TinyBase logoTinyBase

Drawing

In this demo, we build a more complex drawing app, using many of the features of TinyBase together.

Note that this demo does not work particularly well on mobile devices, given its use of drag and drop. Preferably try it out on a desktop browser.

Boilerplate

First, we pull in React, ReactDOM, and TinyBase:

<script src="/umd/react.production.min.js"></script>
<script src="/umd/react-dom.production.min.js"></script>
<script src="/umd/tinybase/index.js"></script>
<script src="/umd/tinybase/persisters/persister-browser/index.js"></script>
<script src="/umd/tinybase/ui-react/index.js"></script>
<script src="/umd/tinybase/ui-react-inspector/index.js"></script>

We'll use a good selection of the TinyBase API and the ui-react module:

const {createCheckpoints, createRelationships, createStore} = TinyBase;
const {
  LinkedRowsView,
  Provider,
  useAddRowCallback,
  useCell,
  useCreateCheckpoints,
  useCreatePersister,
  useCreateRelationships,
  useCreateStore,
  useLinkedRowIds,
  useLocalRowIds,
  useRedoInformation,
  useRemoteRowId,
  useRow,
  useRowListener,
  useSetCellCallback,
  useSetCheckpointCallback,
  useSetPartialRowCallback,
  useStore,
  useUndoInformation,
} = TinyBaseUiReact;
const {
  createContext,
  useCallback,
  useContext,
  useEffect,
  useLayoutEffect,
  useRef,
  useState,
} = React;
const {createLocalPersister} = TinyBasePersisterBrowser;
const {Inspector} = TinyBaseUiReactInspector;

The drawing app relies heavily on being able to drag and drop shapes and their resizing handles. We'll create a React hook called useDraggableObject that returns a ref that can be used to attach this behavior to component. Nothing about this is TinyBase-specific, so let's get it out of the way first.

const useDraggableObject = (
  getInitial,
  onDrag,
  onDragStart = null,
  onDragStop = null,
) => {
  const [start, setStart] = useState();

  const handleMouseDown = useCallback(
    (event) => {
      onDragStart?.();
      setStart({
        x: event.clientX,
        y: event.clientY,
        initial: getInitial(),
      });
      event.stopPropagation();
    },
    [getInitial, onDragStart],
  );
  const handleMouseMove = useCallback(
    (event) => {
      if (start != null) {
        onDrag({
          dx: event.clientX - start.x,
          dy: event.clientY - start.y,
          initial: start.initial,
        });
      }
      event.stopPropagation();
    },
    [onDrag, start],
  );
  const handleMouseUp = useCallback(
    (event) => {
      setStart(null);
      onDragStop?.();
      event.stopPropagation();
    },
    [onDragStop],
  );

  const ref = useRef(null);
  useLayoutEffect(() => {
    const {current} = ref;
    current.addEventListener('mousedown', handleMouseDown);
    return () => current.removeEventListener('mousedown', handleMouseDown);
  }, [ref, handleMouseDown]);
  useEffect(() => {
    if (start != null) {
      addEventListener('mousemove', handleMouseMove);
      addEventListener('mouseup', handleMouseUp);
      return () => {
        removeEventListener('mousemove', handleMouseMove);
        removeEventListener('mouseup', handleMouseUp);
      };
    }
  }, [start, handleMouseMove, handleMouseUp]);

  return ref;
};

Initialization

We have a selection of constants we will use throughout the app. The shapes will live in a TinyBase table called shapes, the background canvas itself will always be shape 0, and shapes have a minimum size and one of two valid types:

const SHAPES = 'shapes';
const CANVAS_ID = '0';
const MIN_WIDTH = 50;
const MIN_HEIGHT = 30;
const TYPES = ['rectangle', 'ellipse'];

We will use mutator listeners to ensure that the type and color of the shapes are always valid if present. These are the two functions to do that:

const constrainType = (store, tableId, rowId, cellId, type) => {
  if (type != null && !TYPES.includes(type)) {
    store.setCell(tableId, rowId, cellId, TYPES[0]);
  }
};

const constrainColor = (store, tableId, rowId, cellId, color) => {
  if (color != null && !/^#[a-f\d]{6}$/.test(color)) {
    store.setCell(tableId, rowId, cellId, '#000000');
  }
};

