chore(release): 3.6.0

Author: asif7774

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -34,7 +34,7 @@ body:
     id: version
     attributes:
       label: OpenGridX version
-      placeholder: "3.5.1"
+      placeholder: "3.6.0"
     validations:
       required: true
   - type: dropdown

CHANGELOG.md

@@ -5,7 +5,7 @@
 
 ---
 
-## [Unreleased]
+## [3.6.0] — 2026-10-12
 
 ### Added
 

CLAUDE.md

@@ -14,7 +14,7 @@ This file is auto-loaded by Claude Code and other AI coding assistants. It provi
 - **Build:** `npm run build:lib` produces the published `dist/opengridx.es.js` / `dist/opengridx.umd.js`. `npm run build` type-checks the lib and builds the **demo site**.
 - **Published:** yes — on npm as `@opencorestack/opengridx` with external consumers. Treat public types and behaviour changes as semver-relevant.
 - **Lint:** `npm run lint` (ESLint). Must pass before every commit.
-- **Current version:** 3.5.1
+- **Current version:** 3.6.0
 
 ---
 
@@ -153,6 +153,20 @@ The GitHub wiki is generated from `docs/` + `wiki/` by `scripts/build-wiki.mjs`
 
 ---
 
+## v3.6.0 — significant changes
+
+- **Sorting** encodes each key into typed arrays (`encodeSortColumn`): rank 0–4 (number, date, boolean, string, other), empty = 5, plus a number per row; strings/other get their position among the distinct texts in collator order. Only keys where an item carries both `comparator` and `params` use the per-pair `callSortComparator` path. Never call the collator inside a comparator over rows
+- `getCellValue` returns `row[field]` without allocating when the column has no `valueGetter` — keep it allocation-free. Built-in aggregations go through `aggregateRows` (row by row, same addition order); `AGGREGATION_FUNCTIONS` stays the value-array API for pivot and cell selection
+- Quick filter: `filterRows` keeps texts by index for the last rows array per `GridColumnLookup` (reused when the same object is at the same index); `createRowFilter` keeps a per-row WeakMap. Joining texts into one string was measured and rejected
+- Row-grouping leaf `GridRowMeta` objects are one per row (sharing per depth was reverted: `params.rowMeta` is public). Equivalence tests with the old code as oracles: `*.equivalence.test.ts`
+- **`dist/opengridx.es.js` starts with `'use client'`** (banner in `vite.config.lib.js`, checked by check-bundle); the UMD build has none
+- **Every optional-peer `import()` (`exceljs`, `jspdf`, `jspdf-autotable`) must stay inside `try`** — webpack and Turbopack only treat it as optional there; a hidden variable specifier breaks apps that install the peers. The `nopeers` smoke fixture (webpack 5, no peers, port 4321, `e2e/nopeers.spec.ts`) guards it; fixtures must keep webpack from resolving the root `node_modules`
+- `validateGridAiState` skips a top-level string `message`; the prompt handler surfaces it as `result.message` for object, JSON-string and fenced input
+- **RTL**: read/write horizontal scroll only through `getScrollStart` / `setScrollStart` (`lib/utils/direction.ts`; scroll state is the offset from the inline start, never negative); map arrows with `logicalArrowKey(key, isRtl(el))`; panels with `getPanelPosition`. `pinnedColumns.left` / `right` are logical (inline start / end): sticky offsets use `insetInlineStart` / `insetInlineEnd` (jsdom: `style.getPropertyValue('inset-inline-start')`), borders `border-inline-*`, box-shadow x-offsets need `:dir(rtl)` mirrors. The column menu pins by screen side. `useGridScrollSync` observes `dir` mutations
+- Testing RTL: Playwright's headless Chromium runs with `--hide-scrollbars`; with classic scrollbars and `scrollbar-gutter: stable`, RTL cannot reach `scrollLeft` 0 — compare against the scroll origin. Reproduce on macOS with `defaults write chrome-headless-shell AppleShowScrollBars Always`. The grid's `scrollbar-width` / `scrollbar-color` disable `::-webkit-scrollbar` in Chromium
+- Header filter strings are in `localeText` (`headerFilter*`, resolved in `lib/utils/headerFilterLocale.ts`)
+- Not in this release: the time-sliced pipeline (Option A of `docs/superpowers/specs/2026-10-12-nonblocking-pipeline-proposal.md`) is held for 3.7.0 until its overhead is within ~20 %
+
 ## v3.5.1 — significant changes
 
 - **Content height is capped to the browser's limit.** Firefox lays out elements up to about 17.9M px and collapses a taller one (it does not clamp like Chromium/WebKit at about 33.5M px), which made large grids unscrollable there. `useGridMaxContentHeight` (hook 32a, before `useGridVirtualization`) measures the limit once per document with an invisible probe (`lib/utils/maxElementHeight.ts`: `getMaxElementHeight`, `countReachableRows`; `undefined` until measured, `Infinity` without a layout engine), and `useGridVirtualization` returns `virtualContainerHeight` capped to it — always render the rows' container with that, never `totalHeight - pinned heights`

demo/App.tsx

@@ -167,7 +167,7 @@ export default function App() {
                         <img src={`${import.meta.env.BASE_URL}logo.png`} alt="OpenGridX Logo" className="app-logo" />
                         <h2 className="app-title">
                             OpenGridX
-                            <span className="app-version">v3.5.1</span>
+                            <span className="app-version">v3.6.0</span>
                         </h2>
                     </div>
 

demo/Home.tsx

@@ -63,7 +63,7 @@ export default function Home(_props: HomeProps) {
                 <div className="home-logo-hero">
                     <img src={`${import.meta.env.BASE_URL}banner.png`} alt="OpenGridX Logo" className="home-banner-image" />
                 </div>
-                <span className="home-badge">OpenGridX v3.5.1</span>
+                <span className="home-badge">OpenGridX v3.6.0</span>
                 <p className="home-subtitle">
                     The elite, high-performance DataGrid for modern React.
                     Built to handle massive data with a premium developer experience.

demo/public/llms.txt

@@ -1,6 +1,6 @@
 # OpenGridX — AI Agent Context
 # Package: @opencorestack/opengridx
-# Version: 3.5.1
+# Version: 3.6.0
 # Docs: https://opencorestack.github.io/OpenGridX/
 # This file is machine-readable. AI developers: start here.
 

llms.txt

@@ -1,6 +1,6 @@
 # OpenGridX — AI Agent Context
 # Package: @opencorestack/opengridx
-# Version: 3.5.1
+# Version: 3.6.0
 # Docs: https://opencorestack.github.io/OpenGridX/
 # This file is machine-readable. AI developers: start here.
 

opengridx-ai-context.md

@@ -8,7 +8,7 @@ that uses the `@opencorestack/opengridx` library. Read it in full before writing
 ## What this library is
 
 `@opencorestack/opengridx` is a zero-dependency, high-performance React DataGrid component.
-Current version: **3.5.1**. It is a full custom implementation — not a wrapper around MUI or any
+Current version: **3.6.0**. It is a full custom implementation — not a wrapper around MUI or any
 other library.
 
 ---

package-lock.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencorestack/opengridx",
-  "version": "3.5.1",
+  "version": "3.6.0",
   "description": "OpenGridX: High-performance React data infrastructure. Unlock advanced Row Grouping, Excel Export, and Column Pinning without the usual \"Pro\" gatekeeping. Built for speed, scale, and complete architectural freedom. Fully open, virtualization-ready, and feature-complete.",
   "type": "module",
   "main": "./dist/opengridx.umd.js",