Scroll down to the script below, click on any sentence (including terminal blocks!) to jump to that spot in the video!
SymfonyCon 2018 Presentation by Tobias Nyholm
Web development is not just about delivering a response. It is also about writing good code. The state pattern will help you move complexity from being all over your code to one or more state machines. This talk will introduce state machines, show how to identify uses of them and implement them in your Symfony application in an object oriented manner using the Symfony Workflow component.
So excellent, excellent. Thank you all for being here. Um, do you know this is a large conference and you know, when you're organizing a large conference like this, you do want to have the best speakers, right? So, and what do you usually do when you're having a conference? You put the best speakers upfront, like the keynote is obviously the most impressive one and then you have the best speakers and then they slowly decline. Um, as you know, Michelle is not here, Michelle was about to do this talk so and her kid got sick and you can't really leave a two year old home in Switzerland by itself. So she had to stay home. So the conference organizers thought, Hey, what should we do? Michelle is sick. So obviously they did the best thing they could. They call the second best speaker, right? Unfortunately that person couldn't make it. So they called the most knowledgeable speaker who knows the Workflow component the best, and that person didn't make it either. Um, however, like five or six calls down the line, they find my name and here I am and I hope I can give you. Yeah, thank you. At least one guy or girl is very polite. I hope I could give you like a, a talk that could match Michelle's talk.
So with uh, with that said, my talk is titled Knowing Your State Machines and I would like to do this talk because I believe web developers only care about printing out HTML as quick as possible. They don't care about fancy algorithms, they don't care about design patterns. They just want to print stuff really, really quick. And so I want to tell you about something that makes your code a little bit more elegant. So I didn't, I didn't bring any state machines. It's nothing I can have in my pocket. It's a state machine is just something that hold states, like any states. It could be the state of a pull request or this small washer on your - small lights on your washer. It could be, yeah, it could be anything. So it's basically a way of thinking, a way to relate to problems. And why state machines? Blah Blah. There's a lot of texts in this. If you work at a company where your manager needs convincing, this is a slide - just to take a photo or show him later. I'm sorry, show him or her later.
Quick about me. I do plenty of work on Happyr. We are a job advert platform - it matches people with confidence, blah blah. I do plenty of Symfony things. I run PHP meetups in Stockholm and I do plenty of open source. I started my real open source career in 2015 on a SymfonyCon, just like this one. The PSR-6 cache had just came out. So me and my friends were like, hey, why don't we implement this. Why don't we make an implementation of PSR-6. So we did. We created the PHP-cache organization and CacheBundle and I thought open source is great. So I started to get involved with HTTPPlug, which is kind of an abstraction over HTTP clients and when you do HTTP clients, you do API clients and then you do mail clients. And I even wrote like, Oh, this is how you should do API clients, this is the boilerplate for FIG. And currently I'm maintaining both Guzzle and Buzz. I've written my PSR-7, I've written PSR standards and stuff. I, I do plenty of things. The most interesting library here, which I like the most is this one, the NSA one. With NSA, you can totally violate class privacy: accessing private properties or methods is super easy with NSA, I do also contribute to a bunch of other libraries: these are only the most fun ones.
So how should we do this about state machine? How should we do this? I would like to start with a lot of theory and then l will put on some examples, examples, and then more examples. How does that sound? Excellent. Because I've got slides for this. How I can barely see you, but how many of you, how many of you feel like you know, state machines by heart? Excellent. It's like 12 of you, which, which is excellent. Then I guess like the rest of you will learn plenty. You are.. anyhow, let's get started.
Let's get: first rule of state machines. We've got to talk about the graph theory and the first rule of graph theory is like these, these are not graphs. These are charts, these are diagrams, these are call them whatever. They are not graphs. A graph. It looks like this. A graph has a node. A more interesting graph got two nodes, and nodes can be connected with edges and edges can either be undirected like this one or it can be directed. An edge can also have a weight. Does this seem familiar for most of you? Excellent. Excellent.
