And so I just work around it. I hate it, but I'll figure it out. yeah, it's super weird.
Dalton McCleery (10:09)
It's clever.
Andy Hinkle (10:13)
Yeah, so the problem is we can't test it. You can't test the functionality with it because, it would be testing Laravel code at that point. You have to test the scheduler, ensure that the schedule of serialization works, which is just weird with Laravel. And you don't want to get that far when you're actually testing the framework to ensure the framework is doing what it should be doing just now. So it's something with the serialization. I didn't see it in the documentation anywhere. Like I tried, there's some mentions of serialization, but
Dalton McCleery (10:22)
Mm-hmm.
Andy Hinkle (10:39)
So what I'm to a point now is I'm trying to figure out how it happened or or like why that's occurring and then maybe even make a note of it in the docs if it's like this is really a thing or if I'm dumb and I just don't know. I've always known for many years past like you don't want you don't want to mess with the constructor too much in jobs and in the scheduler like if you put in the scheduler for example I could have done something like schedule colon colon job and then pass it the class and then comma then gave it like now.
Sub day or or today's update in the actual console scheduler but when if that is cached or if you know that goes through there Where it takes a snapshot of your schedule? It's gonna be that current deployment time. It's not gonna so that that day's gonna be frozen. It's not gonna be the actual sub day if that makes sense, so I know I know that's been an issue like just to be careful on your parameters and stuff so yeah, it's kind of wild but
Dalton McCleery (11:10)
Mm-hmm.
What a nightmare. I think we were even talking about this, was it like three days ago? So this has been like a three day buck. And I don't wanna say anything, but I don't know if I see any gray hairs, but if that was me, there would be a lot of gray hairs after figuring all that out.
Andy Hinkle (11:42)
Yeah
Mm-hmm.
Yeah.
Yeah.
It, it's frustrating because it just, it's one of those things that at the time it could only get it to happen in production. Then you have to do the stupid thing like testing this, testing this, testing this on all your staging environments, just to see if you can actually get it to finally replicate. and finally we got there, but it just, anytime you have to like ship debug statements into production, like log, clone, clone, debug. It's just like, come on, man. And so, and then it's like, you're trying to.
Dalton McCleery (12:22)
Oof.
Andy Hinkle (12:25)
You're trying to not make a release because we have staging in production, trying to make a release for such a dumb thing. Luckily, we're able to replicate it on the staging environment for the production so we don't have to make a release. yeah, man, it's been a headache just trying to figure out the serialization. You know how many times I looked at the constructor? I went letter for letter, like looking like, I missing anything here? And just pulled everything out. then I finally hit, when I just started.
minimizing the whole thing, finally triggered like, okay, it's a serialization that's at issue. I asked to ask cloud, ask chat GPT, nothing got terrible results. It was telling me maybe your app time zone is all like wrong. I'm like, nope, it's UTC. Maybe you try to force the time zone. Okay, I don't know why you would want to be any different, but try that, nope.
Dalton McCleery (13:01)
Such an odd thing,
Andy Hinkle (13:20)
So do you like carbon arrow or like date arrow, use time zone? Yeah, it was a whirlwind, but finally got there.
Dalton McCleery (13:23)
Alright, yeah, yeah, yeah.
What chaos, man. And it's probably just because of like the PHP type hinting, right? That's probably what just jacked you up. Damn PHP.
Andy Hinkle (13:36)
Yeah, I wonder. Yeah, because we talked about well, you know, now you mentioned it like we talked on the show may have been last episode of the one before we talked about how you can you can create a new carbon object from the from the constructor. So which is nice, but now we're we're like fighting the evil side of that. Like, is this the is this what it's doing? So. Yeah, it's coming full circle. Yeah.
Dalton McCleery (13:53)
Yes,
Yeah, hey, that's a good callback. Hm, why didn't think about that? Well, now you know the cons. The cons is a three-day headache. The pros is it looks really nice. Damn PHP.
Andy Hinkle (14:13)
Yeah. Yeah. Three days and two walks, two walks of like maybe, maybe I can move outside the constructor. Would that be the problem? Nope. Tried that. Like moving outside the parentheses of the constructor, like the public promotion, like moving out of the public promotion. Not a thing. So, yep.
Dalton McCleery (14:35)
Wow. What a crazy week you've had.
Andy Hinkle (14:37)
Yeah, so it's been good. Yeah, it's been fun though. It's stuff like that will keep you up at night, but it's always fun once you find a solution and thank goodness it's I found it before the end of day to be in Friday. So that's good.
Dalton McCleery (14:50)
That's true. Yeah. It's like beating a Dark Souls boss. Have you ever beat Dark Souls? A Dark Souls boss? Like the wave of relief when you finally figure something out. Like you can just breathe.
Andy Hinkle (14:55)
I have not played.
I've heard that game is really good because it like, the first when you start out, it'll like, you'll find a boss but you can't defeat it, but it's kind of like, it sets the tone for the game, like you need to, like usually games kind of level up and then you might get to a boss, but it's like level one, here's the boss, or you might have to come back and get this guy later. it kind of sets the tone of like, you have to get better then come back to it. I kind of like that. I've never played it, but my brother tells me all about it.
Dalton McCleery (15:24)
Yeah, you sh- you you should. You should. Get Elden Ring. Elden Ring's good. Yeah, anyway.
Andy Hinkle (15:29)
Yeah, I'm too nerdy for video games. I need to go a level deeper and do programming.
Dalton McCleery (15:35)
Come on.
Program, you know your own video game. Hey, that's a great That's a great like problem like I used to do that. In college, we were actually building a video game as a final project And the video game was to teach you how to program the video game that you're playing And so that's actually a great like final boss the this specific bug that you just had would be the perfect final boss for that video for that like programming video game
Andy Hinkle (16:01)
final boss. Yeah. Yeah, that'd be a good one. Yeah.
Dalton McCleery (16:07)
So, so yeah, like PHP, I think PHP screwed you on that, but there are some nice things about PHP. I don't know if you saw, well, I know you saw, you've seen PHP 8.4, right? Came out. Have you, have you upgraded yet? Any, anything that you like about it?
Andy Hinkle (16:19)
Mm-hmm.
No, yeah. I haven't upgraded yet. I usually wait a little bit, usually about, I don't know, a month or so just to kind of work out the, I know it has like a beta, but I feel like it really gets solidified after a good month or two of rollout. But I've been kind of watching the, you know, just been curious of what's different on it. And I know they have like the property hooks, I think it is, right?
And there's some other things that I haven't seen any benchmarks And years before, I'd be like, hey guys, we have JIT now. Take a look at the difference. And it's like, well, it was pretty incredible. And so, cause I brought up this in our developer meeting here in our agency, both work at. there was like, they're like, what's changed? And I go, not a lot. We got, you know, some array helpers. We got property hooks. got.
