Currently, we only use native `structuredClone` in the browser, falling
back to `JSON.parse(JSON.stringify(...))` elsewhere, despite Node
supporting `structuredClone` [since
v17](https://developer.mozilla.org/en-US/docs/Web/API/structuredClone)
and Cloudflare Workers supporting it [since
2022](https://blog.cloudflare.com/standards-compliant-workers-api/).
This PR adjusts our shim to use the native `structuredClone` on all
platforms, if available.
Additionally, `jsdom` doesn't implement `structuredClone`, a bug [open
since 2022](https://github.com/jsdom/jsdom/issues/3363). This PR patches
`jsdom` environment in all packages/apps that use it for tests.
Also includes a driveby removal of `deepCopy`, a function that is
strictly inferior to `structuredClone`.
### Change Type
<!-- ❗ Please select a 'Scope' label ❗️ -->
- [x] `sdk` — Changes the tldraw SDK
- [x] `dotcom` — Changes the tldraw.com web app
- [ ] `docs` — Changes to the documentation, examples, or templates.
- [ ] `vs code` — Changes to the vscode plugin
- [ ] `internal` — Does not affect user-facing stuff
<!-- ❗ Please select a 'Type' label ❗️ -->
- [ ] `bugfix` — Bug fix
- [ ] `feature` — New feature
- [x] `improvement` — Improving existing features
- [x] `chore` — Updating dependencies, other boring stuff
- [ ] `galaxy brain` — Architectural changes
- [ ] `tests` — Changes to any test code
- [ ] `tools` — Changes to infrastructure, CI, internal scripts,
debugging tools, etc.
- [ ] `dunno` — I don't know
### Test Plan
1. A smoke test would be enough
- [ ] Unit Tests
- [x] End to end tests
Before
After
The probelm was manifesting because our camera updates were not
throttled and our render tick was on a different tick timeline to our
tick manager. Fixing the latter gets rid of the lag without requiring us
to throttle the camera updates.
### Change Type
<!-- ❗ Please select a 'Scope' label ❗️ -->
- [x] `sdk` — Changes the tldraw SDK
- [ ] `dotcom` — Changes the tldraw.com web app
- [ ] `docs` — Changes to the documentation, examples, or templates.
- [ ] `vs code` — Changes to the vscode plugin
- [ ] `internal` — Does not affect user-facing stuff
<!-- ❗ Please select a 'Type' label ❗️ -->
- [x] `bugfix` — Bug fix
### Test Plan
1. Add a step-by-step description of how to test your PR here.
2.
- [ ] Unit Tests
- [ ] End to end tests
### Release Notes
- Add a brief release note for your PR here.
---------
Co-authored-by: Steve Ruiz <steveruizok@gmail.com>
We had a couple regressions in #3110: first a missing `await` was
causing fonts not to get properly embedded in exports. second, some
`readAsText` calls were replaced with `readAsDataURL` calls.
### Change Type
<!-- ❗ Please select a 'Scope' label ❗️ -->
- [x] `sdk` — Changes the tldraw SDK
- [ ] `dotcom` — Changes the tldraw.com web app
- [ ] `docs` — Changes to the documentation, examples, or templates.
- [ ] `vs code` — Changes to the vscode plugin
- [ ] `internal` — Does not affect user-facing stuff
<!-- ❗ Please select a 'Type' label ❗️ -->
- [x] `bugfix` — Bug fix
- [ ] `feature` — New feature
- [ ] `improvement` — Improving existing features
- [ ] `chore` — Updating dependencies, other boring stuff
- [ ] `galaxy brain` — Architectural changes
- [ ] `tests` — Changes to any test code
- [ ] `tools` — Changes to infrastructure, CI, internal scripts,
debugging tools, etc.
- [ ] `dunno` — I don't know
This PR does a few things to help with performance:
1. Instead of doing changes on raf we now do them 60 times per second.
This limits the number of updates on high refresh rate screens like the
iPad. With the current code this only applied to the history updates (so
when you subscribed to the updates), but the next point takes this a bit
futher.
2. We now trigger react updates 60 times per second. This is a change in
`useValue` and `useStateTracking` hooks.
3. We now throttle the inputs (like the `pointerMove`) in state nodes.
This means we batch multiple inputs and only apply them at most 60 times
per second.
We had to adjust our own tests to pass after this change so I marked
this as major as it might require the users of the library to do the
same.