At any given time, either a single shape is selected, or none are. We will set a context for the whole app that will provide the useState value and setter pair for the selected Id across the whole app:

const SelectedIdContext = createContext([null, () => {}]);
const useSelectedIdState = () => useContext(SelectedIdContext);

As for the application itself, we start off by initializing the store in the useCreateStore hook (so that it is memoized across renders), and immediately set its schema. Every shape has two pairs of coordinates, text, a type, colors, and a reference to the 'next' shape so they can be ordered in the z-index with a linked list. We also use the two mutator listeners to programmatically guarantee that types and colors are valid:

const App = () => {
  const store = useCreateStore(() => {
    const store = createStore().setTablesSchema({
      [SHAPES]: {
        x1: {type: 'number', default: 100},
        y1: {type: 'number', default: 100},
        x2: {type: 'number', default: 300},
        y2: {type: 'number', default: 200},
        text: {type: 'string', default: 'text'},
        type: {type: 'string'},
        backColor: {type: 'string', default: '#0077aa'},
        textColor: {type: 'string', default: '#ffffff'},
        nextId: {type: 'string'},
      },
    });
    store.addCellListener(SHAPES, null, 'type', constrainType, true);
    store.addCellListener(SHAPES, null, 'backColor', constrainColor, true);
    store.addCellListener(SHAPES, null, 'textColor', constrainColor, true);
    return store;
  }, []);
  // ...

We want an undo stack, so we create a Checkpoints objects for this store, again memoized with the useCreateCheckpoints hook. We also persist the shapes data to local browser storage, and default it to a single shape on the canvas (which is the start of the linked list of all the shapes, modelled with the Relationships object):

// ...
const checkpoints = useCreateCheckpoints(store, createCheckpoints);

useCreatePersister(
  store,
  (store) => createLocalPersister(store, 'drawing/store'),
  [],
  async (persister) => {
    await persister.startAutoLoad([
      {
        shapes: {
          [CANVAS_ID]: {x1: 0, y1: 0, nextId: '1', text: '[canvas]'},
          1: {},
        },
      },
    ]);
    checkpoints?.clear();
    await persister.startAutoSave();
  },
  [checkpoints],
);

const relationships = useCreateRelationships(store, (store) =>
  createRelationships(store).setRelationshipDefinition(
    'order',
    SHAPES,
    SHAPES,
    'nextId',
  ),
);
// ...

Finally we render the app, comprising a toolbar, the canvas, and the sidebar, all wrapped in a Provider component to make sure our top-level objects are defaulted throughout the app. The SelectedIdContext context provider also passes a useState value and setter pair into the app:

  // ...
  return (
    <Provider
      store={store}
      relationships={relationships}
      checkpoints={checkpoints}
    >
      <SelectedIdContext.Provider value={useState()}>
        <Toolbar />
        <Canvas />
        <Sidebar />
      </SelectedIdContext.Provider>
      <Inspector />
    </Provider>
  );
};

We also added the Inspector component at the end there so you can inspect what is going on with the data during this demo. Simply click the TinyBase logo in the corner.

Anyway, let's mount it into the DOM...

addEventListener('load', () =>
  ReactDOM.createRoot(document.body).render(<App />),
);

...and off we go.

The Toolbar Component

The toolbar across the top of the app contains the undo and redo buttons, the button to add a new shape - and, when a shape is selected, the stack order and delete buttons

const Toolbar = () => {
  const [useSelectedId] = useSelectedIdState();
  return (
    <div id="toolbar">
      <UndoRedo />
      <ShapeAdd />
      {useSelectedId == null ? null : (
        <>
          <ShapeOrder />
          <ShapeDelete />
        </>
      )}
    </div>
  );
};

