tree: d7ec11a99395dfa4376ea5fa1629c8f5f2a068e6 [path history] [tgz]
  1. BUILD.gn
  2. OWNERS
  3. README.md
  4. WhitespaceLayoutObjects.md
  5. abort_controller.cc
  6. abort_controller.h
  7. abort_controller.idl
  8. abort_signal.cc
  9. abort_signal.h
  10. abort_signal.idl
  11. accessibility_role.idl
  12. aria_attributes.idl
  13. attr.cc
  14. attr.h
  15. attr.idl
  16. attr_test.cc
  17. attribute.h
  18. attribute_collection.h
  19. beforeunload_event_listener.cc
  20. beforeunload_event_listener.h
  21. cdata_section.cc
  22. cdata_section.h
  23. cdata_section.idl
  24. character_data.cc
  25. character_data.h
  26. character_data.idl
  27. child_frame_disconnector.cc
  28. child_frame_disconnector.h
  29. child_list_mutation_scope.cc
  30. child_list_mutation_scope.h
  31. child_node.h
  32. child_node.idl
  33. child_node_list.cc
  34. child_node_list.h
  35. class_collection.cc
  36. class_collection.h
  37. collection_index_cache.h
  38. comment.cc
  39. comment.h
  40. comment.idl
  41. common_definitions.idl
  42. container_node.cc
  43. container_node.h
  44. context_features.cc
  45. context_features.h
  46. context_features_client_impl.cc
  47. context_features_client_impl.h
  48. create_element_flags.h
  49. dataset_dom_string_map.cc
  50. dataset_dom_string_map.h
  51. decoded_data_document_parser.cc
  52. decoded_data_document_parser.h
  53. distributed_nodes.cc
  54. distributed_nodes.h
  55. document.cc
  56. document.h
  57. document.idl
  58. document_and_element_event_handlers.h
  59. document_and_element_event_handlers.idl
  60. document_encoding_data.cc
  61. document_encoding_data.h
  62. document_fragment.cc
  63. document_fragment.h
  64. document_fragment.idl
  65. document_init.cc
  66. document_init.h
  67. document_lifecycle.cc
  68. document_lifecycle.h
  69. document_or_shadow_root.h
  70. document_or_shadow_root.idl
  71. document_parser.cc
  72. document_parser.h
  73. document_parser_client.h
  74. document_parser_timing.cc
  75. document_parser_timing.h
  76. document_shutdown_notifier.cc
  77. document_shutdown_notifier.h
  78. document_shutdown_observer.cc
  79. document_shutdown_observer.h
  80. document_statistics_collector.cc
  81. document_statistics_collector.h
  82. document_statistics_collector_test.cc
  83. document_test.cc
  84. document_timing.cc
  85. document_timing.h
  86. document_type.cc
  87. document_type.h
  88. document_type.idl
  89. dom_exception.cc
  90. dom_exception.h
  91. dom_exception.idl
  92. dom_high_res_time_stamp.h
  93. dom_implementation.cc
  94. dom_implementation.h
  95. dom_implementation.idl
  96. dom_implementation_test.cc
  97. dom_node_ids.cc
  98. dom_node_ids.h
  99. dom_node_ids_test.cc
  100. dom_string_list.cc
  101. dom_string_list.h
  102. dom_string_list.idl
  103. dom_string_map.cc
  104. dom_string_map.h
  105. dom_string_map.idl
  106. dom_time_stamp.h
  107. dom_token_list.cc
  108. dom_token_list.h
  109. dom_token_list.idl
  110. element.cc
  111. element.h
  112. element.idl
  113. element_creation_options.idl
  114. element_data.cc
  115. element_data.h
  116. element_data_cache.cc
  117. element_data_cache.h
  118. element_definition_options.idl
  119. element_rare_data.cc
  120. element_rare_data.h
  121. element_registration_options.idl
  122. element_test.cc
  123. element_traversal.h
  124. empty_node_list.cc
  125. empty_node_list.h
  126. events/
  127. first_letter_pseudo_element.cc
  128. first_letter_pseudo_element.h
  129. first_letter_pseudo_element_test.cc
  130. flat_tree_node_data.cc
  131. flat_tree_node_data.h
  132. flat_tree_traversal.cc
  133. flat_tree_traversal.h
  134. flat_tree_traversal_forbidden_scope.h
  135. flat_tree_traversal_test.cc
  136. frame_request_callback.idl
  137. frame_request_callback_collection.cc
  138. frame_request_callback_collection.h
  139. function_string_callback.idl
  140. get_root_node_options.idl
  141. global_event_handlers.h
  142. global_event_handlers.idl
  143. icon_url.cc
  144. icon_url.h
  145. id_target_observer.cc
  146. id_target_observer.h
  147. id_target_observer_registry.cc
  148. id_target_observer_registry.h
  149. idle_deadline.cc
  150. idle_deadline.h
  151. idle_deadline.idl
  152. idle_deadline_test.cc
  153. idle_request_callback.idl
  154. idle_request_options.idl
  155. ignore_opens_during_unload_count_incrementer.h
  156. increment_load_event_delay_count.cc
  157. increment_load_event_delay_count.h
  158. iterator.h
  159. iterator.idl
  160. layout_tree_builder.cc
  161. layout_tree_builder.h
  162. layout_tree_builder_traversal.cc
  163. layout_tree_builder_traversal.h
  164. layout_tree_builder_traversal_test.cc
  165. live_node_list.cc
  166. live_node_list.h
  167. live_node_list_base.cc
