Angular Data Grid Server-Side Data

Server-side data

Delegate sorting, filtering and pagination to your API with a single dataProvider function — the Angular Data Grid fetches one debounced request per state change and drops stale responses.

Client-side mode ends where the browser’s memory does. When your dataset lives in a database with tens of thousands of rows or more, hand the work to your backend: set dataProvider and the grid stops computing row models locally and asks your API for each page instead. This is the single biggest adoption unlocker for large data.

Server-side data

Set dataProvider and the grid switches to server-side mode — every sorting, filtering or pagination change triggers one debounced request (stale responses are dropped automatically), a loading overlay covers the viewport and totals come from your totalRows. Server-side mode implies pagination. This live demo hits CoreUI’s public demo API (apitest.coreui.io/demos/users) — 10,000+ records sorted, filtered and paged server-side.

No items found
0–0 of 0
ts
import { Component } from '@angular/core'
import { DataGridComponent } from '@coreui/angular-data-grid'
import type { DataGridColumn, DataGridDataRequest, DataGridItem } from '@coreui/angular-data-grid'

@Component({
  selector: 'docs-data-grid-server-example',
  imports: [DataGridComponent],
  template: `
    <c-data-grid
      [columns]="columns"
      [itemKey]="itemKey"
      [columnFilters]="true"
      [dataProvider]="dataProvider"
    />
  `
})
export class DataGridServerExample {
  readonly columns: DataGridColumn[] = [
    { key: 'first_name', label: 'First name' },
    { key: 'last_name', label: 'Last name' },
    { key: 'email', label: 'Email' },
    { key: 'country', label: 'Country' },
    { key: 'ip_address', label: 'IP' }
  ]
readonly itemKey = (item: DataGridItem) => String(item.id)
readonly dataProvider = async ({ sorting, columnFilters, pagination }: DataGridDataRequest) => {
    const params = new URLSearchParams({
      offset: String(pagination.pageIndex * pagination.pageSize),
      limit: String(pagination.pageSize)
    })

    for (const { id, value } of columnFilters) {
      params.append(id, String(value))
    }

    const [sort] = sorting
    if (sort) {
      params.append('sort', `${sort.id}%${sort.desc ? 'desc' : 'asc'}`)
    }

    const response = await fetch(`https://apitest.coreui.io/demos/users?${params}`)
    const result = await response.json()
    const totalRows = Number(result.number_of_matching_records)

    return { items: totalRows ? result.records : [], totalRows }
  }
}

The dataProvider contract

<c-data-grid [columns]="columns" [dataProvider]="dataProvider" />
import type { DataGridDataRequest } from '@coreui/angular-data-grid'

readonly dataProvider = async ({ sorting, columnFilters, globalFilter, pagination }: DataGridDataRequest) => {
  // fetch from your API and return the matching page
  return { items, totalRows }
}
  • One request per state change. sorting, columnFilters, globalFilter and pagination arrive as structured state; return the page of items plus the full totalRows count so the pager can render.
  • Debounced & race-safe. Rapid changes coalesce into one request and only the latest response is applied — stale ones are dropped by request id.
  • Loading UX. A .data-grid-loading overlay with a spinner covers the viewport and aria-busy is set while a request is in flight (labels.loading).
  • Errors. A rejected fetch emits the dataError output { error }, shows the empty state (labels.loadError) and leaves the grid interactive.
  • Success. Each load emits the dataLoad output { items, totalRows }.

Selection semantics

rowSelection is keyed by itemKey, so a selection survives page changes by design. The selectionChange output’s selectedItems contains only the items present in the current page’s data — the grid does not cache full objects for pages it has scrolled past.

The same page-bound rule applies to CSV export: in server-side mode every scope (filtered, all, selected) exports the currently loaded page only.

When the filter menu is used, a column’s entry in columnFilters carries the structured value — { conditions: [{ operator, value, value2? }], join: 'and' | 'or' } for typed operators or { operator: 'in', value: [...] } for the set filter — instead of a plain string. Servers should handle both shapes.