Fix internal state consistency of control cards in controlled mode

[gluxon]

Problem

Repro

In an application using React 18 with the createRoot API, a <RadioCard> in controlled mode may not reflect the current controlled checked prop value. Note in the video below how two clicks are required to get the radio button in its checked state.

Problem.mov

Here's the code for the video above.

import { RadioCard } from '@blueprintjs/core'
import React from 'react'

function App() {
  const [testState, setTestState] = React.useState<"toast" | "bread">("toast")

  return (
    <div>
      {testState}
      <RadioCard checked={testState === "toast"} onClick={() => { setTestState("toast"); }}>
        toast
      </RadioCard>
      <RadioCard checked={testState === "bread"} onClick={() => { setTestState("bread"); }}>
        bread
      </RadioCard>
    </div>
  )
}

export default App

Explanation

The repro code above contains buggy code. It's passing a controlled checked prop, but using onClick instead of onChange to toggle that prop.

This causes <RadioCard>'s internal state tracking to be inconsistent, which is arguably still a Blueprint bug despite the buggy usage this component.

With the concurrent renderer

1. The onClick handler on <RadioCard> fires.
2. This changes the props.checked value, and triggers the useEffect here. The effect changes the button's checked state from false to true.
   

   blueprint/packages/core/src/components/control-card/useCheckedControl.ts

   Lines 25 to 30 in daebef2

   	const [checked, setChecked] = React.useState(() => props.defaultChecked ?? false);
   	React.useEffect(() => {
   	if (props.checked !== undefined) {
   	setChecked(props.checked);
   	}
   	}, [props.checked]);

3. The <RadioCard>'s onChange handler then fires, toggling the updated checked state from true back to false. 🤦🏻‍♂️
   

   blueprint/packages/core/src/components/control-card/useCheckedControl.ts

   Lines 31 to 33 in daebef2

   	const onChange = React.useCallback<React.ChangeEventHandler<HTMLInputElement>>(
   	e => {
   	setChecked(c => !c);

Legacy renderer

With the legacy React 17 renderer, this happened to work because the order of these event handlers are reversed.

1. The onClick handler on <RadioCard> fires.
2. The internal onChange handler fires first.
3. Then the useEffect fires second.

I believe the order is different with the concurrent renderer due to an intentional change to make useEffect's timing more consistent.

Summary

In summary of the explanation above, Blueprint's uncontrolled state handling is fighting with the controlled state handling of the application.

Changes

The useEffect here is an example of "you might not need an effect".

blueprint/packages/core/src/components/control-card/useCheckedControl.ts

Lines 26 to 30 in daebef2

	React.useEffect(() => {
	if (props.checked !== undefined) {
	setChecked(props.checked);
	}
	}, [props.checked]);

I think we can simplify this with a value that's "calculated during rendering".

When something can be calculated from the existing props or state, don’t put it in state. Instead, calculate it during rendering. This makes your code faster (you avoid the extra “cascading” updates), simpler (you remove some code), and less error-prone (you avoid bugs caused by different state variables getting out of sync with each other). If this approach feels new to you, Thinking in React explains what should go into state.

const checked = props.checked ?? checkedStateForUncontrolledMode;

[changelog-app]

Generate changelog in packages/core/changelog/@unreleased

What do the change types mean?

• feature: A new feature of the service.
• improvement: An incremental improvement in the functionality or operation of the service.
• fix: Remedies the incorrect behaviour of a component of the service in a backwards-compatible way.
• break: Has the potential to break consumers of this service's API, inclusive of both Palantir services
  and external consumers of the service's API (e.g. customer-written software or integrations).
• deprecation: Advertises the intention to remove service functionality without any change to the
  operation of the service itself.
• manualTask: Requires the possibility of manual intervention (running a script, eyeballing configuration,
  performing database surgery, ...) at the time of upgrade for it to succeed.
• migration: A fully automatic upgrade migration task with no engineer input required.

Note: only one type should be chosen.

How are new versions calculated?

• ❗The break and manual task changelog types will result in a major release!
• 🐛 The fix changelog type will result in a minor release in most cases, and a patch release version for patch branches. This behaviour is configurable in autorelease.
• ✨ All others will result in a minor version release.

Type

• Feature
• Improvement
• Fix
• Break
• Deprecation
• Manual task
• Migration

Description

Fix a case where a control card component (e.g. <RadioCard>) may have rendered with a checked state that does not match its checked prop. This can happen if that prop was changed outside of an onChange handler and in an onClick handler instead.

Check the box to generate changelog(s)

• Generate changelog entry

[svc-palantir-github]

Add generated changelog entries

Build artifact links for this commit: documentation | landing | table | demo

This is an automated comment from the deploy-preview CircleCI job.

[pgoldberg]

We really don't need an effect here. Excellent example of https://react.dev/learn/you-might-not-need-an-effect

Thank you for the detailed description!

[gluxon]

Thanks for the reviews @pgoldberg @styu