Fixes https://github.com/payloadcms/payload/issues/11224
Fixes https://github.com/payloadcms/payload/issues/10492
This PR fixes a few weird behaviours when `validate: true` is set on drafts:
- when autosave is on and you submit an invalid form it would get stuck in an infinite loop
- PreventLeave would not trigger for submitted but invalid forms leading to potential data loss
Changes:
- Adds e2e tests for the above scenarios
- Adds a new `isValid` flag on the `Form` context provider to signal globally if the form is in a valid or invalid state
- Components like Autosave will manage this internally since it manages its own submission flow as well
- Adds PreventLeave to Autosave too for when form is invalid meaning data hasn't been actually saved so we want to prevent the user accidentally losing data by reloading or closing the page
The following tests have been added
If you have multiple blocks that are used in multiple places, this can quickly blow up the size of your Payload Config. This will incur a performance hit, as more data is
1. sent to the client (=> bloated `ClientConfig` and large initial html) and
2. processed on the server (permissions are calculated every single time you navigate to a page - this iterates through all blocks you have defined, even if they're duplicative)
This can be optimized by defining your block **once** in your Payload Config, and just referencing the block slug whenever it's used, instead of passing the entire block config. To do this, the block can be defined in the `blocks` array of the Payload Config. The slug can then be passed to the `blockReferences` array in the Blocks Field - the `blocks` array has to be empty for compatibility reasons.
```ts
import { buildConfig } from 'payload'
import { lexicalEditor, BlocksFeature } from '@payloadcms/richtext-lexical'
// Payload Config
const config = buildConfig({
// Define the block once
blocks: [
{
slug: 'TextBlock',
fields: [
{
name: 'text',
type: 'text',
},
],
},
],
collections: [
{
slug: 'collection1',
fields: [
{
name: 'content',
type: 'blocks',
// Reference the block by slug
blockReferences: ['TextBlock'],
blocks: [], // Required to be empty, for compatibility reasons
},
],
},
{
slug: 'collection2',
fields: [
{
name: 'editor',
type: 'richText',
editor: lexicalEditor({
BlocksFeature({
// Same reference can be reused anywhere, even in the lexical editor, without incurred performance hit
blocks: ['TextBlock'],
})
})
},
],
},
],
})
```
## v4.0 Plans
In 4.0, we will remove the `blockReferences` property, and allow string block references to be passed directly to the blocks `property`. Essentially, we'd remove the `blocks` property and rename `blockReferences` to `blocks`.
The reason we opted to a new property in this PR is to avoid breaking changes. Allowing strings to be passed to the `blocks` property will prevent plugins that iterate through fields / blocks from compiling.
## PR Changes
- Testing: This PR introduces a plugin that automatically converts blocks to block references. This is done in the fields__blocks test suite, to run our existing test suite using block references.
- Block References support: Most changes are similar. Everywhere we iterate through blocks, we have to now do the following:
1. Check if `field.blockReferences` is provided. If so, only iterate through that.
2. Check if the block is an object (= actual block), or string
3. If it's a string, pull the actual block from the Payload Config or from `payload.blocks`.
The exception is config sanitization and block type generations. This PR optimizes them so that each block is only handled once, instead of every time the block is referenced.
## Benchmarks
60 Block fields, each block field having the same 600 Blocks.
### Before:
**Initial HTML:** 195 kB
**Generated types:** takes 11 minutes, 461,209 lines
https://github.com/user-attachments/assets/11d49a4e-5414-4579-8050-e6346e552f56
### After:
**Initial HTML:** 73.6 kB
**Generated types:** takes 2 seconds, 35,810 lines
https://github.com/user-attachments/assets/3eab1a99-6c29-489d-add5-698df67780a3
### After Permissions Optimization (follow-up PR)
Initial HTML: 73.6 kB
https://github.com/user-attachments/assets/a909202e-45a8-4bf6-9a38-8c85813f1312
## Future Plans
1. This PR does not yet deduplicate block references during permissions calculation. We'll optimize that in a separate PR, as this one is already large enough
2. The same optimization can be done to deduplicate fields. One common use-case would be link field groups that may be referenced in multiple entities, outside of blocks. We might explore adding a new `fieldReferences` property, that allows you to reference those same `config.blocks`.