  168. live_node_list_base.h
  169. live_node_list_registry.cc
  170. live_node_list_registry.h
  171. live_node_list_registry_test.cc
  172. mutation_observer.cc
  173. mutation_observer.h
  174. mutation_observer.idl
  175. mutation_observer_init.idl
  176. mutation_observer_interest_group.cc
  177. mutation_observer_interest_group.h
  178. mutation_observer_options.h
  179. mutation_observer_registration.cc
  180. mutation_observer_registration.h
  181. mutation_observer_test.cc
  182. mutation_record.cc
  183. mutation_record.h
  184. mutation_record.idl
  185. name_node_list.cc
  186. name_node_list.h
  187. named_node_map.cc
  188. named_node_map.h
  189. named_node_map.idl
  190. names_map.cc
  191. names_map.h
  192. names_map_test.cc
  193. node.cc
  194. node.h
  195. node.idl
  196. node_child_removal_tracker.cc
  197. node_child_removal_tracker.h
  198. node_computed_style.h
  199. node_filter.idl
  200. node_iterator.cc
  201. node_iterator.h
  202. node_iterator.idl
  203. node_iterator_base.cc
  204. node_iterator_base.h
  205. node_list.h
  206. node_list.idl
  207. node_lists_node_data.cc
  208. node_lists_node_data.h
  209. node_rare_data.cc
  210. node_rare_data.h
  211. node_test.cc
  212. node_traversal.cc
  213. node_traversal.h
  214. node_traversal_strategy.h
  215. node_with_index.h
  216. non_document_type_child_node.h
  217. non_document_type_child_node.idl
  218. non_element_parent_node.h
  219. non_element_parent_node.idl
  220. nonced_element.idl
  221. nth_index_cache.cc
  222. nth_index_cache.h
  223. nth_index_cache_test.cc
  224. parent_node.h
  225. parent_node.idl
  226. parser_content_policy.h
  227. presentation_attribute_style.cc
  228. presentation_attribute_style.h
  229. processing_instruction.cc
  230. processing_instruction.h
  231. processing_instruction.idl
  232. pseudo_element.cc
  233. pseudo_element.h
  234. pseudo_element_data.h
  235. qualified_name.cc
  236. qualified_name.h
  237. range.cc
  238. range.h
  239. range.idl
  240. range_boundary_point.h
  241. range_test.cc
  242. raw_data_document_parser.h
  243. scoped_window_focus_allowed_indicator.h
  244. scriptable_document_parser.cc
  245. scriptable_document_parser.h
  246. scripted_animation_controller.cc
  247. scripted_animation_controller.h
  248. scripted_animation_controller_test.cc
  249. scripted_idle_task_controller.cc
  250. scripted_idle_task_controller.h
  251. scripted_idle_task_controller_test.cc
  252. scripted_task_queue.cc
  253. scripted_task_queue.h
  254. scripted_task_queue.idl
  255. scripted_task_queue_controller.cc
  256. scripted_task_queue_controller.h
  257. scripted_task_queue_controller.idl
  258. shadow_dom_v0_test.cc
  259. shadow_root.cc
  260. shadow_root.h
  261. shadow_root.idl
  262. shadow_root_init.idl
  263. shadow_root_v0.cc
  264. shadow_root_v0.h
  265. sink_document.cc
  266. sink_document.h
  267. slot_assignment.cc
  268. slot_assignment.h
  269. slot_assignment_engine.cc
  270. slot_assignment_engine.h
  271. slot_assignment_recalc_forbidden_scope.h
  272. slot_assignment_test.cc
  273. space_split_string.cc
  274. space_split_string.h
  275. space_split_string_test.cc
  276. static_node_list.h
  277. static_range.cc
  278. static_range.h
  279. static_range.idl
  280. static_range_test.cc
  281. synchronous_mutation_notifier.cc
  282. synchronous_mutation_notifier.h
  283. synchronous_mutation_observer.cc
  284. synchronous_mutation_observer.h
  285. tag_collection.cc
  286. tag_collection.h
  287. task_type_traits.h
  288. template_content_document_fragment.h
  289. text.cc
  290. text.h
  291. text.idl
  292. text_link_colors.cc
  293. text_link_colors.h
  294. text_test.cc
  295. throw_on_dynamic_markup_insertion_count_incrementer.h
  296. transform_source.h
  297. transform_source_libxslt.cc
  298. traversal_range.h
  299. tree_ordered_list.cc
  300. tree_ordered_list.h
  301. tree_ordered_map.cc
  302. tree_ordered_map.h
  303. tree_scope.cc
  304. tree_scope.h
  305. tree_scope_adopter.cc
  306. tree_scope_adopter.h
  307. tree_scope_adopter_test.cc
  308. tree_scope_test.cc
  309. tree_walker.cc
  310. tree_walker.h
  311. tree_walker.idl
  312. user_action_element_set.cc
  313. user_action_element_set.h
  314. user_gesture_indicator.cc
  315. user_gesture_indicator.h
  316. user_gesture_indicator_test.cc
  317. v0_insertion_point.cc
  318. v0_insertion_point.h
  319. visited_link_state.cc
  320. visited_link_state.h
  321. void_function.idl
  322. weak_identifier_map.h
  323. weak_identifier_map_test.cc
  324. whitespace_attacher.cc
  325. whitespace_attacher.h
  326. whitespace_attacher_test.cc
  327. xml_document.cc
  328. xml_document.h
  329. xml_document.idl
third_party/blink/renderer/core/dom/README.md

