ProseMirror Reference manual

Reference manual

This is the reference manual for the ProseMirror rich text editor. It lists and describes the full public API exported by the library.

For more introductory material, please see the guides.

ProseMirror is structured as a number of separate modules. This reference manual describes the exported API per module. We'll usually use the abbreviated module name here, but on NPM they are prefixed with prosemirror-. So if you want to use something from the state module, you have to import it like this:

var EditorState = require("prosemirror-state").EditorState
var state = EditorState.create({schema: mySchema})

Or in ES6 syntax:

import {EditorState} from "prosemirror-state"
let state = EditorState.create({schema: mySchema})

state module

This module implements the state object of a ProseMirror editor, along with the representation of the selection and the plugin abstraction.

Editor State

ProseMirror keeps all editor state (the things, basically, that would be required to create an editor just like the current one) in a single object. That object is updated (creating a new state) by applying transactions to it.

EditorState class

The state of a ProseMirror editor is represented by an object of this type. This is a persistent data structure—it isn't updated, but rather a new state value is computed from an old one with the apply method.

In addition to the built-in state fields, plugins can define additional pieces of state.

doc: Node: The current document.
selection: Selection: The selection.
storedMarks: ?[Mark]: A set of marks to apply to the next character that's typed. Will be null whenever no explicit marks have been set.
schema: Schema: The schema of the state's document.
plugins: [Plugin]: The plugins that are active in this state.
apply(tr: Transaction) → EditorState: Apply the given transaction to produce a new state.
applyTransaction(tr: Transaction) → {state: EditorState, transactions: [Transaction]}: Verbose variant of apply that returns the precise transactions that were applied (which might be influenced by the transaction hooks of plugins) along with the new state.
tr: Transaction: Start a transaction from this state.
reconfigure(config: Object) → EditorState: Create a new state based on this one, but with an adjusted set of active plugins. State fields that exist in both sets of plugins are kept unchanged. Those that no longer exist are dropped, and those that are new are initialized using their init method, passing in the new configuration object..
toJSON(pluginFields: ?Object<Plugin>) → Object: Serialize this state to JSON. If you want to serialize the state of plugins, pass an object mapping property names to use in the resulting JSON object to plugin objects.
static create(config: Object) → EditorState: Create a state. config must be an object containing at least a schema (the schema to use) or doc (the starting document) property. When it has a selection property, that should be a valid selection in the given document, to use as starting selection. Plugins, which are specified as an array in the plugins property, may read additional fields from the config object.
static fromJSON(config: Object, json: Object, pluginFields: ?Object<Plugin>) → EditorState: Deserialize a JSON representation of a state. config should have at least a schema field, and should contain array of plugins to initialize the state with. pluginFields can be used to deserialize the state of plugins, by associating plugin instances with the property names they use in the JSON object.

Transaction class extends `Transform`

An editor state transaction, which can be applied to a state to create an updated state. Use EditorState.tr to create an instance.

Transactions track changes to the document (they are a subclass of Transform), but also other state changes, like selection updates and adjustments of the set of stored marks. In addition, you can store metadata properties in a transaction, which are extra pieces of information that client code or plugins can use to describe what a transacion represents, so that they can update their own state accordingly.

The editor view uses a single metadata property: it will attach a property "pointer" with the value true to selection transactions directly caused by mouse or touch input.

time: number: The timestamp associated with this transaction.
storedMarks: ?[Mark]: The stored marks in this transaction.
selection: Selection: The transform's current selection. This defaults to the editor selection mapped through the steps in this transform, but can be overwritten with setSelection.
setSelection(selection: Selection) → Transaction: Update the transaction's current selection. This will determine the selection that the editor gets when the transaction is applied.
selectionSet: bool: Whether the selection was explicitly updated by this transaction.
setStoredMarks(marks: ?[Mark]) → Transaction: Set the current stored marks.
ensureMarks(marks: [Mark]) → Transaction: Make sure the current stored marks or, if that is null, the marks at the selection, match the given set of marks. Does nothing if this is already the case.
storedMarksSet: bool: Whether the stored marks were explicitly set for this transaction.
setTime(time: number) → Transaction: Update the timestamp for the transaction.
replaceSelection(slice: Slice) → Transaction
replaceSelectionWith(node: Node, inheritMarks: ?bool) → Transaction: Replace the selection with the given node or slice, or delete it if content is null. When inheritMarks is true and the content is inline, it inherits the marks from the place where it is inserted.
deleteSelection() → Transaction: Delete the selection.
insertText(text: string, from: ?number, to: ?number = from) → Transaction: Replace the given range, or the selection if no range is given, with a text node containing the given string.
setMeta(key: string | Plugin | PluginKey, value: any) → Transaction: Store a metadata property in this transaction, keyed either by name or by plugin.
getMeta(key: string | Plugin | PluginKey) → any: Retrieve a metadata property for a given name or plugin.
isGeneric: bool: Returns true if this transaction doesn't contain any metadata, and can thus be safely extended.
scrollIntoView() → Transaction: Indicate that the editor should scroll the selection into view when updated to the state produced by this transaction.
addStoredMark(mark: Mark) → Transaction: Add a mark to the set of stored marks.
removeStoredMark(mark: Mark | MarkType) → Transaction: Remove a mark or mark type from the set of stored marks.

Selection

A ProseMirror selection can be either a classical text selection (of which cursors are a special case), or a node selection, where a specific document node is selected.

Selection class

Superclass for editor selections.

new Selection($anchor: ResolvedPos, $head: ResolvedPos, ranges: ?[SelectionRange]): Initialize a selection with the head and anchor and ranges. If no ranges are given, constructs a single range across $anchor and $head.
ranges: [SelectionRange]: The ranges covered by the selection.
$anchor: ResolvedPos: The resolved anchor of the selection (the side that stays in place when the selection is modified).
$head: ResolvedPos: The resolved head of the selection (the side that moves when the selection is modified).
anchor: number: The selection's immobile side (does not move when shift-selecting).
head: number: The selection's mobile side (the side that moves when shift-selecting).
from: number: The lower bound of the selection's first range.
to: number: The upper bound of the selection's first range.
$from: ResolvedPos: The resolved lower bound of the selection's main range.
$to: ResolvedPos: The resolved upper bound of the selection's main range.
empty: bool: Indicates whether the selection contains any content.
eq(Selection) → bool: Test whether the selection is the same as another selection. The default implementation tests whether they have the same class, head, and anchor.
map(doc: Node, mapping: Mappable) → Selection: Map this selection through a mappable thing. doc should be the new document, to which we are mapping.
content: Slice: Get the content of this selection as a slice.
replace(tr: Transaction, content: ?Slice = Slice.empty): Replace the selection with a slice or, if no slice is given, delete the selection. Will append to the given transaction.
replaceWith(tr: Transaction, node: Node): Replace the selection with the given node, appending the changes to the given transaction.
toJSON() → Object: Convert the selection to a JSON representation. When implementing this for a custom selection class, make sure to give the object a type property whose value matches the ID under which you registered your class. The default implementation adds type, head, and anchor properties.
getBookmark() → SelectionBookmark: Get a bookmark for this selection, which is a value that can be mapped without having access to a current document, and later resolved to a real selection for a given document again. (This is used mostly by the history to track and restore old selections.) The default implementation of this method just converts the selection to a text selection and returns the bookmark for that.
visible: bool: Controls whether, when a selection of this type is active in the browser, the selected range should be visible to the user. Defaults to true.
static findFrom($pos: ResolvedPos, dir: number, textOnly: ?bool) → ?Selection: Find a valid cursor or leaf node selection starting at the given position and searching back if dir is negative, and forward if negative. When textOnly is true, only consider cursor selections.
static near($pos: ResolvedPos, bias: ?number = 1) → Selection: Find a valid cursor or leaf node selection near the given position. Searches forward first by default, but if bias is negative, it will search backwards first.
static atStart(doc: Node, textOnly: ?bool) → ?Selection: Find the cursor or leaf node selection closest to the start of the given document. When textOnly is true, only consider cursor selections.
static atEnd(doc: Node, textOnly: ?bool) → ?Selection: Find the cursor or leaf node selection closest to the end of the given document. When textOnly is true, only consider cursor selections.
static fromJSON(doc: Node, json: Object) → Selection: Deserialize a JSON representation of a selection. Must be implemented for custom classes (as a static class method).
static jsonID(id: string, selectionClass: constructor<Selection>): To be able to deserialize selections from JSON, custom selection classes must register themselves with an ID string, so that they can be disambiguated. Try to pick something that's unlikely to clash with classes from other modules.

TextSelection class extends `Selection`

A text selection represents a classical editor selection, with a head (the moving side) and anchor (immobile side), both of which point into textblock nodes. It can be empty (a regular cursor position).

new TextSelection($anchor: ResolvedPos, $head: ?ResolvedPos = $anchor): Construct a text selection between the given points.
$cursor: ?ResolvedPos: Returns a resolved position if this is a cursor selection (an empty text selection), and null otherwise.
static create(doc: Node, anchor: number, head: ?number = anchor) → TextSelection: Create a text selection from non-resolved positions.
static between($anchor: ResolvedPos, $head: ResolvedPos, bias: ?number) → Selection: Return a text selection that spans the given positions or, if they aren't text positions, find a text selection near them. bias determines whether the method searches forward (default) or backwards (negative number) first. Will fall back to returning a node selection when the document doesn't contain a valid text position.

NodeSelection class extends `Selection`

A node selection is a selection that points at a single node. All nodes marked selectable can be the target of a node selection. In such an object, from and to point directly before and after the selected node.

new NodeSelection($pos: ResolvedPos): Create a node selection. Does not verify the validity of its argument.
node: Node: The selected node.
static create(doc: Node, from: number, ?number) → TextSelection: Create a node selection from non-resolved positions.
static isSelectable(node: Node) → bool: Determines whether the given node may be selected as a node selection.

AllSelection class extends `Selection`

A selection type that represents selecting the whole document (which can not necessarily be expressed with a text selection, when there are for example leaf block nodes at the start or end of the document).

SelectionRange class

Represents a selected range in a document.

new SelectionRange($from: ResolvedPos, $to: ResolvedPos)
$from: ResolvedPos: The lower bound of the range.
$to: ResolvedPos: The upper bound of the range.