Due to nature of server-side rendering, navigation within the admin
panel can lead to slow page response times. This can lead to the feeling
of an unresponsive app after clicking a link, for example, where the
page remains in a stale state while the server is processing. This is
especially noticeable on slow networks when navigating to data heavy or
process intensive pages.
To alleviate the bad UX that this causes, the user needs immediate
visual indication that _something_ is taking place. This PR renders a
progress bar in the admin panel which is immediately displayed when a
user clicks a link, and incrementally grows in size until the new route
has loaded in.
Inspired by https://github.com/vercel/react-transition-progress.
Old:
https://github.com/user-attachments/assets/1820dad1-3aea-417f-a61d-52244b12dc8d
New:
https://github.com/user-attachments/assets/99f4bb82-61d9-4a4c-9bdf-9e379bbafd31
To tie into the progress bar, you'll need to use Payload's new `Link`
component instead of the one provided by Next.js:
```diff
- import { Link } from 'next/link'
+ import { Link } from '@payloadcms/ui'
```
Here's an example:
```tsx
import { Link } from '@payloadcms/ui'
const MyComponent = () => {
return (
<Link href="/somewhere">
Go Somewhere
</Link>
)
}
```
In order to trigger route transitions for a direct router event such as
`router.push`, you'll need to wrap your function calls with the
`startRouteTransition` method provided by the `useRouteTransition` hook.
```ts
'use client'
import React, { useCallback } from 'react'
import { useTransition } from '@payloadcms/ui'
import { useRouter } from 'next/navigation'
const MyComponent: React.FC = () => {
const router = useRouter()
const { startRouteTransition } = useRouteTransition()
const redirectSomewhere = useCallback(() => {
startRouteTransition(() => router.push('/somewhere'))
}, [startRouteTransition, router])
// ...
}
```
In the future [Next.js might provide native support for
this](https://github.com/vercel/next.js/discussions/41934#discussioncomment-12077414),
and if it does, this implementation can likely be simplified.
Of course there are other ways of achieving this, such as with
[Suspense](https://react.dev/reference/react/Suspense), but they all
come with a different set of caveats. For example with Suspense, you
must provide a fallback component. This means that the user might be
able to immediately navigate to the new page, which is good, but they'd
be presented with a skeleton UI while the other parts of the page stream
in. Not necessarily an improvement to UX as there would be multiple
loading states with this approach.
There are other problems with using Suspense as well. Our default
template, for example, contains the app header and sidebar which are not
rendered within the root layout. This means that they need to stream in
every single time. On fast networks, this would also lead to a
noticeable "blink" unless there is some mechanism by which we can detect
and defer the fallback from ever rendering in such cases. Might still be
worth exploring in the future though.
Fixes#10440. When `filterOptions` are set on a relationship field,
those same filters are not applied to the `Filter` component within the
list view. This is because `filterOptions` is not being thread into the
`RelationshipFilter` component responsible for populating the available
options.
To do this, we first need to be resolve the filter options on the server
as they accept functions. Once resolved, they can be prop-drilled into
the proper component and appended onto the client-side "where" query.
Reliant on #11080.
### What?
Initial values should be set from the server when `acceptValues` is
true.
### Why?
This is needed since we take the values from the server after a
successful form submission.
### How?
Add `initialValue` into `serverPropsToAccept` when `acceptValues` is
true.
Fixes https://github.com/payloadcms/payload/issues/10820
---------
Co-authored-by: Alessio Gravili <alessio@gravili.de>
Fixes https://github.com/payloadcms/payload/issues/10940
This PR does the following:
- adds a `useDocumentForm` hook to access the document Form. Useful if
you are within a sub-Form
- ensure the `data` property passed to field conditions, read access
control, validation and filterOptions is always the top-level document
data. Previously, for fields within lexical blocks/links/upload, this
incorrectly was the lexical block-level data.
