YET ANOTHER NVIM CONFIG

English | 中文

A Modular & High-performance set of configurations for Nvim which is written all by hand

Note

This config is used to be streamlined, why the complexity now? As the configuration grew, a flat structure became hard to maintain. The new Modular Architecture separates Policy (what to enable) from Implementation (plugin setup). It is now a data driven configuration.

• lua/modules/profile/: Defines how the editor behaves.
• lua/modules/lang/: Defines what each language provides. This makes the config a framework rather than just a set of dotfiles.

⚡Showcase ⚡

Feature

• Modular Profile System: Switch between different configuration profiles seamlessly. Customize editor behavior on different device/os just by modify a json table.

• Granular Language Control: Fine-grained control over LSP, Treesitter, and Formatter on a per-language basis via environment variables.

• Dynamic Capabilities: Automatic adjustment of features based on the environment (e.g., Neovim version, environment variables).

• Well organized code(also well annotioned) for beginners to understand -- good code is the best document 😈😈😈

Dependency

Some binaries should be installed before launch the configuration.

• make for markdown previewer

• yarn for markdown previewer

• rg for fuzzy finder

• fzf for fuzzy finder

• gcc/clang anyways, a c compiler for tree-sitter parser compilation

• tree-sitter required by nvim-treesitter

• fcitx5-remote Linux,MacOS for ime-switcher

• win32yank.exe Windows for system clipboard support

• cargo full rust toolchain, optionally required by blink.cmp

⚡Try it now ⚡

You can try it immediately without replacing your origin configurations.

./venv # Launch a virtual env shell, it's actually a sanbox shell that isolates plugins, cache and runtime file

vi     # Launch neovim on this config, without make changes to your ~/.local/share

Or, you may download the repo to ~/.config/<name>, and run NVIM_APPNAME=<name> nvim to launch

Implemented functionalities

• Auto completion
• Status line
• Color and comment highlight
• File system explorer
• Markdown preview
• Markdown renderer
• Outline
• Integrated terminal
• Fuzzy finder
• LSP support
• Formatter
• SSH clipboard support(Need tmux extra config)
• Input method auto switch (On Linux and MacOS)
• Workspace patch
• Typst support
• Task runner infrastructure (Overseer)
• Windows Support
• Centralized lang feature switch
• Json-env combined profile system

Optional features

Some features are optional, controlled by a json file -- nvimrc.json, this file should be placed under your config dir. Here're customizable items:

• UI

  • transparent_mode enable transparent mode
  • dashboard_art_name choose an ascii art on dashboard
  • statline_scrollbar_style choose a style for heirline scroll bar, which displays cursor position
  • diagnose_inline do not use virtual lines to display diagnostic messages
  • enable_current_line_blame enable virtual text line blame at the end of line

• Lang module

  • silent_lang_diag do not output log of lang loader
  • lang_blacklist disabled lang configs, default none, split by ','
  • lang_whitelist enabled lang configs, default all, split by ','
  • lang_levels lang feature config, syntax: string c:full;cpp:none;rust:lsp,+ts,-fmt means enable full feature for c, disable all features for cpp, enable tree-sitter and lsp, disable formatter for rust.

• Lsp

  • enable_lsp enable lsp Disable LSP if nvim version <= 0.11
  • use_emmylua_ls use emmylua_ls as lua language server
  • workspace_inject_plugin_path inject plugin path to emmylua_ls workspace config
  • workspace_inject_vim_rt inject vim runtime to emmylua_ls workspace config

• BigFile

  • bigfile_size_byte average byte size
  • bigfile_size_line average line length (useful for minified files)

• Workspace

  • allow_workspace_patch allow editor patch its behavior according to workspace config
  • workspace_patch_always_restrict always enable restrict mode, disable dofile to prevent ACE

• Misc

  • sandbox_mode control sanbox features sesson|undo|shada|swap|wb, wb for writebackup
  • silent_profile_diag do not output log of profile loader
  • disable_im_switch disable auto im switcher
  • blink_use_binary use prebuild binary instead of building

Here are defaults:

local defaults = {
  blink_use_binary = true,
  dashboard_art_name = "Ayanami Rei",
  diagnose_inline = false,
  disable_im_switch = false,
  enable_current_line_blame = false,
  enable_lsp = true,
  lang_blacklist = "all",
  lang_levels = "",
  lang_whitelist = "",
  sandbox_mode = "none",
  silent_lang_diag = false,
  silent_profile_diag = false,
  statline_scrollbar_style = "moon",
  transparent_mode = false,
  use_emmylua_ls = false,
  workspace_inject_plugin_path = false,
  workspace_inject_vim_rt = true
  bigfile_size_byte = 2097152,
  bigfile_size_line = 100000,
  allow_workspace_patch = false,
  workspace_patch_always_restrict = true,
}

Note

Environment variables are still available. They have higher priority than json configured values.

Design philosophy

Unlike standard Neovim distributions that focus on being a "bundle" of features, this configuration is architected as a Configuration Engine. It treats configuration as structured data and the loading process as a controlled pipeline.

1. Separation of Mechanism and Policy

The core of this setup is the Mechanism (the Loader). It doesn't care what is being loaded; it only cares about how to scan directories, resolve module paths, and aggregate data. The Policy (LangSpec or PluginSpec) is pure data.

• Benefit: You can add support for a new language by creating a single declarative file without touching the orchestration logic.

2. Environment-Driven Runtime Masking

This editor should adapt to context without code changes. By utilizing environment variables like NVIM_LANG_WHITELIST and NVIM_LANG_BLACKLIST, the configuration allows for "Runtime Masking."

• The Logic: Whitelists act as a "Permit" that overrides any "Deny All" blacklist.
• Use Case: Debugging a core issue? Execute NVIM_LANG_BLACKLIST=all nvim for a clean slate instantly.

3. Structured Atomicity

Every module (e.g., config.langs.python) is an atomic unit. I use a Recursive Path Stack during loading to ensure that the hierarchy of your filesystem is reflected in the resulting configuration state. This avoids global namespace pollution and makes "Configuration Collections" (one file defining multiple sub-entities) possible and safe.

Comparison: Distro / Typical nvim config vs. This Engine

Feature	Typical Distro	This Configuration
Loading	Procedural & Linear	Scanned & Aggregated
Toggling	Hardcoded Booleans	Semantic Env Masks (+/-)
Data Structure	Imperative Setup Calls	Declarative Specs (Schema-driven)
Philosophy	"Include Everything"	"Load only what's whitelisted"

Maybe wanted features

• Markdown Table Format

• Image preview in markdown(Partly support)

• Search enhance

• Fold range highlighting on unfolding

• AsciiMode -- No nerd font

TODO List

• Latex preview

• Collect assets and remove some hard coded options

This readme is partial generated by generative AI, but highly audited by human