This is a full-fledged graph. This is like the cool graphs can be. And it can represent pretty much whatever thing you like it to represent. If you slap on some numbers on here and maybe you can have a problem, maybe this represents cities and roads between them. Maybe it represents computers in the network, maybe, maybe they represent people and how well they know each other. And say we've got a graph like this and you wonder how much.... say they are computers in the network. How much data can you transfer from a to b? This might be a problem you're having, so this is called the maximum flow problem and there is a solution to this. And say that you want to find the shorter distance between a and b. You can solve this this by Birpartite breadth-first search. And say that you want to... Say that you want to find out what's the cheapest way of visiting all these cities, or these nodes, visiting only every node once. That's a traveling salesman problem and there is a solution to that as well.
Or say that these are cities and you should put out some electricity grid between them. This is the minimum spanning tree problem and there's a solution for that as well. What I want to tell you is if you've got a problem, if you can express this problem as a graph problem, there's already a solution to it. So you don't, you don't really have to solve these problems yourself, you can just solve the problem getting your problem to a graph problem. Does it make sense? Thank you. This, what I just did, I gave you like six months of university math of graphs. So this is basically the big graph theory I just gave you a quick introduction of.
A cool thing about graphs is that you can reduce them. Say, consider this graph. I want to reduce them by giving them a new rule. And say, the rule might be: let's remove all the loops. If I remove all the loops in this graph, I get a tree. And there's plenty of algorithms you can do in trees as well. One of the most difficult to implement is the red then black search tree, which basically is good way to store things if you want to search quickly in it. Anyhow, I want to show you this. You have the big graph theory and inside of it you have the theory of a tree. And you can continue reducing graphs with different set of rules and you will get a workflow.
A workflow also has nodes, but sorry, but we decided, oh also has nodes and it also has edges and we decided that every edge should be directed. We also decided that we call the nodes places and transitions. And a rule in this workflow is them saying: a place can never be connected to another place. All of this has to be a transition between them. I know I'm shooting a lot of rules at you now, but do you kind of feel comfortable? Excellent. And we decided, I mean just so it's easy to see, we draw these transitions like a square and a good example of a workflow is this: if I'm in place A and execute the transition t0, I'll get to place C. If I'm in place B and they execute the transition t1, I get to place C and D. Make sense? Good. If we continue to reduce this workflow, say we want to add rules saying that a transition could only have one input and one output and also the names has to be unique. Then we actually get a state machine that looks something like this. So workflow multiple one input, multiple outputs. A state machine only can have one input and one output, something like this. So if execute t1, if you are in B and execute t1 you get to D. And because I'm drawing these images by hand, I will choose to draw with this annotation. Fair?
So right now I have been talking for eight minutes and we discussed the big graph theory and we discussed the tree and we discussed what a workflow is and that a state machine is just a reduced workflow.
So let's start with some examples. I mean I know this part was the boring part, this part was the theory part. Let's start with some examples of recognize that we recognize. This is a simple state machine. We defined three states and we defined how we can move between the states, right? So it's easy to see that if you are in green we can easily move to yellow and then we can easily move to read. However though if we are in green, we cannot move to red. There's no arrow to red. It's impossible to go direct to red, which is which is a good thing. I've got to keep asking you simple questions until you give me some continuous feedback. Do you understand this?
I work at a company called Happyr and we deal with a lot of job adverts. Basically we put up adverts and we match people and people in common with soft skills, etc, etc. Anyhow, we have a lot of job adverts. So in a business meeting we decided like this, this is the state machine for a job advert. So we started draft and then we can send it to review. While in review, we can trash it, untrash it, and we can send it to review again if you'd like. And from review you can publish and archive it. It is impossible to archive something that is in review or that is something that's in trash. It's impossible. Excellent.
Since you are very confident now, what is this a workflow of? Someone said a pull request issue. Uh, I will say yes. Correct. So basically you start coding your pull request, you submit it to Travis, Travis runs for awhile. You may want to decide when you're waiting for review and then you may want to decide to update it. So we'll go back to Travis running. And maybe someone says, hey, we need some changes. So you go back to coding, then you update it, then send it to review and hopefully it gets merged or else it could be rejected and re-opened again. Seem straightforward?