SelectionBookmark interface

A lightweight, document-independent representation of a selection. You can define a custom bookmark type for a custom selection class to make the history handle it well.

map(mapping: Mapping) → SelectionBookmark: Map the bookmark through a set of changes.
resolve(doc: Node) → Selection: Resolve the bookmark to a real selection again. This may need to do some error checking and may fall back to a default (usually TextSelection.between if mapping made the bookmark invalid.

Plugin System

To make distributing and using extra editor functionality easier, ProseMirror has a plugin system.

PluginSpec: Object

A plugin spec provides a definition for a plugin.

props: ?EditorProps: The view props added by this plugin. Note that the dispatchTransaction and state props can't be defined by plugins, only by the main props object. Props that are functions will be bound to have the plugin instance as their this binding.

Plugin class

Plugins wrap extra functionality that can be added to an editor. They can define new state fields, and add view props.

new Plugin(spec: PluginSpec): Create a plugin.
props: EditorProps: The props exported by this plugin.
spec: Object: The plugin's configuration object.
getState(state: EditorState) → any: Get the state field for this plugin.

StateField interface

A plugin may provide a state field (under its state property) of this type, which describes the state it wants to keep. Functions provided here are always called with the plugin instance as their this binding.

init(config: Object, instance: EditorState) → T: Initialize the value of this field. config will be the object passed to EditorState.create. Note that instance is a half-initialized state instance, and will not have values for any fields initialized after this one.
apply(tr: Transaction, value: T, oldState: EditorState, newState: EditorState) → T: Apply the given transaction to this state field, producing a new field value. Note that the newState argument is a partially constructed state does not yet contain the state from plugins coming after this plugin.
toJSON(value: T) → any: Convert this field to JSON. Optional, can be left off to disable JSON serialization for the field.
fromJSON(config: Object, value: any, state: EditorState) → T: Deserialize the JSON representation of this field. Note that the state argument is again a half-initialized state.

PluginKey class

A key is used to tag plugins in a way that makes it possible to find them, given an editor state. Assigning a key does mean only one plugin of that type can be active in a state.

new PluginKey(name: ?string = "key"): Create a plugin key.
get(state: EditorState) → ?Plugin: Get the active plugin with this key, if any, from an editor state.
getState(state: EditorState) → ?any: Get the plugin's state from an editor state.

view module

ProseMirror's view displays a given editor state in the DOM, and handles user events.

Make sure you load style/prosemirror.css as a stylesheet if you want to instantiate a ProseMirror editor.

EditorView class

An editor view manages the DOM structure that represents an editor. Its state and behavior are determined by its props.

new EditorView(place: ?dom.Node | fn(dom.Node) | {mount: dom.Node}, props: EditorProps): Create a view. place may be a DOM node that the editor should be appended to, a function that will place it into the document, or an object whose mount property holds the node to use. If it is null, the editor will not be added to the document.
state: EditorState: The view's current state.
dom: dom.Element: The editable DOM node containing the document. (You probably should not be directly interfering with its child nodes.)
props: EditorProps: The view's current props.
update(props: EditorProps): Update the view's props. Will immediately cause an update to the view's DOM.
setProps(props: EditorProps): Update the view by updating existing props object with the object given as argument. Equivalent to view.update(Object.assign({}, view.props, props)).
updateState(state: EditorState): Update the editor's state prop, without touching any of the other props.
hasFocus() → bool: Query whether the view has focus.
someProp(propName: string, f: fn(prop: any) → any) → any: Goes over the values of a prop, first those provided directly, then those from plugins (in order), and calls f every time a non-undefined value is found. When f returns a truthy value, that is immediately returned. When f isn't provided, it is treated as the identity function (the prop value is returned directly).
focus(): Focus the editor.
root: dom.Document | dom.DocumentFragment: Get the document root in which the editor exists. This will usually be the top-level document, but might be a shadow DOM root if the editor is inside a shadow DOM.
posAtCoords(coords: {left: number, top: number}) → ?{pos: number, inside: number}: Given a pair of coordinates, return the document position that corresponds to them. May return null if the given coordinates aren't inside of the visible editor. When an object is returned, its pos property is the position nearest to the coordinates, and its inside property holds the position before the inner node that the click happened inside of, or -1 if the click was at the top level.
coordsAtPos(pos: number) → {left: number, right: number, top: number, bottom: number}: Returns the screen rectangle at a given document position. left and right will be the same number, as this returns a flat cursor-ish rectangle.
endOfTextblock(dir: "up" | "down" | "left" | "right" | "forward" | "backward", state: ?EditorState) → bool: Find out whether the selection is at the end of a textblock when moving in a given direction. When, for example, given "left", it will return true if moving left from the current cursor position would leave that position's parent textblock.
destroy(): Removes the editor from the DOM and destroys all node views.
dispatch(tr: Transaction): Dispatch a transaction. Will call the dispatchTransaction when given, and defaults to applying the transaction to the current state and calling updateState otherwise. This method is bound to the view instance, so that it can be easily passed around.

EditorProps interface

The configuration object that can be passed to an editor view. It supports the following properties (only state is required).

The various event-handling functions may all return true to indicate that they handled the given event. The view will then take care to call preventDefault on the event, except with handleDOMEvents, where the handler itself is responsible for that.

Except for state and dispatchTransaction, these may also be present on the props property of plugins. How a prop is resolved depends on the prop. Handler functions are called one at a time, starting with the plugins (in order of appearance), and finally looking at the base props, until one of them returns true. For some props, the first plugin that yields a value gets precedence. For class, all the classes returned are combined.

state: EditorState

The state of the editor.

dispatchTransaction(tr: Transaction)

The callback over which to send transactions (state updates) produced by the view. You'll usually want to make sure this ends up calling the view's updateState method with a new state that has the transaction applied.

handleDOMEvents: ?Object<fn(view: EditorView, event: dom.Event) → bool>

Can be an object mapping DOM event type names to functions that handle them. Such functions will be called before any handling ProseMirror does of events fired on the editable DOM element. Contrary to the other event handling props, when returning true from such a function, you are responsible for calling preventDefault yourself (or not, if you want to allow the default behavior).

handleKeyDown(view: EditorView, event: dom.KeyboardEvent) → bool

Called when the editor receives a keydown event.

handleKeyPress(view: EditorView, event: dom.KeyboardEvent) → bool

Handler for keypress events.

handleTextInput(view: EditorView, from: number, to: number, text: string) → bool

Whenever the user directly input text, this handler is called before the input is applied. If it returns true, the default effect of actually inserting the text is suppressed.

handleClickOn(view: EditorView, pos: number, node: Node, nodePos: number, event: dom.MouseEvent, direct: bool) → bool

Called for each node around a click, from the inside out. The direct flag will be true for the inner node.

handleClick(view: EditorView, pos: number, event: dom.MouseEvent) → bool

Called when the editor is clicked, after handleClickOn handlers have been called.

handleDoubleClickOn(view: EditorView, pos: number, node: Node, nodePos: number, event: dom.MouseEvent, direct: bool) → bool

Called for each node around a double click.

handleDoubleClick(view: EditorView, pos: number, event: dom.MouseEvent) → bool

Called when the editor is double-clicked, after handleDoubleClickOn.

handleTripleClickOn(view: EditorView, pos: number, node: Node, nodePos: number, event: dom.MouseEvent, direct: bool) → bool

Called for each node around a triple click.

handleTripleClick(view: EditorView, pos: number, event: dom.MouseEvent) → bool

Called when the editor is triple-clicked, after handleTripleClickOn.

handleContextMenu(view: EditorView, pos: number, event: dom.MouseEvent) → bool

Called when a context menu event is fired in the editor.

handlePaste(view: EditorView, event: dom.Event, slice: Slice) → bool

Can be used to override the behavior of pasting. slice is the pasted content parsed by the editor, but you can directly access the event to get at the raw content.

handleDrop(view: EditorView, event: dom.Event, slice: Slice, moved: bool) → bool

Called when something is dropped on the editor. moved will be true if this drop moves from the current selection (which should thus be deleted).

onFocus(view: EditorView, event: dom.Event)

Called when the editor is focused.

onBlur(view: EditorView, event: dom.Event)

Called when the editor loses focus.

createSelectionBetween(view: EditorView, anchor: ResolvedPos, head: ResolvedPos) → ?Selection

Can be used to override the selection object created when reading a DOM selection between the given anchor and head.

domParser: ?DOMParser

The parser to use when reading editor changes from the DOM. Defaults to calling DOMParser.fromSchema on the editor's schema.

clipboardParser: ?DOMParser

The parser to use when reading content from the clipboard. When not given, the value of the domParser prop is used.

transformPasted(Slice) → Slice

Can be used to transform pasted content before it is applied to the document.

transformPastedHTML(string) → string

Can be used to transform pasted HTML text, before it is parsed, for example to clean it up.

transformPastedText(string) → string

Transform pasted plain text.

nodeViews: ?Object<fn(node: Node, view: EditorView, getPos: fn() → number, decorations: [Decoration]) → NodeView>

Allows you to pass custom rendering and behavior logic for nodes and marks. Should map node and mark names to constructor function that produce a NodeView object implementing the node's display behavior. getPos is a function that can be called to get the node's current position, which can be useful when creating transactions that update it.

decorations is an array of node or inline decorations that are active around the node. They are automatically drawn in the normal way, and you will usually just want to ignore this, but they can also be used as a way to provide context information to the node view without adding it to the document itself.

clipboardSerializer: ?DOMSerializer

The DOM serializer to use when putting content onto the clipboard. If not given, the result of DOMSerializer.fromSchema will be used.

decorations(EditorState) → ?DecorationSet

A set of document decorations to add to the view.

editable(EditorState) → bool

When this returns false, the content of the view is not directly editable.

attributes: ?Object<string> | fn(EditorState) → ?Object<string>

Control the DOM attributes of the editable element. May be either an object or a function going from an editor state to an object. By default, the element will get a class "ProseMirror", and will have its contentEditable attribute determined by the editable prop. Additional classes provided here will be added to the class. For other attributes, the value provided first (as in someProp) will be used.

scrollThreshold: ?number

Determines the distance (in pixels) between the cursor and the end of the visible viewport at which point, when scrolling the cursor into view, scrolling takes place. Defaults to 0.

scrollMargin: ?number

Determines the extra space (in pixels) that is left above or below the cursor when it is scrolled into view. Defaults to 5.