The undo and redo buttons use the useUndoInformation hook and useRedoInformation hook to disable or enable themselves, and handle the checkpoint moves accordingly:

const UndoRedo = () => {
  const [canUndo, handleUndo, , undoLabel] = useUndoInformation();
  const [canRedo, handleRedo, , redoLabel] = useRedoInformation();

  return (
    <>
      <div
        className={`button undo${canUndo ? '' : ' disabled'}`}
        {...(canUndo ? {onClick: handleUndo, title: `Undo ${undoLabel}`} : {})}
      />
      <div
        className={`button redo${canRedo ? '' : ' disabled'}`}
        {...(canRedo ? {onClick: handleRedo, title: `Redo ${redoLabel}`} : {})}
      />
    </>
  );
};

The button to add a new shape needs to know the current top of the z-index stack (the frontId) upon which a new shape will be added, implemented by adding the new shape's Id as the nextId Cell of the current top shape. It also selects the new shape with the setSelectedId function. The useSetCheckpointCallback hook is used to add the shape creation to the undo stack after the new shape Row has been added via the callback from the useAddRowCallback hook:

const ShapeAdd = () => {
  const frontId = useFrontId();
  const [, setSelectedId] = useSelectedIdState();
  const setCheckpoint = useSetCheckpointCallback(() => 'add shape', []);
  const onAddRow = useCallback(
    (id, store) => {
      store.setCell(SHAPES, frontId, 'nextId', id);
      setSelectedId(id);
      setCheckpoint();
    },
    [frontId, setSelectedId, setCheckpoint],
  );
  const handleClick = useAddRowCallback(
    SHAPES,
    () => ({}),
    [],
    null,
    onAddRow,
    [onAddRow],
  );
  return (
    <div className="button add" onClick={handleClick}>
      Add shape
    </div>
  );
};

We'll have these two hooks for getting the front and back shapes of the ordered stack.

const useBackId = () => useLinkedRowIds('order', CANVAS_ID)[1];
const useFrontId = () => useLinkedRowIds('order', CANVAS_ID).slice(-1)[0];

There are four buttons for changing the shape order: move to back, move backward, move forward, and move to front. Each of these change the order of the linked list created by the nextId pointer in the shapes Table:

const ShapeOrder = () => {
  const [selectedId] = useSelectedIdState();
  const frontId = useFrontId();
  const forwardId = useRemoteRowId('order', selectedId);
  const [previousId] = useLocalRowIds('order', selectedId);
  const [backwardId] = useLocalRowIds('order', previousId);
  const backId = useBackId();
  return [
    ['front', 'To front', frontId, useOrderShape(frontId, 'to front')],
    ['forward', 'Forward', frontId, useOrderShape(forwardId, 'forward')],
    ['backward', 'Backward', backId, useOrderShape(backwardId, 'backward')],
    ['back', 'To back', backId, useOrderShape(CANVAS_ID, 'to back')],
  ].map(([className, label, disabledIfId, handleClick]) => {
    const disabled = selectedId == null || selectedId == disabledIfId;
    return (
      <div
        className={`button ${className} ${disabled ? ' disabled' : ''}`}
        onClick={disabled ? null : handleClick}
        key={className}
      >
        {label}
      </div>
    );
  });
};

These changes are made with the useOrderShape hook. It batches the changes to the nextId Cell values in a single transaction, accounting for edge cases like moving a shape from the top of the stack, and sealing up the linked list after a shape has been moved:

const useOrderShape = (toId, label) => {
  const store = useStore();
  const [selectedId] = useSelectedIdState();
  const [previousId] = useLocalRowIds('order', selectedId);
  const nextId = useRemoteRowId('order', selectedId);
  const nextNextId = useRemoteRowId('order', toId);
  const setCheckpoint = useSetCheckpointCallback(() => `move ${label}`, []);
  return useCallback(() => {
    store.transaction(() => {
      if (nextId != null) {
        store.setCell(SHAPES, previousId, 'nextId', nextId);
      } else {
        store.delCell(SHAPES, previousId, 'nextId');
      }
      if (nextNextId != null) {
        store.setCell(SHAPES, selectedId, 'nextId', nextNextId);
      } else {
        store.delCell(SHAPES, selectedId, 'nextId');
      }
      store.setCell(SHAPES, toId, 'nextId', selectedId);
    });
    setCheckpoint();
  }, [selectedId, toId, store, previousId, nextId, nextNextId, setCheckpoint]);
};

