blink-ripgrep.nvim

Ripgrep source for the blink.cmp completion plugin. Adding it to your configuration offers matching words from your entire project as completions. This can reduce the chance of typos as well as repetitive typing.

Note

By default, a project root is considered to be the nearest ancestor directory containing a .git directory. This can be configured with the project_root_marker option.

Forked here (mikavilpas/blink.cmp) for my own use from niuiic/blink-cmp-rg.nvim.

📦 Installation

The configuration of blink-ripgrep needs to be embedded into the configuration for blink. Example for lazy.nvim:

-- NOTE: you can leave out the type annotations if you don't want to use them

---@module "lazy"
---@type LazySpec
return {
  "saghen/blink.cmp",
  dependencies = {
    "mikavilpas/blink-ripgrep.nvim",
    -- 👆🏻👆🏻 add the dependency here
  },
  ---@module 'blink.cmp'
  ---@type blink.cmp.Config
  opts = {
    sources = {
      default = {
        "buffer",
        "ripgrep", -- 👈🏻 add "ripgrep" here
      },
      providers = {
        -- 👇🏻👇🏻 add the ripgrep provider config below
        ripgrep = {
          module = "blink-ripgrep",
          name = "Ripgrep",
          -- the options below are optional, some default values are shown
          ---@module "blink-ripgrep"
          ---@type blink-ripgrep.Options
          opts = {
            -- For many options, see `rg --help` for an exact description of
            -- the values that ripgrep expects.

            -- the minimum length of the current word to start searching
            -- (if the word is shorter than this, the search will not start)
            prefix_min_len = 3,

            -- The number of lines to show around each match in the preview
            -- (documentation) window. For example, 5 means to show 5 lines
            -- before, then the match, and another 5 lines after the match.
            context_size = 5,

            -- The maximum file size of a file that ripgrep should include in
            -- its search. Useful when your project contains large files that
            -- might cause performance issues.
            -- Examples:
            -- "1024" (bytes by default), "200K", "1M", "1G", which will
            -- exclude files larger than that size.
            max_filesize = "1M",

            -- Specifies how to find the root of the project where the ripgrep
            -- search will start from. Accepts the same options as the marker
            -- given to `:h vim.fs.root()` which offers many possibilities for
            -- configuration. If none can be found, defaults to Neovim's cwd.
            --
            -- Examples:
            -- - ".git" (default)
            -- - { ".git", "package.json", ".root" }
            project_root_marker = ".git",

            -- The casing to use for the search in a format that ripgrep
            -- accepts. Defaults to "--ignore-case". See `rg --help` for all the
            -- available options ripgrep supports, but you can try
            -- "--case-sensitive" or "--smart-case".
            search_casing = "--ignore-case",

            -- (advanced) Any additional options you want to give to ripgrep.
            -- See `rg -h` for a list of all available options. Might be
            -- helpful in adjusting performance in specific situations.
            -- If you have an idea for a default, please open an issue!
            --
            -- Not everything will work (obviously).
            additional_rg_options = {},

            -- When a result is found for a file whose filetype does not have a
            -- treesitter parser installed, fall back to regex based highlighting
            -- that is bundled in Neovim.
            fallback_to_regex_highlighting = true,

            -- Show debug information in `:messages` that can help in
            -- diagnosing issues with the plugin.
            debug = false,
          },
          -- (optional) customize how the results are displayed. Many options
          -- are available - make sure your lua LSP is set up so you get
          -- autocompletion help
          transform_items = function(_, items)
            for _, item in ipairs(items) do
              -- example: append a description to easily distinguish rg results
              item.labelDetails = {
                description = "(rg)",
              }
            end
            return items
          end,
        },
      },
      keymap = {
        -- 👇🏻👇🏻 (optional) add a keymap to invoke the search manually
        ["<c-g>"] = {
          function()
            -- invoke manually, requires blink >v0.8.0
            require("blink-cmp").show({ providers = { "ripgrep" } })
          end,
        },
      },
    },
  },
}

🤔 How it works

When you enter insert mode and start typing a word, blink triggers a search with blink-ripgrep. Only one search is done to save resources. After getting the results, the following keys are used to filter the results.

In this demo (using the option debug = true above), we can see the search starting when the word flashes with a different color. Notice how the word only flashes once:

invocations-working.mov

This is described in much more detail in blink's architecture documentation.

🏁 Performance

Generally performance is very good, but depending on the size of your project and your computer's specifications, the search can be fast or slow. Here are a few things you can do to improve performance:

• Set the prefix_min_len option to a larger number avoid starting a search for very short words. This can prevent unnecessary searches and improve performance.

• Use the max_filesize option to exclude large files from the search. This can prevent performance issues when searching in projects with large files.

• Disable automatic mode and use the manual mode to start the search only when you need it (see below).

• Set the debug = true option, which will log debug information to your :messages in Neovim. You can copy paste these commands to your terminal and try to figure out why the search is slow.

• Use ripgrep's extensive options to exclude/include files (ripgrep docs)

  • ripgrep supports global as well as project/directory specific ignore files. By default, it uses .gitignore, .git/info/exclude, .ignore, and .rgignore files.

• If you still experience performance issues, please open an issue for discussion.

Automatic mode

In this mode, the search starts automatically when typing a word that is at least prefix_min_len in length.

This is enabled by including the ripgrep provider in blink-cmp's providers:

return {
  -- ... other configuration
  ---@module 'blink.cmp'
  ---@type blink.cmp.Config
  opts = {
    sources = {
      default = {
        "lsp",
        "path",
        "snippets",
        "buffer",
        "ripgrep", -- 👈🏻 including this enables automatic search
      },
    }
  }
}

Manual mode

If you prefer to start the search manually, you can use a keymap to invoke the search. The example configuration includes a keymap that invokes the search when pressing Ctrl+g.