ServerSideConfig

Since v0.1.1

Configuration for the server-side data plugin.

Controls how the grid fetches, caches, and paginates data from a remote source. The grid requests data in blocks (contiguous node ranges) as the user scrolls, caching them locally to avoid redundant network requests.

Example

new ServerSidePlugin({
  pageSize: 100,
  maxConcurrentRequests: 2,
})

Properties

Property	Type	Description
pageSize?	number	Number of nodes requested per getRows() call. The same value is surfaced on params.pageSize inside the callback, so backends that expect an explicit page-size query parameter can forward it directly. Larger values reduce request count at the cost of longer per-request latency.
maxConcurrentRequests?	number	Maximum number of concurrent getRows() requests. Limits how many blocks can be fetched simultaneously during fast scrolling. Set to 1 for strict sequential loading; higher values improve perceived performance.
loadThreshold?	number	Prefetch slack in node count (rows). When > 0, blocks are loaded as soon as the user is within loadThreshold rows of an unloaded block, rather than only when the visible window enters it. This reduces the number of placeholder rows the user sees during gentle scrolling at the cost of slightly more eager fetching.
dataSource?	ServerSideDataSource<unknown>	Data source for server-side loading. When provided, the plugin auto-initializes on attach — no need to call setDataSource().
sortMode?	server | local	How sort changes affect the cache. - 'server' (default): purge cache and refetch via getRows({ sortModel }). Use when the backend supports the requested sort shape natively. MultiSortPlugin does not re-sort the returned rows locally — the data source’s order (and its loading placeholders’ block positions) is shown as-is. - 'local': keep the loaded blocks; sort the loaded rows in-place via the active sort plugin (MultiSort or core sort). sortModel is not sent on subsequent block fetches. Loading placeholder rows (see isLoadingPlaceholder()) are pinned to the end regardless of direction.
filterMode?	server | local	How filter changes affect the cache. - 'server' (default): purge cache and refetch via getRows({ filterModel }). FilteringPlugin does not re-filter the returned rows locally — the data source’s result (and its loading placeholders) is shown as-is. - 'local': keep the loaded blocks; filter only the rows currently in cache via FilteringPlugin’s normal pipeline. filterModel is not sent on subsequent block fetches. Loading placeholders (see isLoadingPlaceholder()) always pass the filter until their block loads.
placeholderRendering?	columns | loading-row	How rows whose block is still loading are rendered. - 'loading-row' (default): the grid renders each placeholder itself, as a full-width row in the core row-loading state (.tbw-row-loading, aria-busy="true", the .tbw-row-loading-overlay spinner). Column renderers, format, cellClass, rowClass and a Responsive cardRenderer are not called for it, and context menus (mouse or keyboard) do not open on it. - 'columns': placeholders take the normal cell render path, so your column renderers and callbacks receive them and can draw a custom loading look. Detect them with isLoadingPlaceholder(). v3.9.1+

Property Details

pageSize

Default: 100

---

maxConcurrentRequests

Default: 2

---

loadThreshold

Prefetch slack in node count (rows). When > 0, blocks are loaded as soon as the user is within loadThreshold rows of an unloaded block, rather than only when the visible window enters it. This reduces the number of placeholder rows the user sees during gentle scrolling at the cost of slightly more eager fetching.

The threshold is applied symmetrically: the viewport is expanded by loadThreshold rows in both directions before block coverage is computed. The maxConcurrentRequests cap and per-block dedup still apply, so a large threshold during fast scrolling will not flood the server with requests.

A reasonable starting point is pageSize / 2. Values larger than pageSize will eagerly request 2+ blocks ahead, which can hurt perceived performance with slow backends.

Default: 0 (fetch only when the visible window enters an unloaded block)

---

dataSource

features: {
  serverSide: {
    pageSize: 50,
    dataSource: {
      async getRows(params) {
        const res = await fetch(`/api/data?start=${params.startNode}&end=${params.endNode}`);
        return res.json();
      },
    },
  },
}

---

sortMode

How sort changes affect the cache.

• 'server' (default): purge cache and refetch via getRows({ sortModel }). Use when the backend supports the requested sort shape natively. MultiSortPlugin does not re-sort the returned rows locally — the data source’s order (and its loading placeholders’ block positions) is shown as-is.
• 'local': keep the loaded blocks; sort the loaded rows in-place via the active sort plugin (MultiSort or core sort). sortModel is not sent on subsequent block fetches. Loading placeholder rows (see isLoadingPlaceholder()) are pinned to the end regardless of direction.

Use 'local' when the server cannot sort by every column the user can sort (e.g. APIs that only support a single sort field, or no sort at all). Ignored for the data-src URL source, which always sorts locally.

Default: 'server'

---

filterMode

How filter changes affect the cache.

• 'server' (default): purge cache and refetch via getRows({ filterModel }). FilteringPlugin does not re-filter the returned rows locally — the data source’s result (and its loading placeholders) is shown as-is.
• 'local': keep the loaded blocks; filter only the rows currently in cache via FilteringPlugin’s normal pipeline. filterModel is not sent on subsequent block fetches. Loading placeholders (see isLoadingPlaceholder()) always pass the filter until their block loads.

Ignored for the data-src URL source, which always filters locally.

In 'local' mode the user only filters the currently loaded subset — scrolling further loads more blocks which then re-enter the local filter.

Default: 'server'

---

placeholderRendering

How rows whose block is still loading are rendered.

• 'loading-row' (default): the grid renders each placeholder itself, as a full-width row in the core row-loading state (.tbw-row-loading, aria-busy="true", the .tbw-row-loading-overlay spinner). Column renderers, format, cellClass, rowClass and a Responsive cardRenderer are not called for it, and context menus (mouse or keyboard) do not open on it.
• 'columns': placeholders take the normal cell render path, so your column renderers and callbacks receive them and can draw a custom loading look. Detect them with isLoadingPlaceholder().

In both modes a placeholder gets an internal row id (__loading:<index>), so gridConfig.getRowId and a rowHeight function are never called with one, and valueAccessor / filterValue are never called with one (the value is undefined).

Default: 'loading-row'

---