Few observations:
- The browser calls the raf callbacks when it can. If it gets
overwhelmed it will call them further and further apart. As things call
down it will start calling them more frequently again. You can clearly
see this in the drawing example. When fps gets to a certain level we
start to get fewer updates, then fps can recover a bit. This makes the
experience quite janky. The updates can be kinda ok one second (dropping
frames, but consistently) and then they can completely stop and you have
to let go of the mouse to make them happen again. With the new logic it
seems everything is a lot more consistent.
- We might look into variable refresh rates to prevent this overtaxing
of the browser. Like when we see that the times between our updates are
getting higher we could make the updates less frequent. If we then see
that they are happening more often we could ramp them back up. I had an
[experiment for this
here](4834863966 (diff-318e71563d7c47173f89ec084ca44417cf70fc72faac85b96f48b856a8aec466L30-L35)).
Few tests below. Used 6x slowdown for these.
# Resizing
### Before
https://github.com/tldraw/tldraw/assets/2523721/798a033f-5dfa-419e-9a2d-fd8908272ba0
### After
https://github.com/tldraw/tldraw/assets/2523721/45870a0c-c310-4be0-b63c-6c92c20ca037
# Drawing
Comparison is not 100% fair, we don't store the intermediate inputs
right now. That said, tick should still only produce once update so I do
think we can get a sense of the differences.
### Before
https://github.com/tldraw/tldraw/assets/2523721/2e8ac8c5-bbdf-484b-bb0c-70c967f4541c
### After
https://github.com/tldraw/tldraw/assets/2523721/8f54b7a8-9a0e-4a39-b168-482caceb0149
### Change Type
- [ ] `patch` — Bug fix
- [ ] `minor` — New feature
- [x] `major` — Breaking change
- [ ] `dependencies` — Changes to package dependencies[^1]
- [ ] `documentation` — Changes to the documentation only[^2]
- [ ] `tests` — Changes to any test code only[^2]
- [ ] `internal` — Any other changes that don't affect the published
package[^2]
- [ ] I don't know
[^1]: publishes a `patch` release, for devDependencies use `internal`
[^2]: will not publish a new version
### Release Notes
- Improves the performance of rendering.
---------
Co-authored-by: Steve Ruiz <steveruizok@gmail.com>
Steve tried this in #3043, but we reverted it in #3063. Steve's version
added `JSON.parse`/`JSON.stringify` to the helpers without checking for
where we were already `JSON.parse`ing (or not). In some places we just
store strings directly rather than wanting them jsonified, so in this
version we leave the jsonification to the callers - the helpers just do
the reading/writing and return the string values.
### Change Type
- [x] `patch` — Bug fix
This PR provides some safe wrappers for local storage calls. Local
storage is not available in all environments (for example, a React
Native web view). The PR also adds an eslint rule preventing direct
calls to local / session storage.
### Change Type
- [x] `patch` — Bug fix
### Release Notes
- Fixes a bug that could cause crashes in React Native webviews.
This PR fixes a rogue structuredClone reference in the drawing state.
### Change Type
- [x] `patch` — Bug fix
### Release Notes
- Fixes a reference to structuredClone that caused a crash on older
browsers.
Currently, we type our fractional index keys as `string` and don't have
any validation for them. I'm touching some of this code for my work on
line handles and wanted to change that:
- fractional indexes are now `IndexKey`s, not `string`s. `IndexKey`s
have a brand property so can't be used interchangeably with strings
(like our IDs)
- There's a new `T.indexKey` validator which we can use in our
validations to make sure we don't end up with nonsense keys.
This PR is part of a series - please don't merge it until the things
before it have landed!
1. #2827 (you are here)
2. #2831
3. #2793
4. #2841
5. #2845
### Change Type
- [x] `patch` — Bug fix
### Test Plan
1. Mostly relying on unit & end to end tests here - no user facing
changes.
- [x] Unit Tests
Taking the opportunity for some last-minute low-consequence breaking
changes before 2.0, this diff does some prep work for adding a new
snapping API by making the distinction between the two types of snapping
a bit clearer and cleaning up some naming.
- `SnapManager` has had most of the actual snapping logic moved into two
properties: `shapeBounds` (for snapping shape bounds on translate and
resize) and `handles` (for snapping handles)
- `SnapLine`s are renamed to `SnapIndicator`s. The 'line' name was a bit
confusing because not all of these indicators are lines (the new vertex
snap type will be a single point)
I'm not too worried about this being a breaking change as it touches an
area of the API that I'd be very surprised if more than a couple of
people were using.
