Framework:Javascript Data GridAngular Data GridReact Data GridVue Data Grid

JavaScript Data Grid: Infinite Row Model

If you are an Enterprise user you should consider using the Server-Side Row Model instead of the Infinite Row Model. It offers the same functionality with many more features.

The differences between row models can be found in our row models summary page.

Infinite scrolling allows the grid to lazy-load rows from the server depending on what the scroll position is of the grid. In its simplest form, the more the user scrolls down, the more rows get loaded.

The grid will have an 'auto extending' vertical scroll. That means as the scroll reaches the bottom position, the grid will extend the height to allow scrolling even further down, almost making it impossible for the user to reach the bottom. This will stop happening once the grid has extended the scroll to reach the last record in the table.

How it Works

The grid will ask your application, via a datasource, for the rows in blocks. Each block contains a subset of rows of the entire dataset. The following diagram is a high-level overview.


When the grid scrolls to a position where there is no corresponding block of rows loaded, the model uses the provided datasource to get the rows for the requested block. In the diagram, the datasource is getting the rows from a database in a remote server.

Turning On Infinite Scrolling

To turn on infinite scrolling, you must a) set the grid property rowModelType to 'infinite' and b) provide a datasource.

// before grid initialised
gridOptions.rowModelType = 'infinite';
gridOptions.datasource = myDataSource;

// after grid initialised, you can set or change the datasource


A datasource must be provided to do infinite scrolling. You specify the datasource as a grid property or using the grid API.

// set as a property
gridOptions.datasource = myDatasource;

// or use the api after the grid is initialised

Changing the Datasource

Changing the datasource after the grid is initialised will reset the infinite scrolling in the grid. This is useful if the context of your data changes, i.e. if you want to look at a different set of data.

If you call setDatasource() the grid will act assuming it's a new datasource, resetting the block cache. However you can pass in the same datasource instance. So your application, for example, might have one instance of a datasource that is aware of some external context (e.g. the business date selected for a report, or the 'bank ATM instance' data you are connecting to), and when the context changes, you want to reset, but still keep the same datasource instance. In this case, just call setDatasource() and pass the same datasource in again.

Datasource Interface

In a nutshell, every time the grid wants more rows, it will call getRows() on the datasource. The datasource responds with the rows requested. Your datasource for infinite scrolling should implement the following interface:

// Infinite Scrolling Datasource
interface IDatasource {
    // Callback the grid calls that you implement to fetch rows from the server. See below for params.
    getRows(params: IGetRowsParams): void;

    // optional destroy method, if your datasource has state it needs to clean up
    destroy?(): void;

The getRows() method takes the following parameters:

// Params for the above IDatasource.getRows()
interface IGetRowsParams {
    // The first row index to get.
    startRow: number;

    // The first row index to NOT get.
    endRow: number;

    // If doing server-side sorting, contains the sort model
    sortModel: any,

    // If doing server-side filtering, contains the filter model
    filterModel: any;

    // The grid context object
    context: any;

    // Callback to call when the request is successful.
    successCallback(rowsThisBlock: any[], lastRow?: number): void;

    // Callback to call when the request fails.
    failCallback(): void;


The getRows() function is called by the grid to load a block of rows into the browser-side cache of blocks. It takes the following as parameters:

  • The startRow and endRow define the range expected for the call. For example, if block size is 100, the getRows function will be called with startRow: 0 and endRow: 100 and the grid will expect a result with 100 rows (rows 0 to 99).
  • The successCallback(rowsThisBlock, lastRow) should be called when you successfully receive data from the server. The callback has the following parameters:

    • rowsThisBlock should be the rows you have received for the current block.
    • lastRow should be the index of the last row if known.
  • The failCallback() should be called if the loading failed. Either one of successCallback() or failCallback() should be called exactly once.
  • The filterModel() and sortModel() are passed for doing server-side sorting and filtering.
  • The context is just passed as is and nothing to do with infinite scrolling. It's there if you need it for providing application state to your datasource.

Setting Last Row Index

The success callback parameter lastRow is used to move the grid out of infinite scrolling. If the last row is known, then this should be the index of the last row. If the last row is unknown, then leave blank (undefined, null or -1). This attribute is only used when in infinite scrolling. Once the total record count is known, the lastRow parameter will be ignored.

Under normal operation, you will return null or undefined for lastRow for every time getRows() is called with the exception of when you get to the last block. For example, if block size is 100 and you have 250 rows, when getRows() is called for the third time, you will return back 50 rows in the result and set rowCount to 250. This will then get the grid to set the scrollbar to fit exactly 250 rows and will not ask for any more blocks.

