Mixins
, HOC
, render props,
and Hooks
are four ways to reuse components
Now frontend engineering is more and more important. Although Ctrl+C and Ctrl+V can also be used to complete requirements, once they are modified, it becomes a huge task. Therefore, copying of code is reduced, and the packaging and reuse capabilities are increased to achieve maintainability and reversibility. The code used becomes particularly important.
In React, components are the main unit of code reuse. The combination-based component reuse mechanism is quite elegant, but for more fine-grained logic (state logic, behavior logic, etc.), reuse is not so easy. It is difficult to disassemble the state logic as a reusable function or component. In fact, before the appearance of Hooks, there was a lack of a simple and direct way of component behavior extension, which is considered to be mixins, higher-order components (HOC), and render props. The upper-level model explored under the existing (component mechanism) game rules has not solved the problem of logic reuse between components from the root. This is my 38th Medium article.
Of course, React no longer recommends using mixins as a reuse solution for a long time, but it can still provide support for mixins through create-react-class
. Note that mixins are not supported when declaring components in ES6 classes.
Mixins allow multiple React components to share code. They are very similar to mixins in Python or traits in PHP. The emergence of the mixin solution comes from an OOP intuition. In the early days, it only provided React.createClass()
API to define components. (In React v15.5.0, it is officially abandoned and moved to create-react-class
). Naturally, (class) inheritance has become an intuitive attempt, and in JavaScript
prototype-based extension mode, it is similar to the inherited mixin
scheme. It has become a good solution. Mixin
is mainly used to solve the reuse problem of life cycle logic and state logic, and allows the component life cycle to be extended from the outside. This is especially important in Flux
and other modes, but many defects have also appeared in continuous practice:
- There is an implicit dependency between the component and the
mixin
(Mixin
often depends on the specific method of the component, but the dependency is not known when the component is defined). - There may be conflicts between multiple
mixin
(such as defining the samestate
field). Mixin
tends to add more states, which reduces the predictability of the application and leads to a sharp increase in complexity.- Implicit dependencies lead to opaque dependencies, and maintenance costs and understanding costs are rising rapidly.
- It is difficult to quickly understand the behavior of components, and it is necessary to fully understand all the extension behaviors that rely on
mixin
and their mutual influence. - The method and
state
field of the component itself is afraid to be easily deleted because it is difficult to determine whethermixin
depends on it. Mixin
is also difficult to maintain, becauseMixin
logic will eventually be flattened and merged together, and it is difficult to figure out the input and output of aMixin
.
There is no doubt that these problems are fatal, so Reactv0.13.0
abandoned Mixin
static crosscutting (similar to inherited reuse) and moved to HOC
higher-order components (similar to combined reuse).
Example
The example of the ancient version, a common scenario is: A component needs to be updated regularly. It is easy to do it with setInterval(), but it is very important to cancel the timer when it is not needed to save memory. React provides a lifecycle method to inform the component. The time of creation or destruction, the following Mixin, use setInterval() and ensure that the timer is cleaned up when the component is destroyed.
After Mixin
, HOC high-order components take on the heavy responsibility and become the recommended solution for logical reuse between components. High-order components reveal a high-order atmosphere from their names. In fact, this concept should be derived from high-order functions of JavaScript
. The high-order function is a function that accepts a function as input or output. It can be thought that currying is a higher-order function. The definition of higher-order components is also given in the React
document. Higher-order components receive components and return new components. function. The specific meaning is: High-order components can be seen as an implementation of React
decoration pattern. High-order components are a function, and the function accepts a component as a parameter and returns a new component. It will return an enhanced React
components. High-order components can make our code more reusable, logical and abstract, can hijack the render
method, and can also control props
and state
.
Comparing Mixin
and HOC
, Mixin
is a mixed-in mode. In actual use, Mixin
is still very powerful, allowing us to share the same method in multiple components, but it will also continue to add new methods and attributes to the components. The component itself can not only perceive but also need to do related processing (such as naming conflicts, state maintenance, etc.). Once the mixed modules increase, the entire component becomes difficult to maintain. Mixin
may introduce invisible attributes, such as in the Mixin
method used in the rendering component brings invisible property props
and states
to the component. Mixin
may depend on each other and is coupled with each other, which is not conducive to code maintenance. In addition, the methods in different Mixin
may conflict with each other. Previously React
officially recommended using Mixin
to solve problems related to cross-cutting concerns, but because using Mixin
may cause more trouble, the official recommendation is now to use HOC
. High-order component HOC
belong to the idea of functional programming
. The wrapped components will not be aware of the existence of high-order components, and the components returned by high-order components will have a functional enhancement effect on the original components. Based on this, React
officially recommends the use of high-order components.
Although HOC
does not have so many fatal problems, it also has some minor flaws:
- Scalability restriction:
HOC
cannot completely replaceMixin
. In some scenarios,Mixin
can butHOC
cannot. For example,PureRenderMixin
, becauseHOC
cannot access theState
of subcomponents from the outside, and at the same time filter out unnecessary updates throughshouldComponentUpdate
. Therefore,React
After supportingES6Class
,React.PureComponent
is provided to solve this problem. Ref
transfer problem:Ref
is cut off. The transfer problem ofRef
is quite annoying under the layers of packaging. The functionRef
can alleviate part of it (allowingHOC
to learn about node creation and destruction), so theReact.forwardRef API
API was introduced later.WrapperHell
:HOC
is flooded, andWrapperHell
appears (there is no problem that cannot be solved by one layer, if there is, then two layers). Multi-layer abstraction also increases complexity and cost of understanding. This is the most critical defect. InHOC
mode There is no good solution.
