> For the complete documentation index, see [llms.txt](https://jacksonkasi.gitbook.io/tablecraft/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://jacksonkasi.gitbook.io/tablecraft/guides/performance.md).

# Performance Guide

When dealing with massive datasets (100k+ rows) or high concurrency, application performance can become a bottleneck. TableCraft is designed to help you bypass these bottlenecks safely and effectively.

## Cursor vs. Offset Pagination

TableCraft natively supports both standard Offset and high-performance Cursor pagination. The choice significantly affects query latency on large tables.

### 1. Offset Pagination (`page`, `pageSize`)

When you request `{ page: 1000, pageSize: 20 }`, the SQL database computes an `OFFSET` of 19,980.

**The Problem:** The database engine scans through 19,980 discarded rows before reaching the 20 you requested. As the offset increases, performance linearly degrades.

**Best For:**

* Small to medium datasets (<50k rows).
* Applications where users frequently jump directly to specific pages (e.g., jumping from page 1 to 50).

### 2. Cursor Pagination (`cursor`)

Cursor pagination encodes the sort value (e.g., `{ id: 100 }`) of the last fetched item and sends it back to the engine. The engine then creates an indexed O(1) condition like `WHERE id > 100 LIMIT 20`.

**The Advantage:** The database skips directly to the requested row via index lookups, completely avoiding the offset scanning penalty.

**Best For:**

* Massive datasets (100k+ rows) and infinite scrolling APIs.
* High concurrency environments.

**How to Use:** Simply enable cursor pagination in your config or pass it via the client. If `!!resolvedParams.cursor` evaluates to true on the backend, TableCraft's `CursorPaginationBuilder` takes over entirely.

#### Multi-column sort with cursor pagination

When you sort by more than one field (e.g., `?sort=status,createdAt`), the cursor encodes **all** sort field values from the last row:

```
nextCursor → base64url({ "status": "paid", "createdAt": "2026-01-15T10:00:00Z" })
```

The continuation `WHERE` clause uses a **lexicographic OR-expansion** so rows that share the primary sort value but differ on a secondary value are never skipped:

```sql
-- Correct expansion for ORDER BY status ASC, createdAt ASC
WHERE (status > 'paid')
   OR (status = 'paid' AND createdAt > '2026-01-15T10:00:00Z')
```

A flat `AND` would incorrectly skip rows with the same `status` value but a later `createdAt`. The engine handles this automatically.

{% hint style="info" %}
Add a database index on every column you use in a multi-column sort to keep cursor lookups fast.
{% endhint %}

## Large Dataset Handling

Even with cursor pagination, querying metadata on a massive table can be expensive.

### The `COUNT(*)` Problem

If you request `table.count()` on a table with 10 million rows, the database might perform a full table scan.

* **Solution:** Explicitly disable the total count feature on huge tables if not absolutely necessary. Alternatively, rely solely on `CursorPaginationBuilder`, which checks if a "next page" exists by fetching `pageSize + 1` rows instead of executing a `COUNT(*)`.

## Caching Strategies

A large contributor to UI latency is redundant network requests. Because TableCraft adapters (and React hooks) are entirely deterministic based on query parameters, they are highly cacheable.

### React Query Integration

If you are using `@tablecraft/client/react`, `useTableQuery` natively leverages TanStack React Query.

* **Increase `staleTime`:** Avoid refetching static tables when a user switches tabs or refocuses the window.

```tsx
const { data } = useTableQuery(adapter, {
  staleTime: 60 * 1000, // Data remains "fresh" for 1 minute
});
```

### HTTP Caching for Custom Adapters

If you write a custom REST adapter, ensure your CDN or proxy server recognizes `Cache-Control` headers for idempotent `GET` requests sent by TableCraft. Since filters and sorting are encoded in the URL parameters, identical query strings will safely hit the cache.

## Bundle Size Optimization

To ensure fast load times, TableCraft packages are highly modular:

* Ensure you are tree-shaking your build. Only import the components you need (`DataTable`, `useTableQuery`).
* The `@tablecraft/table` package lazily utilizes generic data structures. If you are deeply optimizing your initial bundle, you can dynamically import the `DataTable` itself using React's `Suspense` and `lazy()` or your framework's equivalent lazy-loading capability.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://jacksonkasi.gitbook.io/tablecraft/guides/performance.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.