NodeView interface

By default, document nodes are rendered using the result of the toDOM method of their spec, and managed entirely by the editor. For some use cases, such as embedded node-specific editing interfaces, when you need more control over the behavior of a node's in-editor representation, and can define a custom node view.

dom: ?dom.Node: The outer DOM node that represents the document node. When not given, the default strategy is used to create a DOM node.
contentDOM: ?dom.Node: The DOM node that should hold the node's content. Only meaningful if the node view also defines a dom property and if its node type is not a leaf node type. When this is present, ProseMirror will take care of rendering the node's children into it. When it is not present, the node view itself is responsible for rendering (or deciding not to render) its child nodes.
update(node: Node, decorations: [Decoration]) → bool: When given, this will be called when the view is updating itself. It will be given a node (possibly of a different type), and an array of active decorations (which are automatically drawn, and the node view may ignore if it isn't interested in them), and should return true if it was able to update to that node, and false otherwise. If the node view has a contentDOM property (or no dom property), updating its child nodes will be handled by ProseMirror.
selectNode(): Can be used to override the way the node's selected status (as a node selection) is displayed.
deselectNode(): When defining a selectNode method, you should also provide a deselectNode method to disable it again.
setSelection(anchor: number, head: number, root: dom.Document): This will be called to handle setting the selection inside the node. By default, a DOM selection will be created between the DOM positions corresponding to the given anchor and head positions, but if you override it you can do something else.
stopEvent(event: dom.Event) → bool: Can be used to prevent the editor view from trying to handle some or all DOM events that bubble up from the node view.
ignoreMutation(dom.MutationRecord) → bool: Called when a DOM mutation happens within the view. Return false if the editor should re-parse the range around the mutation, true if it can safely be ignored.
destroy(): Called when the node view is removed from the editor or the whole editor is detached.

Decoration class

Decorations can be provided to the view (through the decorations prop) to adjust the way the document is drawn. They come in several variants. See the static members of this class for details.

spec: Object

The spec provided when creating this decoration. Can be useful if you've stored extra information in that object.

static widget(pos: number, dom: dom.Node, spec: ?Object) → Decoration

Creates a widget decoration, which is a DOM node that's shown in the document at the given position.

spec: ?Object

These options are supported:

associative: ?string: By default, widgets are right-associative, meaning they end up to the right of content inserted at their position. You can set this to "left" to make it left-associative, so that the inserted content will end up after the widget.

static inline(from: number, to: number, attrs: DecorationAttrs, spec: ?Object) → Decoration

Creates an inline decoration, which adds the given attributes to each inline node between from and to.

spec: ?Object

These options are recognized:

inclusiveStart: ?bool: Determines how the left side of the decoration is mapped when content is inserted directly at that positon. By default, the decoration won't include the new content, but you can set this to true to make it inclusive.

static node(from: number, to: number, attrs: DecorationAttrs, spec: ?Object) → Decoration

Creates a node decoration. from and to should point precisely before and after a node in the document. That node, and only that node, will receive the given attributes.

DecorationAttrs interface

A set of attributes to add to a decorated node. Most properties simply directly correspond to DOM attributes of the same name, which will be set to the property's value. These are exceptions:

class: ?string: A CSS class name or a space-separated set of class names to be added to the classes that the node already had.
style: ?string: A string of CSS to be added to the node's existing style property.
nodeName: ?string: When non-null, the target node is wrapped in a DOM element of this type (and the other attributes are applied to this element).

DecorationSet class

A collection of decorations, organized in such a way that the drawing algorithm can efficiently use and compare them. This is a persistent data structure—it is not modified, updates create a new value.

find(start: ?number, end: ?number) → [Decoration]

Find all decorations in this set which touch the given range (including decorations that start or end directly at the boundaries). When the arguments are omitted, all decorations in the set are collected.

map(mapping: Mapping, doc: Node, options: ?Object) → DecorationSet

Map the set of decorations in response to a change in the document.

options: ?Object

An optional set of options.

onRemove(decorationSpec: Object): When given, this function will be called for each decoration that gets dropped as a result of the mapping, passing the spec of that decoration.

add(doc: Node, decorations: [Decoration]) → DecorationSet

Add the given array of decorations to the ones in the set, producing a new set. Needs access to the current document to create the appropriate tree structure.

remove(decorations: [Decoration]) → DecorationSet

Create a new set that contains the decorations in this set, minus the ones in the given array.

static create(doc: Node, decorations: [Decoration]) → DecorationSet

Create a set of decorations, using the structure of the given document.

static empty: DecorationSet

The empty set of decorations.

model module

This module defines ProseMirror's content model, the data structures used to represent and manipulate documents.

Document Structure

A ProseMirror document is a tree. At each level, a node tags the type of content that's there, and holds a fragment containing its children.

Node class

This class represents a node in the tree that makes up a ProseMirror document. So a document is an instance of Node, with children that are also instances of Node.

Nodes are persistent data structures. Instead of changing them, you create new ones with the content you want. Old ones keep pointing at the old document shape. This is made cheaper by sharing structure between the old and new data as much as possible, which a tree shape like this (without back pointers) makes easy.

Never directly mutate the properties of a Node object. See this guide for more information.

type: NodeType: The type of node that this is.
attrs: Object: An object mapping attribute names to values. The kind of attributes allowed and required are determined by the node type.
content: Fragment: A container holding the node's children.
marks: [Mark]: The marks (things like whether it is emphasized or part of a link) associated with this node.
text: ?string: For text nodes, this contains the node's text content.
nodeSize: number: The size of this node, as defined by the integer-based indexing scheme. For text nodes, this is the amount of characters. For other leaf nodes, it is one. And for non-leaf nodes, it is the size of the content plus two (the start and end token).
childCount: number: The number of children that the node has.
child(index: number) → Node: Get the child node at the given index. Raises an error when the index is out of range.
maybeChild(index: number) → ?Node: Get the child node at the given index, if it exists.
forEach(f: fn(node: Node, offset: number, index: number)): Call f for every child node, passing the node, its offset into this parent node, and its index.
nodesBetween(from: ?number, to: ?number, f: fn(node: Node, pos: number, parent: Node, index: number)): Invoke a callback for all descendant nodes recursively between the given two positions that are relative to start of this node's content. The callback is invoked with the node, its parent-relative position, its parent node, and its child index. If the callback returns false, the current node's children will not be recursed over.
descendants(f: fn(node: Node, pos: number, parent: Node)): Call the given callback for every descendant node.
textContent: string: Concatenates all the text nodes found in this fragment and its children.
textBetween(from: number, to: number, blockSeparator: ?string, leafText: ?string) → string: Get all text between positions from and to. When blockSeparator is given, it will be inserted whenever a new block node is started. When leafText is given, it'll be inserted for every non-text leaf node encountered.
firstChild: ?Node: Returns this node's first child, or null if there are no children.
lastChild: ?Node: Returns this node's last child, or null if there are no children.
eq(other: Node) → bool: Test whether two nodes represent the same content.
sameMarkup(other: Node) → bool: Compare the markup (type, attributes, and marks) of this node to those of another. Returns true if both have the same markup.
hasMarkup(type: NodeType, attrs: ?Object, marks: ?[Mark]) → bool: Check whether this node's markup correspond to the given type, attributes, and marks.
copy(content: ?Fragment = null) → Node: Create a new node with the same markup as this node, containing the given content (or empty, if no content is given).
mark(marks: [Mark]) → Node: Create a copy of this node, with the given set of marks instead of the node's own marks.
cut(from: number, to: ?number) → Node: Create a copy of this node with only the content between the given offsets. If to is not given, it defaults to the end of the node.
slice(from: number, to: ?number = this.content.size) → Slice: Cut out the part of the document between the given positions, and return it as a Slice object.
replace(from: number, to: number, slice: Slice) → Node: Replace the part of the document between the given positions with the given slice. The slice must 'fit', meaning its open sides must be able to connect to the surrounding content, and its content nodes must be valid children for the node they are placed into. If any of this is violated, an error of type ReplaceError is thrown.
nodeAt(pos: number) → ?Node: Find the node after the given position.
childAfter(pos: number) → {node: ?Node, index: number, offset: number}: Find the (direct) child node after the given offset, if any, and return it along with its index and offset relative to this node.
childBefore(pos: number) → {node: ?Node, index: number, offset: number}: Find the (direct) child node before the given offset, if any, and return it along with its index and offset relative to this node.
resolve(pos: number) → ResolvedPos: Resolve the given position in the document, returning an object describing its path through the document.
rangeHasMark(from: ?number, to: ?number, type: MarkType) → bool: Test whether a mark of the given type occurs in this document between the two given positions.
isBlock: bool: True when this is a block (non-inline node)
isTextblock: bool: True when this is a textblock node, a block node with inline content.
inlineContent: bool: True when this node has inline content.
isInline: bool: True when this is an inline node (a text node or a node that can appear among text).
isText: bool: True when this is a text node.
isLeaf: bool: True when this is a leaf node.
isAtom: bool: True when this is an atom, i.e. when it does not have directly editable content. This is usually the same as isLeaf, but can be configured with the leaf property on a node's spec (typically when the node is displayed as an uneditable node view).
toString() → string: Return a string representation of this node for debugging purposes.
contentMatchAt(index: number) → ContentMatch: Get the content match in this node at the given index.
canReplace(from: number, to: number, replacement: ?Fragment, start: ?number, end: ?number) → bool: Test whether replacing the range from to to (by index) with the given replacement fragment (which defaults to the empty fragment) would leave the node's content valid. You can optionally pass start and end indices into the replacement fragment.
canReplaceWith(from: number, to: number, type: NodeType, attrs: ?[Mark]) → bool: Test whether replacing the range from to to (by index) with a node of the given type with the given attributes and marks would be valid.
canAppend(other: Node) → bool: Test whether the given node's content could be appended to this node. If that node is empty, this will only return true if there is at least one node type that can appear in both nodes (to avoid merging completely incompatible nodes).
check(): Check whether this node and its descendants conform to the schema, and raise error when they do not.
toJSON() → Object: Return a JSON-serializeable representation of this node.
static fromJSON(schema: Schema, json: Object) → Node: Deserialize a node from its JSON representation.