- adds a `blockData` property to hooks, field conditions,
read/update/create field access control, validation and filterOptions
for all fields. This allows you to access the data of the nearest parent
block, which is especially useful for lexical sub-fields. Users that
were previously depending on the incorrect behavior of the `data`
property in order to access the data of the lexical block can now switch
to the new `blockData` property
The `useIgnoredEffect` hook is useful in firing an effect only when a _subset_ of dependencies change, despite subscribing to many dependencies. But the previous implementation of `useIgnoredEffect` had a few problems:
- The effect did not receive the updated values of `ignoredDeps` - thus, `useIgnoredEffect` pretty much worked the same way as using `useEffect` and omitting said dependencies from the dependency array. This caused the `ignoredDeps` values to be stale.
- It compared objects by value instead of reference, which is slower and behaves differently than `useEffect` itself.
- Edge cases where the effect does not run even though the dependencies have changed. E.g. if an `ignoredDep` has value `null` and a `dep` changes its value from _something_ to `null`, the effect incorrectly does **not** run, as the current logic detects that said value is part of `ignoredDeps` => no `dep` actually changed.
This PR replaces the `useIgnoredEffect` hook with a new pattern which to combine `useEffect` with a new `useEffectEvent` hook as described here: https://react.dev/learn/separating-events-from-effects#extracting-non-reactive-logic-out-of-effects. While this is not available in React 19 stable, there is a polyfill available that's already used in several big projects (e.g. react-spectrum and bluesky).
Previously, the lexical link drawer did not display any fields if the
`create` permission was false, even though the `update` permission was
true.
The issue was a faulty permission check in `RenderFields` that did not
check the top-level permission operation keys for truthiness. It only
checked if the `permissions` variable itself was `true`, or if the
sub-fields had `create` / `update` permissions set to `true`.
As a result of #9388, the `valid` and `passesCondition` properties no
longer appear in form state. This leads to breaking logic if you were
previously relying on these properties to have explicit values. To fix
this, we simply perform the inverse on these properties before accepting
them into client side form state. In the next major release, we can
accept form state as it is received and instruct users to modify their
logic as needed.
Also comes with a small perf optimization, by keeping the old object
reference of fields if they did not change when server form state comes
back
### What?
In some cases you may want to opt out of using the default access
control that this plugin provides on the tenants collection.
### Why?
Other collections are able to opt out of this already, but the tenants
collection specifically was not configured with an opt out capability.
### How?
Adds new property to the plugin config: `useTenantsCollectionAccess`.
Setting this to `false` allows users to opt out and write their own
access control functions without the plugin merging in its own
constraints for the tenant collection.
Fixes https://github.com/payloadcms/payload/issues/10882
### What?
When hitting the form-state endpoint, previousValue was undefined.
### Why?
It was not being passed through.
### How?
Sets previousValue as the initial value of the field.
Fixes https://github.com/payloadcms/payload/issues/10826
Field validations can be expensive, especially custom validations that
are async or highly complex. This can lead to slow form state response
times when generating form state for many such fields. Ideally, we only
run validations on fields whose values have changed. This is not
possible, however, because field validation functions might reference
_other_ field values with their args, and there is no good way of
detecting exactly which fields should run in this case. The next best
thing here is to only run validations _after the form has been
submitted_, and then every `onChange` event thereafter until a
successful submit has taken place. This is an elegant solution because
we currently don't _render_ field errors until submission anyway.
This change will significantly speed up form state response times, at
least until the form has been submitted. From then on, all field
validations will run regardless, just as they do now. If custom
validations continue to slow down form state response times, there is a
new `event` arg introduced in #10738 that can be used to control whether
heavy operations occur on change or on submit.
Related: #10638
Field paths within hooks are not correct.
For example, an unnamed tab containing a group field and nested text
field should have the path:
- `myGroupField.myTextField`
However, within hooks that path is formatted as:
- `_index-1.myGroupField.myTextField`
The leading index shown above should not exist, as this field is
considered top-level since it is located within an unnamed tab.
This discrepancy is only evident through the APIs themselves, such as
when creating a request with invalid data and reading the validation
errors in the response. Form state contains proper field paths, which is
ultimately why this issue was never caught. This is because within the
admin panel we merge the API response with the current form state,
obscuring the underlying issue. This becomes especially obvious in
#10580, where we no longer initialize validation errors within form
state until the form has been submitted, and instead rely solely on the
API response for the initial error state.
