[SCAL-337048] feat(sdk): support per-action disabledActionReason by jitendra-bhati-ts · Pull Request #711 · thoughtspot/visual-embed-sdk

jitendra-bhati-ts · 2026-10-13T09:08:23Z

What

disabledActionReason can now be a string (unchanged) or a map that gives each disabled action its own tooltip, with an optional GENERIC_REASON fallback.

import { Action } from '@thoughtspot/visual-embed-sdk';

new LiveboardEmbed('#tsEmbed', {
    disabledActions: [Action.Download, Action.Save, Action.Share],
    disabledActionReason: {
        [Action.Download]: 'Downloads require a Pro license',
        [Action.Save]: 'You have read-only access',
        GENERIC_REASON: 'Not available for your account', // optional
    },
});

Why

Today one string is shown on every disabled action. A host that disables Download for licensing and Save for read-only access can only show one generic message for both.

How

Types (types.ts): disabledActionReason?: string | DisabledActionReasonMap; new DisabledActionReasonMap type (Partial<Record<Action | 'GENERIC_REASON', string>>), exported from the main and /react entry points. GENERIC_REASON is a plain key typed in the map, not an exported const, so customers write GENERIC_REASON: '...' and TypeScript still checks it.
Validation (utils/actions.ts, new; lodash isString, isPlainObject, pickBy, isEmpty): a string goes out unchanged. A map keeps only keys that are an Action or GENERIC_REASON with a string value. Anything else is dropped with a logger.warn, and the embed never fails over it. The allow-list also blocks __proto__ / constructor keys. An empty map, or a map where every key was dropped, sends nothing, the same as leaving the option unset.
Transport (ts-embed.ts): the map is never put on the iframe URL. A map can hold a reason for every action, which is too long for a URL, so it is sent in the APP_INIT payload (embedParams.disableHint) on every embed, the same channel excludeConfigFromURL already uses. APP_INIT_PARAMS maps each param to a check on its value (disableHint → isPlainObject), and getUrlQueryParamsObject moves a param to the init data when its check passes. Another param can be kept off the URL by adding one entry with its own check. The string form stays on the URL, so it keeps working on every cluster.
Runtime updates: prerender and UpdateEmbedParams already send the whole config over postMessage, so they carry the map as an object without extra code.

`disabledActionReason`	Sent in
string	iframe URL (`disableHint`), unchanged
map	`APP_INIT` payload (`embedParams.disableHint`), never the URL

Version

The object form is supported from SDK <TBD> (next release) | ThoughtSpot Cloud 26.12.0.cl. The string form is unchanged (SDK 1.6.0).

Backward compatibility

The string form is unchanged, and all existing tests pass as they are. One behaviour change: a non-string, non-object value (number, array, boolean), which the type never allowed, is now dropped with a warning instead of being sent as-is.

App side (merged)

The tooltip is drawn by the ThoughtSpot app, not the SDK. The app side is merged in scaligent #72606 and #72825. The app needs no change for the APP_INIT transport: it already merges APP_INIT embedParams over the URL params on every load. On a cluster without the app change, a map does not render correctly, which is why the object form needs ThoughtSpot Cloud 26.12.0.cl.

Before app side PR	After app side PR	After app side example

SDK map on a cluster without app side changes: tooltip shows [object Object]. (Screenshots were taken while the map still went on the URL; the result is the same with the APP_INIT transport.)

Tests

actions.spec.ts covers the string form, full map, partial map with and without GENERIC_REASON, a map with only GENERIC_REASON, empty map, unknown keys, prototype keys, non-string values and invalid types.
liveboard.spec.ts checks that a map is kept off the iframe URL, and that an empty map sends no disableHint.
ts-embed.spec.ts checks that the map arrives in the APP_INIT payload with and without excludeConfigFromURL (other params stay on the URL), that a string stays on the URL, and the cleaned map in the UpdateEmbedParams payload.
Verified end to end in an embed app against a cluster with the app change: the iframe URL has no disableHint, and each disabled action shows its own reason.

npm run tsc:check, npm run lint, npm run lint-versions and npm run test-sdk all pass (2207 passed, 4 skipped). Bundle size 32.69 kB of the 34 kB limit.

