This feature allows you to specify `collection` for the join field as
array.
This can be useful for example to describe relationship linking like
this:
```ts
{
slug: 'folders',
fields: [
{
type: 'join',
on: 'folder',
collection: ['files', 'documents', 'folders'],
name: 'children',
},
{
type: 'relationship',
relationTo: 'folders',
name: 'folder',
},
],
},
{
slug: 'files',
upload: true,
fields: [
{
type: 'relationship',
relationTo: 'folders',
name: 'folder',
},
],
},
{
slug: 'documents',
fields: [
{
type: 'relationship',
relationTo: 'folders',
name: 'folder',
},
],
},
```
Documents and files can be placed to folders and folders themselves can
be nested to other folders (root folders just have `folder` as `null`).
Output type of `Folder`:
```ts
export interface Folder {
id: string;
children?: {
docs?:
| (
| {
relationTo?: 'files';
value: string | File;
}
| {
relationTo?: 'documents';
value: string | Document;
}
| {
relationTo?: 'folders';
value: string | Folder;
}
)[]
| null;
hasNextPage?: boolean | null;
} | null;
folder?: (string | null) | Folder;
updatedAt: string;
createdAt: string;
}
```
While you could instead have many join fields (for example
`childrenFolders`, `childrenFiles`) etc - this doesn't allow you to
sort/filter and paginate things across many collections, which isn't
trivial. With SQL we use `UNION ALL` query to achieve that.
---------
Co-authored-by: Dan Ribbens <dan.ribbens@gmail.com>
### What?
Previously, in postgres query like:
```ts
const result = await payload.find({
collection: 'blocks',
where: { 'blocks.director.name': { equals: 'Test Director' } },
})
```
where `blocks` is a blocks field, `director` is a relationship field and
`name` is a text field inside `directors`, failed with:
### Why?
The generated query before was a bit wrong.
Before:
```sql
select distinct
"blocks"."id",
"blocks"."created_at",
"blocks"."created_at"
from
"blocks"
left join "directors" "a5ad426a_eda4_4067_af7e_5b294d7f0968" on "a5ad426a_eda4_4067_af7e_5b294d7f0968"."id" = "blocks_blocks_some"."director_id"
left join "blocks_blocks_some" on "blocks"."id" = "blocks_blocks_some"."_parent_id"
where
"a5ad426a_eda4_4067_af7e_5b294d7f0968"."name" = 'Test Director'
order by
"blocks"."created_at" desc
limit
10
```
Notice `left join directors` _before_ join of `blocks_blocks_some`.
`blocks_blocks_some` doesn't exist yet, this PR changes so now we
generate
```sql
select distinct
"blocks"."id",
"blocks"."created_at",
"blocks"."created_at"
from
"blocks"
left join "blocks_blocks_some" on "blocks"."id" = "blocks_blocks_some"."_parent_id"
left join "directors" "a5ad426a_eda4_4067_af7e_5b294d7f0968" on "a5ad426a_eda4_4067_af7e_5b294d7f0968"."id" = "blocks_blocks_some"."director_id"
where
"a5ad426a_eda4_4067_af7e_5b294d7f0968"."name" = 'Test Director'
order by
"blocks"."created_at" desc
limit
10
```
Currently, the join field outputs to its result `hasNextPage: boolean`
and have the `limit` query parameter but lacks `page` which can be
useful. This PR adds it.
The `localized` properly was not stripped out of referenced block fields, if any parent was localized. For normal fields, this is done in sanitizeConfig. As the same referenced block config can be used in both a localized and non-localized config, we are not able to strip it out inside sanitizeConfig by modifying the block config.
Instead, this PR had to bring back tedious logic to handle it everywhere the `field.localized` property is accessed. For backwards-compatibility, we need to keep the existing sanitizeConfig logic. In 4.0, we should remove it to benefit from better test coverage of runtime field.localized handling - for now, this is done for our test suite using the `PAYLOAD_DO_NOT_SANITIZE_LOCALIZED_PROPERTY` flag.
