Framework:Javascript Data GridAngular Data GridReact Data GridVue Data Grid

React Data Grid: SSRM Row Grouping

This section covers Row Grouping in the Server-Side Row Model (SSRM).

Enabling Row Grouping

Row Grouping is enabled in the grid via the rowGroup column definition attribute. The example below shows how to group rows by 'country':

const columnDefs = [
    { field: 'country', rowGroup: true },
    { field: 'sport' },
    { field: 'year' },

<AgGridReact columnDefs={columnDefs}></AgGridReact>

For more configuration details see the section on Row Grouping.

Row Grouping on the Server

The actual grouping of rows is performed on the server when using the SSRM. When the grid needs more rows it makes a request via getRows(params) on the Server-Side Datasource with metadata containing grouping details.

The properties relevant to row grouping in the request are shown below:

// IServerSideGetRowsRequest
    // row group columns
    rowGroupCols: ColumnVO[];

    // what groups the user is viewing
    groupKeys: string[];

    ... // other params

Note in the snippet above the property rowGroupCols contains all the columns (dimensions) the grid is grouping on, e.g. 'Country', 'Year'. The property groupKeys contains the list of group keys selected, e.g. ['Argentina', '2012'].

The example below demonstrates server-side Row Grouping. Note the following:

  • The Partial Store is used, thus Infinite Scrolling is active.
  • The store block size is set to 5 by setting the grid property cacheBlockSize = 5. It can then be observed that rows are loaded in blocks at all levels. For example if you expand United States row, the children rows are loaded in blocks using Partial Scrolling.
  • Country and Sport columns have rowGroup=true defined on their column definitions. This tells the grid there is two levels of grouping, one for Country and one for Sport.
  • The rowGroupCols and groupKeys properties in the request are used by the server to perform grouping.
  • Open the browser's dev console to view the request supplied to the datasource.

The example above also demonstrates Sorting with groups. When the grid sort changes, only impacted rows will be reloaded. For example, if 'Sport' groups are expanded, sorting on the 'Year' column won't cause the top level 'Country' groups to reload, but sorting on the 'Gold' column will.

To avoid this and always refresh top level groups regardless of which column is sorted, set the grid property serverSideSortingAlwaysResets=true.

Grouping and Row Stores

When grouping and a group is expanded, a new Row Store is created to store the child rows of the opened group. The diagram below shows what the Row Stores could look like with Row Grouping, where two top level Row Groups are open.

Node Store - Grouping
Fig 1. Node Store - Grouping

This means there can be any number of Row Stores existing inside the SSRM. Each time a Row Group is expanded, a new Row Store is created for that level. The sections Server-Side Row Grouping explains in detail this topic.

Full vs Partial Store

The Row Grouping mechanics are almost identical with the Full Store and Partial Store. The difference is that when using the Partial Store, data will be requested in blocks and could be requested to have sorting and / or filtering applied.

All the examples presented in this section use the Partial Store as it covers all the semantics found when using both store types.

Configure Stores

By default, each store will have the same configuration (store type, block size etc). This configuration is specified using the grid properties serverSideStoreType, maxBlocksInCache and cacheBlockSize.

It is possible to have different configurations for different stores. For example if grouping, infinite scrolling (using the Partial Store) could be turned off at the top level but turned on at the lower levels.

This is done by implementing the grid callback getServerSideStoreParams(params).

Allows providing different params for different levels of grouping.
getServerSideStoreParams = (
    params: GetServerSideStoreParamsParams
) => ServerSideStoreParams;

interface GetServerSideStoreParamsParams {
  // The level of the store. Top level is 0. 
  level: number;
  // The Row Node for the group that got expanded, or undefined if top level (ie no parent) 
  parentRowNode?: RowNode;
  // Active Row Group Columns, if any. 
  rowGroupColumns: Column[];
  // Active Pivot Columns, if any. 
  pivotColumns: Column[];
  // true if pivot mode is active. 
  pivotMode: boolean;
  // The grid api. 
  api: GridApi;
  // The column api. 
  columnApi: ColumnApi;
  // Application context as set on `gridOptions.context`. 
  context: any;

interface ServerSideStoreParams {
  // What store type to use.
  // If missing, then defaults to grid option `serverSideStoreType`. 
  storeType?: ServerSideStoreType;
  // For Partial Store only.
  // How many blocks to keep in cache.
  // If missing, defaults to grid options `maxBlocksInCache`. 
  maxBlocksInCache?: number;
  // For Partial Store only.
  // Cache block size.
  // If missing, defaults to grid options `cacheBlockSize`. 
  cacheBlockSize?: number;

type ServerSideStoreType = 'full' | 'partial'

