npm install --save @use-gpu/state
yarn add @use-gpu/state
Docs: https://usegpu.live/docs/reference-live-@use-gpu-state
Manipulate JS state declaratively.
Helper library for doing patch
and diff
-based state management.
Can be used to drive undo/redo. Includes cursor hooks for React/Live.
import { patch, diff } from '@use-gpu/state';
// React
import { useUpdateState, useCursor } from '@use-gpu/state/react';
// Live
import { useUpdateState, useCursor } from '@use-gpu/state/live';
patch
will apply an update
to a nested object, without modifying the original.
const value = {hello: 'text', value: 2};
const update = {hello: 'world'};
expect(patch(value, update)).toEqual({hello: 'world', value: 2});
The default behavior is:
update
into value
recursively.To adjust the behavior, e.g. to replace an object instead of merging it, use the included $op
helpers:
e.g.
const value = {hello: {title: 'text', href: '#'}, value: 2};
const update = {hello: $set({title: 'world'});
expect(patch(values, update)).toEqual({ hello: { title: 'world' }, value: 2});
You can use $apply
to make custom patching ops, e.g. to append an item to a list:
const $push = <T>(item: T) => $apply((list: T[]) => [...list, item]);
const newList = patch(list, $push(item));
This can be used anywhere in a patch:
const newState = patch(state, {
nested: {
list: $push('hello'),
}
});
diff
is the complement to patch
.
Given two values A
and B
, it will return an update
so that:
const update = diff(A, B);
expect(patch(A, update)).toEqual(B);
...patching A with it equals the value B (though not the same object(s) as B).
If you diff(A, B)
after a patch(A, ...)
, you get a pure update, without $apply
or $patch
. This can be serialized to JSON.
Note that a diff may contain empty updates such as { }
if an object was cloned. Use getUpdateKeys
to check whether an update contains real changes.
To reverse an update, you can diff(B, A)
:
const B = patch(A, update);
const reversed = diff(B, A);
expect(patch(B, reversed)).toEqual(A);
To optimize and formalize this, revise
is (almost) the same operation.
It is a reverse complement to patch
. Given a value A
, and an update
, it will patch the update
so that:
// Don't need B to reverse
const reversed = revise(A, update);
const B = patch(A, update);
expect(patch(B, reversed)).toEqual(A);
...applying the reversed
update will reverse the original update
along the exact same boundaries.
This can be used to build an automatic undo/redo system that works with any $op
.
Use getUpdateKeys
to check whether a revised update is actually effectful or consists solely of $nop
.
Most UI state is simple, and consists of straight forward "set foo to bar" type actions. When this state lives inside an existing object, this requires a fair amount of boilerplate:
const [state, setState] = useState({
foo: {
// ...
size: 5,
},
// ...
});
const {foo: {size}} = state;
const setSize = (size: number) => {
setState((state) => {
...state,
foo: {
...state.foo,
size,
},
})
};
You can simplify this by relying on patch(…)
to do all your mutating. Instead of a setState(…)
, you now have an updateState(…)
:
import { useCursor, useUpdateState } from '@use-gpu/state/react';
// Create state pair, get cursor
const [state, updateState] = useUpdateState({...});
const stateCursor = useCursor([state, updateState]);
// More compact form
const stateCursor = useCursor(useUpdateState({...}));
Cursors can be traversed just like the original value:
const sizeCursor = stateCursor.foo.size;
To extract a getter/updater pair, call it:
const [size, updateSize] = sizeCursor();
This can be written as stateCursor.foo.size()
.
When you call updateSize(5)
, this is equivalent to updateState({foo: {size: 5}})
. The updater callbacks are auto-generated and all call the same central useUpdateState
.
This works as expected, because useUpdateState
will merge this change into the original state. The argument to updateState
is an Update
, i.e. the argument to patch
.
The merging behavior of an Update
can be precisely controlled, at the individual field level.
Cursors are also available in non-hook form via makeCursor
.
const DEFAULTS = {
foo: { size: 10 },
// ...
};
const stateCursor = useCursor(useUpdateState({...}), DEFAULTS);
useCursor(…)
can accept defaults as a 2nd argument. When it traverses the original value, if it encounters a missing field, it will fill in the one from the default non-destructively.
When it then applies an update, it will first patch in the right default values, and then make the change. This ensures clean partial patches of missing nested fields.
Cursors are immutable: if the value has changed, you get a new cursor instance.
useCursor
is memoized: if the value didn't change, you get the same cursor back.
Lookups cursor.foo.bar
are stable. It's safe to use a derived cursor directly as a hook dependency.
Cursors for unchanged values are stable (if the root updateState
hasn't changed).
i.e. Even if cursor
has changed, cursor.foo.bar
may be reused.
toHash
will hash any JS value to a 10-digit base 64 string.toMurmur53
will hash any JS value to a 53-bit number
.getObjectKey
assigns a unique, incrementing 53-bit ID to each unique object (uses a WeakMap
).makeKey
returns a new unique ID from the same set.Made by Steven Wittens. Part of @use-gpu
.