DOM

Rendered

Author: hayato@chromium.org

The renderer/core/dom directory contains the implementation of DOM.

Basically, this directory should contain only a file which is related to DOM Standard. However, for historical reasons, renderer/core/dom directory has been used as if it were misc directory. As a result, unfortunately, this directory contains a lot of files which are not directly related to DOM.

Please don‘t add unrelated files to this directory any more. We are trying to organize the files so that developers wouldn’t get confused at seeing this directory.

  • See the spreadsheet, as a rough plan to organize Source/core/dom files.

    The classification in the spreadsheet might be wrong. Please update the spreadsheet, and move files if you can, if you know more appropriate places for each file.

  • See crbug.com/738794 for tracking our efforts.

Node and Node Tree

In this README, we draw a tree in left-to-right direction in ascii-art notation. A is the root of the tree.

A
├───B
├───C
   ├───D
   └───E
└───F

Node is a base class of all kinds of nodes in a node tree. Each Node has following 3 pointers (but not limited to):

  • parent_or_shadow_host_node_: Points to the parent (or the shadow host if it is a shadow root; explained later)
  • previous_: Points to the previous sibling
  • next_: Points to the next sibling

ContainerNode, from which Element extends, has additional pointers for its child:

  • first_child_: The meaning is obvious.
  • last_child_: Nit.

That means:

  • Siblings are stored as a linked list. It takes O(N) to access a parent's n-th child.
  • Parent can't tell how many children it has in O(1).

next sibling and previous sibling

Further info:

  • Node, ContainerNode

C++11 range-based for loops for traversing a tree

You can traverse a tree manually:

// In C++

// Traverse a children.
for (Node* child = parent.firstChild(); child; child = child->nextSibling()) {
  ...
}

// ...