Dalton McCleery (17:08)
Cosmo.
Andy Hinkle (17:10)
like versus the getters and setters. So, but I haven't seen any actual, like usually before it's like, Hey, got guys, we got, look at the performance increase. And usually that's kind of been our selling point for clients too. It's like, Hey client, like we're going to upgrade your PHP. We're going to see some performance increases. Now it's just more like, Hey, this is just to say up with compliance. And we're, we're staying with end of life, end of life, you know, stuff, et cetera. So
Dalton McCleery (17:24)
Hmm.
Andy Hinkle (17:37)
But yeah, I'm curious if we can get rid of our getters and setters now. Like we don't have too many of those with Laravel, but like, you know, with property hooks, because for a while now in PHP, like setting a public property is faster than going through a getter and setter to access like a private property.
Dalton McCleery (17:46)
Mm-hmm.
Andy Hinkle (18:00)
So like, I think there was something like if you set a public property instead of going through a getter setter, it's like 60 % slower or something like that. But now they have these new property hooks and I saw something like, like it brings that number up to only a 9 % decrease. And so now it's like, it doesn't really matter if you use a property hook versus getter setter. It's all just, I guess, just what you would prefer. Yeah. So it kind of.
Dalton McCleery (18:19)
All right.
friends.
Andy Hinkle (18:30)
lines up with that, which I just wish there was. know I like the flexibility, but I just wish there was like, just like, everyone, this is the way in Laravel. There's a lot of that. Like, you know, especially with. Yeah. There, you know, there's linters and stuff like that or a PHP stand that will like flag it. But, but yeah, I just wish there was more like, Hey everybody, we, is what we recommend, but it's kind of cool that you can come choose either or, and they're kind of bringing that 60 % slower up to only 9%.
Dalton McCleery (18:42)
Very opinionated.
Andy Hinkle (19:01)
But yeah, I'm not, there's just not, I think in every PHP, yeah, every PHP release in recent years, I'm like, Ooh, that's gonna be really nice about this one. I'm like, okay, you know,
Dalton McCleery (19:04)
small.
Mm-hmm.
Yeah, there's like one, there's probably like two things in it totally that are like, that I thought was like, that's really nice. And it's that one where you don't have to wrap your class in parentheses to call a method anymore. If you just want to do that, that's fine. Love that. That's great. A lot of those like array, there's like an array find or something with a key where you can find a key in an array really quick with a callback. Like that's great. I love those two things. Otherwise.
Andy Hinkle (19:26)
yeah. Yeah.
Mm. Yeah.
Dalton McCleery (19:42)
I haven't looked at anything else to upgrade since, but just those two things. I'm like, all right, that's nice, but nothing's blowing my socks off and it's quite hard to do that. And I've already had my socks blown off like four times this year with the Laravel announcements and cloud and nightwatch and all that other stuff. So this is probably really impressive to like your average PHP developer, but to us, after all these things, this is all right. This is nice.
Andy Hinkle (19:54)
Yeah.
Yeah. Yeah.
Dalton McCleery (20:10)
It's not crazy, it's not bad, it's not it's nice.
Andy Hinkle (20:11)
Yeah.
Yeah, yeah with yeah, I'm kind of curious with PHP and JIT I'm curious like with If they're continue with that. I didn't really see much updates They did something with OP cache like you could set some memory limit thing or something where it's a bit faster with OP cache But just maybe it's just coming more mature now But just in recent releases there hasn't been any
Like with this PHP A4, hasn't been anything game changing or like granular changing with the JIT compiler. So I'm wondering if they're starting to see the limitations there and maybe have to pivot or something. yeah, I'm kind of curious where, you know, where they'll take it with future PHP versions.
Dalton McCleery (20:56)
Yeah, let's see. Let's, they only got to seven four last time. So does that mean eight four is the last one we're going to go to nine or are going to do the apple thing and go, go to 10 and just forget that nine exists? Yeah. Yeah. Nine just.
Andy Hinkle (21:04)
I don't know. Go or do the Laravel thing. Do the Laravel thing. Five, four, five, five, six. I think we had a five, Yeah. And then six, six point seven, eight. Yeah.
Dalton McCleery (21:14)
six, seven, eight, yeah. All right, checkmate PHP maintainers. see if we get an eight, five.
Andy Hinkle (21:21)
Gosh that, that drove. Diversioning on Laravel, like I, I get, I got it from like level maintainability standpoint, but there are so many people like, my gosh, Laravel is just updating so much. I can't keep up. Like when it went from six to seven to eight, it was just so annoying. Yeah. Yeah. It's, it's normal. It's.
Dalton McCleery (21:36)
Right every year Like guys relax Think of it. It's not it's not it's not a crazy like five three of five four like that was crazy Right, and that was just that was like quote unquote minor update I remember that and I was like jesus most of my stuff is broken in five in five four of larvo
Andy Hinkle (21:46)
Yeah.
Mm-hmm. Yeah. Yeah, my one of the sites that I just I changed code on once a year. I did Laravel 10 to 11 and I think I just had to change ENV variables like when they changed the cache driver to cache store or whatever. That was it. And it worked like I didn't have to practically do anything. So, you know, it was really nice.
Dalton McCleery (22:15)
You didn't delete all your config files. That's what I did when I was upgrading to 11. I just got rid of them.
Andy Hinkle (22:18)
No, on on all of them just follow my other apps. Yep. I did that I went through What's it called level shift because it just does all of that? And but yeah other than that that was the only app I didn't do it on just because it's like I I changed like a date on the website when it's like a It's like a local basketball tournament that I help with. I don't make anything I just you know participate and just a fun little thing I do and I just changed some of the context some of the rules and
Dalton McCleery (22:27)
Okay.
Andy Hinkle (22:47)
like the registration link to sign up and that's it. It's a little like landing page, but that's it. And yeah, I just went straight 10, 11. Like I could probably put it like on the next JS app, but it's just like simple. I just love play too much. So it was easy to strike up. yeah, till when it's very simple. So cool.
Dalton McCleery (23:00)
Yeah, blade boys.
Andy Hinkle (23:07)
So, we've been having some downtime, Christmas holidays, which gets a little bit slower. And I've been thinking about, have like a journal or not really a journal. sounds very like sophisticated, but I have like a little notebook that any paper cut I run into with Laravel and I just jot it down. And some of it, it's funny. Some of it, I look at it like, what the heck was I even talking about? You know, it just like, I'll just, I'll write like a bullet point. And so I went back through
the journal and started looking through some things. one thing I mentioned was, publishing stubs. Like when you do PHP artisan stub publish, it gives you everything. I think it's like 50 stubs. so it's like, because they, added a lot more because you can do artisan make, you know, now make cast, et cetera, and you have to have a stub for each one of those. And so,
Dalton McCleery (23:47)
So hot.
