feat(rendering): incremental rendering (#62)

Introduces an incremental rendering that updates only the changed portion(s) of the buffer instead of re-rendering everything on every change. This reduces CPU usages for large sessions.

This PR also incorporate a functional test suite in `tests/manual` where you can record a run snapshot tests.

This is a major revision of the rendering system.


Co-authored-by:  Cameron Ring <[email protected]>

Author: sudo-tee

.emmyrc.json

@@ -0,0 +1,15 @@
+{
+    "$schema": "https://raw.githubusercontent.com/EmmyLuaLs/emmylua-analyzer-rust/refs/heads/main/crates/emmylua_code_analysis/resources/schema.json",
+    "runtime": {
+        "version": "LuaJIT"
+    },
+    "workspace": {
+        "library": [
+            "$VIMRUNTIME",
+            "$HOME/.local/share/nvim/lazy/luvit-meta/library/uv.lua",
+        ],
+        "ignoreGlobs": [
+            "**/*_spec.lua"
+        ]
+    }
+}

.github/workflows/tests.yml

@@ -0,0 +1,59 @@
+# From
+# https://github.com/shortcuts/neovim-plugin-boilerplate/blob/main/.github/workflows/main.yml
+
+name: tests
+
+on:
+  push:
+  pull_request:
+    types: [opened, synchronize]
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    name: lint
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: JohnnyMorganz/stylua-action@v4
+        with:
+          token: ${{ secrets.GITHUB_TOKEN }}
+          version: latest
+          args: --check . -g '*.lua' -g '!deps/'
+
+  test:
+    timeout-minutes: 4
+    strategy:
+      matrix:
+        os: [ubuntu-latest]
+        neovim_version: ["v0.10.3", "v0.11.4", "nightly"]
+        include:
+          - os: macos-latest
+            neovim_version: v0.11.4
+    runs-on: ${{ matrix.os }}
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: setup neovim
+        uses: rhysd/action-setup-vim@v1
+        with:
+          neovim: true
+          version: ${{ matrix.neovim_version }}
+
+      # only needed if testing in ssh but doesn't hurt
+      - name: Add nvim to PATH
+        run: echo "${{ steps.setup_nvim.outputs.executable }}" >> $GITHUB_PATH
+
+      - name: Install plenary
+        run: |
+          git clone --depth 1 https://github.com/nvim-lua/plenary.nvim ~/.local/share/nvim/site/pack/vendor/start/plenary.nvim
+
+      - name: Run tests
+        run: ./run_tests.sh
+
+      # # For sshing in to debug GH actions
+      # - name: Setup tmate session
+      #   uses: mxschmitt/action-tmate@v3
+      #   with:
+      #     detached: true

.luarc.json

@@ -1,16 +1,24 @@
 # AGENTS.md
 
 ## Build, Lint, and Test
+
 - **Run all tests:** `./run_tests.sh`
 - **Minimal tests:**
   `nvim --headless -u tests/minimal/init.lua -c "lua require('plenary.test_harness').test_directory('./tests/minimal', {minimal_init = './tests/minimal/init.lua', sequential = true})"`
 - **Unit tests:**
   `nvim --headless -u tests/minimal/init.lua -c "lua require('plenary.test_harness').test_directory('./tests/unit', {minimal_init = './tests/minimal/init.lua'})"`
 - **Run a single test:** Replace the directory in the above command with the test file path, e.g.:
-  `nvim --headless -u tests/minimal/init.lua -c "lua require('plenary.test_harness').test_directory('./tests/unit/job_spec.lua', {minimal_init = './tests/minimal/init.lua'})"`
+  `nvim --headless -u tests/manual/init.lua -c "lua require('plenary.test_harness').test_directory('./tests/unit/job_spec.lua', {minimal_init = './tests/minimal/init.lua'})"`
+- **Manual/Visual tests:** `./tests/manual/run_replay.sh` - Replay captured event data for visual testing
+- **Debug rendering in headless mode:**
+  `nvim --headless -u tests/manual/init_replay.lua "+ReplayHeadless" "+ReplayLoad tests/data/FILE.json" "+ReplayAll 0" "+qa"`
+  This will replay events and dump the output buffer to stdout, useful for debugging rendering issues without a UI.
+  You can also run to just a specific message # with (e.g. message # 12):
+  `nvim --headless -u tests/manual/init_replay.lua "+ReplayHeadless" "+ReplayLoad tests/data/message-removal.json" "+ReplayNext 12" "+qa"`
 - **Lint:** No explicit lint command; follow Lua best practices.
 
 ## Code Style Guidelines
+
 - **Imports:** `local mod = require('mod')` at the top. Group standard, then project imports.
 - **Formatting:** 2 spaces per indent. No trailing whitespace. Lines ≤ 100 chars.
 - **Types:** Use Lua annotations (`---@class`, `---@field`, etc.) for public APIs/config.
@@ -19,6 +27,6 @@
 - **Comments:** Only when necessary for clarity. Prefer self-explanatory code.
 - **Functions:** Prefer local functions. Use `M.func` for module exports.
 - **Config:** Centralize in `config.lua`. Use deep merge for user overrides.
-- **Tests:** Place in `tests/minimal/` or `tests/unit/`.
+- **Tests:** Place in `tests/minimal/` or `tests/unit/`. Manual/visual tests in `tests/manual/`.
 
 _Agentic coding agents must follow these conventions strictly for consistency and reliability._

AGENTS.md

@@ -175,6 +175,10 @@ require('opencode').setup({
       tools = {
         show_output = true, -- Show tools output [diffs, cmd output, etc.] (default: true)
       },
+      rendering = {
+        markdown_debounce_ms = 250, -- Debounce time for markdown rendering on new data (default: 250ms)
+        on_data_rendered = nil, -- Called when new data is rendered; set to false to disable default RenderMarkdown/Markview behavior
+      },
     },
     input = {
       text = {
@@ -530,7 +534,7 @@ The plugin defines several highlight groups that can be customized to match your
 - `OpencodeAgentBuild`: Agent indicator in winbar for Build mode (default: #616161 background)
 - `OpencodeAgentCustom`: Agent indicator in winbar for custom modes (default: #3b4261 background)
 - `OpencodeContestualAction`: Highlight for contextual actions in the output window (default: #3b4261 background)
-- `OpencodeInpuutLegend`: Highlight for input window legend (default: #CCCCCC background)
+- `OpencodeInputLegend`: Highlight for input window legend (default: #CCCCCC background)
 - `OpencodeHint`: Highlight for hinting messages in input window and token info in output window footer (linked to `Comment`)
 
 ## 🔧 Setting up Opencode