Here's comprehensive example of how field paths _should_ be formatted:
```
{
// ...
fields: [
{
// path: 'topLevelNamedField'
// schemaPath: 'topLevelNamedField'
// indexPath: ''
name: 'topLevelNamedField',
type: 'text',
},
{
// path: 'array'
// schemaPath: 'array'
// indexPath: ''
name: 'array',
type: 'array',
fields: [
{
// path: 'array.[n].fieldWithinArray'
// schemaPath: 'array.fieldWithinArray'
// indexPath: ''
name: 'fieldWithinArray',
type: 'text',
},
{
// path: 'array.[n].nestedArray'
// schemaPath: 'array.nestedArray'
// indexPath: ''
name: 'nestedArray',
type: 'array',
fields: [
{
// path: 'array.[n].nestedArray.[n].fieldWithinNestedArray'
// schemaPath: 'array.nestedArray.fieldWithinNestedArray'
// indexPath: ''
name: 'fieldWithinNestedArray',
type: 'text',
},
],
},
{
// path: 'array.[n]._index-2'
// schemaPath: 'array._index-2'
// indexPath: '2'
type: 'row',
fields: [
{
// path: 'array.[n].fieldWithinRowWithinArray'
// schemaPath: 'array._index-2.fieldWithinRowWithinArray'
// indexPath: ''
name: 'fieldWithinRowWithinArray',
type: 'text',
},
],
},
],
},
{
// path: '_index-2'
// schemaPath: '_index-2'
// indexPath: '2'
type: 'row',
fields: [
{
// path: 'fieldWithinRow'
// schemaPath: '_index-2.fieldWithinRow'
// indexPath: ''
name: 'fieldWithinRow',
type: 'text',
},
],
},
{
// path: '_index-3'
// schemaPath: '_index-3'
// indexPath: '3'
type: 'tabs',
tabs: [
{
// path: '_index-3-0'
// schemaPath: '_index-3-0'
// indexPath: '3-0'
label: 'Unnamed Tab',
fields: [
{
// path: 'fieldWithinUnnamedTab'
// schemaPath: '_index-3-0.fieldWithinUnnamedTab'
// indexPath: ''
name: 'fieldWithinUnnamedTab',
type: 'text',
},
{
// path: '_index-3-0-1'
// schemaPath: '_index-3-0-1'
// indexPath: '3-0-1'
type: 'tabs',
tabs: [
{
// path: '_index-3-0-1-0'
// schemaPath: '_index-3-0-1-0'
// indexPath: '3-0-1-0'
label: 'Nested Unnamed Tab',
fields: [
{
// path: 'fieldWithinNestedUnnamedTab'
// schemaPath: '_index-3-0-1-0.fieldWithinNestedUnnamedTab'
// indexPath: ''
name: 'fieldWithinNestedUnnamedTab',
type: 'text',
},
],
},
],
},
],
},
{
// path: 'namedTab'
// schemaPath: '_index-3.namedTab'
// indexPath: ''
label: 'Named Tab',
name: 'namedTab',
fields: [
{
// path: 'namedTab.fieldWithinNamedTab'
// schemaPath: '_index-3.namedTab.fieldWithinNamedTab'
// indexPath: ''
name: 'fieldWithinNamedTab',
type: 'text',
},
],
},
],
},
]
}
```
Field validations currently run very often, such as within form state on
type. This can lead to serious performance implications within the admin
panel if those validation functions are async, especially if they
perform expensive database queries. One glaring example of this is how
all relationship and upload fields perform a database lookup in order to
evaluate that the given value(s) satisfy the defined filter options. If
the field is polymorphic, this can happen multiple times over, one for
each collection. Similarly, custom validation functions might also
perform expensive tasks, something that Payload has no control over.
The fix here is two-fold. First, we now provide a new `event` arg to all
`validate` functions that allow you to opt-in to performing expensive
operations _only when documents are submitted_, and fallback to
significantly more performant validations as form state is generated.
