Feature architecture

Read core first for the headless model and stable inputs.

Register the behavior the table uses

<!-- skill-snippet:check -->

import {
  createSortedRowModel,
  rowSortingFeature,
  sortFn_alphanumeric,
  tableFeatures,
} from '@tanstack/table-core'

export const features = tableFeatures({
  rowSortingFeature,
  sortedRowModel: createSortedRowModel(),
  sortFns: { alphanumeric: sortFn_alphanumeric },
})

Pass the stable features object to the installed adapter constructor. The concrete registry determines optional APIs, options, state slices, and their types. The core row model is automatic; a custom core model uses coreRowModel.

Registration rules

• Register each prerequisite feature before its dependent slot in the same tableFeatures call. A processing factory alone cannot install that feature's APIs or state.
• Row-model factories use create*RowModel() feature slots. Their names and prerequisites are in the feature reference. They take no function-registry arguments.
• filterFns, sortFns, and aggregationFns are feature slots, not table options. They require columnFilteringFeature, rowSortingFeature, and rowAggregationFeature, respectively.
• Import individual built-ins such as sortFn_alphanumeric and register their conventional keys. Those keys become typed string names; 'auto' can resolve only registered functions. Pass a function directly when a column needs no named registry entry. Full registry exports bundle every built-in.
• columnResizingFeature requires columnSizingFeature; globalFilteringFeature requires columnFilteringFeature. Check installed FeatureSlotPrereqs for other dependencies.
• Use explicit features for normal construction. stockFeatures is a deliberate all-features convenience or temporary migration aid, with the corresponding bundle cost.