Andy Hinkle (23:59)
They've added several new ones probably I want to say around 30 maybe less probably around 20 to be realistic but new steps, but I found it overwhelming especially like if I need if I need to a stub for My migrations for example, I don't like to do a down method which I'd like to talk about that I'll show but you know, I I don't like to do like down methods So I don't have to want to have to remove it every time put it in a stub remove it and then if I
Dalton McCleery (24:18)
We will get to that.
Andy Hinkle (24:28)
But if I need it, you know, you can add in later, et cetera. But the stuff is really nice. It just, it keeps you focused on what you need to do. But if I do that, like every single step is published. And then if Laravel goes in, add something new or changes something I've run into this where I'll upgrade app and then I'll have stubs back on from like Laravel six. And so I have to go figure out and like, okay, I want you to force these stuff changes. And you just have to remember to do that. And I was like, I'd rather just rely on Laravel's defaults and
Dalton McCleery (24:48)
Oof.
Andy Hinkle (24:57)
if it changes, whatever, you know, so anyway, yeah. So one thing I wrote down like stub publish, like, why is that so frustrating? Like it just gives you everything. It should give you something maybe like a flag of like dash dash only, like stub publish dash dash only, and you just tell it what you want. And so I started looking at doing stub published dash dash only.
And I got thinking about it and we have like when you do PHP artists and make controller, if you don't pass it anything, it gives you like this really nice interface of like, Hey, what do want to name this? what kind of options do you want to have? especially when you're setting up a new Laravel app, it's really interactive that way. And so that's using Laravel prompts. And I was thinking like, what would it be like if I just did a, like a select in there, you can select which one you want. And that just expanded into, okay, well,
What if I do a multi-select? Okay, so I made it to where you could do a multi-select when you do stop publish. And then I went one further, I'm like, what if you could type something in there, like migrations and it just filtered or typing controllers and just showed you everything. And you could just select which ones you want on your keyboard, hit enter and it just publishes those steps. So made a PR and so.
Dalton McCleery (26:09)
Mm-hmm.
Andy Hinkle (26:10)
And what that will do is you can, you can tell it which steps you want to publish directly from that selection menu. was explaining. So it's easier to target the exactly the stuff that you need. Use this Laravel prompts. He has that fuzzy search. Like I was talking about, and it has an all option. So if you want to go back to the way it was before, if you do stub publish, which you're an animal, if you want to do this, like if you just, I just want everything, but you can just do stub published dash dash all. So, I felt really good about it. I sent the PR in.
Then I started looking at like, I totally blank. Cause usually when I think of this stuff, I kind of see if somebody else has done this in the past. Cause I've run into other things of like, Ooh, this would be nice. And then I go look up and PRs and somebody else has done it. And then Taylor just didn't like it. Maybe even the, the implementation or if it's just something about it, you know, maybe it just wasn't the right time, et cetera. And so I've totally forgot to do that. And I went back and it was submitted by other people a couple other times. One was by Luke Downing, which is a.
Laracast instructor, really cool dude. really nice guy met him at Laricon this past year. So, you know, he submitted the, the only option like, stub publish dash dash only. And then Taylor came back and said, Hey, what about if we did like layer of prompts or something? And that's why he closed it. And I was like, Hey, I just didn't realize that. So I put a comment. was like, Hey, Hey guys. Like, yeah, I didn't realize that, that's the direction they're looking to go. Anyway, I just happened to stumble upon it. Just trying to.
Dalton McCleery (27:10)
Mm-hmm.
Cha-ching.
Andy Hinkle (27:37)
Like what if we could just do more out of this and ended up coming into that. And there was a couple other PRs where people were kind of doing something similar, you know, just trying to do a match. But anyway, submitted in, Taylor came back and he wanted the names to sound a bit more like human readable. And so after reviewing them, I realized like the names, didn't really sound natural anyway. It'd be like migration.create.stub.
be like one of the options you'd have to select through. It wasn't really friendly. And so I totally got that angle. So, and I also found out like when I just did, when I use like the string helper, like migration, create, it would just be a dot stuff. would just be migration, create. I'm like, well, that doesn't really make sense. So I'd do something like I'd use a secondary descriptor. So and if controller invocable, would be invocable controller, same thing like create migration, you know, through that list.
So I did that, brought lowercase API to all uppercase API. know, just all the, just little things in there, just yeah, handling strings, just like you would expect doing that. But then what caught me by surprise is Taylor also said, he's like, think what the most let's trim down the section to only include the most commonly used steps published ups. I'm like, what are the most common ones? Well, as I was like, well,
Dalton McCleery (28:39)
the nice cities.
Mm-hmm.
Andy Hinkle (29:01)
So as I went through all of our apps at our agency, just kind of like jotting down which ones are used. And I just kind of thought of like as a community, like which ones, and I just like, couldn't come up with a great answer. So I just like, what are, what are just kind of the most commonly ones and end up still being quite the list. So, I better review all of them and just, just kind of focus on the essentials, but there's just so many, cause you never know with your type of application. feel like not having it there is kind of critical, but.
I went through and it was, I did, I got my list here at models, controllers, migrations, Seeders, factories, tests, resources, notifications, mailables and jobs. No, that's a lot, but I feel like an, at least in our agency, I've been in every, like all the stubs in those situations I may have interacted like cues. For example, I hate that the, I know Pint could fix this.
Dalton McCleery (29:48)
and it's still kind of a lot.
Andy Hinkle (30:00)
But I hate how like the traits are all on one, like use this, use that, use this like on when you create a job. So like we have a stub just to make it like just one use and it's comma separated. yeah, anyway, yeah. So I went through, just try to find, and it brought the list, the stub late, like when you, the ones you can produce from 52, I think it was to around 30. So then I was like, this is, I just told Taylor, like, this is my.
best opinion, I don't know, but just some thoughts there. And there were some things in like enum stubs and like, those probably aren't going to be used, you know? So yeah, but those are brought the list down. and I also was asking like, maybe you should create, like, we should add like a list all flag. Like you can do stub publish list all, and it gives you those enums, but it's just kind of weird as a developer. Like you got to know that.
Dalton McCleery (30:47)
yeah.
Andy Hinkle (30:52)
Why isn't you know I'm stubbed in here? Why is it not in here if you go a Polish dub and it's not there? So I'm kind of I don't know it kind of go I can see Taylor kind of probably going back and forth on it because usually he's very prompt on his PR so like let you know and it's been like a week or anything so So it's been it's been open. Yeah, I've got a few like, you know I asked some feedback on Twitter asked and blue sky and people seem to really like it at least because it's been it's it's not it's not
Dalton McCleery (31:08)
I was gonna ask about the follow-up.
