I am trying to use the final-form calculator to clear a field whenever another field has changed.
In my example, I have two fields. Whenever the first field changes, the second field is cleared. This is expected.
However, A problem arises when the parent component of the form is re-rendered. Each time the parent component calls to it's render function, the second field is cleared even though the first field has not changed. This can be observed by clicking the forceUpdate button at the top.
Is it possible to prevent the second field from clearing like this? Preferably without using shouldComponentUpdate
I have been able to resolve it by moving the decorators array outside of the component.
const decorators = [calculator]; // declared outside of App
And reference the value in the form props
<Form
decorators={decorators}
...
I'm new to React, sorry if my questions are trivial.
In React, if I write:
<input value="" />
then it will be an read-only field, and React forces me to add an onChange handler as:
<input value={this.state.greeting} onChange={this.handleChange} />
where handleChange() update component's state.greeting, therefore value gets updated too
Below are my questions:
Q1-why <input value="" /> is not read-only in plain html but read-only in React? How does React make it read-only
Q2-I found that if I write the code as below it also works, the input is not read-only
<input onChange={this.handleChange} />
and isn't that this approach better and more concise? because the internal value will get updated automatically in the browser, therefore we don't need to include an value attribute to read the data back from the state, which is unnecessary in most of times, so why I always see code like:
<input value={this.state.greeting} onChange={this.handleChange} />
Additional info:
some says it is controlled form elements that needs to have value attribute, but why we need to have a value attribute to read from the state? and when we type sth into the input, the onChange handler already updates the state, so it is still controlled.
To know how React makes your tags readonly, you will need to study the source-code that runs at your end and/or view the generated HTML. If still unsure about it, then you might want to send your first question to the authors of the tool.
The state is not on the server, unless you are polling or doing something of the like. It's in your browser as well. The value property specifies the initial value of your HTML element, that is, before you do anything your tag will have a value. In your case, your tag is controlled by React, but you need to initialize it. Benefits:
you will have the initial value
you will have a more readable code
your code will be written in the React-way, so you will not need to worry of unpleasant surprises
I have some checkboxes that I'm just using as filters for data that need to run a function on click. Usually, and all tutorials ive read, have suggested to handle checkbox checked attribute using state. Is it possible to do this without using state? I don't really want to declare all the different checkboxes in my state just to toggle them on and off. I was hoping I could do something like this
<input type="checkbox" checked={true} onChange={(e) => e.target.checked = false } />
Sure you can, using uncontrolled components, then using refs to get the values. It's considered "quick and dirty", but you can do it no problem. See this article for more detailed information: https://reactjs.org/docs/uncontrolled-components.html.
Sorry for the long explanation, the question is at the end.
There is an example given on the reactjs site ( https://reactjs.org/docs/forms.html ) that looks like this:
handleChange(event) {
this.setState({value: event.target.value.toUpperCase()});
}
These tutorials usually say that in controlled forms (say in an input field) when you press a key, then onChange is called, that calls the handleChange event, that calls setState, and that re-renders the component, that will show the changed value. But this is not the whole truth. By the time the native DOM input field calls onChange, the input field was already "rendered" by the browser, and React does not re-render it. Actually it compares the state of the (re-rendered) virtual DOM element with the state of the native DOM element, and then it finds out that there is no difference, so it does nothing. In this case, the action has no side effect (other than changing the component's state).
I know that for a beginner, in the beginning of the tutorial, it may be confusing to tell the whole story at once. But this "toUpperCase()" example in the tutorial is misleading, because it suggests that by changing the value to value.toUpperCase() in handleChange will make the component only accept captialized letters, without side effects. But this is not true, and anyone can try this: just select a part of the text in the middle of the input field, and press a key. It will surely be converted to upper case, but in this case React will re-set the value attribute of the input, causing the text cursor (insertion point) to move to the end of the text. Which is a side effect, that can be very distrubing. For example, if the user wanted to replace a word in the middle of the text, then he just can't do it in any practical way. He will say that this is wrong, it is not working - and he will be damn right. In this particular case, I know that I can set text-transform:uppercase and store the lower case version in the store, but that is faulty too: if I need it to be upper case, then of course it should be in upper case in the store. The behaviour of the view must not determine the data representation in the store. Views should only be views, and nothing else.
I have checked many React related tutorials, and all of them have this "flaw". For example, here is a diagram from reactjs.org that shows Flux in action:
It suggests that the View is rendered after the action went through the dispatcher, the stores, then the controller-view. But the reality is that the View not only emits an action. It will also change its appearance before the action is emitted. React wants to see DOM elements as views that emit actions, but in reality they do other things.