So how about this? Oh yeah, pull request. So how about this? What is it the workflow of? What is the state machine of? People usually yell out like, like you, sorry, like you did. You said doors. I'm like no, no, not doors, it's elevator doors. But you're equally correct. It's doors. So basically this is the four states that a set of doors can be in. Either the doors can be opened and then you press the close button and then be closing. While the closing we can press open button so they go to opening. It's impossible for doors to be open and then in one, the next instance be closed. They have to be closing, right?
I think. I think we kind of know, know state machines now. Maybe business a dictates that... Maybe business a business dictates that you should have a form like this, like maybe a sign up form or whatever. So you can have a state machine to know where the user is in the form. So if you leave this process, you can, you can get back to where where he or she left off. And if your extra nice you can have back buttons so we can travel back and forth in this process. If you want to implement this, there's basically two implementations. I'm gonna can talk to them with them bot. No, I'm not going to talk with them. I'm gonna talk with you about the both of them - the implementations. The first implementation is the state pattern and that's using the Moore state machine. And a Moore state machine only cares about the current state. The next state only cares about, the next state is defined by the current state. The other state machine is the Mealy state machine and the next state depends on the current state and some input.
I just want to say this out loud now and then I'm gonna repeat it later. Let's start with the state pattern and the Moore state machine. This is basically what you easily read on wikipedia. You... a state. I'll click this. Say you want to implement, something like this. We want to implement a... something that reminds you to complete your profile, say a user signs up and you want to say like "Hey, you should add your name" and we remind them twice about adding their name. If in order to add your name, like we should, hey you should update your image or you should add your resume. So implementing this with a state pattern means that each of these five states have a class and that class itself only have one f unction. It has a Send function. And, and that calls itself decides what to do and where to go next. Makes sense? So the idea here is that we have, have, we have a worker like a cron job saying like: Hey, give me all the state machine from database somehow and start yourself. I don't really care what you're doing - you just send the email, send the correct emails.
So in the implementing this, I'm basically looping over all the states and if the, if the state machine says "I'm done", then we're done. So any given state just looks something like this. I, this is, I mean, you send an email and then I say go to next state. Makes sense? We should obviously do a little, we should modify this a little bit. We can maybe say like, if the user doesn't have a name, if the user do have a name, go to add your image state. So the state itself says where to go next. The sates itself decides everything what to do. And the state is a single class. We should also be a little bit more friendly. This would actually flood the user's email address, so uh, email inbox. So we should: you should be in this state for at least seven days until we send another email.
I know this might feel weird. I know this might feel like that's not really PHP. It feels more like Java or something else. It feels weird and I will agree with you. This feels a bit weird. Using this and, for the simple state machine I showed you the five states, you have a bunch of classes like this. And they all pretty much look the same. Each of this state just look, oh, they look like this.
So I... with this I would like you to get an idea how the state pattern works and some of you guys are nodding politely. Excellent. You obviously have issues with this. I mean you have issues: how do you store the state machines in this database? How do you keep track of the state? How do you store the createdAt entity... or data. Also it feels kind of weird that you have one class with state.
So let's, let's do something different. Like now we know the Moore state machine, we know the state pattern, so let's look at something different. Let's look at something that feels more like PHP. And I'm going to stress this multiple times: the Mealy state machine, the next state depends on the current state and some input. So it's obviously a little bit more flexible. Let's try. Let's consider this state machine again, you all know this by heart now. If you would write an implementation of this, if we like PHP developers would write an implementation of this, how would we do? Well, what's the first? We obviously need a class, right? We need a state machine - traffic light state machine class. And we kind of want to have a function that applies new state. And I imagine it to be good to have like a function that "can I apply this state?" Good. So let's. So I create my state machine looking something like this. I define all my states as constants and I also choose to have a private property called state. And then I just, via a switch statement say: if someone say: "Can I do transition to yellow?" I check if I'm currently in state green and someone says "can (I go) to yellow?" Then sure, that's allowed. If I'm in state red and someone says "to green", then it's obviously no, I return false. And I do a similar, very similar thing for apply. I just do something like this.