Fragment class

Fragment is the type used to represent a node's collection of child nodes.

Fragments are persistent data structures. That means you should not mutate them or their content, but create new instances whenever needed. The API tries to make this easy.

nodesBetween(from: number, to: number, f: fn(node: Node, start: number, parent: Node, index: number) → ?bool): Invoke a callback for all descendant nodes between the given two positions (relative to start of this fragment).
descendants(f: fn(node: Node, pos: number, parent: Node)): Call the given callback for every descendant node.
append(other: Fragment) → Fragment: Create a new fragment containing the content of this fragment and other.
cut(from: number, to: ?number) → Fragment: Cut out the sub-fragment between the two given positions.
replaceChild(index: number, node: Node) → Fragment: Create a new fragment in which the node at the given index is replaced by the given node.
eq(other: Fragment) → bool: Compare this fragment to another one.
firstChild: ?Node: The first child of the fragment, or null if it is empty.
lastChild: ?Node: The last child of the fragment, or null if it is empty.
childCount: number: The number of child nodes in this fragment.
child(index: number) → Node: Get the child node at the given index. Raise an error when the index is out of range.
offsetAt(index: number) → number: Get the offset at (size of children before) the given index.
maybeChild(index: number) → ?Node: Get the child node at the given index, if it exists.
forEach(f: fn(node: Node, offset: number, index: number)): Call f for every child node, passing the node, its offset into this parent node, and its index.
findDiffStart(other: Fragment) → ?number: Find the first position at which this fragment and another fragment differ, or null if they are the same.
findDiffEnd(other: Node) → ?{a: number, b: number}: Find the first position, searching from the end, at which this fragment and the given fragment differ, or null if they are the same. Since this position will not be the same in both nodes, an object with two separate positions is returned.
toString() → string: Return a debugging string that describes this fragment.
toJSON() → ?Object: Create a JSON-serializeable representation of this fragment.
static fromJSON(schema: Schema, value: ?Object) → Fragment: Deserialize a fragment from its JSON representation.
static fromArray(array: [Node]) → Fragment: Build a fragment from an array of nodes. Ensures that adjacent text nodes with the same style are joined together.
static from(nodes: ?Fragment | Node | [Node]) → Fragment: Create a fragment from something that can be interpreted as a set of nodes. For null, it returns the empty fragment. For a fragment, the fragment itself. For a node or array of nodes, a fragment containing those nodes.
static empty: Fragment: An empty fragment. Intended to be reused whenever a node doesn't contain anything (rather than allocating a new empty fragment for each leaf node).

Mark class

A mark is a piece of information that can be attached to a node, such as it being emphasized, in code font, or a link. It has a type and optionally a set of attributes that provide further information (such as the target of the link). Marks are created through a Schema, which controls which types exist and which attributes they have.

type: MarkType: The type of this mark.
attrs: Object: The attributes associated with this mark.
addToSet(set: [Mark]) → [Mark]: Given a set of marks, create a new set which contains this one as well, in the right position. If this mark is already in the set, the set itself is returned. If a mark of this type with different attributes is already in the set, a set in which it is replaced by this one is returned.
removeFromSet(set: [Mark]) → [Mark]: Remove this mark from the given set, returning a new set. If this mark is not in the set, the set itself is returned.
isInSet(set: [Mark]) → bool: Test whether this mark is in the given set of marks.
eq(other: Mark) → bool: Test whether this mark has the same type and attributes as another mark.
toJSON() → Object: Convert this mark to a JSON-serializeable representation.
static fromJSON(schema: Schema, json: Object) → Mark
static sameSet(a: [Mark], b: [Mark]) → bool: Test whether two sets of marks are identical.
static setFrom(marks: ?Mark | [Mark]) → [Mark]: Create a properly sorted mark set from null, a single mark, or an unsorted array of marks.
static none: [Mark]: The empty set of marks.

Slice class

A slice represents a piece cut out of a larger document. It stores not only a fragment, but also the depth up to which nodes on both side are 'open' / cut through.

new Slice(content: Fragment, openLeft: number, openRight: number)
content: Fragment: The slice's content nodes.
openLeft: number: The open depth at the start.
openRight: number: The open depth at the end.
size: number: The size this slice would add when inserted into a document.
eq(other: Slice) → bool: Tests whether this slice is equal to another slice.
toJSON() → ?Object: Convert a slice to a JSON-serializable representation.
static fromJSON(schema: Schema, json: ?Object) → Slice: Deserialize a slice from its JSON representation.
static maxOpen(fragment: Fragment) → Slice: Create a slice from a fragment by taking the maximum possible open value on both side of the fragment.
static empty: Slice: The empty slice.

ReplaceError class extends `Error`

Error type raised by Node.replace when given an invalid replacement.

Resolved Positions

Positions in a document can be represented as integer offsets. But you'll often want to use a more convenient representation.

ResolvedPos class

You'll often have to 'resolve' a position to get the context you need. Objects of this class represent such a resolved position, providing various pieces of context information and helper methods.

Throughout this interface, methods that take an optional depth parameter will interpret undefined as this.depth and negative numbers as this.depth + value.

pos: number: The position that was resolved.
depth: number: The number of levels the parent node is from the root. If this position points directly into the root, it is 0. If it points into a top-level paragraph, 1, and so on.
parentOffset: number: The offset this position has into its parent node.
parent: Node: The parent node that the position points into. Note that even if a position points into a text node, that node is not considered the parent—text nodes are 'flat' in this model.
node(depth: ?number) → Node: The ancestor node at the given level. p.node(p.depth) is the same as p.parent.
index(depth: ?number) → number: The index into the ancestor at the given level. If this points at the 3rd node in the 2nd paragraph on the top level, for example, p.index(0) is 2 and p.index(1) is 3.
indexAfter(depth: ?number) → number: The index pointing after this position into the ancestor at the given level.
start(depth: ?number) → number: The (absolute) position at the start of the node at the given level.
end(depth: ?number) → number: The (absolute) position at the end of the node at the given level.
before(depth: ?number) → number: The (absolute) position directly before the node at the given level, or, when level is this.level + 1, the original position.
after(depth: ?number) → number: The (absolute) position directly after the node at the given level, or, when level is this.level + 1, the original position.
textOffset: number: When this position points into a text node, this returns the distance between the position and the start of the text node. Will be zero for positions that point between nodes.
nodeAfter: ?Node: Get the node directly after the position, if any. If the position points into a text node, only the part of that node after the position is returned.
nodeBefore: ?Node: Get the node directly before the position, if any. If the position points into a text node, only the part of that node before the position is returned.
marks(after: ?bool) → [Mark]: Get the marks at this position, factoring in the surrounding marks' inclusive property. If the position is at the start of a non-empty node, or after is true, the marks of the node after it (if any) are returned.
sharedDepth(pos: number) → number: The depth up to which this position and the given (non-resolved) position share the same parent nodes.
blockRange(other: ?ResolvedPos = this, pred: ?fn(Node) → bool) → ?NodeRange: Returns a range based on the place where this position and the given position diverge around block content. If both point into the same textblock, for example, a range around that textblock will be returned. If they point into different blocks, the range around those blocks or their ancestors in their common ancestor is returned. You can pass in an optional predicate that will be called with a parent node to see if a range into that parent is acceptable.
sameParent(other: ResolvedPos) → bool: Query whether the given position shares the same parent node.
max(other: ResolvedPos) → ResolvedPos: Return the greater of this and the given position.
min(other: ResolvedPos) → ResolvedPos: Return the smaller of this and the given position.

NodeRange class

Represents a flat range of content.

$from: ResolvedPos: A resolved position along the start of the content. May have a depth greater than this object's depth property, since these are the positions that were used to compute the range, not re-resolved positions directly at its boundaries.
$to: ResolvedPos: A position along the end of the content. See caveat for $from.
depth: number: The depth of the node that this range points into.
start: number: The position at the start of the range.
end: number: The position at the end of the range.
parent: Node: The parent node that the range points into.
startIndex: number: The start index of the range in the parent node.
endIndex: number: The end index of the range in the parent node.

Document Schema

The schema to which a document must conform is another data structure. It describes the nodes and marks that may appear in a document, and the places at which they may appear.

Schema class

A document schema.

new Schema(spec: SchemaSpec): Construct a schema from a specification.
spec: SchemaSpec: The spec on which the schema is based, with the added guarantee that its nodes and marks properties are OrderedMap instances (not raw objects or null).
nodes: Object<NodeType>: An object mapping the schema's node names to node type objects.
marks: Object<MarkType>: A map from mark names to mark type objects.
cached: Object: An object for storing whatever values modules may want to compute and cache per schema. (If you want to store something in it, try to use property names unlikely to clash.)
topNodeType: NodeType: The type of the default top node for this schema.
node(type: string | NodeType, attrs: ?Object, content: ?Fragment | Node | [Node], marks: ?[Mark]) → Node: Create a node in this schema. The type may be a string or a NodeType instance. Attributes will be extended with defaults, content may be a Fragment, null, a Node, or an array of nodes.
text(text: string, marks: ?[Mark]) → Node: Create a text node in the schema. Empty text nodes are not allowed.
mark(type: string | MarkType, attrs: ?Object) → Mark: Create a mark with the given type and attributes.
nodeFromJSON(json: Object) → Node: Deserialize a node from its JSON representation. This method is bound.
markFromJSON(json: Object) → Mark: Deserialize a mark from its JSON representation. This method is bound.

SchemaSpec interface

An object describing a schema, as passed to the Schema constructor.

nodes: Object<NodeSpec> | OrderedMap<NodeSpec>: The node types in this schema. Maps names to NodeSpec objects describing the node to be associated with that name. Their order is significant
marks: ?Object<MarkSpec> | OrderedMap<MarkSpec>: The mark types that exist in this schema.
topNode: ?string: The name of the default top-level node for the schema. Defaults to "doc".