// Traverse nodes in tree order, depth-first traversal.
void foo(const Node& node) {
  ...
  for (Node* child = node.firstChild(); child; child = child->nextSibling()) {
    foo(*child);  // Recursively
  }
}

Tree order is:

tree order

However, traversing a tree in this way might be error-prone. Instead, you can use NodeTraversal and ElementTraversal. They provides a C++11's range-based for loops, such as:

// In C++
for (Node& child : NodeTraversal::childrenOf(parent) {
  ...
}

e.g. Given a parent A, this traverses B, C, and F in this order.

// In C++
for (Node& node : NodeTraversal::startsAt(root)) {
  ...
}

e.g. Given the root A, this traverses A, B, C, D, E, and F in this order.There are several other useful range-based for loops for each purpose. The cost of using range-based for loops is zero because everything can be inlined.

Further info:

  • NodeTraversal and ElementTraversal (more type-safe version)
  • The CL, which introduced these range-based for loops.

Shadow Tree

A shadow tree is a node tree whose root is a ShadowRoot. From web developer's perspective, a shadow root can be created by calling element.attachShadow{ ... } API. The element here is called a shadow host, or just a host if the context is clear.

  • A shadow root is always attached to another node tree through its host. A shadow tree is therefore never alone.
  • The node tree of a shadow root’s host is sometimes referred to as the light tree.

shadow tree

For example, given the example node tree:

A
├───B
├───C
   ├───D
   └───E
└───F

Web developers can create a shadow root, and manipulate the shadow tree in the following way:

// In JavaScript
const b = document.querySelector('#B');
const shadowRoot = b.attachShadow({ mode: 'open' });
const sb = document.createElement('div');
shadowRoot.appendChild(sb);

The resulting shadow tree would be:

shadowRoot
└── sb

The shadowRoot has one child, sb. This shadow tree is being attached to B:

A
└── B
    ├──/shadowRoot
       └── sb
    ├── C
       ├── D
       └── E
    └── F

In this README, a notation (──/) is used to represent a shadowhost-shadowroot relationship, in a composed tree. A composed tree will be explained later. A shadowhost-shadowroot is 1:1 relationship.

Though a shadow root has always a corresponding shadow host element, a light tree and a shadow tree should be considered separately, from a node tree's perspective. (──/) is NOT a parent-child relationship in a node tree.

For example, even though B hosts the shadow tree, shadowRoot is not considered as a child of B. The means the following traversal:

// In C++
for (Node& node : NodeTraversal::startsAt(A)) {
  ...
}

traverses only A, B, C, D, E and F nodes. It never visits shadowRoot nor sb. NodeTraversal never cross a shadow boundary, ──/.

Further info:

  • ShadowRoot
  • Element#attachShadow

TreeScope

Document and ShadowRoot are always the root of a node tree. BothDocument and ShadowRoot implements TreeScope.

TreeScope maintains a lot of information about the underlying tree for efficiency. For example, TreeScope has a id-to-element mapping, as TreeOrderedMap, so that querySelector('#foo') can find an element whose id attribute is “foo” in O(1). In other words, root.querySelector('#foo') can be slow if that is used in a node tree whose root is not TreeScope.

Each Node has tree_scope_ pointer, which points to:

  • The root node: if the node's root is either Document or ShadowRoot.
  • owner document, otherwise.

The means tree_scope_ pointer is always non-null (except for while in a DOM mutation), but it doesn‘t always point to the node’s root.

Since each node doesn't have a pointer which always points to the root, Node::getRootNode(...) may take O(N) if the node is neither in a document tree nor in a shadow tree. If the node is in TreeScope (Node#IsInTreeScope() can tell it), we can get the root in O(1).

Each node has flags, which is updated in DOM mutation, so that we can tell whether the node is in a document tree, in a shadow tree, or in none of them, by using Node::IsInDocumentTree() and/or Node::IsInShadowTree().

If you want to add new features to Document, Document might be a wrong place to add. Instead, please consider to add functionality to TreeScope. We want to treat a document tree and a shadow tree equally as much as possible.

