Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Profiler RFC #51

Merged
merged 12 commits into from
Jul 25, 2019
137 changes: 137 additions & 0 deletions text/0000-profiler.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,137 @@
- Start Date: 2018-05-22
- RFC PR: (leave this empty)
- React Issue: (leave this empty)

# Summary

New React profiling component that collects timing information in order to measure the "cost" of rendering.

This component will also integrate with the [experimental `interaction-tracking` API](https://github.com/facebook/react/pull/13234) so that tracked interactions can be correlated with the render(s) they cause. This enables the calculation of "wall time" (elapsed real time) from when e.g. a user clicks a form button until when the DOM is updated in response. It also enables long-running renders to be more easily attributed and reproduced.

Note that an experimental release of the profiler component is available in version 16.4 as `React.unstable_Profiler`. It does not yet support interactions as that package has not been released.

# Usage example

`Profiler` can be declared anywhere within a React tree to measure the cost of rendering that portion of the tree.

For example, to profile a `Navigation` component and its descendants:
```js
render(
<App>
<Profiler id="Navigation" onRender={callback}>
<Navigation {...props} />
</Profiler>
<Main {...props} />
</App>
);
```

Multiple `Profiler` components can be used to measure different parts of an application:
```js
render(
<App>
<Profiler id="Navigation" onRender={callback}>
<Navigation {...props} />
</Profiler>
<Profiler id="Main" onRender={callback}>
<Main {...props} />
</Profiler>
</App>
);
```

`Profiler` components can also be nested to measure different components within the same subtree:
```js
render(
<App>
<Profiler id="Panel" onRender={callback}>
<Panel {...props}>
<Profiler id="Content" onRender={callback}>
<Content {...props} />
</Profiler>
<Profiler id="PreviewPane" onRender={callback}>
<PreviewPane {...props} />
</Profiler>
</Panel>
</Profiler>
</App>
);
```

Although `Profiler` is a light-weight component, it should be used only when necessary; each use adds CPU and memory overhead to an application.

# Motivation

It is important that render timing metrics work properly with React's experimental async rendering mode. When asynchronously rendering, React may yield periodically so that an app remains responsive even on low-powered devices. This yielded time (when React is not running) should not be included when considering the "cost" of a render. This distinction is not possible to implement in user space.

Timing measurements should also be significantly lighter weight than the current User Timing API so that they can be gathered in production without negatively impacting user experience. (The User Timing API is currently disabled for production because it is slow.) In addition to a faster implementation, we can further limit the impact on existing apps by creating a new production + profiling bundle. This way, apps that don't make use of the `Profiler` component (or wish to disable it globally) will not incur any additional overhead. (The `Profiler` component will render its children in production mode but its `onRender` callback will not be called.)

# Detailed design

The `onRender` callback is called each time a component within the `Profiler` renders. It receives the following parameters:
```js
function onRenderCallback(
id: string,
phase: "mount" | "update",
actualDuration: number,
baseDuration: number,
startTime: number,
commitTime: number,
interactions: Array<{ name: string, timestamp: number }>,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shape changed slightly and matches in 16.8.4 (guess this originates in the scheduler?)

Suggested change
interactions: Array<{ name: string, timestamp: number }>,
interactions: Set<{ id: number; name: string, timestamp: number }>,

Will the implementation change or is the proposal outdated?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, yeah. This RFC is a bit outdated.

): void {
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why not pass all of this as a single object so people can pick out the keys they need?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reasonable question!

I'd say the reasons are that I'm following precedent (we don't pass named parameters anywhere else that I can think of off the top of my head) and avoiding allocating a wrapper Object during commit.

I'd be interested to hear what others think about this aspect.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd vote for an object, despite the fact that it breaks with existing React API precedent.

The order of these timing arguments is going to be tough to memorize, and I can imagine only being interested in a subset of them. Using an object also enables you to add additional timing data down the road.

I'd view it as somewhat analogous to an event object, which has a variety of keys, only some of which are of interest for any given listener.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why would you need to memorize it? I imagine you'd only use Profiler in a few places in the app, and each time could consult the docs.

The need to avoid allocations is pretty important because adding GC pressure can skew the profiling results.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Even if you use Profiler in more than one place, the callback you pass it is likely shared- (this is why the id parameter exists)- so you would only need to write these params (in the correct order) in a single place.

Copy link
Collaborator Author

@bvaughn bvaughn May 23, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yup, this concern makes sense. And I agree that we wouldn't be allocating too many new objects for this, because it would only be one per Profiler per commit. I was just sharing rationale for why it is currently the way it is.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just my personal opinion here, but an object looks good from my point of view as long functions with more that 2/3 args are always hard to remember, you tend to start adding null for the values that you might not want to use, I'm looking at you JSON.stringify(foo, null, 2) 😅 , you also need to remember the order and it's harder to refactor as you impact anyone already using that order.

Plus with the actual syntax for destructuring the function signature looks pretty much the same but with curly braces 😁, the best of two worlds!

onRenderCallback({ id, phase, actualTime, baseTime, startTime, commitTime })

vs

onRenderCallback(id, phase, actualTime, baseTime, startTime, commitTime)

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Now that the Profiling API is out in the 16.4.1 release, I assume you decided to take no action on this?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The unstable_Profiler component was introduced in 16.4.0. The only thing that's new in 16.4.1 is a production+profiling build.

Unstable APIs can change. We haven't decided one way or another. This is kind of an open thread for discussion.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just circling back on this particular thread. Sebastian and I chatted about this yesterday, and we've decided to avoid named parameters because the overhead of the wrapper objects (however small each individual one is) will add up in larger applications.

// Aggregate or log render timings...
}