So would this state machine. I can easily, I can easily implement my... This, this class is easily implement my traffic light state machine, right? There's an issue though: this state machine has a private state. So I need to store the state machine in the database somehow. This state machine is only for one traffic light. If I have multiple traffic lights in my system, I have to have multiple state machines which each needs to be stored in the database, because of the state property. What if I made my state machines more stateless? Like I removed the private state and put that on the, on the traffic light entity? So I did something like this. Can this specific traffic light do this transition? So now I don't need to store the state machine in the database anymore. I can use this state machine for all the traffic lights.
It feels quite, feels quite good, right? Yes. It feels quite good. I diddothe same thing with apply function. I just say
$trafficLight->getCurrentState() and I have a
$trafficLight->setState(). Very similar, very small changes. And I feel pretty good about the state machine at this point and whenever I'm using this, it might be an idea of mine to have a function that allows more traffic. I mean if, if, if the state is currently in the red state and I want to allow more traffic, I just, whatever state they're currently in, I just move it towards green. So I might attempt to do something like this. So I know if I call allow more traffic twice. I will definitely be in green.
You look confused and you should. Because this was like my trickery, but I didn't trick you. If I want to make this state machine more generic, more reusable, obviously this is a horrible idea. Right? So if I'm going to make it more reusable, I may want to consider instead of having the states all over the place, I may want to put them in a, uh, in a property like this. So with this weird array looking thing, it's actually me trying to be a bit clever again. So I can just say: can I do the transition? I was checking if this is, if this property is set, if this array key exists. I do the same kind of wizardry in apply. Do you feel confident with this? You kind of understand it at least that's, that's the most important thing. However, say you want to make this even more generic. I mean, I shouldn't, I mean this is only for traffic lights. I said, traffic light here. What if I do something different? What if I... Oh, sorry. Oh, look, the states, it should be states here. What if I inject the states instead in a constructor like this?
Then I have those states - I can do this for any kind of state machine and any kind of traffic light. And now, oh no, I have traffic lights over here. What if I did something like this to have the
StateAwareInterface instead? And this
StateAwareInterface just have a
setState() method. With this class of 15 lines. This is actually a pretty good, pretty generic state machine, right? And it all feels (like) PHP, all in one class and I can configure it however like and use an object that I like. I hope we feel like we understand this. Does anyone have... oh no, I can't ask you for issues. Does everyone feel like you understand this class? Excellent. Excellent.
I'll tell you a little bit of a secret. This is actually how the Workflow component works. This is the workflow of the Workflow component. And it was very true like two or three years ago, this was exactly the Workflow component. And nowadays there's a lot of edge cases, but in principle it works exactly like this. The Workflow component all have some more functions we showed. I showed an implementation of "can we do the transition?" And "apply transition". The workflow component has two more methods saying "like what state are we currently in?" And what are my valid transitions?
So let's show some examples. Same old job advert workflow again. In the Workflow component, this might look like something like this. The red circle is the current state, so I can say: give me the state machine for job adverts. And I say give me the current marking for this entity and give me the places. And marking here sounds weird. It's because it's basically the abstraction of getting properties from the, from the Advert. If you ignore that, it's, this is very simple, right? So I get an array back and I see I'm in Draft. I can continue showing examples saying: state machine, can I do publish? No, I can't, but can I do "to review"? Yes, that's true I can. So I continue. I go to review and when I execute this line, it actually moves the circle away and from here I can say go to trash, go to draft and go to review again. And in this state I can say which are my enabled translations, and I get an array back to say I can go to publish and trash.
I just, you just understand everything in the Workflow component. The Workflow component itself got like 15 classes and this is the most complex one. And you understand it fully now, which I think is quite... not impressive that you understand it, but it's impressive how simple Symfony components are. Uh, this is also, since this is Symfony, this is also available in Twig. Imagine, imagine when you're editing an advert on Happyr, imagine it looks exactly like, like Wordpress. You have this big text box in the middle and then you have a sidebar with like trash, publish and stuff button like this. This is the code for the sidebar. So I'm checking if the workflow "can" go to review, then show the link for review. If the, the, the, this advert can be published, then show the link for published. So the same sidebar will look differently on different adverts. And if you are, if you're smart with the naming of your work for transitions and you're naming of the routes, you can do something like this to print the sidebar. I want to show this because it's, it's possible, but you probably shouldn't do this. I do want to say that... I mean, I think this is pretty cool, right? You, sorry.