Example

document
└── a1
    ├──/shadowRoot1
       └── s1
    └── a2
        └── a3

document-fragment
└── b1
    ├──/shadowRoot2
       └── t2
    └── b2
        └── b3
  • Here, there are 4 node trees; The root node of each tree is document, shadowRoot1, document-fragment, and shadowRoot2.
  • Suppose that each node is created by document.createElement(...) (except for Document and ShadowRoot). That means each node's owner document is document.
nodenode's rootnode's _tree_scope points to:
documentdocument (self)document (self)
a1documentdocument
a2documentdocument
a3documentdocument
shadowRoot1shadowRoot1 (self)shadowRoot1 (self)
s1shadowRoot1shadowRoot1
document-fragmentdocument-fragment (self)document
b1document-fragmentdocument
b2document-fragmentdocument
b3document-fragmentdocument
shadowRoot2shadowRoot2 (self)shadowRoot2 (self)
t1shadowRoot2shadowRoot2

Further Info:

Composed Tree (a tree of node trees)

In the previous picture, you might think that more than one node trees, a document tree and a shadow tree, were connected to each other. That is true in some sense. We call this super tree as composed tree, which is a tree of trees.

super tree

The following is a complex example:

document
├── a1 (host)
   ├──/shadowRoot1
      └── b1
   └── a2 (host)
       ├──/shadowRoot2
          ├── c1
             ├── c2
             └── c3
          └── c4
       ├── a3
       └── a4
└── a5
    └── a6 (host)
        └──/shadowRoot3
            └── d1
                ├── d2
                ├── d3 (host)
                   └──/shadowRoot4
                       ├── e1
                       └── e2
                └── d4 (host)
                    └──/shadowRoot5
                        ├── f1
                        └── f2

If you see this carefully, you can notice that this composed tree is composed of 6 node trees; 1 document tree and 5 shadow trees:

  • document tree

    document
    ├── a1 (host)
       └── a2 (host)
           ├── a3
           └── a4
    └── a5
        └── a6 (host)
    
  • shadow tree 1

    shadowRoot1
    └── b1
    
  • shadow tree 2

    shadowRoot2
    ├── c1
       ├── c2
       └── c3
    └── c4
    
  • shadow tree 3

    shadowRoot3
    └── d1
        ├── d2
        ├── d3 (host)
        └── d4 (host)
    
  • shadow tree 4

    shadowRoot4
    ├── e1
    └── e2
    
  • shadow tree 5

    shadowRoot5
    ├── f1
    └── f2
    

If we consider each node tree as node of a super-tree, we can draw a super-tree as such:

document
├── shadowRoot1
├── shadowRoot2
└── shadowRoot3
    ├── shadowRoot4
    └── shadowRoot5

Here, a root node is used as a representative of each node tree; A root node and a node tree itself can be sometimes exchangeable in explanations.

We call this kind of a super-tree (a tree of node trees) a composed tree. The concept of a composed tree is very useful to understand how Shadow DOM's encapsulation works.

DOM Standard defines the following terminologies:

For example,

  • d1's shadow-including ancestor nodes are shadowRoot3, a6, a5, and document
  • d1's shadow-including descendant nodes are d2, d3, shadowRoot4, e1, e2, d4, shadowRoot5, f1, and f2.

To honor Shadow DOM's encapsulation, we have a concept of visibility relationship between two nodes.

In the following table, “-” means that “node A is visible from node B”.

A \ Bdocumenta1a2b1c1d1d2e1f1
document---------
a1---------
a2---------
b1hiddenhiddenhidden-hiddenhiddenhiddenhiddenhidden
c1hiddenhiddenhiddenhidden-hiddenhiddenhiddenhidden
d1hiddenhiddenhiddenhiddenhidden----
d2hiddenhiddenhiddenhiddenhidden----
e1hiddenhiddenhiddenhiddenhiddenhiddenhidden-hidden
f1hiddenhiddenhiddenhiddenhiddenhiddenhiddenhidden-

For example, document is visible from any nodes.

To understand visibility relationship easily, here is a rule of thumb:

  • If node B can reach node A by traversing an edge (in the first picture of this section), recursively, A is visible from B.
  • However, an edge of (──/) ( shadowhost-shadowroot relationship) is one-directional:
    • From a shadow root to the shadow host -> Okay
    • From a shadow host to the shadow root -> Forbidden

In other words, a node in an inner tree can see a node in an outer tree in a composed tree, but the opposite is not true.