The button to delete a shape needs to account for a shape being removed from the linked list (and making its next shape the previous shape's next shape instead), but otherwise it simply uses the delRow method to remove the record from the table.

const ShapeDelete = () => {
  const store = useStore();
  const [selectedId, setSelectedId] = useSelectedIdState();
  const [previousId] = useLocalRowIds('order', selectedId);
  const nextId = useRemoteRowId('order', selectedId);
  const setCheckpoint = useSetCheckpointCallback(() => 'delete', []);
  const handleClick = useCallback(() => {
    store.transaction(() => {
      if (nextId == null) {
        store.delCell(SHAPES, previousId, 'nextId');
      } else {
        store.setCell(SHAPES, previousId, 'nextId', nextId);
      }
      store.delRow(SHAPES, selectedId);
    });
    setCheckpoint();
    setSelectedId();
  }, [store, selectedId, setSelectedId, previousId, nextId, setCheckpoint]);
  return (
    <div className="button delete" onClick={handleClick}>
      Delete
    </div>
  );
};

That's the top toolbar taken care of. Try selecting, unselecting, and moving a few shapes around to get a sense for what is happening with each of these components. You can observe the resulting data structures by inspecting your local storage with your browser developer tools.

The Sidebar Component

The sidebar comprises controls to configure a shape's type, color and position when selected:

const Sidebar = () => {
  const [selectedId] = useSelectedIdState();
  return (
    <div id="sidebar">
      {selectedId == null ? null : (
        <>
          <SidebarTypeCell />
          <SidebarColorCell label="Text" cellId="textColor" />
          <SidebarColorCell label="Back" cellId="backColor" />
          <SidebarNumberCell label="Left" cellId="x1" />
          <SidebarNumberCell label="Top" cellId="y1" />
          <SidebarNumberCell label="Right" cellId="x2" />
          <SidebarNumberCell label="Bottom" cellId="y2" />
        </>
      )}
    </div>
  );
};

Each of these controls will be nested in a component to allow the CSS to lay them out correctly:

const SidebarCell = ({label, children}) => (
  <div className="cell">
    {label}: {children}
  </div>
);

First up, the control to configure a shape's type to be either a rectangle or an ellipse. This is a dropdown of the TYPES values, which when selected, fires a setCell method wrapped in a callback from the useSetCellCallback hook. The value to set the type Cell is extracted from the event's target.value, and the change is added to the undo stack after the change has been made, with the callback from the useSetCheckpointCallback hook.

const SidebarTypeCell = () => {
  const [selectedId] = useSelectedIdState();
  const setCheckpoint = useSetCheckpointCallback(() => 'change of type', []);
  return (
    <SidebarCell label="Shape">
      <select
        value={useCell(SHAPES, selectedId, 'type')}
        onChange={useSetCellCallback(
          SHAPES,
          selectedId,
          'type',
          (e) => e.target.value,
          [],
          null,
          setCheckpoint,
        )}
      >
        {TYPES.map((type) => (
          <option key={type}>{type}</option>
        ))}
      </select>
    </SidebarCell>
  );
};

Setting the shape's two colors (stored in the textColor and backColor Cell values) is done with the SidebarColorCell component. It also uses the useSetCellCallback hook to get a callback that in this case uses the color picker event's target.value - and again, the useSetCheckpointCallback hook to add it to the undo stack after the change has been made:

const SidebarColorCell = ({label, cellId}) => {
  const [selectedId] = useSelectedIdState();
  const setCheckpoint = useSetCheckpointCallback(
    () => `change of '${label.toLowerCase()}' color`,
    [label],
  );
  return (
    <SidebarCell label={label}>
      <input
        type="color"
        value={useCell(SHAPES, selectedId, cellId)}
        onChange={useSetCellCallback(
          SHAPES,
          selectedId,
          cellId,
          (e) => e.target.value,
          [],
          null,
          setCheckpoint,
        )}
      />
    </SidebarCell>
  );
};

And finally (and similarly), the SidebarNumberCell component displays and lets you nudge the four numeric x1, y1, x2, and y2 Cell values. One interesting difference here is that the useSetCellCallback hook is passed not an absolute number for those values, but a function that will nudge the value.