Example
Specifically, a high-order component is a function whose parameter is a component and the return value is a new component. A component converts props
into a UI
but a high-order component converts a component into another component. HOC
is very common in React
third-party libraries, such as Redux
’s connect
and Relay
’s createFragmentContainer
.
Attention should be paid here, do not try to modify the component prototype in the HOC
in any way, but should use the combination method to realize the function by packaging the component in the container component. Under normal circumstances, there are two ways to implement high-order components:
- Property agent
Props Proxy
. - Reverse inheritance
Inheritance Inversion
.
Property Agent
For example, we can add a stored id
attribute value to the incoming component. We can add a props
to this component through high-order components. Of course, we can also operate on the props
in the WrappedComponent
component in JSX
. Note that it is not to manipulate the incoming WrappedComponent
class, we should not directly modify the incoming component, but can operate on it in the process of combination.
We can also use high-order components to load the state of new components into the packaged components. For example, we can use high-order components to convert uncontrolled components into controlled components.
Or our purpose is to wrap it with other components to achieve the purpose of layout or style.
Reverse inheritance
Reverse inheritance means that the returned component inherits the previous component. In reverse inheritance, we can do a lot of operations, modify state
, props
and even flip the Element Tree
. There is an important point in the reverse inheritance that reverse inheritance cannot ensure that the complete sub-component tree is parsed. That means if the parsed element tree contains components (function
type or Class
type), the sub-components of the component can no longer be manipulated.
When we use reverse inheritance to implement high-order components, we can control rendering through rendering hijacking. Specifically, we can consciously control the rendering process of WrappedComponent
to control the results of rendering control. For example, we can decide whether to render components according to some parameters.
We can even hijack the life cycle of the original component by rewriting.
Since it is actually an inheritance relationship, we can read the props
and state
of the component. If necessary, we can even add, modify, and delete the props
and state
. Of course, the premise is that the risks caused by the modification need to be controlled by yourself. In some cases, we may need to pass in some parameters for the high-order attributes, then we can pass in the parameters in the form of currying, and cooperate with the high-order components to complete the operation similar to the closure of the component.
note
Don’t change the original components
Don’t try to modify the component prototype in HOC
, or change it in other ways.
Doing so will have some undesirable consequences. One is that the input component can no longer be used as before the HOC
enhancement. What is more serious is that if you use another HOC
that also modifies componentDidUpdate
to enhance it, the previous HOC
will be invalid, and this HOC
cannot be applied to functional components that have no life cycle.
Modifying the HOC
of the incoming component is a bad abstraction, and the caller must know how they are implemented to avoid conflicts with other HOC
. HOC
should not modify the incoming components, but should use a combination of components to achieve functions by packaging the components in container components.
Filter props
HOC
adds features to components and should not significantly change the convention itself. The components returned by HOC
should maintain similar interfaces with the original components. HOC
should transparently transmit props
that have nothing to do with itself, and most HOC
should include a render
method similar to the following.
Maximum composability
Not all HOCs
are the same. Sometimes it only accepts one parameter, which is the packaged component.
const NavbarWithRouter = withRouter(Navbar);
HOC
can usually receive multiple parameters. For example, in Relay
, HOC additionally receives a configuration object to specify the data dependency of the component.
const CommentWithRelay = Relay.createContainer(Comment, config);
The most common HOC signatures are as follows, connect is a higher-order function that returns higher-order components.
This form may seem confusing or unnecessary, but it has a useful property, like the single-parameter HOC
returned by the connect
function has the signature Component => Component
, and functions with the same output type and input type can be easily combined. The same attributes also allow connect
and other HOCs
to assume the role of decorator. In addition, many third-party libraries provide compose tool functions, including lodash
, Redux
, and Ramda
.
Don’t use HOC in the render method
React
’s diff
algorithm uses the component identifier to determine whether it should update the existing subtree or discard it and mount the new subtree. If the component returned from the render
is the same as the component in the previous render ===
, React
passes The subtree is distinguished from the new subtree to recursively update the subtree, and if they are not equal, the previous subtree is completely unloaded.
Usually, you don’t need to consider this when using it, but it is very important for HOC
, because it means that you should not apply HOC
to a component in the render
method of the component.
This is not just a performance issue. Re-mounting the component will cause the state of the component and all its subcomponents to be lost. If the HOC
is created outside the component, the component will only be created once. So every time you render
it will be the same component. Generally speaking, this is consistent with your expected performance. In rare cases, you need to call HOC
dynamically, you can call it in the component’s lifecycle method or its constructor.
Be sure to copy static methods
Sometimes it is useful to define static methods on React
components. For example, the Relay
container exposes a static method getFragment
to facilitate the composition of GraphQL
fragments. But when you apply HOC
to a component, the original component will be packaged with a container component, which means that the new component does not have any static methods of the original component.
To solve this problem, you can copy these methods to the container component before returning.
But to do this, you need to know which methods should be copied. You can use hoist-non-react-statics
to automatically copy all non-React
static methods.
In addition to exporting components, another feasible solution is to additionally export this static method.
Refs will not be passed
Although the convention of high-level components is to pass all props
to the packaged component, this does not apply to refs
, because ref
is not actually a prop
, just like a key
, it is specifically handled by React
. If the ref
is added to the return component of the HOC
, the ref
reference points to the container component, not the packaged component. This problem can be explicitly forwarded to the internal component through the React.forwardRefAPI
refs
.