| .github/workflows | ||
| .gitignore | ||
| eshell-atuin.el | ||
| LICENSE | ||
| README.org | ||
eshell-atuin
atuin stores shell history in a database, which allows for having the same history across multiple shells, sessions, and optionally across different machines. This package provides functionality to store and browse eshell history in atuin.
Installation
The package isn't yet available anywhere but in this repository. My preferred way for such cases is use-package and straight.el.
(use-package atuin
:straight (:host github :repo "SqrtMinusOne/eshell-atuin")
:after eshell
:config
(eshell-atuin-mode))
Alternatively, clone the repository, add it to the load-path, and require the package.
Configuration
If your atuin binary is located in a place unknown to executable-find, set the atuin-executable variable.
If you are using a vertical completion system such as Ivy, Selectrum, etc., you can configure the completion interface, e.g.:
(setq eshell-atuin-search-fields '(time duration command))
(setq eshell-atuin-history-format "%-160c %t + %d")
The available flags are:
| Flag | atuin field (see help atuin search) |
Required |
|---|---|---|
| %t | time |
+ |
| %c | command |
+ |
| %e | exit |
|
| %d | duration |
|
| %i | directory |
|
| %u | user |
|
| %h | host |
|
| %r | relativetime |
See (emacs) Custom Format Strings for information on the general format-spec syntax.
I suspect the package might be slow if your history has a lot of records (I haven't checked yet). In this case, it might be worth setting a limit:
(setq eshell-atuin-search-options '("--exit" "0" "--limit" "10000"))
Usage
Enable eshell-atuin-mode to turn on storing eshell commands in atuin.
Run eshell-atuin-history inside an eshell buffer to browse the saved history. Accepting the completion will insert the command.
Implementation notes
I may have overengineered the package a bit to scale on lots of records.
The package caches the results of atuin search in eshell-atuin--history-cache (which see on the algorithm), and updates the cache incrementally. A formatted string for each entry is created at the moment of addition; entries are additionally "indexed" by a hashmap to lookup "raw" commands by their formatted versions.
So, the only places I see with the computational complexity of O(N), where N is the number of unique commands in atuin, are:
- populating the cache at the first run of
M-x eshell-atuin-history; - feeding the entirety of the cache to
completing-readon each run ofM-x eshell-atuin-history.