const nudgeUp = (cell) => cell + 1;
const nudgeDown = (cell) => cell - 1;

const SidebarNumberCell = ({label, cellId}) => {
  const [selectedId] = useSelectedIdState();
  const setCheckpoint = useSetCheckpointCallback(
    () => `nudge of '${label.toLowerCase()}' value`,
    [label],
  );
  const handleDown = useSetCellCallback(
    SHAPES,
    selectedId,
    cellId,
    () => nudgeDown,
    [nudgeDown],
    null,
    setCheckpoint,
  );
  const handleUp = useSetCellCallback(
    SHAPES,
    selectedId,
    cellId,
    () => nudgeUp,
    [nudgeUp],
    null,
    setCheckpoint,
  );

  return (
    <SidebarCell label={label}>
      <div className="spin">
        <div className="button" onClick={handleDown}>
          -
        </div>
        {useCell(SHAPES, selectedId, cellId)}
        <div className="button" onClick={handleUp}>
          +
        </div>
      </div>{' '}
    </SidebarCell>
  );
};

The Canvas Component

The canvas component is the main part of the application upon which the other shapes appear. Its core implementation is essentially to render a LinkedRowsView component using ordered view of the shapes according to the nextId-based linked list.

However, it also includes additional complexity arising from the fact that we enforce shapes to fit within the bounds of the canvas. This means two things:

  • Firstly the canvas itself attaches mutator listeners to the Store with the useRowListener hook, so that shapes can never have x and y coordinates greater than its bounds when they change. The getShapeDimensions function also enforces the minimum width and height constraints for each shape.
  • Secondly, the canvas must respond to DOM changes that affect its own size. This is done with the updateDimensions function, tracking the DOM with a ResizeObserver. Changes to the canvas size change both the mutator listeners for future shape moves, but also does a one-off sweep through the current objects to fit them into canvas in case it shrinks.

While this may seem a little complicated to describe, it's easier to see in action. Try moving an object (or its handles) off the edge of the canvas. You'll discover that you can't, since the mutator listeners are stopping any values from exceeding the bounds. And then if you move a shape to the far-right of the canvas and shrink your browser window, you will see the smaller canvas pulling the shape in to fit.

Finally, the canvas listens to the handleMouseDown event so that if you click anywhere on the app background, the current shape will get deselected.

const Canvas = () => {
  const ref = useRef(null);
  const store = useStore();

  const [canvasDimensions, setCanvasDimensions] = useState([0, 0]);

  const getShapeDimensions = useCallback(
    (id, maxX, maxY) => {
      const {x1, x2, y1, y2} = store.getRow(SHAPES, id);
      const w = Math.max(x2 - x1, Math.min(MIN_WIDTH, maxX));
      const h = Math.max(y2 - y1, Math.min(MIN_HEIGHT, maxY));
      return {x1, x2, y1, y2, w, h};
    },
    [store],
  );

  useRowListener(
    SHAPES,
    null,
    (store, _tableId, rowId, getCellChange) => {
      const [maxX, maxY] = canvasDimensions;
      if (maxX == 0 || maxY == 0) {
        return;
      }

      const [x1Changed] = getCellChange(SHAPES, rowId, 'x1');
      const [x2Changed] = getCellChange(SHAPES, rowId, 'x2');
      const [y1Changed] = getCellChange(SHAPES, rowId, 'y1');
      const [y2Changed] = getCellChange(SHAPES, rowId, 'y2');
      if (
        (x1Changed || x2Changed || y1Changed || y2Changed) &&
        rowId != CANVAS_ID
      ) {
        const {x1, x2, y1, y2, w, h} = getShapeDimensions(rowId, maxX, maxY);
        if (x1Changed && x1 != null) {
          store.setCell(
            SHAPES,
            rowId,
            'x1',
            between(x1, 0, Math.min(x2, maxX) - w),
          );
        }
        if (x2Changed && x2 != null) {
          store.setCell(
            SHAPES,
            rowId,
            'x2',
            between(x2, Math.max(x1, 0) + w, maxX),
          );
        }
        if (y1Changed && y1 != null) {
          store.setCell(
            SHAPES,
            rowId,
            'y1',
            between(y1, 0, Math.min(y2, maxY) - h),
          );
        }
        if (y2Changed && y2 != null) {
          store.setCell(
            SHAPES,
            rowId,
            'y2',
            between(y2, Math.max(y1, 0) + h, maxY),
          );
        }
      }
    },
    [...canvasDimensions, getShapeDimensions],
    true,
  );