Jira: https://thoughtspot.atlassian.net/browse/SCAL-337048

gemini-code-assist

Code Review

This pull request introduces support for a per-action disabledActionReason map (DisabledActionReasonMap) in the Visual Embed SDK, allowing developers to specify custom tooltip reasons for different disabled actions, with an optional GENERIC_REASON fallback. It includes validation utilities, warning logs for invalid types or keys, updated TypeScript types, and comprehensive unit tests. The feedback suggests a more robust implementation of the isPlainObject utility to prevent built-in object types like Map, Set, Date, and RegExp from bypassing the type check and being silently ignored.

pkg-pr-new · 2026-10-13T09:14:14Z

Open in StackBlitz

npm i https://pkg.pr.new/@thoughtspot/visual-embed-sdk@711

commit: 34c1063

disabledActionReason accepted only one string, shown on every disabled action, so a host could not explain why each action was disabled. It now also accepts a map keyed by Action, with an optional GENERIC_REASON fallback. A string is sent unchanged. A map is cleaned before it is sent: keys that are not an Action or GENERIC_REASON, and values that are not strings, are dropped with a console warning, so a bad entry never fails the embed. What is left goes out JSON-encoded in the existing disableHint param; an empty result sends nothing, the same as leaving the option unset. GENERIC_REASON and DisabledActionReasonMap are exported from the main and React entry points.

…-337048) With excludeConfigFromURL, disableHint is held back from the iframe URL and sent in the APP_INIT embedParams instead. Add tests that a map arrives there as the cleaned object, without the unknown key, and that a string arrives unchanged. Also detect a plain object by its built-in type, so a Map, Date or other built-in gets the invalid-type warning, named by its real type, instead of being treated as an empty map.

…337048) Use lodash isPlainObject and isString, which the SDK already uses, instead of a hand-written type check. isPlainObject is false for arrays, Maps, Dates and other built-ins, so they still get the invalid-type warning. The warning no longer names the value's type.

…AL-337048) 1.53.1 is already released, so the map form ships in the next release. Use the <TBD> placeholder, which the release bump fills in, and point the disabledActionReason doc at DisabledActionReasonMap for the SDK version instead of repeating it.

…7048) A general name, so other action helpers can live in the same file.

shivam-kumar-ts · 2026-10-27T12:00:31Z

this last few comment this will be the breakign changes for the 26.11 or previous version

… (SCAL-337048) A per-action map can hold a reason for every action, which is too long for the iframe URL. Send the map in the APP_INIT payload's embedParams on every embed, the same place excludeConfigFromURL already sends it; the app merges those embedParams over the URL params, so it needs no change. The string form stays on the URL, so it keeps working on every cluster. Prerender and UpdateEmbedParams already carry the map as an object. Filter the map with lodash pickBy and walk the query params with lodash forEach, matching the lodash helpers the SDK already uses.

… (SCAL-337048) Map each param to a check on its value instead of hard-coding the object check in the loop, so a future param can be kept off the URL by adding one entry with its own condition. No behaviour change.

shivam-kumar-ts

LGTM

…y (SCAL-337048) GENERIC_REASON was the only const in types.ts. Type the map key as Action | 'GENERIC_REASON' instead, so customers write GENERIC_REASON as a plain key and TypeScript still checks it. The key on the wire is unchanged, so the merged app side needs nothing. The constant was never released.

jitendra-bhati-ts requested a review from a team as a code owner October 13, 2026 09:08

gemini-code-assist Bot reviewed Oct 13, 2026

View reviewed changes

Comment thread src/utils/disabled-action-reason.ts Outdated

shivam-kumar-ts reviewed Oct 14, 2026

View reviewed changes

Comment thread src/utils/actions.ts Outdated

jitendra-bhati-ts requested a review from yinstardev October 20, 2026 05:30