Skip to content

Latest commit

 

History

History
99 lines (65 loc) · 3.87 KB

README.md

File metadata and controls

99 lines (65 loc) · 3.87 KB
Raw

company-tabnine

MELPA

TabNine is the all-language autocompleter. It uses machine learning to provide responsive, reliable, and relevant suggestions.

company-tabnine provides TabNine completion backend for company-mode. It takes care of TabNine binaries, so installation is easy.

screenshot screenshot-deep-1 screenshot-deep-2 screenshot-deep-3

Installation

  1. Make sure company-mode is installed and configured.

  2. Install company-tabnine. This package is part of MELPA.

    Note: See https://melpa.org/#/getting-started for MELPA usage. Make sure to use the "bleeding-edge" repository instead of MELPA stable.

    • With use-package

      Put the following in your config:

      (use-package company-tabnine :ensure t)

    • With package.el (built-in)

      Install the package:

      M-x package-install RET company-tabnine RET

      Put the following in your config:

      (require 'company-tabnine)

  3. Add company-tabnine to company-backends

    (add-to-list 'company-backends #'company-tabnine)

  4. Run M-x company-tabnine-install-binary to install the TabNine binary for your system.

Recommended Configuration

Below are some recommended company-mode configuration that works well with company-tabnine.

;; Trigger completion immediately.
(setq company-idle-delay 0)

;; Number the candidates (use M-1, M-2 etc to select completions).
(setq company-show-numbers t)

Usage

company-tabnine should work out of the box.

See M-x customize-group RET company-tabnine RET for customizations.

Auto-balance parentheses

TabNine can automatically balance parentheses, by removing and adding closing parentheses after the cursor. See the examples here.

Note: The automatically-balancing happens in company's post-completion hook. However, company-tng-frontend actually suppresses this hook. In order to use automatic parentheses balancing, you need to manually call company-complete-selection or similar commands in this case, which will almost always happen if you do not use company-tng-frontend.

Known Issues

  • company-transformers or plugins that use it (such as company-flx-mode) can interfere with TabNine's sorting. If this happens, put the following temporary workaround in your config:

    ;; workaround for company-transformers
(setq company-tabnine--disable-next-transform nil)
(defun my-company--transform-candidates (func &rest args)
  (if (not company-tabnine--disable-next-transform)
      (apply func args)
    (setq company-tabnine--disable-next-transform nil)
    (car args)))

(defun my-company-tabnine (func &rest args)
  (when (eq (car args) 'candidates)
    (setq company-tabnine--disable-next-transform t))
  (apply func args))

(advice-add #'company--transform-candidates :around #'my-company--transform-candidates)
(advice-add #'company-tabnine :around #'my-company-tabnine)

  • Spacemacs configurations can override the settings for company-backends.

  • Conflict with ESS: See emacs-ess/ESS#955

  • TabNine's local deep learning completion might be enabled by default. It is very CPU-intensive if your device can't handle it. You can check by typing "TabNine::config" in any buffer (your browser should then automatically open to TabNine's config page) and disable Deep TabNine Local (you will lose local deep learning completion).