  const updateDimensions = useCallback(
    (current) => {
      const {clientWidth: maxX, clientHeight: maxY} = current;
      setCanvasDimensions([maxX, maxY]);
      store.forEachRow(SHAPES, (id) => {
        if (id != CANVAS_ID) {
          const {x2, y2, w, h} = getShapeDimensions(id, maxX, maxY);
          if (x2 > maxX) {
            store.setPartialRow(SHAPES, id, {
              x1: Math.max(0, maxX - w),
              x2: maxX,
            });
          }
          if (y2 > maxY) {
            store.setPartialRow(SHAPES, id, {
              y1: Math.max(0, maxY - h),
              y2: maxY,
            });
          }
        }
      });
    },
    [store, getShapeDimensions],
  );

  useEffect(() => {
    const {current} = ref;
    const observer = new ResizeObserver(() => updateDimensions(current));
    observer.observe(current);
    updateDimensions(current);
    return () => observer.disconnect();
  }, [ref, store, updateDimensions]);

  const [, setSelectedId] = useSelectedIdState();
  const getRowComponentProps = useCallback((id) => ({id}), []);
  const handleMouseDown = useCallback(() => setSelectedId(), [setSelectedId]);
  const backId = useBackId();
  return (
    <div id="canvas" onMouseDown={handleMouseDown} ref={ref}>
      {backId == null ? null : (
        <LinkedRowsView
          relationshipId="order"
          firstRowId={backId}
          rowComponent={Shape}
          getRowComponentProps={getRowComponentProps}
        />
      )}
    </div>
  );
};

const between = (value, min, max) =>
  value < min ? min : value > max ? max : value;

The Shape Component

Relatively speaking, each shape is quite a simple component. It uses the useSelectedIdState hook to identify if it is selected (and show the ShapeGrips resize handles if so).

It also uses the useDraggableObject hook we introduced at the start of the demo to make itself movable via the handleDrag callback. When you finish a drag, the handleDragStop callback records a checkpoint so that you can undo a whole dragging movement.

const Shape = ({id}) => {
  const [selectedId, setSelectedId] = useSelectedIdState();
  const selected = id == selectedId;
  const {x1, y1, x2, y2, backColor, type} = useRow(SHAPES, id);

  const store = useStore();
  const getInitial = useCallback(() => store.getRow(SHAPES, id), [store, id]);

  const handleDrag = useSetPartialRowCallback(
    SHAPES,
    id,
    ({dx, dy, initial}) => ({
      x1: initial.x1 + dx,
      y1: initial.y1 + dy,
      x2: initial.x2 + dx,
      y2: initial.y2 + dy,
    }),
    [],
  );
  const handleDragStart = useCallback(
    () => setSelectedId(id),
    [setSelectedId, id],
  );
  const handleDragStop = useSetCheckpointCallback(() => 'drag', []);
  const ref = useDraggableObject(
    getInitial,
    handleDrag,
    handleDragStart,
    handleDragStop,
  );

  const style = {
    left: `${x1}px`,
    top: `${y1}px`,
    width: `${x2 - x1}px`,
    height: `${y2 - y1}px`,
    background: backColor,
  };

  return (
    <>
      <div
        ref={ref}
        className={`shape ${type}${selected ? ' selected' : ''}`}
        style={style}
      >
        <ShapeText id={id} />
      </div>
      {selected ? <ShapeGrips id={id} /> : null}
    </>
  );
};

Inside the shape, the ShapeText component shows a text label, which, when double-clicked, sets the editing state to true, and becomes editable by turning into an <input> component. And once again, the useSetCheckpointCallback hook provides a callback to add text changes to the undo stack.