We have designed (or re-designed) a bunch of Web-facing APIs to honor this basic principle. If you add a new API to the web platform and Blink, please consider this rule and don't leak a node which should be hidden to web developers.

Warning: Unfortunately, a composed tree had a different meaning in the past; it was used to specify a flat tree (which will be explained later). If you find a wrong usage of a composed tree in Blink, please fix it.

Further Info:

  • TreeScope::ParentTreeScope()
  • Node::IsConnected()
  • DOM Standard: connected
  • DOM Standard: retarget

Flat tree

A composed tree itself can‘t be rendered as is. From the rendering’s perspective, Blink has to construct a layout tree, which would be used as an input to the paint phase. A layout tree is a tree whose node is LayoutObject, which points to Node in a node tree, plus additional calculated layout information.

Before the Web Platform got Shadow DOM, the structure of a layout tree is almost similar to the structure of a document tree; where only one node tree, document tree, is being involved there.

Since the Web Platform got Shadow DOM, we now have a composed tree which is composed of multiple node trees, instead of a single node tree. That means We have to flatten the composed tree to the one node tree, called a flat tree, from which a layout tree is constructed.

flat tree

For example, given the following composed tree,

document
├── a1 (host)
   ├──/shadowRoot1
      └── b1
   └── a2 (host)
       ├──/shadowRoot2
          ├── c1
             ├── c2
             └── c3
          └── c4
       ├── a3
       └── a4
└── a5
    └── a6 (host)
        └──/shadowRoot3
            └── d1
                ├── d2
                ├── d3 (host)
                   └──/shadowRoot4
                       ├── e1
                       └── e2
                └── d4 (host)
                    └──/shadowRoot5
                        ├── f1
                        └── f2

This composed tree would be flattened into the following flat tree (assuming there are not <slot> elements there):

document
├── a1 (host)
   └── b1
└── a5
    └── a6 (host)
        └── d1
            ├── d2
            ├── d3 (host)
               ├── e1
               └── e2
            └── d4 (host)
                ├── f1
                └── f2

We can't explain the exact algorithm how to flatten a composed tree into a flat tree until I explain the concept of slots and node distribution If we are ignoring the effect of <slot>, we can have the following simple definition. A flat tree can be defined as:

  • A root of a flat tree: document
  • Given node A which is in a flat tree, its children are defined, recursively, as follows:
    • If A is a shadow host, its shadow root's children
    • Otherwise, A's children

Slots and node assignments

Please see this nice article how <slot> elements work in general.

Slots are placeholders inside your component that users can fill with their own markup.

Here, I'll show some examples.

Example 1

Given the following composed tree and slot assignments,

Composed tree:

A
├──/shadowRoot1
   ├── slot1
   └── slot2
├── B
└── C

Slot Assignments:

slotslot's assigned nodes
slot1[C]
slot2[B]

The flat tree would be:

A
├── slot1
   └── C
└── slot2
    └── B

Example 2

More complex example is here.

Composed tree:

A
├──/shadowRoot1
   ├── B
      └── slot1
   ├── slot2
      └── C
   ├── D
   └── slot3
       ├── E
       └── F
├── G
├── H
├── I
└── J

Slot Assignments:

slotslot's assigned nodes
slot1[H]
slot2[G, I]
slot3[] (nothing is assigned)

The flat tree would be:

A
├── B
   └── slot1
       └── H
├── slot2
   ├── G
   └── I
├── D
└── slot3
    ├── E
    └── F
  • slot2's child, C, is not shown in this flat tree because slot2 has non-empty assigned nodes, [G, I], which are used as slot2's children in the flat tree.
  • If a slots doesn‘t have any assigned nodes, the slot’s children are used as fallback contents in the flat tree. e.g. slot3s children in the flat tree are E and F.
  • If a host's child node is assigned to nowhere, the child is not used. e.g. J

Example 3

A slot itself can be assigned to another slot.

For example, if we attach a shadow root to B, and put a <slot>, slot4, inside of the shadow tree.

