Skip to Content
harness-fme

πŸ“„ Paging

Page size limits, collectAll, and the listAll helpers that walk every page for you.

List endpoints return a Page<T>: { objects: T[]; totalCount?: number }. Each list* helper fetches a single page; each listAll* helper walks them all.

Limits

The admin API caps every list endpoint at 50 objects per page, exported as MAX_PAGE_SIZE. Single-page helpers such as listFeatureFlags, listWorkspaces and listSegments default to limit = 20 and offset = 0:

const page = await listFeatureFlags(client, {
  workspaceId,
  limit: 50,
  offset: 100,
  tag: "checkout",
})

The page size is sent as limit. The flag listing also accepts tag, rolloutStatusId and name (a partial match) as filters.

collectAll

collectAll(fetchPage, { pageSize }) is the generic walker. Give it a function that fetches one page at an offset and it returns every object:

import { collectAll, listSegments } from "@kud/harness-fme"

const all = await collectAll((offset, limit) =>
  listSegments(client, { workspaceId, limit, offset }),
)
if (all.ok) console.log(all.data.length)

It is careful about how it stops:

  • The offset advances by the number of objects actually returned, not the number asked for, so a server that caps pages below the requested size still yields every object exactly once.
  • It ends on totalCount when the API provides one, or on a short page when it does not.
  • An empty page always ends the walk.
  • The first failed page ends it too, returning that failure as the result.

pageSize defaults to MAX_PAGE_SIZE.

The listAll* helpers

HelperWalks
getAllWorkspaces(client)every workspace
listAllFeatureFlags(client, { workspaceId, … })every flag, with the same filters
listAllFlagDefinitions(client, { workspaceId, environmentId })every definition in an environment
const flags = await listAllFeatureFlags(client, {
  workspaceId,
  tag: "checkout",
})

Fetching at the maximum page size matters: a workspace of a thousand flags takes 20 requests rather than 50. A bug where the page size was sent under the wrong parameter name, silently leaving every page at 20, is the reason the helpers all go through one tested path.

Counting cheaply

countFlags(client, workspaceId) asks for a single-object page and reads totalCount, so you learn the size of a workspace with one tiny request. buildFlagRows uses it to decide between reading definitions flag by flag and listing a whole environment, see Flags, definitions and segments.