Andy Hinkle (31:21)
groundbreaking or anything, but it is an ICD. It's just something that's always kind of bugged me. Yeah. And so, I got some good feedback. hope maybe it might be accepted or continue just kind of poking at it. But yeah, we'll see. But it is kind of nice to, when you do your stub publishing it, you can choose which ones you want. Shouldn't get the whole kitchen sink. So what do think of that? Is that, are you a stub guy? You stub guy. sounds so weird to say you stub guy.
Dalton McCleery (31:24)
Yeah, quality of life. Yeah.
Yeah, because I don't want to have go back in and delete a bunch of stuff.
Well, Andy, let me tell you No, not I I think I told you I think the only like stub that I Routinely want to update is the migrations and that's that's that's usually my only use case for it because This is weird and and it's totally my preference is I hate the comments in the migration So I'll always remove the comments from the up and the down method
Andy Hinkle (31:51)
You
Yeah. Yeah.
Dalton McCleery (32:17)
Most times I'll remove the down method too. So it's literally just an up. It's the leanest file that I can possibly get it to be. Because I want it to be so simple. So that's usually my only use case that I'll do it. I'll always keep the default factories, models, controllers, whatever Laravel thinks is great, I'll use that. I'm not much for the customization part.
Andy Hinkle (32:24)
Mm-hmm.
Yeah, and to your point, I know when we work on projects together, like in a model, for example, like if you have the public function cast or public property cast and up at top, there's a comment, the attributes that should be cast, like, yes, yes, yes. Yeah, and so we set a stub to remove those. You could probably do something through Pint or something, but then I found out through, it's not really reliable through the PHP CS fixture of like.
Dalton McCleery (32:52)
Hmm.
Yeah, get those out of there.
Andy Hinkle (33:06)
I don't want this particular one, but there's some other times I do want to comment so you can't do the either or. So we just strip it in stubs and works really nice. People are going to know. People are going to know that's the cast. Come on. Yeah.
Dalton McCleery (33:15)
Yeah, I, yeah, yeah, I mean, it's cast, it's public, or I guess it's protected, right? Protected cast, or you use the public function, cast. You know what it is. yeah, yeah. You know what it is, right? Like, I don't need a seven line comment of like, this is where you're gonna cast things into different classes and et cetera, et cetera. So yeah, those are my only use cases. I guess, models and migrations are my big issues. I'm curious though if...
Andy Hinkle (33:22)
Yeah, but if you use public good, yeah, be careful.
Mm-hmm.
Dalton McCleery (33:44)
You said you posted it on Twitter and Blue Sky. Did you ask people what their most common stubs that they use? Is that what you use to determine sort of quote unquote popular ones? Like, hey, of these 52 classes, how many do you guys normally publish? Did you ask stuff like that?
Andy Hinkle (34:02)
Yeah. Yeah, I did. And I got some weird responses. Like, why are you asking? Like, that's such a weird day. Like, what, you know, what, what does it matter? And I was like, yeah, I had to kind of explain. but I did ask, it was on discord asked a few, just like, like Laravel community members. Like when you guys do a stop publish, like, what do you do? And most of them just say, they just publish everything. They just don't, I guess they just never really questioned it. And they just like, well, if I, if I, just publish it, but if I ever need to like,
change that, just go change it, you know, cause it's already published. doesn't really matter. But my side of like after upgrading Laravel apps over the years, I just found that those stubs get left behind and I don't know if Laravel shift fixes that. think it was good suggestion for Jason if, if it doesn't, but, I think it does. I don't know, but some of the apps that we've upgraded, can like, man, this is using the old factory stub, you know, stuff like that. And we just have to,
Just ensure that it's updated and we get confused of why we do a create factory and it's using the old stub, you know, or something or one from the five or six days.
Dalton McCleery (35:07)
Yeah, those factory classes back in the day, woof. They're so much better now, but they were pretty rough back in the day. Yeah, because I mean, the last thing you're going to do when you're upgrading from like 10 to 11 or using shift is I need to republish all my steps. And at that point, you have to republish them all, right? Because now if you make changes to the migration thing, you basically override it you have to go back and sort of undo the republish.
Andy Hinkle (35:11)
Yeah.
Yeah.
Who thinks of that? Nobody. Yeah.
Mm-hmm.
Yeah,
Dalton McCleery (35:36)
Do you, it's, That's a bad paper.
Andy Hinkle (35:38)
Yeah, maybe I'm just weird and I really like just to the small things of like, want my application to match like the Laravel slash Laravel repo, like the skeleton repo, like even the code comments to the slide as things are any, and Laravel shift does an amazing job at that. And otherwise you have to manually do it and stuff, but even like for stubs or for anything, I want it to be like, I don't have to worry about like something weird. Like when I'm focused on a feature, when I'm focused on like writing code, I don't want this like.
what's going on here? This is from Laravel 5 or Laravel 6 and they to figure out how to baby it along, you know, going forward. It's always nice just to when you do have, you know, everything's up to date, everything's fresh. It doesn't have to be on the latest version. I don't have to be on PHP 8.4 today. But I'm just saying when I do finally make that upgrade, I really expect to know my code to kind of reflect that. And I don't have all this legacy code hanging around.
Dalton McCleery (36:34)
Yeah, the less, the less that we have to maintain the better because it just gets insane. I mean, we're, think last I looked, I think I had like 22 larval sites that I manage now. I mean, everything that we have to change times it by 22, you know, on a given day, I don't want to have to republish and edit 22 different sites worth of, know, stub files, whatever 22 times 52 is that's a lot. That's a lot of files.
Andy Hinkle (36:52)
Mm-hmm. Yeah.
Right.
Dalton McCleery (37:04)
I don't want to have to do that.
Andy Hinkle (37:05)
Right, right. Or if we're handing off the Laravel app to somebody else or, you know, I'm having somebody else work on a feature, I don't want them have to like realize, this is a weird thing from Laravel six or Laravel seven and old, old component pattern. Don't worry about it. You know, it just doesn't make sense for sure.
Dalton McCleery (37:25)
Yeah, I should, I should put one of those old factory classes from five four and just one random Laravel site just to see what happens. Just, and then I'm going to go on PTO and I'm, and I'm just going to see if somebody loses their mind. Like what in the hell is this? Sorry, Jam. No, it's bad. I remember it was like.
Andy Hinkle (37:33)
Just confuse people. What's this do?
I don't even remember what that looks like.
I'm trying to picture that.
Dalton McCleery (37:48)
It wasn't even like a class, right? It was just like a PHP file, and I think you gave it a model class, and it was just like an array of just garbage, right? It was just like a... Yeah, it was basically like a config file, if I remember right. It was just like it returned an array almost, and it wasn't like an actual class that you could define methods and stuff in, if I remember right. It was garbage.
Andy Hinkle (37:52)
Mm-hmm.
wouldn't it like user factory colon colon, then you have to do it. Yeah.