NodeSpec interface

content: ?string

The content expression for this node, as described in the schema guide. When not given, the node does not allow any content.

group: ?string

The group or space-separated groups to which this node belongs, as referred to in the content expressions for the schema.

inline: ?bool

Should be set to a truthy value for inline nodes. (Implied for text nodes.)

atom: ?bool

Can be set to true to indicate that, though this isn't a leaf node, it doesn't have directly editable content and should be treated as a single unit in the view.

attrs: ?Object<AttributeSpec>

The attributes that nodes of this type get.

selectable: ?bool

Controls whether nodes of this type can be selected (as a node selection). Defaults to true for non-text nodes.

draggable: ?bool

Determines whether nodes of this type can be dragged without being selected. Defaults to false.

code: ?bool

Can be used to indicate that this node contains code, which causes some commands to behave differently.

defining: ?bool

Determines whether this node is considered an important parent node during replace operations (such as paste). Non-defining (the default) nodes get dropped when their entire content is replaced, whereas defining nodes persist and wrap the inserted content. Likewise, the the inserted content, when not inserting into a textblock, the defining parents of the content are preserved. Typically, non-default-paragraph textblock types, and possible list items, are marked as defining.

toDOM(Node) → DOMOutputSpec

Defines the default way a node of this type should be serialized to DOM/HTML (as used by DOMSerializer.fromSchema). Should return an array structure that describes the resulting DOM structure, with an optional number zero (“hole”) in it to indicate where the node's content should be inserted.

For text nodes, the default is to create a text DOM node. Though it is possible to create a serializer where text is rendered differently, this is not supported inside the editor, so you shouldn't override that in your text node spec.

parseDOM: ?[ParseRule]

Associates DOM parser information with this node, which can be used by DOMParser.fromSchema to automatically derive a parser. The node field in the rules is implied (the name of this node will be filled in automatically). If you supply your own parser, you do not need to also specify parsing rules in your schema.

MarkSpec interface

attrs: ?Object<AttributeSpec>

The attributes that marks of this type get.

inclusive: ?bool

Whether this mark should be active when the cursor is positioned at the start or end boundary of the mark. Defaults to true.

excludes: ?string

Determines which other marks this mark can coexist with. Should be a space-separated strings naming other marks or groups of marks. When a mark is added to a set, all marks that it excludes are removed in the process. If the set contains any mark that excludes the new mark but is not, itself, excluded by the new mark, the mark can not be added an the set. You can use the value "_" to indicate that the mark excludes all marks in the schema.

Defaults to only being exclusive with marks of the same type. You can set it to an empty string (or any string not containing the mark's own name) to allow multiple marks of a given type to coexist (as long as they have different attributes).

group: ?string

The group or space-separated groups to which this node belongs.

toDOM(mark: Mark) → DOMOutputSpec

Defines the default way marks of this type should be serialized to DOM/HTML.

parseDOM: ?[ParseRule]

Associates DOM parser information with this mark (see the corresponding node spec field). The mark field in the rules is implied.

AttributeSpec interface

Used to define attributes. Attributes that have no default or compute property must be provided whenever a node or mark of a type that has them is created.

The following fields are supported:

default: ?any: The default value for this attribute, to choose when no explicit value is provided.
compute() → any: A function that computes a default value for the attribute.

NodeType class

Node types are objects allocated once per Schema and used to tag Node instances with a type. They contain information about the node type, such as its name and what kind of node it represents.

name: string: The name the node type has in this schema.
schema: Schema: A link back to the Schema the node type belongs to.
spec: NodeSpec: The spec that this type is based on
isBlock: bool: True if this is a block type
isText: bool: True if this is the text node type.
isInline: bool: True if this is an inline type.
isTextblock: bool: True if this is a textblock type, a block that contains inline content.
inlineContent: bool: True if this node type has inline content.
isLeaf: bool: True for node types that allow no content.
isAtom: bool: True when this node is an atom, i.e. when it does not have directly editable content.
create(attrs: ?Object, content: ?Fragment | Node | [Node], marks: ?[Mark]) → Node: Create a Node of this type. The given attributes are checked and defaulted (you can pass null to use the type's defaults entirely, if no required attributes exist). content may be a Fragment, a node, an array of nodes, or null. Similarly marks may be null to default to the empty set of marks.
createChecked(attrs: ?Object, content: ?Fragment | Node | [Node], marks: ?[Mark]) → Node: Like create, but check the given content against the node type's content restrictions, and throw an error if it doesn't match.
createAndFill(attrs: ?Object, content: ?Fragment | Node | [Node], marks: ?[Mark]) → ?Node: Like create, but see if it is necessary to add nodes to the start or end of the given fragment to make it fit the node. If no fitting wrapping can be found, return null. Note that, due to the fact that required nodes can always be created, this will always succeed if you pass null or Fragment.empty as content.
validContent(content: Fragment, attrs: ?Object) → bool: Returns true if the given fragment is valid content for this node type with the given attributes.

MarkType class

Like nodes, marks (which are associated with nodes to signify things like emphasis or being part of a link) are tagged with type objects, which are instantiated once per Schema.

name: string: The name of the mark type.
schema: Schema: The schema that this mark type instance is part of.
spec: MarkSpec: The spec on which the type is based.
create(attrs: ?Object) → Mark: Create a mark of this type. attrs may be null or an object containing only some of the mark's attributes. The others, if they have defaults, will be added.
removeFromSet(set: [Mark]) → [Mark]: When there is a mark of this type in the given set, a new set without it is returned. Otherwise, the input set is returned.
isInSet(set: [Mark]) → ?Mark: Tests whether there is a mark of this type in the given set.
excludes: MarkType: → bool

ContentMatch class

Represents a partial match of a node type's content expression, and can be used to find out whether further content matches here, and whether a given position is a valid end of the parent node.

matchNode(node: Node) → ?ContentMatch: Match a node, returning a new match after the node if successful.
matchType(type: NodeType, attrs: ?Object, marks: ?[Mark] = Mark.none) → ?ContentMatch: Match a node type and marks, returning an match after that node if successful.
matchFragment(fragment: Fragment, from: ?number = 0, to: ?number = fragment.childCount) → ?ContentMatch | bool: Try to match a fragment. Returns a new match when successful, null when it ran into a required element it couldn't fit, and false if it reached the end of the expression without matching all nodes.
matchToEnd(fragment: Fragment, start: ?number, end: ?number) → bool: Returns true only if the fragment matches here, and reaches all the way to the end of the content expression.
validEnd() → bool: Returns true if this position represents a valid end of the expression (no required content follows after it).
fillBefore(after: Fragment, toEnd: bool, startIndex: ?number) → ?Fragment: Try to match the given fragment, and if that fails, see if it can be made to match by inserting nodes in front of it. When successful, return a fragment of inserted nodes (which may be empty if nothing had to be inserted). When toEnd is true, only return a fragment if the resulting match goes to the end of the content expression.
allowsMark(markType: MarkType) → bool: Check whether a node with the given mark type is allowed after this position.
findWrapping(target: NodeType, targetAttrs: ?Object, targetMarks: ?[Mark]) → ?[{type: NodeType, attrs: Object}]: Find a set of wrapping node types that would allow a node of type target with attributes targetAttrs to appear at this position. The result may be empty (when it fits directly) and will be null when no such wrapping exists.
findWrappingFor(node: Node) → ?[{type: NodeType, attrs: Object}]: Call findWrapping with the properties of the given node.

DOM Representation

Because representing a document as a tree of DOM nodes is central to the way ProseMirror operates, DOM parsing and serializing is integrated with the model.

(But note that you do not need to have a DOM implementation loaded to load this module.)

DOMParser class

A DOM parser represents a strategy for parsing DOM content into a ProseMirror document conforming to a given schema. Its behavior is defined by an array of rules.

new DOMParser(schema: Schema, rules: [ParseRule])

Create a parser that targets the given schema, using the given parsing rules.

schema: Schema

rules: [ParseRule]

parse(dom: dom.Node, options: ?Object = {}) → Node

Parse a document from the content of a DOM node.

options: ?Object

Configuration options.

preserveWhitespace: ?bool | "full": By default, whitespace is collapsed as per HTML's rules. Pass true to preserve whitespace, but normalize newlines to spaces, and "full" to preserve whitespace entirely.

parseSlice(dom: dom.Node, options: ?Object = {}) → Slice

Parses the content of the given DOM node, like parse, and takes the same set of options. But unlike that method, which produces a whole node, this one returns a slice that is open at the sides, meaning that the schema constraints aren't applied to the start of nodes to the left of the input and the end of nodes at the end.

static schemaRules(schema: Schema) → [ParseRule]

Extract the parse rules listed in a schema's node specs.

static fromSchema(schema: Schema) → DOMParser

Construct a DOM parser using the parsing rules listed in a schema's node specs.

ParseRule interface

A value that describes how to parse a given DOM node or inline style as a ProseMirror node or mark.

tag: ?string

A CSS selector describing the kind of DOM elements to match. A single rule should have either a tag or a style property.

style: ?string

A CSS property name to match. When given, this rule matches inline styles that list that property.

context: ?string

When given, restricts this rule to only match when the current context—the parent nodes into which the content is being parsed—matches this expression. Should contain one or more node names or node group names followed by single or double slashes. For example "paragraph/" means the rule only matches when the parent node is a paragraph, "blockquote/paragraph/" restricts it to be in a paragraph that is inside a blockquote, and "section//" matches any position inside a section—a double slash matches any sequence of ancestor nodes.

node: ?string

The name of the node type to create when this rule matches. Only valid for rules with a tag property, not for style rules. Each rule should have one of a node, mark, or ignore property (except when it appears in a node or mark spec, in which case the node or mark property will be derived from its position).

mark: ?string

The name of the mark type to wrap the matched content in.

priority: ?number

Can be used to change the order in which the parse rules in a schema are tried. Those with higher priority come first. Rules without a priority are counted as having priority 50. This property is only meaningful in a schema—when directly constructing a parser, the order of the rule array is used.

ignore: ?bool

When true, ignore content that matches this rule.

skip: ?bool

When true, ignore the node that matches this rule, but do parse its content.

attrs: ?Object

Attributes for the node or mark created by this rule. When getAttrs is provided, it takes precedence.

getAttrs(dom.Node | string) → ?bool | Object

A function used to compute the attributes for the node or mark created by this rule. Can also be used to describe further conditions the DOM element or style must match. When it returns false, the rule won't match. When it returns null or undefined, that is interpreted as an empty/default set of attributes.