### Change Type
- [x] `major` — Breaking change
### Test Plan
- No user-facing changes.
### Release Notes
- `SnapLine`s are now called `SnapIndicator`s
- Snapping methods moved from `editor.snaps` to
`editor.snaps.shapeBounds` and `editor.snaps.handles` depending on the
type of snapping you're trying to do.
(gif is with 6x CPU throttling to make the effect more visible)
This is the first of a few diffs I'm working on to make dropping images
onto the canvas feel a lot faster.
There are three main changes here:
1. We operate on `Blob`s and `File`s rather than data urls. This saves a
fair bit on converting to/from base64 all the time. I've updated our
`MediaHelper` APIs to encourage the same in consumers.
2. We only check the max canvas size (slow) if images are above a
certain dimension that we consider "safe" (8k x 8k)
3. Switching from the `downscale` npm library to canvas native
downscaling. that library claims to give better results than the
browser, but hasn't been updated in ~7 years. in modern browsers, we can
opt-in to native high-quality image smoothing to achieve similar results
much faster than with an algorithm implemented in pure JS.
I want to follow this up with a system to show image placeholders whilst
we're waiting for long-running operations like resizing etc but i'm
going to split that out into its own diff as it'll involve some fairly
complex changes to the history management API.
### Change Type
- [x] `major` — Breaking change
[^1]: publishes a `patch` release, for devDependencies use `internal`
[^2]: will not publish a new version
### Test Plan
1. Tested manually, unit tests & end-to-end tests pass
Adds validation for urls we use for our shapes and assets. This PR
includes a migration so we should check that existing rooms still load
correctly. There might be some that won't, but that means that they had
invalid url set.
### Change Type
- [x] `patch` — Bug fix
- [ ] `minor` — New feature
- [ ] `major` — Breaking change
- [ ] `dependencies` — Changes to package dependencies[^1]
- [ ] `documentation` — Changes to the documentation only[^2]
- [ ] `tests` — Changes to any test code only[^2]
- [ ] `internal` — Any other changes that don't affect the published
package[^2]
- [ ] I don't know
[^1]: publishes a `patch` release, for devDependencies use `internal`
[^2]: will not publish a new version
### Test Plan
1. Existing rooms should still load correctly (there should be no
validation errors).
2. Adding new images and videos should also work (test both local and
multiplayer rooms as they handle assets differently).
- [x] Unit Tests
- [ ] End to end tests
### Release Notes
- Add validation to urls.
---------
Co-authored-by: alex <alex@dytry.ch>
This is an attempt at #1989. The big issue there is when `shapeUtils`
change when you're relying on tldraw to provide you with the store
instead of providing your own. Our `useTLStore` component had a bug
where it would rely on effects & a ref to detect when its options had
changed whilst still scheduling updates. Fresh opts would come in, but
they'd be different from the ones in the ref, so we'd schedule an
update, so the opts would come in again, but they'd still be different
as we hadn't run effects yet, and we'd schedule an update again (and so
on).
This diff fixes that by storing the previous opts in state instead of a
ref, so they're updating in lockstep with the store itself. this
prevents the update loop.
There are still situations where we can get into loops if the developer
is passing in custom tools, shapeUtils, or components but not memoising
them or defining them outside of react. As a DX improvement, we do some
auto-memoisation of these values using shallow equality to help with
this issue.
### Change Type
- [x] `patch` — Bug fix
### Test Plan
- [x] Unit Tests
### Release Notes
- Fixes a bug that could cause crashes due to a re-render loop with HMR
#1989
follow up to #2189
adds runtime warnings for deprecated fields. cleans up remaining fields
and usages. Adds a lint rule to prevent access to deprecated fields.
Adds a lint rule to prevent using getters.
### Change Type
- [x] `patch` — Bug fix
This PR moves code between our packages so that:
- @tldraw/editor is a “core” library with the engine and canvas but no
shapes, tools, or other things
- @tldraw/tldraw contains everything particular to the experience we’ve
built for tldraw
At first look, this might seem like a step away from customization and
configuration, however I believe it greatly increases the configuration
potential of the @tldraw/editor while also providing a more accurate
reflection of what configuration options actually exist for
@tldraw/tldraw.
## Library changes
@tldraw/editor re-exports its dependencies and @tldraw/tldraw re-exports
@tldraw/editor.
- users of @tldraw/editor WITHOUT @tldraw/tldraw should almost always
only import things from @tldraw/editor.