The fix, added in https://github.com/payloadcms/payload/pull/11096
wasn't sufficient enough. It did handle the case when the same query
path / table was joined twice and caused incorrect `totalDocs`, but it
didn't handle the case when `JOIN` returns more than 1 rows, which 2
added new assertions here check.
Now, we use `COUNT(*)` only if we don't have any joined tables. If we
do, instead of using `SELECT (COUNT DISTINCT id)` which as described in
the previous PR is _very slow_ for large tables, we use the following
query:
```sql
SELECT COUNT(1) OVER() as count -- window function, executes for each row only once
FROM users
LEFT JOIN -- ... here additional rows are added
WHERE -- ...
GROUP BY users.id -- this ensures we're counting only users without additional rows from joins.
LIMIT 1 -- Since COUNT(1) OVER() executes and resolves before doing LIMIT, we can safely apply LIMIT 1.
```
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`.
Fixes https://github.com/payloadcms/payload/issues/10810
This was caused by using `COUNT(*)` aggregation instead of
`COUNT(DISTINCT table.id)`. However, we want to use `COUNT(*)` because
`COUNT(DISTINCT table.id)` is slow on large tables. Now we fallback to
`COUNT(DISTINCT table.id)` only when `COUNT(*)` cannot work properly.
Example of a query that leads to incorrect `totalDocs`:
```ts
const res = await payload.find({
collection: 'directors',
limit: 10,
where: {
or: [
{
movies: {
equals: movie2.id,
},
},
{
movies: {
equals: movie1.id,
},
},
{
movies: {
equals: movie1.id,
},
},
],
},
})
```
### What?
If you had multiple operator constraints on a single field, the last one
defined would be the only one used.
Example:
```ts
where: {
id: {
in: [doc2.id],
not_in: [], // <-- only respected this operator constraint
},
}
```
and
```ts
where: {
id: {
not_in: [],
in: [doc2.id], // <-- only respected this operator constraint
},
}
```
They would yield different results.
### Why?
The results were not merged into an `$and` query inside parseParams.
### How?
Merges the results within an `$and` constraint.
Fixes https://github.com/payloadcms/payload/issues/10944
Supersedes https://github.com/payloadcms/payload/pull/11011
Previously, data for globals was inconsistent across database adapters.
In Postgres, globals didn't store correct `createdAt`, `updatedAt`
fields and the `updateGlobal` lacked the `globalType` field. This PR
solves that without introducing schema changes.
Fixes https://github.com/payloadcms/payload/issues/10780
Previously, with enabled versions, nested select `hasMany: true` fields
weren't working with SQL database adapters. This was due to wrongly
passed `parent` to select rows data because we store arrays and blocks
in versions a bit differently, using both, `id` and `_uuid` (which
contains the normal Object ID) columns. And unlike with non versions
`_uuid` column isn't actually applicable here as it's not unique, thus
we need to save blocks/arrays first and then map their ObjectIDs to
generated by the database IDs and use them for select fields `parent`
data
- reduces unnecessary shallow copying within operations by removing
unnecessary spreads or .map()'s
- removes unnecessary `deleteMany` call in `deleteUserPreferences` for
auth-enabled collections
- replaces all instances of `validOperators.includes` with
`validOperatorMap[]`. O(n) => O(1)
- optimizes the `sanitizeInternalFields` function. Previously, it was
doing a **lot** of shallow copying
This PR modifies `tsconfig.base.json` by setting the following
strictness properties to true: `strict`, `noUncheckedIndexedAccess` and
`noImplicitOverride`.
In packages where compilation errors were observed, these settings were
opted out, and TODO comments were added to make it easier to track the
roadmap for converting everything to strict mode.
The following packages now have increased strictness, which prevents new
errors from being accidentally introduced:
- storage-vercel-blob
- storage-s3*
- storage-gcs
- plugin-sentry
- payload-cloud*
- email-resend*
- email-nodemailer*
*These packages already had `strict: true`, but now have
`noUncheckedIndexedAccess` and `noImplicitOverride`.
Note that this only affects the `/packages` folder, but not
`/templates`, `/test` or `/examples` which have a different `tsconfig`.