const ShapeText = ({id}) => {
  const {text, textColor} = useRow(SHAPES, id);
  const ref = useRef();
  const [editing, setEditing] = useState(false);
  const setCheckpoint = useSetCheckpointCallback(() => 'edit text', []);
  const handleDoubleClick = useCallback(() => setEditing(true), []);
  const handleBlur = useCallback(() => {
    setEditing(false);
    setCheckpoint();
  }, [setCheckpoint]);
  const handleChange = useSetCellCallback(
    SHAPES,
    id,
    'text',
    (e) => e.target.value,
    [],
  );
  const handleKeyDown = useCallback((e) => {
    if (e.which == 13) {
      e.target.blur();
    }
  }, []);

  useEffect(() => {
    if (editing) {
      ref.current.focus();
    }
  }, [editing, ref]);

  const style = {color: textColor};

  return editing ? (
    <input
      ref={ref}
      style={style}
      value={text}
      onChange={handleChange}
      onKeyDown={handleKeyDown}
      onBlur={handleBlur}
    />
  ) : (
    <span style={style} onDoubleClick={handleDoubleClick}>
      {text != '' ? text : '\xa0'}
    </span>
  );
};

The Shape Grips

When selected, a shape has eight handles for resizing it. Clockwise from top left, these adjust the x1,y1 coordinates, the y1 coordinate alone, the y1,x2 coordinates, the y2 coordinate alone, and so on. Each also has a different resizing cursor style to make it evident how they can move:

const ShapeGrips = ({id}) => {
  const {x1, y1, x2, y2} = useRow(SHAPES, id);
  const xm = (x1 + x2) / 2;
  const ym = (y1 + y2) / 2;
  return (
    <>
      <Grip m={[1, 1, 0, 0]} id={id} x={x1} y={y1} d="nwse" />
      <Grip m={[0, 1, 0, 0]} id={id} x={xm} y={y1} d="ns" />
      <Grip m={[0, 1, 1, 0]} id={id} x={x2} y={y1} d="nesw" />
      <Grip m={[0, 0, 1, 0]} id={id} x={x2} y={ym} d="ew" />
      <Grip m={[0, 0, 1, 1]} id={id} x={x2} y={y2} d="nwse" />
      <Grip m={[0, 0, 0, 1]} id={id} x={xm} y={y2} d="ns" />
      <Grip m={[1, 0, 0, 1]} id={id} x={x1} y={y2} d="nesw" />
      <Grip m={[1, 0, 0, 0]} id={id} x={x1} y={ym} d="ew" />
    </>
  );
};

Each grip itself is a small <div> element, made draggable with the useDraggableObject hook, and affecting the underlying shape's coordinates according to the 4-item m array passed in from the ShapeGrips component. Of course, the useSetCheckpointCallback hook also means that resizes get added to the undo stack:

const Grip = ({m: [mx1, my1, mx2, my2], id, x, y, d}) => {
  const store = useStore();
  const getInitial = useCallback(() => store.getRow(SHAPES, id), [store, id]);

  const handleDrag = useSetPartialRowCallback(
    SHAPES,
    id,
    ({dx, dy, initial}) => ({
      x1: initial.x1 + dx * mx1,
      y1: initial.y1 + dy * my1,
      x2: initial.x2 + dx * mx2,
      y2: initial.y2 + dy * my2,
    }),
    [mx1, my1, mx2, my2],
  );
  const handleDragStop = useSetCheckpointCallback(() => 'resize', []);

  return (
    <div
      ref={useDraggableObject(getInitial, handleDrag, null, handleDragStop)}
      className="grip"
      style={{left: `${x}px`, top: `${y}px`, cursor: `${d}-resize`}}
    />
  );
};

The Styling

Finally we style the app with some LESS. This should be mostly self-explanatory, including the SVG for the button icons.