- users of @tldraw/tldraw should almost always only import things from
@tldraw/tldraw.
- @tldraw/polyfills is merged into @tldraw/editor
- @tldraw/indices is merged into @tldraw/editor
- @tldraw/primitives is merged mostly into @tldraw/editor, partially
into @tldraw/tldraw
- @tldraw/file-format is merged into @tldraw/tldraw
- @tldraw/ui is merged into @tldraw/tldraw
Many (many) utils and other code is moved from the editor to tldraw. For
example, embeds now are entirely an feature of @tldraw/tldraw. The only
big chunk of code left in core is related to arrow handling.
## API Changes
The editor can now be used without tldraw's assets. We load them in
@tldraw/tldraw instead, so feel free to use whatever fonts or images or
whatever that you like with the editor.
All tools and shapes (except for the `Group` shape) are moved to
@tldraw/tldraw. This includes the `select` tool.
You should use the editor with at least one tool, however, so you now
also need to send in an `initialState` prop to the Editor /
<TldrawEditor> component indicating which state the editor should begin
in.
The `components` prop now also accepts `SelectionForeground`.
The complex selection component that we use for tldraw is moved to
@tldraw/tldraw. The default component is quite basic but can easily be
replaced via the `components` prop. We pass down our tldraw-flavored
SelectionFg via `components`.
Likewise with the `Scribble` component: the `DefaultScribble` no longer
uses our freehand tech and is a simple path instead. We pass down the
tldraw-flavored scribble via `components`.
The `ExternalContentManager` (`Editor.externalContentManager`) is
removed and replaced with a mapping of types to handlers.
- Register new content handlers with
`Editor.registerExternalContentHandler`.
- Register new asset creation handlers (for files and URLs) with
`Editor.registerExternalAssetHandler`
### Change Type
- [x] `major` — Breaking change
### Test Plan
- [x] Unit Tests
- [x] End to end tests
### Release Notes
- [@tldraw/editor] lots, wip
- [@tldraw/ui] gone, merged to tldraw/tldraw
- [@tldraw/polyfills] gone, merged to tldraw/editor
- [@tldraw/primitives] gone, merged to tldraw/editor / tldraw/tldraw
- [@tldraw/indices] gone, merged to tldraw/editor
- [@tldraw/file-format] gone, merged to tldraw/tldraw
---------
Co-authored-by: alex <alex@dytry.ch>
This PR adds a `meta` property to shapes and other records.
It adds it to:
- asset
- camera
- document
- instance
- instancePageState
- instancePresence
- page
- pointer
- rootShape
## Setting meta
This data can generally be added wherever you would normally update the
corresponding record.
An exception exists for shapes, which can be updated using a partial of
the `meta` in the same way that we update shapes with a partial of
`props`.
```ts
this.updateShapes([{
id: myShape.id,
type: "geo",
meta: {
nemesis: "steve",
special: true
}
])
```
## `Editor.getInitialMetaForShape`
The `Editor.getInitialMetaForShape` method is kind of a hack to set the
initial meta property for newly created shapes. You can set it
externally. Escape hatch!
### Change Type
- [x] `minor` — New feature
### Test Plan
todo
- [ ] Unit Tests (todo)
### Release Notes
- todo
Removes `propsForNextShape` and replaces it with the new styles API.
Changes in here:
- New custom style example
- `setProp` is now `setStyle` and takes a `StyleProp` instead of a
string
- `Editor.props` and `Editor.opacity` are now `Editor.sharedStyles` and
`Editor.sharedOpacity`
- They return an object that flags mixed vs shared types instead of
using null to signal mixed types
- `Editor.styles` returns a `SharedStyleMap` - keyed on `StyleProp`
instead of `string`
- `StateNode.shapeType` is now the shape util rather than just a string.
This lets us pull the styles from the shape type directly.
- `color` is no longer a core part of the editor set on the shape
parent. Individual child shapes have to use color directly.
- `propsForNextShape` is now `stylesForNextShape`
- `InstanceRecordType` is created at runtime in the same way
`ShapeRecordType` is. This is so it can pull style validators out of
shape defs for `stylesForNextShape`
- Shape type are now defined by their props rather than having separate
validators & type defs
### Change Type
- [x] `major` — Breaking change
### Test Plan
1. Big time regression testing around styles!
2. Check UI works as intended for all shape/style/tool combos
- [x] Unit Tests
- [ ] End to end tests
### Release Notes
-
---------
Co-authored-by: Steve Ruiz <steveruizok@gmail.com>
Based on #1549, but with a lot of code-structure related changes backed
out. Shape schemas are still defined in tlschemas with this diff.