This new pattern will be the new default for relationship and upload
fields, however, any custom validation functions will need to be
implemented in this way in order to take advantage of it. Here's what
that might look like:
```
[
// ...
{
name: 'text'
type: 'text',
validate: async (value, { event }) => {
if (event === 'onChange') {
// Do something highly performant here
return true
}
// Do something more expensive here
return true
}
}
]
```
The second part of this is to only run validations _after the form as
been submitted_, and then every change event thereafter. This work is
being done in #10580.
Custom block row labels defined on `admin.components.Label` were not
rendering despite existing in the config. Instead, if a custom label
component was defined on the _top-level_ blocks field itself, it was
incorrectly replacing each blocks label _in addition_ to the field's
label. Now, custom labels defined at the field-level now only replace
the field's label as expected, and custom labels defined at the
block-level are now supported as the types suggest.
The `access.read` function executed within form state was missing the
`id` arg, and was also incorrectly setting `data` as `doc`. When
building form state, there is no concept of a "doc" because it is
possible to build form state using only a subset of fields. There is
"data", however, which represents the schema path at the entry point of
the function. Similarly, when building form state on within an
`onChange` function, for example, we do not send the original doc
through the request, which is what "doc" would represent. Instead, we
send either `data` or `formState`, both of which could represent a
_modified_ doc. This particular invocation of read access does not
effect the visibility of fields themselves, but rather their return
values from the form state endpoint. Field visibility is determined at
the request level.
This significantly optimizes the form state, reducing its size by up to
more than 3x and improving overall response times. This change also has
rolling effects on initial page size as well, where the initial state
for the entire form is sent through the request. To achieve this, we do
the following:
- Remove `$undefined` strings that are potentially attached to
properties like `value`, `initialValue`, `fieldSchema`, etc.
- Remove unnecessary properties like empty `errorPaths` arrays and empty
`customComponents` objects, which only need to exist if used
- Remove unnecessary properties like `valid`, `passesCondition`, etc.
which only need to be returned if explicitly `false`
- Remove unused properties like `isSidebar`, which simply don't need to
exist at all, as they can be easily calculated during render
## Results
The following results were gathered by booting up each test suite listed
below using the existing seed data, navigating to a document in the
relevant collection, then typing a single letter into the noted field in
order to invoke new form-state. The result is then saved to the file
system for comparison.
| Test Suite | Collection | Field | Before | After | Percentage Change |
|------|------|---------|--------|--------|--------|
| `field-perf` | `blocks-collection` | `layout.0.field1` | 227kB | 110
kB | ~52% smaller |
| `fields` | `array-fields` | `items.0.text` | 14 kB | 4 kB | ~72%
smaller |
| `fields` | `block-fields` | `blocks.0.richText` | 25 kB | 14 kB | ~44%
smaller |
Fixes#9882 and #9691
In 2.0, we would accept data coming back from an update operation and
then reflect those changes in UI.
However, in 3.0, we did not do that anymore - meaning you could change a
document with hooks in `beforeChange` or `afterChange`, but then not see
the changes made on the server.
This PR updates the way that `mergeServerFormState` works, and adds a
property to optionally allow values from server form state - which can
then be used in the `onSuccess` form handler which may need to define
new field values.
Whenever form state fails, like when field conditions, validations, or
default value functions throw errors, blocks and array rows are stuck
within an infinite loading state. Examples of this might be when
accessing properties of undefined within these functions, etc. Although
these errors are logged to the server console, the UI is be misleading,
where the user often waits for the request to resolve rather than
understanding that an underlying API error has occurred. Now, we safely
execute these functions within a `try...catch` block and handle their
failures accordingly. On the client, form state will resolve as expected
using the default return values for these functions.
Fixes#10070. Adding new blocks or array rows can randomly get stuck
within an infinite loading state. This was because the abort controllers
responsible for disregarding duplicate `onChange` and `onSave` events
was not properly resetting its refs across invocations. This caused
subsequent event handlers to incorrectly abort themselves, leading to
unresolved requests and a `null` form state. Similarly, the cleanup
effects responsible for aborting these requests on component unmount
were also referencing its `current` property directly off the refs,
which can possible be stale if not first set as a variable outside the
return function.