I need to reverse, blah blah, this is a sidebar, decisions... This is pretty cool, right? Your sidebar really, it changes depending on the states. And what's cool here though is on a business meeting, we sat down and someone asked, how come you can't trash something that's in draft? It's obviously no arrow here, but what's the reason? And I'm like, yeah, there is no real reason. So let's, add this arrow. And here's, here's the kicker. I only changed the whiteboard. I want to change how the, how the state machine looks like. I did not change any of my models. I did not change any of my controller. I did not change any of my services, I did not change Twig layer, I didn't change anything but my configuration. And still with this change, I will see the trash in the sidebar when the advert is in draft. And that's pretty cool. A year ago, I did this presentation and when I said that's pretty cool. People like, yeah, that's cool. Good job, and they were kind of cheering. But you become the words, not only to be excited with me.
So I want to change the configuration And the configuration looked like this. It's super simple, This is my, this is my core of my business. This is my business logic. If you're really, really paying attention, you might spot a weird thing here. I told you that a workflow can have, no sorry, I told you the the state machine only can have one input and one output. And clearly I have two inputs here. This Symfony configuration is smart enough to know like, oh, I can make two transition of this. So it, it says, it looks like only one, but it's actually gonna to become two transitions.
And also quickly want to show you a multistep signup form. Business may dictate that this is the way we sign up to our service. So have an intro, you add your name, you add your email, you add your twitter, and then you're done.
And doing this workflow in the configuration is simple, you just do it like this. You just say that this is a state machine, this is my places, and this is how you move between places. And business might say this is good and all, but we want to have back buttons. We want to have a workflow like this where you go back. And we said, sure, no worries. We just do something like this. However, this is not valid YAML, right? You cannot have multiple keys on the same level which is called
back. So what we need to do, we need to do something like this instead. We add arbitrary numbers or arbitrary names and then we have the name. So this how we get around the issue with YAML. So with this you can actually please the business to have back buttons.
And I know I've been talking a lot about state machines now and I haven't really... This is called the Workflow component. And it says workflow like all over the place here, right? So what is the difference between a state machine and workflow? In the beginning I mentioned the state machine is a reduced workflow, but what is the actual, what is the actual difference? As an example, I want to show you this: this an example of a workflow. Previously we've only seen state machines. This is the example of a workflow. And the main difference in a workflow and a state machine is that in a workflow, you may be in two places at once. It might feel weird, but it's actually valid. A workflow more or less describes a process. A state machine describes traffic lights, when you actually go back to the beginning and stuff - how you move. So this is a workflow for a blog post: it's first a draft, then you request reviews and then you get in, you get in the places for waiting for spellchecker and waiting for the journalist at the same time. So there's an example where a transition that makes you to be in two places at once. And spellchecker is probably automatic, right? So that's pretty quick, but you have to wait for the journalist. And it might be tempting to try to publish this blog post right now. But you can't execute the publishe transition because both its inputs are not fulfilled: we don't have one of the inputs, right? So we have to wait for the journalists to catch up and then you can eventually publish it.
I will also say a workflow rarely contains loops. It's rare that you go back in your workflow. However, state machines, loops are encouraged... kind of. People used to ask me: when should I use workflow when I use state machine? And the answer is boring. It's obviously: it depends. I tend to use workflow when there's a clear process. I use state machine for everything else. It's just how I like it. The most important thing is that you know which one you're using. You need to know how to relate to them. It's like, for 99 percent of the problems, it doesn't really matter which one you're using. But you should know which one you're using. Does it make sense?
