Carets are the vertical line which indicates your editing position in a document. When there are multiple viewers of a document, each viewer's current editing position would be represented by different carets. You can click the mouse to reposition the caret in a new location, or use various OS-specific shortcuts to move the caret by increments (like by word or line).

When a document is loaded, the client inserts two carets for the user: an "anchor" caret, and a "focus" caret. Usually these carets are located at the same position. When the anchor and focus carets are one more caret positions apart, the editor treats this as a selection. For example, when you click and drag your mouse to create a selection, the anchor caret stays in place and the focus caret follows your mouse.

Carets in edit-text are stored as part of the document itself. The data structure for carets looks like the following:

enum Attrs {
    Caret {
        client_id: String,
        focus: boolean,
    }
    ...
}

The client_id property represents the client for whom the caret belongs. (To see your client ID, click the debug menu in the editor.) The focus property is false if this is the anchor caret, and is true if this is the focus caret.

Carets are represented as inline elements in HTML, and must exist only inside a block-level element.

All editing operations that delete carets must also recreate the carets (possibly in a new location) to preserve there being one anchor and one focus caret for each client at all times.

A related concept in edit-text is a cursor, represented by the CurSpan type, which specifies a single discrete position inside of the document (DocSpan type). When a user clicks their mouse, this is translated to a specific cursor inside the document. This can then be converted into a position to insert a caret (see the cur_to_caret function). A caret can have fewer valid positions than there are cursor positions, so this conversion is lossy.