Called with a DOM Element for tag rules, and with a string (the style's value) for style rules.

contentElement: ?string

For tag rules that produce non-leaf nodes or marks, by default the content of the DOM element is parsed as content of the mark or node. If the child nodes are in a descendent node, this may be a CSS selector string that the parser must use to find the actual content element.

getContent(dom.Node) → Fragment

Can be used to override the content of a matched node. Will be called, and its result used, instead of parsing the node's child nodes.

preserveWhitespace: ?bool | "full"

Controls whether whitespace should be preserved when parsing the content inside the matched element. false means whitespace may be collapsed, true means that whitespace should be preserved but newlines normalized to spaces, and "full" means that newlines should also be preserved.

DOMSerializer class

A DOM serializer knows how to convert ProseMirror nodes and marks of various types to DOM nodes.

new DOMSerializer(nodes: Object<fn(node: Node) → DOMOutputSpec>, marks: Object<fn(mark: Mark) → DOMOutputSpec>): Create a serializer. nodes should map node names to functions that take a node and return a description of the corresponding DOM. marks does the same for mark names.
nodes: Object<fn(node: Node) → DOMOutputSpec>
marks: Object<fn(mark: Mark) → DOMOutputSpec>
serializeFragment(fragment: Fragment, options: ?Object = {}) → dom.DocumentFragment: Serialize the content of this fragment to a DOM fragment. When not in the browser, the document option, containing a DOM document, should be passed so that the serializer can create nodes.
serializeNode(node: Node, options: ?Object = {}) → dom.Node: Serialize this node to a DOM node. This can be useful when you need to serialize a part of a document, as opposed to the whole document. To serialize a whole document, use serializeFragment on its content.
static renderSpec(doc: dom.Document, structure: DOMOutputSpec) → {dom: dom.Node, contentDOM: ?dom.Node}: Render an output spec.
static fromSchema(schema: Schema) → DOMSerializer: Build a serializer using the toDOM properties in a schema's node and mark specs.
static nodesFromSchema(schema: Schema) → Object<fn(node: Node) → DOMOutputSpec>: Gather the serializers in a schema's node specs into an object. This can be useful as a base to build a custom serializer from.
static marksFromSchema(schema: Schema) → Object<fn(mark: Mark) → DOMOutputSpec>: Gather the serializers in a schema's mark specs into an object.

DOMOutputSpec interface

A description of a DOM structure. Can be either a string, which is interpreted as a text node, a DOM node, which is interpreted as itself, or an array.

An array describes a DOM element. The first element in the array should be a string, and is the name of the DOM element. If the second element is a non-Array, non-DOM node object, it is interpreted as an object providing the DOM element's attributes. Any elements after that (including the 2nd if it's not an attribute object) are interpreted as children of the DOM elements, and must either be valid DOMOutputSpec values, or the number zero.

The number zero (pronounced “hole”) is used to indicate the place where a ProseMirror node's content should be inserted.

transform module

This module defines a way to transform documents. You can read more about transformations in this guide.

Steps

Transforming happens in Steps, which are atomic, well-defined modifications to a document. Applying a step produces a new document.

Each step provides a change map that maps positions in the old document to position in the transformed document. Steps can be inverted to create a step that undoes their effect, and chained together in a convenience object called a Transform.

Step class

A step object wraps an atomic operation. It generally applies only to the document it was created for, since the positions associated with it will only make sense for that document.

New steps are defined by creating classes that extend Step, overriding the apply, invert, map, getMap and fromJSON methods, and registering your class with a unique JSON-serialization identifier using Step.jsonID.

apply(doc: Node) → StepResult: Applies this step to the given document, returning a result object that either indicates failure, if the step can not be applied to this document, or indicates success by containing a transformed document.
getMap() → StepMap: Get the step map that represents the changes made by this step.
invert(doc: Node) → Step: Create an inverted version of this step. Needs the document as it was before the step as argument.
map(mapping: Mappable) → ?Step: Map this step through a mappable thing, returning either a version of that step with its positions adjusted, or null if the step was entirely deleted by the mapping.
merge(other: Step) → ?Step: Try to merge this step with another one, to be applied directly after it. Returns the merged step when possible, null if the steps can't be merged.
offset(n: number) → Step: Returns a copy of this step in which all positions have n added to them. The main use for this is to take a step in one document, and make it apply to a sub-document, or a larger document that the original document is a part of.
toJSON() → Object: Create a JSON-serializeable representation of this step. By default, it'll create an object with the step's JSON id, and each of the steps's own properties, automatically calling toJSON on the property values that have such a method.
static fromJSON(schema: Schema, json: Object) → Step: Deserialize a step from its JSON representation. Will call through to the step class' own implementation of this method.
static jsonID(id: string, stepClass: constructor<Step>): To be able to serialize steps to JSON, each step needs a string ID to attach to its JSON representation. Use this method to register an ID for your step classes. Try to pick something that's unlikely to clash with steps from other modules.

StepResult class

The result of applying a step. Contains either a new document or a failure value.

doc: ?Node: The transformed document.
failed: ?string: Text providing information about a failed step.
static ok(doc: Node) → StepResult: Create a successful step result.
static fail(message: string) → StepResult: Create a failed step result.
static fromReplace(doc: Node, from: number, to: number, slice: Slice) → StepResult: Call Node.replace with the given arguments. Create a successful result if it succeeds, and a failed one if it throws a ReplaceError.

ReplaceStep class extends `Step`

Replace a part of the document with a slice of new content.

new ReplaceStep(from: number, to: number, slice: Slice, structure: ?bool): The given slice should fit the 'gap' between from and to—the depths must line up, and the surrounding nodes must be able to be joined with the open sides of the slice. When structure is true, the step will fail if the content between from and to is not just a sequence of closing and then opening tokens (this is to guard against rebased replace steps overwriting something they weren't supposed to).

ReplaceAroundStep class extends `Step`

Replace a part of the document with a slice of content, but preserve a range of the replaced content by moving it into the slice.

new ReplaceAroundStep(from: number, to: number, gapFrom: number, gapTo: number, slice: Slice, insert: number, structure: ?bool): Create a replace-wrap step with the given range and gap. insert should be the point in the slice into which the gap should be moved. structure has the same meaning as it has in the ReplaceStep class.

AddMarkStep class extends `Step`

Add a mark to all inline content between two positions.

new AddMarkStep(from: number, to: number, mark: Mark)

RemoveMarkStep class extends `Step`

Remove a mark from all inline content between two positions.

new RemoveMarkStep(from: number, to: number, mark: Mark)

Position Mapping

Mapping positions from one document to another by running through the replacements produced by steps is a fundamental operation in ProseMirror.

Mappable interface

There are several things that positions can be mapped through. We'll denote those as 'mappable'.

map(pos: number, assoc: ?number) → number: Map a position through this object. When given, assoc (should be -1 or 1, defaults to 1) determines with which side the position is associated, which determines in which direction to move when a chunk of content is inserted at the mapped position, and when to consider the position to be deleted.
mapResult(pos: number, assoc: ?number) → MapResult: Map a position, and return an object containing additional information about the mapping. The result's deleted field tells you whether the position was deleted (completely enclosed in a replaced range) during the mapping.

MapResult class

An object representing a mapped position with extra information.

pos: number: The mapped version of the position.
deleted: bool: Tells you whether the position was deleted, that is, whether the step removed its surroundings from the document.

StepMap class

A map describing the deletions and insertions made by a step, which can be used to find the correspondence between positions in the pre-step version of a document and the same position in the post-step version. This class implements Mappable.

new StepMap(ranges: [number]): Create a position map. The modifications to the document are represented as an array of numbers, in which each group of three represents a modified chunk as [start, oldSize, newSize].
mapResult(pos: number, assoc: ?number) → MapResult: Map the given position through this map. The assoc parameter can be used to control what happens when the transform inserted content at (or around) this position—if assoc is negative, the a position before the inserted content will be returned, if it is positive, a position after the insertion is returned.
map(pos: number, assoc: ?number) → number: Map the given position through this map, returning only the mapped position.
forEach(f: fn(oldStart: number, oldEnd: number, newStart: number, newEnd: number)): Calls the given function on each of the changed ranges denoted by this map.
invert() → StepMap: Create an inverted version of this map. The result can be used to map positions in the post-step document to the pre-step document.

Mapping class

A mapping represents a pipeline of zero or more step maps. It has special provisions for losslessly handling mapping positions through a series of steps in which some steps are inverted versions of earlier steps. (This comes up when ‘rebasing’ steps for collaboration or history management.) This class implements Mappable.

new Mapping(maps: ?[StepMap]): Create a new mapping with the given position maps.
maps: [StepMap]: The step maps in this mapping.
from: number: The starting position in the maps array, used when map or mapResult is called.
to: number: The end positions in the maps array.
slice(from: ?number = 0, to: ?number = this.maps.length) → Mapping: Create a mapping that maps only through a part of this one.
appendMap(map: StepMap, mirrors: ?number): Add a step map to the end of this mapping. If mirrors is given, it should be the index of the step map that is the mirror image of this one.
appendMapping(mapping: Mapping): Add all the step maps in a given mapping to this one (preserving mirroring information).
appendMappingInverted(mapping: Mapping): Append the inverse of the given mapping to this one.
map(pos: number, assoc: ?number) → number: Map a position through this mapping.
mapResult(pos: number, assoc: ?number) → MapResult: Map a position through this mapping, returning a mapping result.

Transform Helpers

Because you often need to collect a number of steps together to effect a composite change, ProseMirror provides an abstraction to make this easy. A value of this class is also the payload in the transform action.

Transform class

Abstraction to build up and track such an array of steps.

The high-level transforming methods return the Transform object itself, so that they can be chained.

new Transform(doc: Node)

Create a transformation that starts with the given document.

addMark(from: number, to: number, mark: Mark) → Transform

Add the given mark to the inline content between from and to.

removeMark(from: number, to: number, mark: ?Mark | MarkType = null) → Transform

Remove the given mark, or all marks of the given type, from inline nodes between from and to.

clearMarkup(from: number, to: number) → Transform

Remove all marks and non-text inline nodes from the given range.

replaceRange(from: number, to: number, slice: Slice) → Transform

Replace a range of the document with a given slice, using from, to, and the slice's openLeft property as hints, rather than fixed start and end points. This method may grow the replaced area or close open nodes in the slice in order to get a fit that is more in line with WYSIWYG expectations, by dropping fully covered parent nodes of the replaced region when they are marked non-defining, or including an open parent node from the slice that is marked as defining.