A
├──/shadowRoot1
   ├── B
      ├──/shadowRoot2
         └── K
             └── slot4
      └── slot1
   ├── slot2
      └── C
   ├── D
   └── slot3
       ├── E
       └── F
├── G
├── H
├── I
└── J
slotslot's assigned nodes
slot1[H]
slot2[G, I]
slot3[] (nothing is assigned)
slot4[slot1]

The flat tree would be:

A
├── B
   └── K
       └── slot4
           └── slot1
               └── H
├── slot2
   ├── G
   └── I
├── D
└── slot3
    ├── E
    └── F

Slot Assignment Recalc

Please see Incremental Shadow DOM to know how assignments are recalc-ed.

FlatTreeTraversal

Blink doesn't store nor maintain a flat tree data structure in the memory. Instead, Blink provides a utility class, FlatTreeTraversal, which traverses a composed tree in a flat tree order.

e.g. in the above example 3,

  • FlatTreeTraversal::firstChild(slot1) returns H
  • FlatTreeTraversal::parent(H) returns slot1
  • FlatTreeTraversal::nextSibling(G) returns I
  • FlatTreeTraversal::previousSibling(I) returns G

The APIs which FlatTreeTraversal provides are very similar to ones other traversal utility classes provide, such as NodeTraversal and ElementTraversal.

Warning

For historical reasons, Blink still supports Shadow DOM v0, where the different node distribution mechanism is still used. To support v0, you need to call Node::UpdateDistributionForFlatTreeTraversal before calling any function of FlatTreeTraversal.

If you use FlatTreeTraversal without updating distribution, you would hit DCHECK. :(

Since Node::UpdateDistributionForFlatTreeTraversal can take O(N) in the worst case (even if the distribution flag is clean!), you should be careful not to call it in hot code paths. If you are not sure, please contact dom-dev@chromium.org, or add hayato@chromium.org to reviewers.

Once Blink removes Shadow DOM v0 in the future, you don‘t need to call Node::UpdateDistributionForFlatTreeTraversal before using FlatTreeTraversal beforehand in most cases, however, that wouldn’t happen soon.

Event path and Event Retargeting

DOM Standard defines how an event should be dispatched here, including how event path should be calculated, however, I wouldn't be surprised if the steps described there might look a kind of cryptogram to you.

In this README, I'll explain how an event is dispatched and how its event path is calculated briefly by using some relatively-understandable examples.

Basically, an event is dispatched across shadow trees.

event dispatch

Let me show more complex example composed tree, involving a slot:

A
└── B
    ├──/shadowroot-C
       └── D
           ├──/shadowroot-E
              └── F
                  └── slot-G
           └── H
               └── I
                   ├──/shadowroot-J
                      └── K
                          ├──/shadowroot-L
                             └── M
                                 ├──/shadowroot-N
                                    └── slot-O
                                 └── slot-P
                          └── Q
                              └── slot-R
                   └── slot-S
    └── T
        └── U

Slot Assignments:

slotslot's assigned nodes
slot-G[H]
slot-O[slot-P]
slot-P[O]
slot-R[slot-S]
slot-S[T]

Given that, suppose that an event is fired on U, an event path would be (in reverse order):

[U => T => slot-S => slot-R => Q => slot-P => slot-O => shadowroot-N => M
=> shadowroot-L => K => shadowroot-J => I => H => slot-G => F => shadowroot-E
=> D => shadowroot-C => B => A]

Roughly speaking, an event's parent (the next node in event path) is calculated as follows:

  • If a node is assigned to a slot, the parent is the node's asigned slot.
  • If a node is a shadow root, the parent is its shadow host.
  • In other cases, the parent is node's parent.

In the above case, event.target, U, doesn't change in its lifetime because U can be seen from every nodes there. However, if an event is fired on node Q, for example, event.target would be adjusted for some nodes in event path to honer encapsulation. That is called event re-targeting.

Here is an event path for an event which is fired on Q :

event.currenttarget(re-targeted) event.target
QQ
slot-PQ
slot-OQ
shadowroot-NQ
MQ
shadowroot-LQ
KQ
shadowroot-JQ
II
HI
slot-GI
FI
shadowroot-EI
DI
shadowroot-CI
BB
AB

Design goal of event path calculation

TODO(hayato): Explain.

Composed events

TODO(hayato): Explain.

Event path and related targets

TODO(hayato): Explain.

DOM mutations

TODO(hayato): Explain.

Related flags

TODO(hayato): Explain.