@accentColor: #d81b60;
@font-face {
  font-family: Inter;
  src: url(https://tinybase.org/fonts/inter.woff2) format('woff2');
}
* {
  box-sizing: border-box;
  outline-color: @accentColor;
}

body {
  user-select: none;
  display: grid;
  grid-template-rows: auto 1fr;
  grid-template-columns: 1fr 10rem;
  font-family: Inter, sans-serif;
  letter-spacing: -0.04rem;
  font-size: 0.8rem;
  margin: 0;
  height: 100vh;

  #toolbar {
    z-index: 2;
    grid-column: span 2;
    background: #ddd;
    display: flex;
    align-items: center;
    border-bottom: 1px solid #aaa;
    > .button {
      cursor: pointer;
      line-height: 1rem;
      white-space: nowrap;
      border-right: 1px solid #aaa;
      padding: 0.5rem;
      &:hover {
        background: #ccc;
      }
      &::before {
        vertical-align: top;
        width: 1rem;
        height: 1rem;
        display: inline-block;
      }
      &.undo::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><path fill="none" d="M25 50a42 42 0 0 1 60 0" /><path d="M14 41v20 h20z" /></svg>');
      }
      &.redo::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><path fill="none" d="M15 50a42 42 0 0 1 60 0" /><path d="M86 41v20 h-20z" /></svg>');
      }
      &.add::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><rect x="20" y="20" width="60" height="60" fill="white"/><path d="M50 30v40M30 50h40" /></svg>');
      }
      &.front::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><rect x="2" y="2" width="40" height="40" fill="grey"/><rect x="58" y="58" width="40" height="40" fill="grey"/><rect x="20" y="20" width="60" height="60" fill="white"/></svg>');
      }
      &.forward::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><rect x="11" y="11" width="60" height="60" fill="grey"/><rect x="29" y="29" width="60" height="60" fill="white"/></svg>');
      }
      &.backward::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><rect x="11" y="11" width="60" height="60" fill="white"/><rect x="29" y="29" width="60" height="60" fill="grey"/></svg>');
      }
      &.back::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><rect x="20" y="20" width="60" height="60" fill="white"/><rect x="2" y="2" width="40" height="40" fill="grey"/><rect x="58" y="58" width="40" height="40" fill="grey"/></svg>');
      }
      &.delete::before {
        content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" height="1rem" viewBox="0 0 100 100" stroke-width="4" stroke="black"><rect x="20" y="20" width="60" height="60" fill="white"/><path d="M30 30l40 40M30 70l40-40" /></svg>');
      }
      &.disabled {
        opacity: 0.5;
        cursor: default;
        &:hover {
          background: none;
        }
      }
    }
  }

  #sidebar {
    z-index: 1;
    background: #eee;
    padding: 0.5rem;
    border-left: 1px solid #aaa;
    .cell {
      height: 2rem;
      text-align: right;
      select,
      input,
      .spin {
        font: inherit;
        letter-spacing: -0.05rem;
        margin-left: 5px;
        width: 4.5rem;
        height: 1.4rem;
        display: inline-flex;
        justify-content: space-between;
        align-items: center;
        .button {
          cursor: pointer;
          border: 1px solid #aaa;
          font: inherit;
          width: 1rem;
          text-align: center;
          &:hover {
            background: #ccc;
          }
        }
      }
    }
  }

  #canvas {
    position: relative;
    z-index: 0;
    .shape {
      align-items: center;
      background: white;
      border: 1px solid black;
      box-sizing: border-box;
      display: flex;
      justify-content: center;
      position: absolute;
      overflow: hidden;
      white-space: nowrap;
      z-index: 1;
      &.ellipse {
        border-radius: 50%;
      }
      &.selected {
        cursor: move;
      }
      input,
      span {
        background: transparent;
        border: none;
        font: inherit;
        width: 100%;
        text-align: center;
        margin: 0.5rem;
        text-overflow: ellipsis;
        overflow: hidden;
      }
    }
    .grip {
      background: white;
      border: 1px solid @accentColor;
      box-sizing: border-box;
      height: 6px;
      margin: -3px;
      position: absolute;
      width: 6px;
      z-index: 2;
    }
  }
}

That's It!

That's the drawing app.

Admittedly, there was a lot going on in here: a demonstration of a Relationships-based linked list, an undo stack, plenty of callbacks, and React hooks galore. On the other hand, we have a fast, and functional drawing application that makes the most of the structured data store underneath. Hopefully that all made sense.