The example below demonstrates the getServerSideStoreParams(params) callback. Note the following:

  • The grid is configured differently depending on whether grouping is active or not by implementing the getServerSideStoreParams(params) callback. The callback logs its results to the dev console.
  • When grouping is active, the stores are configured as follows:

    • Level 0 - Full Store (no infinite scrolling)
    • Level 1 - Partial Store (infinite scrolling) with block size of 5
    • Level 2 - Partial Store (infinite scrolling) with block size of 2

    To observe, expand different levels of the data and notice when rows are read back in blocks.

  • When no grouping is active, the store is configured to use infinite scroll and only keeps two blocks of rows in the store.

    To observe this, remove all grouping and scroll down to load more blocks. Then scroll back up to observe the initial blocks getting reloaded.

Store State & Store Info

For debugging purposes, the grid has the API getServerSideStoreState() which returns info on all existing Row Stores. This is good for learning purposes, as you can see details about the store such as the store type and it's route.

Returns info on all server side stores.
getServerSideStoreState = () => ServerSideStoreState[];

interface ServerSideStoreState {
  // Store type, `partial` or `full` 
  type: ServerSideStoreType;
  // The route that identifies this store. 
  route: string[];
  // How many rows the store has. This includes 'loading rows'. 
  rowCount: number;
  // Partial store only.
  // Whether the last row index is know. 
  lastRowIndexKnown?: boolean;
  // Any extra info provided to the store, when data was loaded. 
  info?: any;
  // Partial store only.
  // Max blocks allowed in the store. 
  maxBlocksInCache?: number;
  // Partial store only.
  // The size (number of rows) of each block. 
  cacheBlockSize?: number;

type ServerSideStoreType = 'full' | 'partial'

Inspecting the Store State can be useful, for example when wanting to know what Route to use when providing Transactions or doing a Store Refresh.

It is also possible to attach info to each store as data is loaded. This is done through the success() callback when rows are fetched.

const createDatasource = server => {
    return {
        // called by the grid when more rows are required
        getRows: params => {
            // get data for request from server
            const rows = server.getData(params.request);

            // pass rows back along with any additional store info
                rowData: rows,
                storeInfo: {a: 22, b: 55}

The info object is merged into the Store Info (which is initially an empty object) and then available in the following locations:

  1. Included in the Store State returned from getServerSideStoreState().
  2. Included in the params to isApplyServerSideTransaction(). This method is explained in Cancelling Transactions.

If rows are loaded multiple times into the Store, then the Store Info values will over write existing values as they are merged on top of the existing values. Rows can be loaded multiple times if a) the store is Refreshed or b) Partial Store is used (as each block load will get the opportunity to add info data).

The example below shows Store Info in action.

Open by Default

It is possible to have rows open as soon as they are loaded. To do this implement the grid callback isServerSideGroupOpenByDefault.

Allows groups to be open by default.
isServerSideGroupOpenByDefault = (
    params: IsServerSideGroupOpenByDefaultParams
) => boolean;

interface IsServerSideGroupOpenByDefaultParams {
  data: any;
  rowNode: RowNode;
  // The grid api. 
  api: GridApi;
  // The column api. 
  columnApi: ColumnApi;
  // Application context as set on `gridOptions.context`. 
  context: any;
// Example implementation
function isServerSideGroupOpenByDefault(params) {
    var rowNode = params.rowNode;
    var isZimbabwe = rowNode.field == 'country' && rowNode.key == 'Zimbabwe';
    return isZimbabwe;

It may also be helpful to use the Row Node API getRoute() to inspect the route of a row node.