Based on https://github.com/payloadcms/payload/pull/10154
If the actual database schema is not changed (no new columns, enums,
indexes, tables) - skip calling Drizzle push. This, potentially can
significantly reduce overhead on reloads in development mode especially
when using remote databases.
If for whatever reason you need to preserve the current behavior you can
use `PAYLOAD_FORCE_DRIZZLE_PUSH=true` env flag.
Previously we had been downgrading rimraf to v3 simply to handle clean
with glob patterns across platforms. In v4 and newer of rimraf you can
add `-g` to use glob patterns.
This change updates rimraf and adds the flag to handle globs in our
package scripts to be windows compatible.
### What?
With Postgres, before join to self like:
```ts
import type { CollectionConfig } from 'payload'
export const SelfJoins: CollectionConfig = {
slug: 'self-joins',
fields: [
{
name: 'rel',
type: 'relationship',
relationTo: 'self-joins',
},
{
name: 'joins',
type: 'join',
on: 'rel',
collection: 'self-joins',
},
],
}
```
wasn't possible, even though it's a valid usage.
### How?
Now, to differentiate parent `self_joins` and children `self_joins` we
do additional alias for the nested select -
`"4d3cf2b6_1adf_46a8_b6d2_3e1c3809d737"`:
```sql
select
"id",
"rel_id",
"updated_at",
"created_at",
(
select
coalesce(
json_agg(
json_build_object('id', "joins_alias".id)
),
'[]' :: json
)
from
(
select
"created_at",
"rel_id",
"id"
from
"self_joins" "4d3cf2b6_1adf_46a8_b6d2_3e1c3809d737"
where
"4d3cf2b6_1adf_46a8_b6d2_3e1c3809d737"."rel_id" = "self_joins"."id"
order by
"4d3cf2b6_1adf_46a8_b6d2_3e1c3809d737"."created_at" desc
limit
$1
) "joins_alias"
) as "joins_alias"
from
"self_joins"
where
"self_joins"."id" = $2
order by
"self_joins"."created_at" desc
limit
$3
```
Fixes https://github.com/payloadcms/payload/issues/10144
-->
### What?
Exposes ability to enable
[AUTOINCREMENT](https://www.sqlite.org/autoinc.html) for Primary Keys
which ensures that the same ID cannot be reused from previously deleted
rows.
```ts
sqliteAdapter({
autoIncrement: true
})
```
### Why?
This may be essential for some systems. Enabled `autoIncrement: true`
also for the SQLite Adapter in our tests, which can be useful when
testing whether the doc was deleted or not when you also have other
create operations.
### How?
Uses Drizzle's `autoIncrement` option.
WARNING:
This cannot be enabled in an existing project without a custom
migration, as it completely changes how primary keys are stored in the
database.
* Avoids additional file system writes (1 for `await writeFile` and then
`npx prettier --write`) instead prettier now formats the javascript
string directly. Went from 650 MS to 250 MS for the prettify block.
* Disables database connection, since the `db.generateSchema` doesn't
need connection, this also disables Drizzle schema push.
* Properly exits the bin script process.
### What?
The join field had a limitation imposed that prevents it from targeting
polymorphic relationship fields. With this change we can support any
relationship fields.
### Why?
Improves the functionality of join field.
### How?
Extended the database adapters and removed the config sanitization that
would throw an error when polymorphic relationships were used.
Fixes #
This PR allows to have full type safety on `payload.drizzle` with a
single command
```sh
pnpm payload generate:db-schema
```
Which generates TypeScript code with Drizzle declarations based on the
current database schema.
Example of generated file with the website template:
https://gist.github.com/r1tsuu/b8687f211b51d9a3a7e78ba41e8fbf03
Video that shows the power:
https://github.com/user-attachments/assets/3ced958b-ec1d-49f5-9f51-d859d5fae236
We also now proxy drizzle package the same way we do for Lexical so you
don't have to install it (and you shouldn't because you may have version
mismatch).