Excellent. Excellent. So back to the Workflow component. If you have been around in the Symfony community for awhile, you might notice Symfony have like 30 components and the components are very decoupled from each other, but they play really, really well with each other. And the secret there is events. Events is the thing that makes them play well with each other. And the workflow component is no exception. We have events or, the Workflow component has events for everything. We, all the events looks like this: you have a generic event, say
workflow.leave. Then you have an event, say workflow - the job advert - when they leave. And you have events for, workflow, job_advert, leave, draft node.
And all these, you have all these three types of events for everything. So we have workflow leave, when you're leaving a node. You have transition that would actually transition from one node to another. And when you enter the node, when you have been entered the node, and when a transition is completed. And then you have announce saying: Hey, I'm all done. It might feel weird that we have so many events. I mean these add up to like 21 events for each transition, which is a lot of events. However, there, there's a reason for all of them. Like, I know I've reviewed all the pull requests adding more events and they all makes sense in this specific weird edge case. You should only care about the one you care about. I mean, you shouldn't... I only use like leave and enter... and that's fine.
So there's plenty of events. There's also this even more events. There's one event called the Guard event and the Guard event is something that you use for, adding extra logic, like external dependencies. So: can I publish this blog post? Like say we want to, we want to have a rule saying: you can only publish this blog post if it's between3 and 4 PM. Or if the database is online or if the manager is in the office or if you have the role
ROLE_PUBLISHER. So a Guard, a Guard event is how you add this extra knowledge to your state machine. And I think you all are comfortable with events subscribers. I basically say at the bottom, I say: this is the event I'm listening to -
workflow.job_advert.guard.publish - like the "got published" transition. Whenever that event is dispatch called the
guardPublish() function and run this three lines in the middle. I would, I will assume that you are comfortable with this.
Except for the Guard event, we may, we can register normal event listeners. Say you want to log every time you move from one node to another. So I say, the workflow, when the job advert leaves any node -
workflow.job_advert.leave - then I just to do a logging. I say advert with ID performed transition from one node to another. It might be useful to have this kind of log message.
I'm checking - you're still with me? It looks like most of you are with me. I... no surprise: I love the Workflow Component. I use the Workflow component all over the place. If I have... the Symfony Security components have the concept of voters, which are basically: if someone wants to watch, view an advert, I got to make sure that advert is published. And I'm using the Workflow component to make sure I am in state
published. So here's a simple security voter and I say: if someone tries to view then I check this: if the state machine got the marking for the advert and marking is published, like if I'm in published state, then you're allowed to view it. You should not be allowed to view something that's in draft. So I used my state machine all over the place.
I also use the state machine for domain events. I know for sure that no advert in my application will ever be published unless the
workflow.job_advert.enter.published event is dispatched. I can trust that for sure. That's why I use this event for my domain knowledge. Like say business say: you should email all your users whenever an advert is published. And this is a good, is a good place to hook into this. So I never dispatch any events on my own, I just use the workflow events. Obviously we don't email all our users when adverts are published, I just wanted to say, but it's, an example how you use your domain logic with these events?
Um, whenever, whenever I'm in business meetings and we, the business people say we need this and we need that and we gotta figure out a new way to do stuff. We write the workflow on the whiteboard. And we change our minds, we go back and we add new states. We, because business people, they tend to like graphs and images, they tend to like, big circles and squares and arrows between them. So it's easy to work with, work with a whiteboard. However, when the whiteboard meeting is done, I want to make sure, I go back to coding and I got to make sure that I have configured this properly. So I add my configuration and then I run this command, the dump command and I'd get an image. I mean the dump command just prints out some text and I use this program called dots or this script called dot which converts, takes to an image, and I can get this image. So then I compare the whiteboard with my image and more often than not I made a mistake in my configuration.
So this is a good way to make sure you configure this properly. And so I usually dump two or three times and then I, then I know I got the right conflagration.
And I know I talked a lot, and state machines they sure are great. I gives us, I've told you so many benefits, but how do... I mean where... so I want to give you some hints. I want to give you some: if you ever have, if you ever have code like this, this is code, I've written, I wrote a cCRMm or whatever it's called, basically for salespeople and they have a quotation and I defined the Quotation can be in these three states, four states. I even have a property called status. And whatever you see code like this in your application, you should know like, hey, maybe I should use a state machine for this.