This PR also carries over some missing `onSave` logic from the default
edit view into the live preview view. In the future the logic between
these two views should be standardized, as they're nearly identical but
often become out of sync. This can likely be done through the use of
reusable hooks, such as `useOnSave`, `useOnChange`, etc. Same with the
document locking functionality which is complex and deeply integrated
into each of these views.
Fixes#9888. Field permissions were not being passed into custom
components. This led to custom components, such as arrays and blocks,
unable to render default Payload fields. This was because their props
lacked the permissions object required for rendering. For example:
```ts
'use client'
import type { ArrayFieldClientComponent } from 'payload'
import { ArrayField } from '@payloadcms/ui'
export const MyArray: ArrayFieldClientComponent = (props) => <ArrayField {...props} />
```
In this example the array field itself would render, but the fields
within each row would not, because the array field did not pass its
permissions down to the rows.
Extension of #9933. Custom components are returned by form state,
meaning that if we don't wait for form state to return before rendering
row labels, the default row label component will render in briefly
before being swapped by a custom component (if applicable). Using the
new `isLoading` prop on array and block rows, we can conditionally
render them just as we currently do for the row fields themselves.
Rework of https://github.com/payloadcms/payload/pull/5912
### What?
Now, when `defaultValue` is defined as function you can receive the
`req` argument:
```ts
{
name: 'defaultValueFromReq',
type: 'text',
defaultValue: async ({ req, user, locale }) => {
return Promise.resolve(req.context.defaultValue)
},
},
```
`user` and `locale` even though are repeated in `req`, this potentially
leaves some room to add more args in the future without removing them
now.
This also improves type for `defaultValue`:
```ts
type SerializableValue = boolean | number | object | string
export type DefaultValue =
| ((args: {
locale?: TypedLocale
req: PayloadRequest
user: PayloadRequest['user']
}) => SerializableValue)
| SerializableValue
```
### Why?
To access the current URL / search params / Local API and other things
directly in `defaultValue`.
### How?
Passes `req` through everywhere where we call `defaultValue()`
### What?
Removes the `localized` property from typescript suggestion for row and
collapsible fields.
### Why?
Currently, this property doesn't do anything for them. This may be
changed when/if we support `name` for them, but it'll work only with
`name`.
Fixes https://github.com/payloadcms/payload/issues/4720
Currently, custom components do not respect `admin.condition` unless
manually wrapped with the `withCondition` HOC, like all default fields
currently do. This should not be a requirement of component authors.
Instead, we can automatically detect custom client and server fields and
wrap them with the underlying `WatchCondition` component which will
subscribe to the `passesCondition` property within client-side form
state.
For my future self: there are potentially multiple instances where
fields subscribe to conditions duplicately, such as when rendering a
default Payload field within a custom field component. This was always a
problem and it is non-breaking, but needs to be reevaluated and removed
in the future for performance. Only the default fields that Payload
renders client-side need to subscribe to field conditions in this way.
When importing a Payload field into your custom field component, for
example, it should not include the HOC, because custom components now
watch conditions themselves.
When a condition exists on a field and it resolves to `false`, it
currently "blinks" in and out when rendered within an array or block
row. This is because when add rows to form state, we iterate over the
_fields_ of that row and render their respective components. Then when
conditions are checked for that field, we're expecting `passesCondition`
to be explicitly `false`, ultimately _rendering_ the field for a brief
moment before form state returns with evaluated conditions. The fix is
to set these fields into local form state with a new `isLoading: true`
prop, then display a loader within the row until form state returns with
its proper conditions.
### What?
It became possible for fields to reset to a defined `defaultValue` when
bulk editing from the `edit-many` drawer.
### Why?
The form-state of all fields were being considered during a bulk edit -
this also meant using their initial states - this meant any fields with
default values or nested fields (`arrays`) would be overwritten with
their initial states
I.e. empty values or default values.
### How?
Now - we only send through the form data of the fields specifically
being edited in the edit-many drawer and ignore all other fields.
Leaving all other fields stay their current values.