```

#### `id: string`
The `id` value of the `Profiler` tag that was measured. This value can change between renders if e.g. it is derived from `state` or `props`.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm a bit confused about that part. Why would the id change between renders? It sounds like the second part was meant for an additional parameter e.g. something like reason?
Was a bit confused about this part. It sounded like the id might include some metadata why onRender was called but what this refers to is the value of the id prop e.g. <Profiler id={'tweets-' + props.userId} onRender={handleRenderCallBackWithChangingIds} />


#### `phase: "mount" | "update"`
Identifies whether this component has just been mounted or re-rendered due to a change in `state` or `props`.

#### `actualDuration: number`
Time spent rendering the `Profiler` and its descendants for the current (most recent recent) update. This time tells us how well the subtree makes use of `shouldComponentUpdate` for memoization.

Ideally, this time should decrease significantly after the initial mount as many of the descendants will only need to re-render if their specific `props` change.

Note that in async mode, under certain conditions, React might render the same component more than once as part of a single commit. (In this event, the "actual" time for an update might be larger than the initial time.)

#### `baseDuration: number`
Duration of the most recent `render` time for each individual component within the `Profiler` tree. In other words, this value will only change when a component is re-rendered. It reflects a worst-case cost of rendering (e.g. the initial mount or no `shouldComponentUpdate` memoization).

#### `startTime: number`
Start time identifies when a particular commit started rendering. Although insufficient to determine the cause of the render, it can at least be used to rule out certain interactions (e.g. mouse click, Flux action). This may be helpful if you are also collecting other types of interactions and trying to correlate them with renders.

Start time isn't just the commit time less the "actual" time, because in async rendering mode React may yield during a render. This "yielded time" (when React was not doing work) is not included in either the "actual" or "base" time measurements.

#### `commitTime: number`
Commit time could be roughly determined using e.g. `performance.now()` within the `onRender` callback, but multiple `Profiler` components would end up with slightly different times for a single commit. Instead, an explicit time is provided (shared between all `Profiler`s in the commit) enabling them to be grouped if desirable.

#### `interactions: Array<{ name: string, timestamp: number }>`
An array of interactions that were being tracked (via the `interaction-tracking` package) when this commit was initially scheduled (e.g. when `render` or `setState` were called).

In the event of a cascading render (e.g. an update scheduled from `componentDidMount` or `componentDidUpdate`) React will forward these interactions along to the subsequent `onRender` calls.

# Drawbacks

Overuse of this component might negatively impact application performance.

# Alternatives

None considered.

# Adoption strategy

This is an entirely new component. Adoption can be organic and gradual.

# How we teach this

A reactjs.org blog post would be a good initial start.

Perhaps we could provide some sort of discoverability within React DevTools.

# Related proposals

* [facebook/react/pull/13253](https://github.com/facebook/react/pull/13253): Integration with the proposed `interaction-tracking` package
* [facebook/react-devtools/pull/1069](https://github.com/facebook/react-devtools/pull/1069): Integration with React DevTools