Couple differences between this and #1549:
- This tightens up the relationship between store schemas and editor
schemas a bit
- Reduces the number of places we need to remember to include core
shapes
- Only `<TLdrawEditor />` sets default shapes by default. If you're
doing something funky with lower-level APIs, you need to specify
`defaultShapes` manually
- Replaces `validator` with `props` for shapes
### Change Type
- [x] `major` — Breaking Change
### Test Plan
1. Add a step-by-step description of how to test your PR here.
2.
- [x] Unit Tests
- [ ] Webdriver tests
### Release Notes
[dev-facing, notes to come]
This PR does the following:
- Add `selfHosted.js`, which is a great option for users that wish to
self host the assets. Works well for both self hosting from the public
folder or via a CDN.
- Updates the docs for assets. We now have a dedicated page for assets
where all the options are more clearly explained. I also removed the
assets explanation from the main docs as the unpkg option should work
out of the box and setting up the assets is no longer necessary.
- Cleaned up the `refresh-assets` script. We now use common `types.d.ts`
file to define our types. All the other options then reuse them.
- Pulled out the `formatAssetUrl` into it's own file. It's now static an
no longer generated.
- `urls.d.ts`, `import.d.ts`, and newly added `selfhosted.d.ts` are now
also no longer generated as we can import the types from `types.d.ts`.
- You can now pass a subset of `assetUrls` to `<Tldraw />` and it will
override the default option with the passed in overrides. This makes it
easy to only customizes certain assets (only change the draw font as an
example).
### Change Type
- [x] `patch` — Bug Fix
We noticed that when inferring the type of a shape from its ID, it was
getting inferred as `any` which was hiding some issues. This diff
switches `BaseRecord`'s automatic ID to an explicit one, which lets us
pass in our correct `TLShapeId` definition and still have it play nicely
with other places.
### Change Type
- [x] `patch` — Bug Fix
### Release Notes
[internal only, covered by #1432 changelog]
---------
Co-authored-by: Steve Ruiz <steveruizok@gmail.com>
This PR moves our "fractical indices" library into its own package.
- [x] `major` — Breaking Change
### Release Notes
- [@tldraw/editor] Remove fractional indices code into `@tldraw/indices`
- [@tldraw/indices] Create library for fractional indices code
This fixes a bug where creating a page would fail if there were multiple
pages with the same index.
This also changes the store to use a throttled version of
requestAnimationFrame. This should be good for relieving backpressure in
situations where the store is updated many times in quick succession. It
also makes testing a lot easier since it has the mocking logic built in.
### Change Type
- [x] `patch` — Bug Fix
### Release Notes
- Fix a bug where creating a page could throw an error in some
multiplayer contexts.
This PR adds
- A new `TLInstancePresence` record type, to collect info about the
presence state in a particular instance of the editor. This will
eventually be used to sync presence data instead of sending
instance-only state across the wire.
- **Record Scopes**
`RecordType` now has a `scope` property which can be one of three
things:
- `document`: the record belongs to the document and should be synced
and persisted freely. Currently: `TLDocument`, `TLPage`, `TLShape`, and
`TLAsset`
- `instance`: the record belongs to a single instance of the store and
should not be synced at all. It should not be persisted directly in most
cases, but rather compiled into a kind of 'instance configuration' to
store alongside the local document data so that when reopening the
associated document it can remember some of the previous instance state.
Currently: `TLInstance`, `TLInstancePageState`, `TLCamera`, `TLUser`,
`TLUserDocument`, `TLUserPresence`
- `presence`: the record belongs to a single instance of the store and
should not be persisted, but may be synced using the special presence
sync protocol. Currently just `TLInstancePresence`
This sets us up for the following changes, which are gonna be pretty
high-impact in terms of integrating tldraw into existing systems:
- Removing `instanceId` as a config option. Each instance gets a
randomly generated ID.
- We'd replace it with an `instanceConfig` option that has stuff like
selectedIds, camera positions, and so on. Then it's up to library users
to get and reinstate the instance config at persistence boundaries.
- Removing `userId` as config option, and removing the `TLUser` type
altogether.
- We might need to revisit when doing auth-enabled features like locking
shapes, but I suspect that will be separate.