Fixes#9590
---------
Co-authored-by: Dan Ribbens <dan.ribbens@gmail.com>
### What?
Previously, `initCollapsed: true` `array` fields would auto collapse
when typing in their respective inputs while in the create new view.
### Why?
This was due to the fact that we were only checking if `preferences`
existed in `form state` to handle the current state of the array row and
then falling back on the `initCollapsed` prop if `preferences` didn't
exist.
This was a problem because during create - `preferences` do not exist
yet. As a result, the state of the array row would keep falling back to
collapsed if `initCollapsed` was set to `true`.
### How?
To fix this, we now check the actual form state first before falling
back to preferences and then falling back to the initCollapsed prop
value.
Fixes#9775
### What?
Fixes issue with stale locale from searchParams
### Why?
Bad use of useEffect/useState inside our useSearchParams provider.
### How?
Memoize the locale instead of relying on the useEffect which was causing
unnecessary renders with stale values.
### What?
When the document is saved the formState was not being reset from the
server.
### Why?
getFormState was not being called onSuccess of the form submission
### How?
The `Form` onSuccess function now allows for an optional return type of
`FormState` if the functions returns formState then we check to see if
that differs from the current formState on the client. If it does then
we dispatch the `REPLACE_STATE` action with the newState.
Fixes https://github.com/payloadcms/payload/issues/9423
If you had a lot of fields and collections, createClientConfig would be
extremely slow, as it was copying a lot of memory. In my test config
with a lot of fields and collections, it took 4 seconds(!!).
And not only that, it also ran between every single page navigation.
This PR significantly speeds up the createClientConfig function. In my
test config, its execution speed went from 4 seconds to 50 ms.
Additionally, createClientConfig is now properly cached in both dev &
prod. It no longer runs between every single page navigation. Even if
you trigger a full page reload, createClientConfig will be cached and
not run again. Despite that, HMR remains fully-functional.
This will make payload feel noticeably faster for large configs -
especially if it contains a lot of richtext fields, as it was previously
deep-copying the relatively large richText editor configs over and over
again.
## Before - 40 sec navigation speed
https://github.com/user-attachments/assets/fe6b707a-459b-44c6-982a-b277f6cbb73f
## After - 1 sec navigation speed
https://github.com/user-attachments/assets/384fba63-dc32-4396-b3c2-0353fcac6639
## Todo
- [x] Implement ClientSchemaMap and cache it, to remove
createClientField call in our form state endpoint
- [x] Enable schemaMap caching for dev
- [x] Cache lexical clientField generation, or add it to the parent
clientConfig
## Lexical changes
Red: old / removed
Green: new
### Speed up version queries
This PR comes with performance optimizations for fetching versions
before a document is loaded. Not only does it use the new select API to
limit the fields it queries, it also completely skips a database query
if the current document is published.
### Speed up lexical init
Removes a bunch of unnecessary deep copying of lexical objects which
caused higher memory usage and slower load times. Additionally, the
lexical default config sanitization now happens less often.
Fixes https://github.com/payloadcms/payload/issues/9363
This fixes the following issues that caused fields to be either hidden,
or incorrectly set to readOnly in certain configurations:
- In some cases, permissions were sanitized incorrectly. This PR
rewrites the sanitizePermissions function and adds new unit tests
- after a document save, the client was receiving unsanitized
permissions. Moving the sanitization logic to the endpoint fixes this
- Various incorrect handling of permissions in our form state endpoints
/ RenderFields
Fixes#9264. When externally updating array or block rows through the
`addFieldRow` or `replaceFieldRow` methods, nested rich text fields
along with any custom components within them are never rendered. This is
because unless the form is explicitly set to modified, as the default
array and blocks fields currently do, the newly generated form-state
will skip the rendering step. Now, the underlying callbacks themselves
automatically set the form to modified to trigger rendering.
The biggest difference comes from calling `RenderServerComponent` as a
function, instead of rendering it by using `<RenderServerComponent`.