Interesting. Yeah. Yeah. I'm trying to remember that too. It's, it's crazy. It wasn't that long ago, but I just, like now I'm like, how could we ever done anything any differently? thing for level components. I how did we ever, how did we do it without this back in just the NVC days? Like everything's a model view controller.
Dalton McCleery (38:31)
A lot of includes in those blade files. Lots of includes. Extend, I haven't used extend in forever, right? Extend in sections, I remember all that stuff, ooh.
Andy Hinkle (38:33)
yeah. Man, I can't remember the last time I wrote in include.
So another one, yeah, another one I've been, I went upon is like, I was doing a lot of modular stuff, like modular, development. So, I forget who that package is by he's a community member. It's a Chris, call him boss Chris, but camera is it Chris morale boss Chris is yeah, it's, it's his like, it's his like Laravel community member named, Daniel Colburn calls him boss Chris cause he used to work for him. So.
Dalton McCleery (38:46)
strip down memory lane right there.
boss Chris? Is that what you said?
Andy Hinkle (39:16)
Really smart guy met him at Liricon this year and he also has a podcast But he has this package out there allows you to develop level apps in a modular way and you've helped me on few of these and you've kind of questioned it and like I've tried to get you to bite in on a few of it, but without and get into the whole modular debate There's another area of where I'm looking to help for this too. And that's also package development but so one thing I would like to
talking about stubs and just file generation in general. One thing I've been looking at to PRing into Laravel is this idea of in your app service provider having like an enforced file map. So hear me out, do like a make, PHP artisan make model and maybe you don't want it to be in your app models folder. Maybe you're tailored out well and you just like it in your app folder. He was like, he didn't change that for a long time. I feel like he...
Dalton McCleery (40:04)
Mm-hmm.
it was long time.
Andy Hinkle (40:13)
Yeah, I feel like he heard enough from the community and like, we need Taylor. need to app models folder. But if you, if you prefer that way, right. you can do like, so just like, hate, I feel like I've talked a lot about like, talk like code, what it looks like in on audio podcasts, which is not helpful, but just pretend you're in your, your app service provider. You type in, you type in generator command, colon, colon, and then you have enforce.
File map and open parentheses. So this is very similar to the eloquent enforce morph map. If you're familiar with that. Yep. So inside there you can do, you can open. So you have enforced file map, open parentheses. Then you have an array for this array iteration inside the array. The key of the array is the stub name. So it would be like command dot stub or model dot stub
Dalton McCleery (40:44)
Okay, yep.
Morph map?
Okay.
Andy Hinkle (41:10)
function. So like it just like where you want it to go. So basically you're making a callback of where you want the file to go. So it'd be like the key would be command dot stub arrow function. Function name is like the original name. Like, this is where the name of the file is going to be like you inside the function. So then you can give it a return path of a string. So can do like app path slash custom slash.
blank or named at PHP so you can move your when you do PHP artists and make model is going to put it in the custom folder like the literally the name folder called custom or you can just do app path and just not give it anything and then it's just will deliberately drop it in your app path. Mike Taylor likes to have his models back in the day. This is really helpful for modular development and also like package development like there's been a few times where
Dalton McCleery (41:45)
Mm-hmm. Custom.
Andy Hinkle (42:06)
You can't test something that's package related in Laravel. So what you do is you make like a folder in your directory and just basically copy your your package out of there and And just work from an actual like a level application sitting on top, you know, does that make sense? Yep. Yep. I think everyone's been there Yeah, everyone's been there because there's been sometimes you just can't do it like you may you might have to see a webpage or something or whatever the case is you have to yeah, you
Dalton McCleery (42:23)
That's me. That's me. That's me.
Andy Hinkle (42:34)
need to actually physically see it. So what you do is you drop your package and maybe your app directory and just reference it in your composer file or something. But anyway, composer this if you have like a level app set up just for that, you can do like come in your app service provider, generator command, colon colon, enforce file map, open that up, and then you can do model dot stub. And then so when you create a model, you can put it inside that package. And so when you do like make model, it's like a CLI basically for your package development.
Dalton McCleery (42:40)
Mm-hmm.
Andy Hinkle (43:04)
makes it so much easier, but helps with like domain driven development. can, it's a function, so you can do whatever you want in there. So if you do like function name, you can just do like a match statement based on the name. So there's two parameters I put in there. I was thinking about is like when you do function name, so that's the name of the file that's coming in. And then comma is the default file space. So if you want to do like a string helper to kind of tweak that or whatever.
Or if you want to just slightly add one more director on top of that or just change it whenever in my case, like for my tests, for example, like one of my big use cases for this, my tests, I like to reflect it to be exactly like the application folder. So instead of like feature HTTP controllers, I want it to be like feature app, HTTP controllers, et cetera, et cetera. So that way it reflects exactly parabatums. That way I know exactly like the file that's testing. So getting a little long winded here, but that's kind of, I have.
I've cut up a proof concept, haven't submitted a pull request yet, but one thing I ran into, I'll just kind of stop there. I'm just kind of curious what your thoughts.
Dalton McCleery (44:14)
Yeah, the thing that I would use it the most for is package development, but I'm also, curious on when you were talking about modular, like if I want to make a model for a specific feature and it has a specific folder, how can I like dynamically tell it where to go? Like I want to make a model. So I do PHP artists and make model. And I've got this, this enforce a thing in my service provider and I have, I want models, know, model create stub to go to the app.
model's directory. How can I tell that to go to, I want it to go into this feature A, and then now the next time I do that, I want it to go into feature B. You know what I mean? I want to make a model for this feature. Now, a week later, I'm going to make another model for a different feature. How do I change the directories on the fly without altering the actual make command to have an extra argument? Does that make sense?
Andy Hinkle (44:55)
Mm Yeah.
Yeah, that's, yeah, yeah, it does. And that's actually what the modular package does is it overrides. It basically extends that class and then hooks into your app service provider and then just changes how it works. So basically when you do make model, there's a flag in there called dash dash module. So you can do like dash dash module administration. And then it will go in your modular and it knows.
Dalton McCleery (45:34)
Uh-huh.
Andy Hinkle (45:38)
this is going to be in the administration area. And so that's how that package works. It just extends that class. So it works fine today. This wouldn't be able to fix that per se, but it would allow them to kind of lean on Laravel a bit more to kind of construct that of how it should be placed instead of having to extend it and then override it. then same thing for package developers. It can be helpful. And then if you have any business rules, business logic, or just something like really specific that you want this
You want to do make controller, but you want the controller to be in somewhere very specific. It gets that point. One thing that you, I'm glad you brought that up because one thing I've been thinking about is like something like that. You want to do, maybe you want to pass a modular flag or just some, any type of flag that you want. Maybe, inside the method we can like, I told you, we have, have, we gave it the callback on it. We give them the name of the file and the default namespace. Maybe, maybe the third thing is just like.
the options that was provided. Yeah, like so that way they at least have context. They can do like a match or like a return based on like the options that were given. So that way they at least know. Yeah, it gets really kind of complicated, just complicated in scope. And so I'm trying to figure out like the best way to run to kind of set this up. I was going to mention.