This is the method, for example, to handle paste. The similar replace method is a more primitive tool which will not move the start and end of its given range, and is useful in situations where you need more precise control over what happens.

replaceRangeWith(from: number, to: number, node: Node) → Transform

Replace the given range with a node, but use from and to as hints, rather than precise positions. When from and to are the same and are at the start or end of a parent node in which the given node doesn't fit, this method may move them out towards a parent that does allow the given node to be placed. When the given range completely covers a parent node, this method may completely replace that parent node.

deleteRange(from: number, to: number) → Transform

Delete the given range, expanding it to cover fully covered parent nodes until a valid replace is found.

delete(from: number, to: number) → Transform

Delete the content between the given positions.

replace(from: number, to: ?number = from, slice: ?Slice = Slice.empty) → Transform

Replace the part of the document between from and to with the given slice.

replaceWith(from: number, to: number, content: Fragment | Node | [Node]) → Transform

Replace the given range with the given content, which may be a fragment, node, or array of nodes.

insert(pos: number, content: Fragment | Node | [Node]) → Transform

Insert the given content at the given position.

lift(range: NodeRange, target: number) → Transform

Split the content in the given range off from its parent, if there is sibling content before or after it, and move it up the tree to the depth specified by target. You'll probably want to use liftTarget to compute target, in order to be sure the lift is valid.

wrap(range: NodeRange, wrappers: [{type: NodeType, attrs: ?Object}]) → Transform

Wrap the given range in the given set of wrappers. The wrappers are assumed to be valid in this position, and should probably be computed with findWrapping.

setBlockType(from: number, to: ?number = from, type: NodeType, attrs: ?Object) → Transform

Set the type of all textblocks (partly) between from and to to the given node type with the given attributes.

setNodeType(pos: number, type: ?NodeType, attrs: ?Object, marks: ?[Mark]) → Transform

Change the type and attributes of the node after pos.

split(pos: number, depth: ?number = 1, typesAfter: ?[?{type: NodeType, attrs: ?Object}]) → Transform

Split the node at the given position, and optionally, if depth is greater than one, any number of nodes above that. By default, the parts split off will inherit the node type of the original node. This can be changed by passing an array of types and attributes to use after the split.

join(pos: number, depth: ?number = 1, ?bool) → Transform

Join the blocks around the given position. If depth is 2, their last and first siblings are also joined, and so on.

doc: Node

The current document (the result of applying the steps in the transform).

steps: [Step]

The steps in this transform.

docs: [Node]

The documents before each of the steps.

mapping: Mapping

A mapping with the maps for each of the steps in this transform.

before: Node

The document at the start of the transformation.

step(step: Step) → Transform

Apply a new step in this transformation, saving the result. Throws an error when the step fails.

maybeStep(step: Step) → StepResult

Try to apply a step in this transformation, ignoring it if it fails. Returns the step result.

docChanged: bool

True when this transaction changes the document.

The following helper functions can be useful when creating transformations or determining whether they are even possible.

replaceStep(doc: Node, from: number, to: ?number = from, slice: ?Slice = Slice.empty) → ?Step

"Fit" a slice into a given position in the document, producing a step that inserts it.

liftTarget(range: NodeRange) → ?number

Try to find a target depth to which the content in the given range can be lifted.

findWrapping(range: NodeRange, nodeType: NodeType, attrs: ?Object) → ?[{type: NodeType, attrs: ?Object}]

Try to find a valid way to wrap the content in the given range in a node of the given type. May introduce extra nodes around and inside the wrapper node, if necessary. Returns null if no valid wrapping could be found.

canSplit(doc: Node, pos: number, depth: ?[?{type: NodeType, attrs: ?Object}] = 1) → bool

Check whether splitting at the given position is allowed.

canJoin(doc: Node, pos: number) → bool

Test whether the blocks before and after a given position can be joined.

joinPoint(doc: Node, pos: number, dir: ?number = -1) → ?number

Find an ancestor of the given position that can be joined to the block before (or after if dir is positive). Returns the joinable point, if any.

insertPoint(doc: Node, pos: number, nodeType: NodeType, attrs: ?Object) → ?number

Try to find a point where a node of the given type can be inserted near pos, by searching up the node hierarchy when pos itself isn't a valid place but is at the start or end of a node. Return null if no position was found.

commands module

This module exports a number of ‘commands‘, which are building block functions that encapsulate an editing action. A command function takes an editor state and optionally an onAction function that it can use to take an action. It should return a boolean that indicates whether it could perform any action. When no onAction callback is passed, the command should do a 'dry run', determining whether it is applicable, but not actually doing anything.

These are mostly used to bind keys to, and to define menu items.

chainCommands(...commands: [fn(EditorState, ?fn(tr: Transaction)) → bool]) → fn(EditorState, ?fn(tr: Transaction)) → bool

Combine a number of command functions into a single function (which calls them one by one until one returns true).

deleteSelection(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Delete the selection, if there is one.

joinBackward(state: EditorState, dispatch: ?fn(tr: Transaction), view: ?EditorView) → bool

If the selection is empty and at the start of a textblock, move that block closer to the block before it, by lifting it out of its parent or, if it has no parent it doesn't share with the node before it, moving it into a parent of that node, or joining it with that. Will use the view for accurate start-of-textblock detection if given.

joinForward(state: EditorState, dispatch: ?fn(tr: Transaction), view: ?EditorView) → bool

If the selection is empty and the cursor is at the end of a textblock, move the node after it closer to the node with the cursor (lifting it out of parents that aren't shared, moving it into parents of the cursor block, or joining the two when they are siblings). Will use the view for accurate start-of-textblock detection if given.

joinUp(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Join the selected block or, if there is a text selection, the closest ancestor block of the selection that can be joined, with the sibling above it.

joinDown(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Join the selected block, or the closest ancestor of the selection that can be joined, with the sibling after it.

lift(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Lift the selected block, or the closest ancestor block of the selection that can be lifted, out of its parent node.

newlineInCode(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

If the selection is in a node whose type has a truthy code property in its spec, replace the selection with a newline character.

exitCode(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

When the selection is in a node with a truthy code property in its spec, create a default block after the code block, and move the cursor there.

createParagraphNear(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

If a block node is selected, create an empty paragraph before (if it is its parent's first child) or after it.

liftEmptyBlock(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

If the cursor is in an empty textblock that can be lifted, lift the block.

splitBlock(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Split the parent block of the selection. If the selection is a text selection, also delete its content.

splitBlockKeepMarks(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Acts like splitBlock, but without resetting the set of active marks at the cursor.

selectParentNode(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Move the selection to the node wrapping the current selection, if any. (Will not select the document node.)

selectAll(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Select the whole document.

wrapIn(nodeType: NodeType, attrs: ?Object) → fn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Wrap the selection in a node of the given type with the given attributes.

setBlockType(nodeType: NodeType, attrs: ?Object) → fn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Returns a command that tries to set the textblock around the selection to the given node type with the given attributes.

toggleMark(markType: MarkType, attrs: ?Object) → fn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Create a command function that toggles the given mark with the given attributes. Will return false when the current selection doesn't support that mark. This will remove the mark if any marks of that type exist in the selection, or add it otherwise. If the selection is empty, this applies to the stored marks instead of a range of the document.

autoJoin(command: fn(state: EditorState, ?fn(tr: Transaction)) → bool, isJoinable: fn(before: Node, after: Node) → bool | [string]) → fn(state: EditorState, ?fn(tr: Transaction)) → bool

Wrap a command so that, when it produces a transform that causes two joinable nodes to end up next to each other, those are joined. Nodes are considered joinable when they are of the same type and when the isJoinable predicate returns true for them or, if an array of strings was passed, if their node type name is in that array.

baseKeymap: Object

A basic keymap containing bindings not specific to any schema. Binds the following keys (when multiple commands are listed, they are chained with chainCommands:

Enter to newlineInCode, createParagraphNear, liftEmptyBlock, splitBlock
Mod-Enter to exitCode
Backspace to deleteSelection, joinBackward
Mod-Backspace to deleteSelection, joinBackward
Delete to deleteSelection, joinForward
Mod-Delete to deleteSelection, joinForward
Alt-ArrowUp to joinUp
Alt-ArrowDown to joinDown
Mod-BracketLeft to lift
Escape to selectParentNode

history module

An implementation of undo/redo history for ProseMirror.

history(config: ?Object) → Plugin

Returns a plugin that enables the undo history for an editor. The plugin will track undo and redo stacks, which the undo and redo commands can use to move the state back and forward.

Note that this implementation doesn't implement history by simply resetting back to some previous state. In order to support collaborative editing (as well as some other use cases), it selectively rolls back some transactions, but not other (for example, not the changes made by other users). You can set an "addToHistory" metadata property of false on a transaction to prevent it from being rolled back by undo.

config: ?Object

Supports the following configuration options:

depth: ?number: The amount of history events that are collected before the oldest events are discarded. Defaults to 100.

undo(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

A command function that undoes the last change, if any.

redo(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

A command function that redoes the last undone change, if any.

undoDepth(state: EditorState) → number

The amount of undoable events available in a given state.

redoDepth(state: EditorState) → number

The amount of redoable events available in a given editor state.

closeHistory(tr: Transaction) → Transaction

Set a flag on the given transaction that will prevent further steps from being appended to an existing history event (so that they require a separate undo command to undo).

collab module

This module implements an API into which a communication channel for collaborative editing can be hooked. See this guide for more details and an example.

collab(config: ?Object = {}) → Plugin

Creates a plugin that enables the collaborative editing framework for the editor.

config: ?Object

An optional set of options

version: ?number: The starting version number of the collaborative editing. Defaults to 0.

getVersion(state: EditorState) → number

Get the version up to which the collab plugin has synced with the central authority.

receiveTransaction(state: EditorState, steps: [Step], clientIDs: [number | string]) → Transaction

Create a transaction that represents a set of new steps received from the authority. Applying this transaction moves the state forward to adjust to the authority's view of the document.

sendableSteps(state: EditorState) → ?{version: number, steps: [Step], clientID: number | string, origins: [Transaction]}

Provides the data describing the editor's unconfirmed steps, which you'd send to the central authority. Returns null when there is nothing to send.

origins holds the original transactions that produced each steps. This can be useful for looking up time stamps and other metadata for the steps, but note that the steps may have been rebased, whereas the origin transactions are still the old, unchanged objects.

keymap module

A plugin for conveniently defining key bindings.

keymap(bindings: Object) → Plugin

Create a keymap plugin for the given set of bindings.