Block Cache

The grid keeps the blocks in a cache. You have the choice to never expire the blocks, or to set a limit to the number of blocks kept. If you set a limit, then as you scroll down, previous blocks will be discarded and will be loaded again if the user scrolls back up. The maximum blocks to keep in the cache is set using the maxBlocksInCache property.

Block Size

The block size is set using the grid property cacheBlockSize. This is how many rows each block in the cache should contain. Each call to your datasource will be for one block.

Debounce Block Loading

It is also possible to debounce the loading to prevent blocks loading until scrolling has stopped. This can be configured using: blockLoadDebounceMillis.

Aggregation and Grouping

Aggregation and grouping are not available in infinite scrolling. This is because to do so would require the grid knowing the entire dataset, which is not possible when using the Infinite Row Model. If you need aggregation and / or grouping for large datasets, check the Server-Side Row Model for doing aggregations on the server-side.

Sorting & Filtering

The grid cannot do sorting or filtering for you, as it does not have all of the data. Sorting or filtering must be done on the server-side. For this reason, if the sort or filter changes, the grid will use the datasource to get the data again and provide the sort and filter state to you.

Simple Example: No Sorting or Filtering

The example below makes use of infinite scrolling and caching. Notice that the grid will load more data when you bring the scroll all the way to the bottom.


Selection works on the rows in infinite scrolling by using the ID of the row node. If you do not provide IDs for the row nodes, the index of the row node will be used. Using the index of the row breaks down when (server-side) filtering or sorting, as these change the index of the rows. For this reason, if you do not provide your own IDs, then selection is cleared if sort or filter is changed.

To provide your own IDs, implement the method getRowNodeId(data), which takes the data and should return the ID for the data.

gridOptions.getRowNodeId: function(item) {
    // the ID can be any string, as long as it's unique within your dataset

Once you have getRowNodeId() implemented, selection will persist across sorts and filters.

Example: Sorting, Filtering and Selection

The following example extends the example above by adding server-side sorting, filtering and persistent selection.

Any column can be sorted by clicking the header. When this happens, the datasource is called again with the new sort options.

The columns Age, Country and Year can be filtered. When this happens, the datasource is called again with the new filtering options.

When a row is selected, the selection will remain inside the grid, even if the grid gets sorted or filtered. Notice that when the grid loads a selected row (e.g. select first row, scroll down so the first block is removed from cache, then scroll back up again) the row is not highlighted until the row is loaded from the server. This is because the grid is waiting to see what the ID is of the row to be loaded.

The example below uses AG Grid Enterprise, to demonstrate the set filter with server-side filtering. AG Grid Enterprise is not required for infinite scrolling.

When performing multiple row selections using shift-click, it is possible that not all rows are available in memory if the configured value of maxBlocksInCache doesn't cover the range. In this case multiple selections will not be allowed.

Specify Selectable Rows

It is also possible to specify which rows can be selected via the gridOptions.isRowSelectable() callback function.

For instance if we only wanted to allow rows where the property is the 'United States' we could implement the following:

gridOptions.isRowSelectable: function(data) {
    return === 'United States';

Note that in the above example we have also included an optional checkbox to help highlight which rows are selectable.

Configuring a Bit Differently

The examples above use old-style JavaScript objects for the datasource. This example turns things around slightly and creates a datasource Class. The example also just generates data on the fly.

Loading Spinner

The examples on this page use a loading spinner to show if the row is waiting for its data to be loaded. The grid does not provide this, rather it is a simple rendering technique used in the examples. If the data is loading, then the rowNode will have no ID, so if you use the ID as the value, the cell will get refreshed when the ID is set.

loadingSpinnerColumn = {
    // use a value getter to have the node ID as this column's value
    valueGetter: '',

    // then a custom cellRenderer
    cellRenderer: function(params) {
        if (params.value === undefined) {
            // when no node id, display the spinner image
            return '<img src="loading.gif" />';
        } else {
            // otherwise just display node ID (or whatever you wish for this column)
            return params.value;

Refer to section Cell Rendering for how to build cell renderers.

More Control via Properties and API

Infinite scrolling has a cache working behind the scenes. The following properties and API are provided to give you control of the cache.


