::info Update June 2023: added a section outlining what my current approach is. :::
What is state? #
- UI: state that is used by a single, or a set of UI components. It is used to control what we can see, how we interact (e.g. input fields) on a detail level. UI state can exist on a global level as well (e.g. dark-mode).
- Remote: state from the server that is cached on the client for quick and easy access for all UI components. The remote state should not deviate from the server, except when applying optimistic UI.
- URL: information stored in the URL, like object IDs or filter information, that can be used to determine what to render, or what information to retrieve from the cache/server on (initial) rendering of the page.
- Meta: also known as 'state about state'. A common example is a loading state around fetch requests.
State management #
So what about state management? For many, state management feels like a black box. What is happening within Redux? Why does it feel so complicated? I look at it this way: State management is nothing more than patterns to make it "easier" for us to change state.
State management is nothing more than patterns to make it "easier" for us to change state
In modern front-end there are several different patterns that are popular to use these days.
- Flux: the popular pattern behind the "Redux" package. It centralises your state and allows you to define "actions" that mutate state. The UI has the ability to dispatch these actions onto the store.
- Proxy: a layer that acts as an intermediate on top of your state. It allows you to predefine possible mutations and add observability to give your state an reactive nature.
- Signal: signal is a value that changes over time and whose change events can trigger side effects. It is reactive by nature.
- State machines: restricts the possible states something can be in, and limits the possibilities to move from one state to another (for more info check here).
What to use? #
Different types of state require different solutions. There is no single state management solution that works best for all. The framework or library that you are using limits or helps you as well with this decision. React offers a big ecosystem and a lot of choices. Like a lot. But Svelte on the other hand has a tremendous implementation right out of the gate.
My own approach is a mixture of different patterns for different purposes. For remote state I would use service workers or create/use a library around 'state-while-revalidate'. This s a great example where state machines In the background these are powered by state machines to avoid unwanted caching side-effects.
For things like application state or meta state, it highly depends on the complexity of the challenge. If it is something simple, a small reactive store like a signal or a flat proxy suffices. However, it the challenge becomes more complex I will choice one of two routes:
- If there are many actual 'states' I would model everything in a reactive state machine.
- If the goal is data manipulation in many ways, I would go for a reactive proxy that allows me to define a state API. These are functions defining how I can interact with the state (e.g. the chifferobe package).
The common theme in my approach is reactivity. Even an "SWR-like" library is reactive. The thing is, in front-end development we always want something to change when our state changes.
Wrapping up #
Now you should have a better understanding of some fundamentals of state management. Knowing the different types of state and how to manage state is the start. With proper state management, you can get a long way in complex web applications. But it is the start. There are many (more) ways to manage data that are important in client-side applications. When you master state, go dive into persistent storage or caching.