Bindings should map key names to command-style functions, which will be called with (EditorState, dispatch, EditorView) arguments, and should return true when they've handled the key. Note that the view argument isn't part of the command protocol, but can be used as an escape hatch if a binding needs to directly interact with the UI.

Key names may be strings like "Shift-Ctrl-Enter", a key identifier prefixed with zero or more modifiers. Key identifiers are based on the strings that can appear in KeyEvent.key. Use lowercase letters to refer to letter keys (or uppercase letters if you want shift to be held). You may use "Space" as an alias for the " " name.

Modifiers can be given in any order. Shift- (or s-), Alt- (or a-), Ctrl- (or c- or Control-) and Cmd- (or m- or Meta-) are recognized. For characters that are created by holding shift, the Shift- prefix is implied, and should not be added explicitly.

You can use Mod- as a shorthand for Cmd- on Mac and Ctrl- on other platforms.

You can add multiple keymap plugins to an editor. The order in which they appear determines their precedence (the ones early in the array get to dispatch first).

keydownHandler(bindings: Object) → fn(view: EditorView, event: dom.Event) → bool

Given a keymap, return a keydown handler that implements the bindings for that map, using the same rules as keymap.

inputrules module

This module defines a plugin for attaching ‘input rules’ to an editor, which can react to or transform text typed by the user. It also comes with a bunch of default rules that can be enabled in this plugin.

InputRule class

Input rules are regular expressions describing a piece of text that, when typed, causes something to happen. This might be changing two dashes into an emdash, wrapping a paragraph starting with "> " into a blockquote, or something entirely different.

new InputRule(match: RegExp, handler: string | fn(state: EditorState, match: [string], start: number, end: number) → ?Transaction)

Create an input rule. The rule applies when the user typed something and the text directly in front of the cursor matches match, which should probably end with $.

The handler can be a string, in which case the matched text, or the first matched group in the regexp, simply be replaced by that string.

Or a it can be a function, which will be called with the match array produced by RegExp.exec, as well as the start and end of the matched range, and which can return a transaction that describes the rule's effect, or null to indicate the input was not handled.

inputRules(config: {rules: [InputRule]}) → Plugin

Create an input rules plugin. When enabled, it will cause text input that matches any of the given rules to trigger the rule's action, and binds the backspace key, when applied directly after an input rule triggered, to undo the rule's effect.

undoInputRule(state: EditorState, dispatch: ?fn(Transaction)) → bool

Command that will undo an input rule, if it applied to the last thing that the user did.

The module comes with a number of predefined rules:

emDash: InputRule

Converts double dashes to an emdash.

ellipsis: InputRule

Converts three dots to an ellipsis character.

openDoubleQuote: InputRule

“Smart” opening double quotes.

closeDoubleQuote: InputRule

“Smart” closing double quotes.

openSingleQuote: InputRule

“Smart” opening single quotes.

closeSingleQuote: InputRule

“Smart” closing single quotes.

smartQuotes: [InputRule]

Smart-quote related input rules.

allInputRules: [InputRule]

All schema-independent input rules defined in this module.

These utility functions take schema-specific parameters and create input rules specific to that schema.

wrappingInputRule(regexp: RegExp, nodeType: NodeType, getAttrs: ?Object | fn([string]) → ?Object, joinPredicate: ?fn([string], Node) → bool) → InputRule

Build an input rule for automatically wrapping a textblock when a given string is typed. The regexp argument is directly passed through to the InputRule constructor. You'll probably want the regexp to start with ^, so that the pattern can only occur at the start of a textblock.

nodeType is the type of node to wrap in. If it needs attributes, you can either pass them directly, or pass a function that will compute them from the regular expression match.

By default, if there's a node with the same type above the newly wrapped node, the rule will try to join those two nodes. You can pass a join predicate, which takes a regular expression match and the node before the wrapped node, and can return a boolean to indicate whether a join should happen.

textblockTypeInputRule(regexp: RegExp, nodeType: NodeType, getAttrs: ?Object | fn([string]) → ?Object) → InputRule

Build an input rule that changes the type of a textblock when the matched text is typed into it. You'll usually want to start your regexp with ^ to that it is only matched at the start of a textblock. The optional getAttrs parameter can be used to compute the new node's attributes, and works the same as in the wrappingInputRule function.

blockQuoteRule(nodeType: NodeType) → InputRule

Given a blockquote node type, returns an input rule that turns "> " at the start of a textblock into a blockquote.

orderedListRule(nodeType: NodeType) → InputRule

Given a list node type, returns an input rule that turns a number followed by a dot at the start of a textblock into an ordered list.

bulletListRule(nodeType: NodeType) → InputRule

Given a list node type, returns an input rule that turns a bullet (dash, plush, or asterisk) at the start of a textblock into a bullet list.

codeBlockRule(nodeType: NodeType) → InputRule

Given a code block node type, returns an input rule that turns a textblock starting with three backticks into a code block.

headingRule(nodeType: NodeType, maxLevel: number) → InputRule

Given a node type and a maximum level, creates an input rule that turns up to that number of # characters followed by a space at the start of a textblock into a heading whose level corresponds to the number of # signs.

schema-basic module

This module defines a simple schema, to use directly, extend, or to reuse pieces in a schema you assemble.

schema: Schema

This schema rougly corresponds to the document schema used by CommonMark, minus the list elements, which are defined in the schema-list module.

To reuse elements from this schema, extend or read from its spec.nodes and spec.marks properties.

nodes: Object

doc: NodeSpec: The top level document node.

marks: Object

link: MarkSpec: A link. Has href and title attributes. title defaults to the empty string.

schema-list module

This module exports list-related schema elements and commands. The commands assume which assume lists to be nestable, but with the restriction that the first child of a list item is a plain paragraph.

These are the node specs:

orderedList: NodeSpec

An ordered list node type spec. Has a single attribute, order, which determines the number at which the list starts counting, and defaults to 1.

bulletList: NodeSpec

A bullet list node spec.

listItem: NodeSpec

A list item node spec.

You can extend a schema with this helper function.

addListNodes(nodes: OrderedMap, itemContent: string, listGroup: ?string) → OrderedMap

Convenience function for adding list-related node types to a map describing the nodes in a schema. Adds OrderedList as "ordered_list", BulletList as "bullet_list", and ListItem as "list_item". itemContent determines the content expression for the list items. If you want the commands defined in this module to apply to your list structure, it should have a shape like "paragraph block*", a plain textblock type followed by zero or more arbitrary nodes. listGroup can be given to assign a group name to the list node types, for example "block".

Using this would look something like this:

const mySchema = new Schema({
  nodes: addListNodes(baseSchema.nodeSpec, "paragraph block*", "block"),
  marks: baseSchema.markSpec
})

The following functions are commands:

wrapInList(listType: NodeType, attrs: ?Object) → fn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Returns a command function that wraps the selection in a list with the given type an attributes. If apply is false, only return a value to indicate whether this is possible, but don't actually perform the change.

splitListItem(itemType: NodeType) → fn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Build a command that splits a non-empty textblock at the top level of a list item by also splitting that list item.

liftListItem(itemType: NodeType) → fn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Create a command to lift the list item around the selection up into a wrapping list.

sinkListItem(itemType: NodeType) → fn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Create a command to sink the list item around the selection down into an inner list.

schema-table module

This module defines schema elements and commands to integrate tables in your editor.

Note that Firefox will, by default, add various kinds of controls to editable tables, even though those don't work in ProseMirror. The only way to turn these off is globally, which you might want to do with the following code:

document.execCommand("enableObjectResizing", false, "false")
document.execCommand("enableInlineTableEditing", false, "false")

These are the node specs for basic table support:

table: NodeSpec

A table node spec. Has one attribute, columns, which holds a number indicating the amount of columns in the table.

tableRow: NodeSpec

A table row node spec. Has one attribute, columns, which holds a number indicating the amount of columns in the table.

tableCell: NodeSpec

A table cell node spec.

Two special step implementations are necessary to atomically add or remove columns. You probably don't have to interact with these directly.

AddColumnStep class extends `Step`

A Step subclass for adding a column to a table in a single atomic step.

static create(doc: Node, tablePos: number, columnIndex: number, cellType: NodeType, cellAttrs: ?Object) → AddColumnStep: Create a step that inserts a column into the table after tablePos, at the index given by columnIndex, using cells with the given type and attributes.

RemoveColumnStep class extends `Step`

A subclass of Step that removes a column from a table.

static create(doc: Node, tablePos: number, columnIndex: number) → RemoveColumnStep: Create a step that deletes the column at columnIndex in the table after tablePos.

And some utility functions:

addTableNodes(nodes: OrderedMap, cellContent: string, tableGroup: ?string) → OrderedMap

Convenience function for adding table-related node types to a map describing the nodes in a schema. Adds Table as "table", TableRow as "table_row", and TableCell as "table_cell". cellContent should be a content expression describing what may occur inside cells.

createTable(nodeType: NodeType, rows: number, columns: number, attrs: ?Object) → Node

Create a table node with the given number of rows and columns.

And a number of table-related commands:

addColumnBefore(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Command function that adds a column before the column with the selection.

addColumnAfter(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Command function that adds a column after the column with the selection.

removeColumn(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Command function that removes the column with the selection.

addRowBefore(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Command function that adds a row after the row with the selection.

addRowAfter(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Command function that adds a row before the row with the selection.

removeRow(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Command function that removes the row with the selection.

selectNextCell(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Move to the next cell in the current table, if there is one.

selectPreviousCell(state: EditorState, dispatch: ?fn(tr: Transaction)) → bool

Move to the previous cell in the current table, if there is one.