The very same mistake appears in the reactjs tutorial, with the select element. ( https://reactjs.org/docs/forms.html ) They just say that by changing the "multiple={true}" attribute, you can have a multi select. No you can't, it just does not work. When you select multiple items in a multi select, the browser will call the onChange event multiple times: a separate call for every item selected. As a result, it will be impossible to set the state of the component to the set of selected items (at least not from the event handler). I COULD setup a timer, collect all changes emitted from onChange, and the finally call setState() with the collected items. But that is against React's rules, because there should be a single source of truth. And in this case, there will be multiple sources (at least for a while, until all events are collected). And anyone can see that in a complex application with many async handlers, this will eventually lead to bugs which are hard to debug.
A very simple thing ("only accept upper case letters in an input box") appears to me a hard thing to do correctly. Either I have to implement my own input box in JavaScript, or write a custom rendering method for input fields that keeps the cursor position at the correct place when value changes. Both seem to be bad ideas.
After Googling around, I have noticed that there are tons of components re-implementing native elements. There are literally hundreds of select and multiselect components written for React, and I suspect that this is partly because of how native DOM elements behave. (E.g. they are not pure views.)
Question: does anyone use react with native DOM components? And if so, how do you achieve the true "controlled" behaviour without the side effects? Is it possible? Or should I not care, and choose some kind of UI toolkit that was written specifically for React? (But even with those, implementing an "upper case input field" may require extra programming.)
This should be sufficient:
handleChange(event) {
const input = event.target;
const start = input.selectionStart;
const end = input.selectionEnd;
this.setState(
{value: input.value.toUpperCase()},
() => input.setSelectionRange(start, end)
);
}
Also check: http://blog.vishalon.net/javascript-getting-and-setting-caret-position-in-textarea
I have written simple toInputUppercase function which override e.target.value and applied attribute onInput to <input /> element which I want to be capitalize.
import React, { useState } from "react";
import ReactDOM from "react-dom";
const toInputUppercase = e => {
e.target.value = ("" + e.target.value).toUpperCase();
};
const App = () => {
const [name, setName] = useState("");
return (
<input
value={name}
onChange={e => setName(e.target.value)}
onInput={toInputUppercase} // apply on input which do you want to be capitalize
/>
);
};
ReactDOM.render(<App />, document.getElementById("root"));
Hope it will work for you.
You can simply add this property to the input field :
onInput={(e) => e.target.value = ("" + e.target.value).toUpperCase()}
Hope it works :)
You are right in what you observe but the issue does not come from React, it comes from the browser itself.
Consider the following example:
function handler() {
const elt = document.getElementById('test');
elt.value = elt.value.toUpperCase();
};
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width">
<title>JS Bin</title>
</head>
<body>
<input id="test" type="text" value="default value" onKeyPress="handler();" />
</body>
</html>
How would you make it work even before talking about React ? React can't override the browser behavior.
If you want to uppercase everything while keeping the cursor position, you could look at rich text editing with draft js for example.
easy: css + normalization (toUpperCase()) after submit
text-transform: uppercase;
I'm creating a form in react that has conditional flow. For example two input fields might be hidden when some select option is chosen. Is it a good idea to create states based on the select options and make the visibility of the input elements to depend on the current state. Or shall I add refs to the input elements and make the select cause a change event which will be used to manipulate the visibility of the inputs?
Unless you have any complicated logic for deciding whether to draw an input or not I think it's perfectly fine, even encouraged, to simply tie it to some state variable. Here's an example using Reacts two-way binding addon:
var MyForm = React.createClass({
mixins: [React.addons.LinkedStateMixin],
getInitialState: function() {
return {hasName: false};
},
render: function() {
var nameInput = null;
if (this.state.hasName) {
nameInput = <label>
Name:
<input type="text" />
</label>;
}
return <form>
<label>
<input type="checkbox" checkedLink={this.linkState('hasName')} />
Do you have a name?
</label>
<br/>
{nameInput}
</form>;
}
});
React.render(<MyForm/>, document.body);
http://jsfiddle.net/p4u1qhym/
And when you're using event handlers you should still set a state variable for the input: Even though React is smart enough to keep any DOM state intact during re-rendering (e.g. a visibility styling option), it won't always work. Imagine for example a situation where you serialize the forms state and want to use it to initialize another form later on. React won't be able to infer that the name input should not be displayed, and renders it.
There still exist situations when it's useful to use this.refs however, for example when setting the focus to a particular input, as described here.
From the title of your question I assume you are worrying about performance: Yes, the entire component will be re-rendered, but only in virtual DOM. React then finds all the differences to the actual DOM and only applies the changes required for the two to match. You can read about it here. As this whole process is very efficient, there should be virtually no difference in performance to setting some style attribute on the actual DOM node using this.refs. And it's a lot less verbose!