Returns the route of the row node. If the Row Node is a group, it returns the route to that Row Node. If the Row Node is not a group, it returns undefined.
getRoute = () => string[] | undefined;

Below shows isServerSideGroupOpenByDefault() and getRoute in action. Note the following:

  • The callback opens the following routes as soon as those routes are loaded:

    • [Zimbabwe]
    • [Zimbabwe, Swimming]
    • [United States, Swimming]
  • Note [Zimbabwe] and [Zimbabwe, Swimming] are visibly open by default.
  • Note [United States, Swimming] is not visibly open by default, as the parent group 'United States' is not open. However open 'United States' is open, it's 'Swimming' group is open.
  • Selecting a row and clicking 'Route of Selected' will print the route to the selected node.
  • The grid is configured with both Partial Store (top level) and Full Store (all other levels) to demonstrate the feature working with both.

Expand All / Collapse All

It is possible to expand and collapse all group rows using the expandAll() and collapseAll() grid API's.

// Expand all group rows

// Collapse all group rows

Calling expandAll() and collapseAll() will impact all loaded group nodes, including those not visible due to their containing group been closed. This means there could potentially be a huge number of groups expanded, so this method should be used very wisely to not create massive amount of server requests and loading a large amount of data.

Calling expandAll() and collapseAll() will have no impact on rows yet to be loaded.

To open only specific groups, e.g. only groups at the top level, then use the forEachNode() callback and open / close the row using setExpanded() as follows:

// Expand all top level row nodes
gridApi.forEachNode(node => {
    if ( && node.level == 0) {

The example below demonstrates these techniques. Note the following:

  • Clicking 'Expand All' will expand all loaded group rows. Doing this when the grid initially loads will expand all Year groups. Clicking it a second time (after Year groups have loaded) will cause all Year groups as well as their children Country groups to be expanded - this is a heaver operation with 100's of rows to expand.
  • Clicking 'Collapse All' will collapse all rows.
  • Clicking 'Expand Top Level Only' will expand Years only, even if more group rows are loaded..

Providing Child Counts

By default, the grid will not show row counts beside the group names. If you do want row counts, you need to implement the getChildCount(dataItem) callback for the grid. The callback provides you with the row data; it is your application's responsibility to know what the child row count is. The suggestion is you set this information into the row data item you provide to the grid.

Allows setting the child count for a group row.
getChildCount = (dataItem: any) => number;
const getChildCount = data => {
    // here child count is stored in the 'childCount' property
    return data.childCount;

<AgGridReact getChildCount={getChildCount}></AgGridReact>


When a sort is applied to a grouped grid using the SSRM, the grid will behave differently depending on what store is used. How it behaves is as follows:

  • Full Store

    The Full Store always sorts inside the grid. The rows are never reloaded due to a sort.

  • Partial Store

    • Non-group levels always refresh - all rows are loaded again from the server.
    • Group levels refresh (reload from server) if the sort was changed in:

      • Any column with a value active (ie colDef.aggFunc='something')
      • Any secondary column (ie you are pivoting and sort a pivot value column)
      • A Column used for this levels group (eg you are grouping by 'Country' and you sort by 'Country').

It is possible to force the grid to always refresh (reload data) after a sort changes. Do this by setting grid property serverSideSortingAlwaysResets=true.


When a filter is applied to a grouped grid using the SSRM, the grid will behave differently depending on what store is used. How it behaves is as follows:

  • Full Store

    The Full Store always filters inside the grid. The rows are never reloaded due to a filter change. Filtering is only possible at the lowest level (ie not group levels).

  • Partial Store

    Changing the filter on any column will always refresh the Partial Store. Rows will be loaded again from the server with the new filter information.

It is possible to force the grid to always refresh (reload data) after a filter changes. Do this by setting grid property serverSideFilteringAlwaysResets=true.

Complex Columns

It is possible the data provided has composite objects, in which case it's more difficult for the grid to extract group names. This can be worked with using value getters or embedded fields (i.e. the field attribute has dot notation).

In the example below, all rows are modified so that the rows look something like this:

// sample contents of row data
    // country field is complex object
    country: {
        name: 'Ireland',
        code: 'IRE'

    // year field is complex object
    year: {
        name: '2012',
        shortName: "'12"

    // other fields as normal

Then the columns are set up so that country uses a valueGetter that uses the field with dot notation, i.e.

Next Up

Continue to the next section to learn how to perform Data Refresh.