Also, when you're looking at your database and you see something like this. Like have boolean for active, a boolean for ended, a boolean for closed. This is also a good sign, like "hey maybe a state machine will be better in this scenario" And people asking me: "Oh, Tobias, should you really have state machines if you only have three or four states?" And the answer is obviously it depends. Are you sure you never ever would add more states? How are the rules between those states? How do you move? And I would say, if you want to make sure you don't have any bugs. Yes, a state machine might be a good idea. Obviously if you're only have two states, I mean it all depends and I understand, I'm sure you're getting the gist.
So how about processes? If you don't have a quotation, but a process. What if you're doing e-commerce, it's obviously a process here. You obviously create an order somehow and items are being added to that order and then you need the user to fill in the delivery address and then you need them to add the payment method and then you need to call or somehow notify the delivery team to make sure that that order is delivered. And this is also like if you have this strict workflow, you should think of the Workflow component. You should think: "Hey, maybe, maybe I should use a workflow here." And then you start drawing on the whiteboard: you draw the order is created, and then somehow we add an item
So we need to add an extra arrow here. The user can add more than one item so it can buy more stuff from you, right? And this, it's pretty simple: you just go through the workflow and the item added, sure, add another item, add more items, add shipping, and then order confirmed. And the cool thing here is: this will, each of these steps will dispatch an event. And on this event I have an event listener because this is my domain logic. I can trust that this event listener is executed each time an order is confirmed. And I just asked the delivery service to take care of this order somehow. Makes Sense? So using the state machine, I also know that no order will ever be completed or confirmed unless I added the shipping address. It's literally impossible to complete an order without adding a shipping address. So what I basically want to say is, it all comes down to this: Using the Symfony Workflow component will make your write fewer bugs. And I hope this short presentation of mine have given you a new way of thinking, a new way to relate to your problems.
Thank very much. Thank you.
I will share my slides on joind.in. And so if you, if you're interested in anything, you please feel free. If you have any questions, I think we have a few minutes and I will throw this cube at you. Anyone want to have a cube? There's a question over there, please help me. Two rows back. Three rows back. Did you activate the cube? It was a mistake. Oh, okay, mistake question. Someone else - just throw it to someone waving over there. Whoever has the laser pointer in their eye.
Sorry, can you hear me? Okay, great. So a question is really simple: where is e-commerce here?
Where's the camera? I don't know where the camera is - oh there's the camera.
I'm saying this was callex using the workflow component for E-commerce.
Ah, where is the e-commerce? Yeah. I assume that you missed my five minutes beginning. Michelle sadly got sick. So I did a stand-in talk about Workflow component. So yes, I missed a lot of e-commerce stuff and we can only blame Michelle's three year old kid.
There's one more question and then I'm gonna let you all go.
You hear me?
Yes I do.
Uh, so my question is how to properly connect workflows with the security. So you have two ways: you can write a guard listener, uh, to prevent some transitions based on security. Or you can, as you showed, you can do a voter and there decide based on workflow whether you can do something or not. So what's the correct way?
It depends what you're trying to do. I show the voters, which, I will try to explain, I used for handling my app security. I used the state machine as a substitute, as a help. If you're going to add security, saying like you want to make sure... if you want to make sure that you have the correct role for doing such as this action, use the Guard event. Okay. So you do something like this.
So both ways are correct?
Both are correct depending on what you're actually trying to solve.
Okay. Thank you.
I will be around for the rest of the conference and I also told you the best speakers in the front. I'm doing a second talk at the very, very end of the conference for some reason. So if you're interested in Symfony 4 internals, please come and visit me then. Please review, rate all the speakers on joind.in
I think I found the related repo on GitHub, see https://github.com/Nyholm/w...
Let us know if it helps you ;)
You're welcome! Good question btw! Might be useful for others.
"Houston: no signs of life"
Start the conversation!
Is there any repository on github to see real examples?