Dalton McCleery (46:52)
Yeah, yeah, that's...
Andy Hinkle (47:03)
The one of the things I ran into was namespaces. So Laravel kind of guesses that the namespace based at like what's provided in the context. So if you do like make controller slash administration slash user admin controller, it's going to like ensure that the file space is correct, that it's including administration in your namespace and stuff like that. And it's going to put that new folder in there called administration. So it works really well with that. With my change,
it has a thing in there in the code is if it knows your callback, so you can either do a return of a string and it just knows that this is the direct path of where the file needs to go. If you return an array that there's another method, if you return an array, can give it inside the array, can give either, you can give a path in a namespace. So you can give it the path of where you want it to go.
And then the next key is the namespace. So if you want to have this really bizarre custom namespace, all good. So, so that's one, like it's really useful for packages really is what that kind of got me thinking there.
Dalton McCleery (47:58)
Hmm
Andy Hinkle (48:03)
that was just so to enforce file generation assignments would be kind of a neat thing to have. I don't know. It might be in scope and possible issues. It like, it may arise from it. if
I know Laravel looks for controllers and stuff they know in the app directory and so if you put something weird and you put it your public domain or you know just something that a junior might you know run into if you're not really careful but then again you could really screw yourself up with morph map if you don't know what you're doing there so I don't know.
Dalton McCleery (48:33)
Yeah, now that's interesting. I mean, that gets my wheels turning on what I would use it for, specifically for package development. That would be nice. Modular development would be cool. I mean, they've got a package for that already. So you would just sort of like another version of that within Laravel that you can do without including that package. Yeah, there's a lot of weird things that you could.
Andy Hinkle (48:42)
Mm-hmm.
Yeah, they kind of solved that, right? Mm-hmm.
Dalton McCleery (48:59)
I mean it's dynamic enough to where I could probably give it a callback and say, based on an argument one, I want it to put into this directory and I want it to follow this namespace. I don't want it to be app, admin, whatever. I want it to be admin something, HTTP controllers. Or.
Andy Hinkle (49:12)
Mm-hmm.
Dalton McCleery (49:20)
If it's not admin and it's user, I do want it in the app directory and I wanted an app models namespace. So it, opens me up for more dynamic stuff for complex Laravel applications. feel like normal Laravel applications, probably a bit overkill. So yeah, I'm curious on how many big, curious on how many big, big applications would, would use something like that instead of a modular approach or just doing.
Andy Hinkle (49:37)
Yeah, I don't, yeah.
Dalton McCleery (49:50)
Just doing it all by hand where they just do, make model. And then I'll just manually move it into this directory and change the namespace myself. Right. So yeah, it's very specific. It's very specific paper cut that I think, I think it has a use cases in complex Laravel applications. I don't know how the, I don't know how the hell you would, test all of those scenarios. You know what I mean? That's, that's, that's a lot.
Andy Hinkle (49:58)
Yeah, domain driven development, yeah.
Mm-hmm, yeah.
Yeah, yeah, well, yeah. I've run into some problems with Laravel itself because they have the generator command, is what just about most commands they use for generating things. Sometimes they overwrite that. If you do make model dash dash f, it's gonna make you a model migration factory. it's actually, that command is the make model.
command explicitly is like it's listening for those dash dash F and it's overriding those methods that hook into this generator command, if that makes sense. So basically I have to, I basically have to rework it into a trait and just call it on several different commands in the Laravel framework to make this even work properly. And then that's even when it's possibly even close to being considerable for PR.
But I've been really curious, just from other people's opinions, mostly with big, level apps, do you run into this when you want your file generated? Most of time, it's probably just like Stubbs. When I ask people's opinion there, they're just like, I just publish it, I don't care, it's there. But I bet you would, when you do make model or make controller, if it's not in the right place. So I'm just gonna click and drag the file over and call today. It takes you two extra seconds, but.
I guess for just for compliance and for things you don't really have to think about it when things just work, you know, and based on your business requirements, it's just something you set up in your app service provider. It makes things really handy,
Dalton McCleery (51:50)
Yeah, on I mean onboarding if you're working with multiple developers on a team, right? Like everybody's making models here and there. It's just gonna throw here. Mike forgot to move his model to this directory because he just forgot.
Andy Hinkle (52:03)
Yeah. yeah. Well, another one, I don't know if this would work for live wire, but for this example, but, when I was, I don't know if they fixed it, but when you were in modular and you're working with live wire, if you did like make live wire and you gave it modular flag, it just didn't work. Like make like a live our component dash dash modular. It didn't work because it was like, it was something in the light, you know, the live wire generation is its own separate thing.
So I don't know if it'll work for that. just something, I just want it to be kind of fluid across the board and kind of help. I don't know if that's still a thing now in the modular package. I don't know if you can, when you do make live wire, like if it, if it's still like dash dash modular, does that work now? But I'm curious. But anyway, at least when I was working on it, it did not, but that was about a year ago.
Dalton McCleery (52:56)
it's it's there's some stuff to chew on there and I think it almost makes me think that I would like I would like some sort of a skeleton for package development. Like I normally just create a Laravel app and I'll do exactly what you said. I'll put a folder in there and I'll put all of my files in it and I'll start just dragging dropping stuff just to make sure that it works. I know if there's just like a slimmer version of Laravel that I can just ignore those extra files and I don't know.
Andy Hinkle (52:58)
Mm-hmm.
Yeah. What I usually do is I do the spotty. they have a package skeleton. I use that, but I still have to create like a blank Laravel app and drag it in, you know, like in. Yeah. Right. Yeah. You don't have the CLI or anything. he, works for Laravel now, but, there, there was some package out there where you could test very granular things. And it's like, it was like a test sandbox, but for Laravel development packages, it was really good.
Dalton McCleery (53:30)
Yeah, because you don't have the artisan file.
Yeah.
Andy Hinkle (53:48)
can't remember the name of it. He worked on Laravel Nova for a while. anyway, he's actually, think, the new Dries. He kind of handles gap issues for Laravel. anyway, he manages a package that helps with package development. But still, I don't think you can look at an HTTP page on your browser or anything and to be able to test something, you have to deliberately create a Laravel app and poke around. So like for filament extensions and stuff, that's primarily where my headspace is going is.
Anytime I created a filament extension or stuff, you can't do that in just a package. Yeah.
Dalton McCleery (54:18)
Dude.
For real, for real. That headache, I've had that so often in the last like six months. It's insane how crazy that is to do. Because the way that I do it is in filament in a package for Laravel. So now I have to have the package and I have to have the filament and then I have to have Laravel with filament installed just to test to see if this damn link shows up in the dashboard that I want it to be in the correct order. Like I want to order it so that
Andy Hinkle (54:49)
Mm-hmm.
