And just like that we've hooked up useEditable to our editorRef, which points to the

element that is being rendered, and to setCode which drives our state containing some code.
Browser Compatibility
This library has been tested against and should work properly using:
- Chrome
- Safari
- iOS Safari
- Firefox
There are known issues in IE 11 due to the MutationObserver method being unable to
read text nodes that have been removed via the contenteditable.
FAQ
$3
Traditionally, there have been three options when choosing editing surfaces in React. Either one
could go for a large project like ProseMirror / CodeMirror or similar which take control over much
of the editing and rendering events and are hence rather opinionated, or it's possible to just
use contenteditable and render to raw HTML that is replaced in the element's content, or lastly one
could combine a textarea with an overlapping div that renders stylised content.
All three options don't allow much customisation in terms of what actually gets rendered or put
unreasonable restrictions on how easy it is to render and manage an editable's content.
So what makes rendering to a contenteditable element so hard?
Typically this is tough because they edit the DOM directly. This causes most rendering libraries, like
React and Preact to be confused, since their underlying Virtual DOMs don't match up with the actual
DOM structure anymore. To prevent this issue use-editable creates a MutationObserver, which watches
over all changes that are made to the contenteditable element. Before it reports these changes to
React it first rolls back all changes to the DOM so that React sees what it expects.
Furthermore it also preserves the current position of the caret, the selection, and restores it once
React has updated the DOM itself. This is a rather common technique for contenteditable editors, but
the MutationObserver addition is what enables use-editable to let another view library update the element's
content.
$3
Currently either the rendered elements' text content has to eventually exactly match the code input,
or your implementation must be able to convert the rendered text content back into what you're using
as state. This is a limitation of how contenteditable's work, since they'll only capture the actual
DOM content. Since use-editable doesn't aim to be a full component that manages the render cycle, it
doesn't have to keep any extra state, but will only pass the DOM's text back to the onChange callback.
Using the onChange callback you'll also receive a Position object describing the cursor position,
the current line number, and the line's contents up until the cursor, which is useful for auto-suggestions,
which could then be applied with the update function that useEditable returns to update the cursor
position.
API
$3
The first argument is elementRef and accepts a ref object of type RefObject which
points to the element that should become editable. This ref is allowed to be null or change during
the runtime of the hook. As long as the changes of the ref are triggered by React, everything should
behave as expected.
The second argument is onChange and accepts a callback of type (text: string, pos: Position) => void
that's called whenever the content of the contenteditable changes. This needs to be set up so that
it'll trigger a rerender of the element's contents.
The text that onChange receives is just the textual representation of the element's contents, while the
Position it receives contains the current position of the cursor, the line number (zero-indexed), and
the content of the current line up until the cursor, which is useful for autosuggestions.
The third argument is an optional options object. This accepts currently two options to change
the editing behavior of the hook:
- The disabled option disables editing on the editable by removing the contentEditable attribute from
  it again.
- The indentation option may be a number of displayed spaces for indentation. This also enables the
  improved Tab key behavior, which will indent the current line or dedent the current line when shift is
  held (Be aware that this will make the editor act as a focus trap!)
When options.indentation is set then useEditable will prevent the insertion of tab characters and
will instead insert the specified amount of whitespaces, which makes handling of columns much easier.
Additionally the useEditable hook returns an Edit handle with several methods, as documented below.
#### Edit.update
Edit.update(content: string): void
Replaces the entire content of the editable while adjusting the caret position.
This will shift the caret by the difference in length between the current content and the passed content.
#### Edit.insert
Edit.insert(append: string, offset?: number): void
Inserts new text at the caret position while deleting text in range of the offset (which accepts negative offsets).
For example, when offset is set to -1 then a single character is deleted to the left of the caret before
inserting any new text. When it's set to 2 then two characters to the right of the carets are deleted.
The append text may also be set to an empty string to only apply deletions without inserting any text.
When any text is selected then it's simply erased first and offset is ignored.
#### Edit.move
Edit.move(pos: number | { row: number; column: number }): void
This moves the caret to the specified position. The position may either be a character index (a number)
or coordinates specifying a row and column separately.
#### Edit.getState
Edit.getState(): { text: string; position: Position }
This method allows getting the current state of the editable, which is the same as what onChange usually
receives. This is useful when adding custom editing actions in a key down handler or when programmatically
imitating onChange otherwise, while the editable is selected.
Acknowledgments
- react-live, which I've worked on
  had one of the early tiny contenteditable editors. (But with raw HTML updates)
- react-simple-code-editor was the first (?) library to use a split textarea
  and rendering surface implementation, which presented what a nice editing API should look like.
- codejar contains the best tricks to manage selections, although it lacks some
  Firefox workarounds. It also uses raw HTML highlighting / updating.
- codemirror.next` is an invaluable source to see different techniques when
  handling text input and DOM update tricks.