(Partial Store only) How many rows for each block in the store, i.e. how many rows returned from the server at a time.
Default: 100
(Partial Store only) How many blocks to keep in the store. Default is no limit, so every requested block is kept. Use this if you have memory concerns, and blocks that were least recently viewed will be purged when the limit is hit. The grid will additionally make sure it has all the blocks needed to display what is currently visible, in case this property is set to a low value.
How many requests to hit the server with concurrently. If the max is reached, requests are queued.
Default: 1


Marks all the currently loaded blocks in the cache for reload. If you have 10 blocks in the cache, all 10 will be marked for reload. The old data will continue to be displayed until the new data is loaded.
Purges the cache. The grid is then told to refresh. Only the blocks required to display the current data on screen are fetched (typically no more than 2). The grid will display nothing while the new blocks are loaded. Use this to immediately remove the old data from the user.
The row count defines how many rows the grid allows scrolling to.
Returns true if grid allows for scrolling past the last row to load more rows, thus providing infinite scroll.
Sets the rowCount and lastRowIndexKnown properties. The second parameter, lastRowIndexKnown, is optional and if left out, only rowCount is set. Set rowCount to adjust the height of the vertical scroll. Set lastRowIndexKnown to enable / disable searching for more rows. Use this method if you add or remove rows into the dataset and need to reset the number of rows or put the data back into 'look for data' mode.
Returns an object representing the state of the cache. This is useful for debugging and understanding how the cache is working.

Adding / removing rows directly in the grid for infinite scrolling is not recommended as it will complicate your application. It will make your life easier if you update the data on the server and refresh the block cache.

Example: Using Cache API Methods

Below demonstrates the different API methods via the buttons. The example outputs a lot of debugging items to the console because the grid property debug=true is set. The buttons are as follows:

  • Insert Rows: Inserts 5 rows at row index 2 from the server, then refreshes the grid.
  • Delete Rows: Deletes 10 rows at row index 3 from the server, then refreshes the grid.
  • Set Row Count: Sets the row count to 200. This adjusts the vertical scroll to show 200 rows. If the scroll is positioned at the end, this results in the grid automatically re-adjusting as it seeks ahead for the next block of data.
  • Print Info: Prints rowCount and maxFound to the console.
  • Jump to 500: Positions the grid so that row 500 is displayed.
  • Print Cache State: Debugging method, to see the state of the cache.
  • Set Prices High & Set Prices Low: Sets the prices on the server-side to either high or low prices. This will not impact the grid until after a block cache is loaded. Use these buttons to then further test the refresh and purge methods.
  • Refresh Cache: Calls for the cache to be refreshed.
  • Purge Cache: Calls for the cache to be purged.

The example also makes each Honda row bold - demonstrating that the callbacks getRowStyle and getRowClass get called after the data is set as well as when the row is created (when the data may not yet be available).

Changing Columns

Changing columns is possible using infinite scroll and it does not require the data getting fetched again from the server. If the change of columns impacts the sort or filter (i.e. a column with a sort or filter applied is removed), the grid will fetch data again similar to how data is fetched again after the user changes the sort or filter explicitly.

The example below demonstrates changing columns on the infinite row model. The following can be noted:

  • Hit the buttons 'Show Year' and 'Hide Year'. Notice that the data is not re-fetched.
  • Add a sort or filter to Age column. When the sort or filter is applied the data is re-fetched. However once fetched, you can add and remove the Year column without re-fetching the data.
  • Add a sort or filter to the Year column. When the sort or filter is applied the data is re-fetched. Now remove the Year column and the data is re-fetched again as the sort or filter has changed.


As with all row models, it is possible to enable pagination with infinite scrolling. With infinite scrolling, it is possible to mix and match with the configuration to achieve different effects. The following examples are presented:

ExamplePage SizeBlock SizeComment
Example 1AutoLargeMost Recommended
Example 2EqualEqualRecommended Sometimes

Having smaller infinite blocks size than your pagination page size is not supported |

You must have infinite block size greater than or equal to the pagination page size. If you have a smaller block size, the grid will not fetch enough rows to display one page. This breaks how infinite scrolling works and is not supported.

Example 1: Auto Pagination Page Size, Large Infinite Block Size

This example is the recommended approach. The infinite block size is larger than the pages size, so the grid loads data for a few pages, allowing the user to hit 'next' a few times before a server sided call is needed.

Example 2: Equal Pagination Page Size and Large Infinite Block Size

This example demonstrates having the page and block sizes equal. Here the server is hit every time a new page is navigated to.


The infinite row model does not use overlays like the Client-Side Row Model. It does not use 'loading' overlay as rows load in blocks and it would be wrong to hide all the grid because some rows are getting loaded. The grid does not use 'no rows' overlay as the 'no rows' could be because you have a filter set, and a grid with a filter shows an empty grid when no rows pass the filter.

If you do want to show overlays, then please see overlays section for details on how to show the overlays manually.