Dalton McCleery (54:51)
My widget's always the first widget on the dashboard. Wild. That's probably where this receding hairline came from. I don't know if you can see it. Lot of years.
Andy Hinkle (54:56)
Yeah, yeah. Let's talk about, we talked about earlier, talk about down methods. So we talked about you being a stub guy, you being a, are you a down method guy?
Dalton McCleery (55:10)
I think I think we've Yeah, I think we've had this exact argument before And I say argument you loosely loosely because I don't think we've actually argued with each other. But yeah, I'm I'm a down I'm a down guy. I like to have it there just To be there because I think before I can't remember what you use. I think you use migrate refresh seed dash dash seed I think that's what you use and I was using Or maybe maybe you're doing migrate
Andy Hinkle (55:12)
trying to remember in your PRs if I see DOM methods.
Parking lot.
Okay.
Dalton McCleery (55:40)
reset or something like that. I was using a different one and you were using one that ignored the down table. just like basically just deleted your database and started over. And the command that I was using was, was, was rollback, right? It reset all the migrations and then run them all in order again. So I like to have the down methods specifically for that so I can go back and forth. Whereas I think you before was just like, just delete everything, start from the top. So I like to have down methods just in case like somebody comes in and they do migrate rollback.
Andy Hinkle (55:48)
Yep, yep.
Yep, blow it away.
Dalton McCleery (56:10)
I'll put them in there, I'll keep them auto-generated. I'll do create posts table and it'll automatically have a schema create and a schema drop. And I'll leave it as that. I'll delete the comments and do my stuff and move on. I have down methods.
Andy Hinkle (56:10)
Okay, yeah.
Yeah. Yeah. So, yeah, hear me out on this. I think down methods are dangerous. Yeah. So the reason, reason, I know you're getting all heated over there. So I've, I find like a complex down method, like especially in production when you need to roll back and it's has not been tested and we don't know if it's been tested and obviously it can't be tested. If you're testing this and CI and stuff, that's stupid. So
Dalton McCleery (56:29)
Okay. Okay.
I'm getting angry now.
Andy Hinkle (56:54)
You're not going to testing your down methods to see how to ensure they work unless you're, I don't know. But anyway, yeah, the, the, a complex down method that's especially it hasn't been tested is more dangerous than no down method method at all. And, it kind of, anytime you have the, you have to be really careful with your, your actual migration method. Like that's going in your insert that's going in and not having that sense of consideration of like,
Dalton McCleery (56:59)
crazy.
Andy Hinkle (57:24)
I could just roll back. You know, things go wrong, you know, I can see that take of like just making sure it's correct. And, there have been times where I do want to roll back. So I do have, I do have down methods on existing data tables, database tables. I will say that. yeah, yeah. But if it's a new, if it's a new feature, new method, like, no, like we're either going to blow it away or we're not going to have, like, I don't need to have rolled back this
Dalton McCleery (57:41)
You will admit that you have down methods.
Andy Hinkle (57:54)
model like if it's I always just like to continue going forward like if there's something wrong I'll just create another migration to fix it you know and there's been times where you know the the down method is like deleting a table but or sorry the insert is deleting a table so the down method needs to be recreating that table there's been so many migrations since then maybe I don't even know what the table fully looks like and so I have to
kind of look at table plus to kind of architect that out of like, well, this has an index and this is over here. And then you have to look at foreign keys. It's just a mess. I'm like, I'm like just down to it down. Forget about it. You know, either we're, know for sure when this deploys, that table is gone. Like back it up. You know, if you, yeah, if you want it, if you want that thing to come back, you gotta have some backups, but yeah, usually not. So if I'm creating a new table,
or if I'm deleting a table, no down method. If I'm changing something, absolutely. If I'm changing something from a string, VAR card to like a text, sure. Like if I need more, yeah, I'll absolutely do a down method. But yeah, it's kind of my hot take on that. Just too much to maintain. They don't really work when you want, like you don't ever use them. When you do use them, it probably won't work because it's kind of testy. You make sure it does work because you want to, it hasn't been thoroughly tested.
You might break something else and then you're worse in the pickle than just creating a new migration to fix it. You have some opinions, I can tell.
Dalton McCleery (59:24)
Hmm. no, I mean, I agree with everything that you're saying. Especially if I'm changing columns, like if I'm going to drop a column and add a column, I'll usually have a down method for that. That way, if I wanted to roll back, but yeah, I'm curious, like people listening, how often do you put a down method? Like, are you using migrate refresh? Are you using migrate rollback? Like how often, I don't think I've ever ran a rollback in production. I don't think I've ever done it.
Andy Hinkle (59:34)
Mm.
Dalton McCleery (59:52)
I've done it in local. I don't think I've, I don't think.
Andy Hinkle (59:52)
I've done it probably I could count on my hands like how many times but it's very very rare like shoot you know it's usually a Linux versus Mac issue like when it does happen like when it does it's usually some discrepancy that something weird you know because we usually see I will always catch it you know test will always catch it
Dalton McCleery (59:59)
Yeah.
Yeah, I mean it's great. I see both sides of our arguments. I'm curious on what other people use because I can't fault you for what you said. That sounds right. I would use that for those use cases. I use the down method just because that's what they come with when I create a migration file. Maybe I just need to publish those stubs and get rid of them completely.
Andy Hinkle (1:00:36)
Yeah, yeah, and just get rid of that down method. My stuff, my, my steps for, my personal websites do not have down methods, but they're nothing. They're nothing. you know, I don't need to, it's continuous, you know, so I'm not going to sit there and, you know, create those, but I always continue to go forward. I always see migrations as going forward. one thing I don't get too often tangent with this, but like one thing I've always wanted ever since the beginning, when I was first learning Laravel, Symphony had this, I believe.
but I really want model based migrations. Like that would just be so nice. Do you know where I'm going with this? Yeah, we got to talk about that sometime. But basically in your model, your user model inside there, imagine having like, you type in a command and something with constraints that you set up in your user model and it just creates a migration for you.
Dalton McCleery (1:01:09)
Bro, yeah, that's yeah, that's that's a whole podcast right there Agreed. All right, add it to the list
Andy Hinkle (1:01:33)
based on your schema that you have in your user model table. So you can have your schema set up in your user model and it just, you type in like generate migration and it's loop and it generates a migration. The tricky, I've started going down this road. The tricky part is, like, we're getting into this topic, but the tricky, I'll keep it short, but I'll keep it short and sweet. But the tricky part is, like Laravel is continuous. It's going forward. It, Laravel can't do, it can't go back and, decide and roll through the method and decide where you're, where you're.
tables are at the current state. so basically it, Laravel doesn't know, it doesn't have the ability right now to know you need to change this from a text to a VAR card. I need to generate migration, what that looks like. You have to, you have to like think of your own paradigm for that, like, and how you're gonna solve that. But we'll get into that topic next show. Good, yeah. As I dove right into it, yeah. But one thing, yeah, cause we're also wanted to.