Instead, you can import from Drizzle like this:
```ts
import {
pgTable,
index,
foreignKey,
integer,
text,
varchar,
jsonb,
boolean,
numeric,
serial,
timestamp,
uniqueIndex,
pgEnum,
} from '@payloadcms/db-postgres/drizzle/pg-core'
import { sql } from '@payloadcms/db-postgres/drizzle'
import { relations } from '@payloadcms/db-postgres/drizzle/relations'
```
Fixes https://github.com/payloadcms/payload/discussions/4318
In the future we can also support types generation for mongoose / raw
mongodb results.
### What?
Previously, the `req` argument:
In database operations (e.g `payload.db`) was required and you needed to
pass the whole `req` with all the properties. This is confusing because
in database operations we never use its properties outside of
`req.transactionID` and `req.t`, both of which should be optional as
well.
Now, you don't have to do that cast:
```ts
payload.db.findOne({
collection: 'posts',
req: {} as PayloadRequest,
where: {
id: {
equals: 1,
},
},
})
```
Becomes:
```ts
payload.db.findOne({
collection: 'posts',
where: {
id: {
equals: 1,
},
},
})
```
If you need to use transactions, you're not required to do the `as` cast
as well now, as the `req` not only optional but also partial -
`Partial<PayloadRequest>`.
`initTransaction`, `commitTransaction`, `killTransaction` utilities are
typed better now as well. They do not require to you pass all the
properties of `req`, but only `payload` -
`MarkRequired<Partial<PayloadRequest>, 'payload'>`
```ts
const req = { payload }
await initTransaction(req)
await payload.db.create({
collection: "posts",
data: {},
req
})
await commitTransaction(req)
```
The same for the Local API. Internal operations (for example
`packages/payload/src/collections/operations/find.ts`) still accept the
whole `req`, but local ones
(`packages/payload/src/collections/operations/local/find.ts`) which are
used through `payload.` now accept `Partial<PayloadRequest>`, as they
pass it through to internal operations with `createLocalReq`.
So now, this is also valid, while previously you had to do `as` cast for
`req`.
```ts
const req = { payload }
await initTransaction(req)
await payload.create({
collection: "posts",
data: {},
req
})
await commitTransaction(req)
```
Marked as deprecated `PayloadRequest['transactionIDPromise']` to remove
in the next major version. It was never used anywhere.
Refactored `withSession` that returns an object to `getSession` that
returns just `ClientSession`. Better type safety for arguments
Deduplicated in all drizzle operations to `getTransaction(this, req)`
utility:
```ts
const db = this.sessions[await req?.transactionID]?.db || this.drizzle
```
Added fallback for throwing unique validation errors in database
operations when `req.t` is not available.
In migration `up` and `down` functions our `req` is not partial, while
we used to passed `req` with only 2 properties - `payload` and
`transactionID`. This is misleading and you can't access for example
`req.t`.
Now, to achieve "real" full `req` - we generate it with `createLocalReq`
in all migration functions.
This all is backwards compatible. In all public API places where you
expect the full `req` (like hooks) you still have it.
### Why?
Better DX, more expected types, less errors because of types casting.
### What?
Querying by nested to rows fields in has many relationships like this:
```ts
const result = await payload.find({
collection: 'relationship-fields',
where: {
'relationToRowMany.title': { equals: 'some-title' },
},
})
```
Where the related collection:
```ts
const RowFields: CollectionConfig = {
slug: rowFieldsSlug,
fields: [
{
type: 'row',
fields: [
{
name: 'title',
label: 'Title within a row',
type: 'text',
required: true,
},
],
},
],
}
```
was broken
### Why?
We migrated to use `flattenedFields`, but not in this specific case.
This error would be caught earlier we used `noImplictAny` typescript
rule. https://www.typescriptlang.org/tsconfig/#noImplicitAny which
wouldn't allow us to create variable like this:
```ts
let relationshipFields // relationshipFields is any here
```
Instead, we should write:
```ts
let relationshipFields: FlattenedField[]
```
We should migrate to it and `strictNullChecks` as well.
Fixes https://github.com/payloadcms/payload/issues/9534
