@atlaskit/editor-plugin-interactivity
Version:
Interactivity plugin for @atlaskit/editor-core
132 lines (131 loc) • 6.11 kB
TypeScript
/**
* The shape of what the `editor interactivity` event carries: one snapshot, the groups inside
* it, and the values its fields can take.
*
* This is the contract with the analytics pipeline, which the collector fills in and the
* payload wraps, and the one place in the package where types are shared. A type serving a
* single module stays with that module.
*/
/**
* What produced a snapshot.
*
* `navigation` and `modeChange` end a session while the editor stays mounted: Confluence
* live pages keep one editor across page transitions and switch between reading and
* editing with a command. A session covers one document in one mode, so either change
* closes it and opens the next.
*/
export type SnapshotReason = 'timer' | 'hidden' | 'pagehide' | 'unmount' | 'navigation' | 'modeChange';
/**
* What the editor was being used for at the time of the snapshot: `editing` is the
* editable editor, `reading` is a live page being viewed with the editor still mounted.
*/
export type SessionMode = 'editing' | 'reading';
/**
* The editor groups, which are also the fields their histograms are reported in.
*
* `editorOther` is the remainder — an interaction the browser counts that is neither typing nor
* pointing — so the three together cover whatever the browser calls an interaction.
*/
export type EditorInteractionGroupName = 'editorOther' | 'editorPointer' | 'editorTyping';
/**
* The group a record is attributed to. Unlike the `page` histogram, which counts the editor's
* interactions as well, `outsideEditor` is only the interactions that are not the editor's.
*/
export type SlowInteractionGroup = EditorInteractionGroupName | 'outsideEditor';
/**
* One of the slowest interactions of the session: which target was slow, and where the time went,
* neither of which the histograms can answer.
*
* The three phases divide `durationMs` and add up to it within rounding, and so do the four totals.
* Everything but the phases comes from the Long Animation Frames the interaction ran in, so all of
* it is absent when the browser reported none — and the fields describing one script are absent as
* well when no script of those frames overlapped the interaction.
*/
export type SlowInteraction = {
durationMs: number;
/** Function the interaction's slowest script ran in. */
functionName?: string;
group: SlowInteractionGroup;
/** Time between the event arriving and its handlers starting to run. */
inputDelayMs?: number;
/**
* What ran that script, as the browser names it: `event-listener`, `user-callback`,
* `resolve-promise`, `classic-script` and so on.
*/
invokerType?: string;
/** How much of the interaction's slowest script fell inside the interaction. */
longestScriptMs?: number;
/** Type of the event the latency was measured on, which for a pointer press is usually `click`. */
name: string;
/** Time between the handlers finishing and the next frame being presented. */
presentationDelayMs?: number;
/** Time spent running the event's handlers. */
processingMs?: number;
/** The bundle that script came from, without its origin or query. */
scriptName?: string;
/** The phase of the interaction that script ran in. */
scriptSubpart?: 'inputDelay' | 'presentationDelay' | 'processing';
/** A short DOM path. Absent once the element has been removed from the document. */
target?: string;
/** Time between the last frame of the interaction ending and the screen updating. */
totalPaintDurationMs?: number;
/**
* Script time inside the interaction across its frames, the part a script forced into style and
* layout excluded.
*/
totalScriptDurationMs?: number;
/** Style and layout across the frames of the interaction, the part forced from a script included. */
totalStyleAndLayoutDurationMs?: number;
/**
* The part of the latency the frames account for nothing in — the main thread was busy with
* something no Long Animation Frame attributed to a script, to style and layout, or to paint.
* Not idle time: the browser reports no frame under 50 ms, so the work of those lands here too.
*/
totalUnattributedDurationMs?: number;
};
/**
* Session-to-date latency distribution for one group of interactions.
*
* `totalCount` counts every interaction, including those below the Event Timing reporting
* threshold, so `totalCount - observedCount` is the sub-threshold count. `buckets` is keyed by
* each bucket's upper boundary in milliseconds and is not cumulative; empty buckets are
* omitted, so a missing bucket means zero.
*/
export type InteractionGroupSnapshot = {
buckets: Record<string, number>;
maxMs: number;
observedCount: number;
/**
* Temporary. Percentiles of the same interactions, keyed by percentile and exact to the 8 ms
* Event Timing reports durations at, to confirm that a percentile read off `buckets` lands
* where the latencies actually are. Goes once that is established.
*/
percentilesMs: Record<string, number>;
sumMs: number;
totalCount: number;
};
/**
* One session-to-date snapshot. Consumers take the highest `seq` per
* `interactivitySessionId` and then sum bucket counts across sessions.
*/
export type InteractivitySnapshot = {
activeMs: number;
editorDomSize?: number;
/** Interactions inside the editor that are neither typing nor pointing. */
editorOther: InteractionGroupSnapshot;
editorPointer: InteractionGroupSnapshot;
editorTyping: InteractionGroupSnapshot;
hiddenMs: number;
interactivitySessionId: string;
nodeSize?: number;
objectId?: string;
/** Every interaction on the page, the editor groups included. */
page: InteractionGroupSnapshot;
reason: SnapshotReason;
schema: number;
seq: number;
/** Fixed for the whole session: a mode change closes it and opens the next. */
sessionMode?: SessionMode;
/** Slowest first. Absent when nothing was slow enough to record. */
slowest?: SlowInteraction[];
};