This gets rid of wasteful blocks of codes sent to the client that look
like this:
HTML size comparison:
## Admin test suite
| View | Before | After |
|------|---------|--------|
| Dashboard | 331 kB | 83 kB |
| collections/custom-views-one Edit | 285 kB | 76.6 kB |
## Fields test suite
| View | Before | After |
|------|---------|--------|
| collections/lexical Edit | 189 kB | 94.4 kB |
| collections/lexical List | 152 kB | 62.9 kB |
## Community test suite
| View | Before | After |
|------|---------|--------|
| Dashboard | 78.9 kB | 43.1 kB |
Custom field description functions were being duplicately called in both
the Client Config and form state. Static field descriptions were also
being rendered in form state unnecessarily. Now, field description
functions are only executed once within form state, and static
descriptions are deferred to the client for rendering.
Supports bi-directional import/export between MDX <=> Lexical. JSX will
be mapped to lexical blocks back and forth.
This will allow editing our mdx docs in payload while keeping mdx as the
source of truth
---------
Co-authored-by: Germán Jabloñski <43938777+GermanJablo@users.noreply.github.com>
The field RSC now provides an initial state for all lexical blocks. This
completely obliterates any flashes and lexical block loading states when
loading or saving a document.
Previously, when a document is loaded or saved, every lexical block was
sending a network request in order to fetch their form state. Now, this
is batched and handled in the lexical server component. All lexical
block form states are sent to the client together with the parent
lexical field, and are thus available immediately.
We also do the same with block collapsed preferences. Thus, there are no
loading states or layout shifts/flashes of blocks anymore.
Additionally, when saving a document while your cursor is inside a
lexical field, the cursor position is preserved. Previously, a document
save would kick your cursor out of the lexical field.
## Look at how nice this is:
https://github.com/user-attachments/assets/21d736d4-8f80-4df0-a782-7509edd993da
**BREAKING:**
This removes the `feature.hooks.load` and `feature.hooks.save`
interfaces from custom lexical features, as they weren't used internally
and added unnecessary, additional overhead.
If you have custom features that use those, you can migrate to using
normal payload hooks that run on the server instead of the client.
With this PR, you can now customize the way that `blocks` and
`inlineBlocks` are rendered within Lexical's `BlocksFeature` by passing
your own React components.
This is super helpful when you need to create "previews" or more
accurate UI for your Lexical blocks.
For example, let's say you have a `gallery` block where your admins
select a bunch of images. By default, Lexical would just render a
collapsible with your block's fields in it. But now you can customize
the `admin.components.Block` property on your `block` config by passing
it a custom React component for us to render instead.
So using that, with this `gallery` example, you could make a dynamic
gallery React component that shows the images to your editors - and then
render our built-in `BlockEditButton` to allow your editors to manage
your gallery in a drawer.
Here is an example where the BlockEditButton is added to the default
Block Collapsible/Header:
---------
Co-authored-by: James <james@trbl.design>
Protects the `/api/access` endpoint behind authentication and sanitizes
the result, making it more secure and significantly smaller. To do this:
1. The `permission` keyword is completely omitted from the result
2. Only _truthy_ access results are returned
3. All nested permissions are consolidated when possible
---------
Co-authored-by: Dan Ribbens <dan.ribbens@gmail.com>
Co-authored-by: Jacob Fletcher <jacobsfletch@gmail.com>
Co-authored-by: James <james@trbl.design>
### What?
Aligns types for HiddenField and the WatchCondition component with the
rest of the fields. Since path is required when rendering a Field
component, there is no need to keep it optional in the WatchCondition
component.
### Why?
Hidden fields were requiring the `field` property to be passed, but the
only reason it needed it was to allow the path to fallback to name if
path was not passed. But path is required so there is no need for this
anymore.
This makes using the HiddenField simpler now.
### How?
Adjusts type on the HiddenField and the WatchCondition component.
### What?
Ensures `path` is required and only present on the fields that expect it
(all fields except row).
Deprecates `useFieldComponents` and `FieldComponentsProvider` and
instead extends the RenderField component to account for all field
types. This also improves type safety within `RenderField`.
### Why?
`path` being optional just adds DX overhead and annoyance.
### How?
Added `FieldPaths` type which is added to iterable field types. Placed
`path` back onto the ClientFieldBase type.