Dalton McCleery (1:02:05)
Hmm.
That's a teaser right there.
Strong opinions.
Andy Hinkle (1:02:29)
we threw in here was migrations. There's a lot of debate here. Migrations versus Seeders. And we had this come up not too long ago where I think it was fairly recent where you comment on a PR and you're like, why is this a migration?
Dalton McCleery (1:02:47)
Yeah, you put like some, don't, I think it was an insert method. You had an insert method in a migration. You built a table and then you inserted values directly into that table, just like in the next line. I was like, why, Tell me why.
Andy Hinkle (1:02:53)
Yeah.
You're like what?
Well, I think this goes from my early days because back in the day, back in like the layer of all four and five days, you could not have factories and Seeders in production. It, just wouldn't like, if you ran user factory, it would just be like, this doesn't exist. Cause I think it was actually like a local development dependency. So I think that's where my mind started, like the first initial thought. But then once you get past that, it's just like, when you have like permissions and stuff as fundamental as like a schema.
like your app won't function without permissions. I think it's so much easier to do a migration. The tricky part is if I develop a new seeder, like I have a new permission coming in or I have a change coming in, I have to create this seeder, like I have to figure out how to make that, put that seeder into the deployment pipeline to like when my app deploys, I have to make sure it runs this random seeder in there.
And perhaps you can get something into our runs your Seeders every time and does it to check like if this doesn't exist and create it if this is a great but that's just a nightmare. I just do migrations. Yeah, I just do migrations. And another thing is this tester easier you can do you can like set up the migration or you can set up the Seeders in your tests and do all that but I hate that because if you're doing something like user controller you test you
Dalton McCleery (1:04:08)
That's extra, extra maintainability.
Andy Hinkle (1:04:28)
You assert that the controller can create a user right, but if you have these Seeders are already creating Authenticated admins in your app is then it's like so if you assert that One user has been created you might have like 10 users and then you have this weird Quirk in your test code of like well we were doing 10 here because the Seeder creates nine users and it's just ridiculous So I'm just so tests are easier if you do migration
And if permissions change, you maybe want it tracked in the migration. You can do a down method. You know, you can't do down methods in Seeders. So I've done down methods before where if something's weird about this new permission, or if I change a permission from one to another, I need it to happen immediately. don't need to like once the migration fit, I need to ensure that the thing has changed. And if something goes wrong, can.
run a down commander, you know, if I needed to. So yeah, that's my cases for, I love migrations really like Seeders for local development. Like we have a very large app that obviously, you know, I never want to bring in production data. it just run Seeders in there, sets it up locally. Don't have to, but Seeders in production. know. So, we're in the page.
Dalton McCleery (1:05:33)
Bingo. Bingo.
Mm-hmm.
No, no, no, no. Yeah, I'm in a hundred percent agreeance. I think Seeders are great for local development. Like I'm coming into your site. I need to be able to get a user that doesn't have any billing and information. need a user that has billing information. I need someone who has outdated billing information. I need those three pieces so I can test if a bug comes up for one of those three. I need a Seeder for that. I'm gonna run PHP, Artisan, Migrate, not rollback, but refresh, dash dash seed.
Andy Hinkle (1:05:49)
Well, we agree. Yeah.
Dalton McCleery (1:06:14)
I want a fresh database, all the seeders run. I do not run seeders in production ever because yeah, I'll have to add them to the deployment script and then remove them or I'll wait for it to deploy and then manually run that command through Forge. But if it's stuff like permissions, there's like 15, 30 seconds between the deployment being done, the migrations being run, and then me manually running a seeder command. I forgot, know, something like that and it's broken.
Andy Hinkle (1:06:21)
Yeah.
I need to that.
Dalton McCleery (1:06:43)
So yeah, I'm in 100 % agreeance. After I saw that, I had the perfect use case for it like a week ago. And I added a migration that created a table. And then within that table, I was able to seed, you know, four or five default information or default values into that table. And boom, it just, it worked. All you had to do is PHP artists and migrate, done. You got the stuff, it's ready to go. Yeah, I'm at 100 % agreeance.
I think Seeders are great for local only.
Andy Hinkle (1:07:12)
All of the solid bros or separation of concerns, bros that, know, all it needs to be clearly, you know, they'll say differently, just man, when real world comes to some of this stuff, it's just when you, especially permissions, and that's my example, like when primarily use it, you just need things to happen.
as the application deploys, there's not at the moment, there's not like a streamlined process for Laravel of like, hey, on this next deployment, I want to make sure this artisan command runs, but just this one time, don't do it again. You know, like just run it this one instance of try to fix something. There's not anything like that. So it's, there might be a package or something that does it, but hey, migrations, just use those. It makes it easy. So, yeah.
Dalton McCleery (1:07:58)
They're there and listen, the solid bros, if you've got a problem, you can add us. It's Andy, whatever your Twitter handle is, Andy, Andy Hink, Andy H and K and MindStopMcClair, can at me. That's fine. I don't care. All I'm going to tell you to do is you can write your cedar and then just call the cedar in the migration file. Boom, we're both right. Gotcha.
Andy Hinkle (1:08:02)
Solid bros.
Yeah, HNK, yep.
Yep. How did that be? Great. I didn't think of that. could just call your cedar. Yeah. Well, technically if you're depends on how your application, if you're in migration down, it's going to fuss at you. But yeah, for sure. Yeah.
Dalton McCleery (1:08:27)
Call your cedar in the migration. It's just an extra.
That's just like an extra file, you know, I wouldn't do that. Why would I why would I? abstract that to another file and I just call it right in that migration file Whatever man Add us
Andy Hinkle (1:08:42)
Yeah, for sure. All right, man, it's been really good. Yeah, Atis, you want to close this out?
Dalton McCleery (1:08:49)
Yeah, yeah, so I think this will be the last episode we do this year, so Merry Christmas, Happy Holidays, Happy Kwanzaa, Hanukkah, all the other stuff, sorry if I forgot. We want to thank you for listening. you've made it this far, double extra thank you to you.
Andy Hinkle (1:08:54)
Merry Christmas!
You, I bless you, you are very technical. If you've made it this far, please reach out because you're obviously you made it, you stuck with us. You are a very technical person because I bet just five minutes in this show, like we jumped off the deep end. So appreciate that.
Dalton McCleery (1:09:19)
I just glazed over and like, whoa. Yeah, so if you made it this far, extra special, extra special thank you to you. We really appreciate you and we'll be back in 2025 for more. So stay tuned. Bye.
Andy Hinkle (1:09:22)
This is a very technical show, but that's what I like about it.