Emacs's Command Frequency

Xah Lee, 2007-08

This page lists emacs's commands in the order of their frequency of use.

In emacs, each key press actually invokes a command. (For example, typing “a” actually invokes the command “self-insert-command”. Typing Ctrl+f invokes the command forward-char. Typing Alt+x invokes execute-extended-command. Typing the right arrow key “→” invokes forward-char.)

The motive of this study, is in designing a A Ergonomic Keyboard Shortcut Layout For Emacs. I needed to know which command are actually more frequently used, in order to give it a easier-to-press keybinding.

Preliminary

The following frequency list is based 3 users. Xah Lee (me), Marc Shapiro, “rgb” (rbielaws_@_i1_._net). The total count of all command calls is 203118. The following is a break down of each's data contribution:

Name          cmd calls         Percentage
-------------------------------------------
rgb             93 534          46%
Xah             82 090          40%
Marc            27 494          13%
-------------------------------------------
total          203 118         100%

The total count of self-insert-command calls is 65 496. The following is a break down by that command.

Name         number of         Percentage
             self-insert-     
             command call     
-------------------------------------------
Xah             29 758          45%
Marc            20 026          31%
rgb             15 712          24%
-------------------------------------------
total           65 496         100%

Following is a table showing the proportion of each person's use of data-entry commands and non-data-entry commands. “data-entry” commands are the “self-insert-command” and “newline” commands. (The “newline” is bound to the Enter/Return key.) All other commands are considered “non-data-entry”.

Name           Percentage of non-data-entry commands
---------------------------
rgb             83%
Xah Lee         63%
Marc Shapiro    26%

A higher percentage of non-data-entry commands probably means that the person is doing a lot editing, as opposed to writing. (if a person uses emacs for email or IRC chat, she's likely to have high percentage of data-entry commands.)

Commands that are executed by the same key are grouped together. For example, the commands “delete-backward-char”, “backward-delete-char-untabify”, “python-backspace”, “cperl-electric-backspace” are all just the “Backspace” key.

The following is the complete list of commands that are merged.

command names                 merged into this glyph
---------------------------------------------
next-line                     ↓
dired-next-line               ↓
next-history-element          ↓
                             
previous-line                 ↑
dired-previous-line           ↑
previous-history-element      ↑
                             
delete-backward-char          ⌫
backward-delete-char-untabify ⌫
python-backspace              ⌫
cperl-electric-backspace      ⌫

scroll-up                     ▼
cua-scroll-up                 ▼

scroll-down                   ▲
cua-scroll-down               ▲

isearch-forward                isearch-→
isearch-repeat-forward         isearch-→

isearch-backward               isearch-←
isearch-repeat-backward        isearch-←

For clarity, some commands are replaced by a glyph. Here's the complete list:

backward-char           ←
forward-char            →
backward-word           ←w
forward-word            →w
backward-sentence       ←s
forward-sentence        →s
backward-paragraph       ↑¶
forward-paragraph        ↓¶

move-beginning-of-line  |←
move-end-of-line         →|
beginning-of-buffer     |◀
end-of-buffer           ▶|

delete-char            ⌦
kill-word              ⌦w
backward-kill-word     ⌫w
kill-line              ⌦l
kill-sentence          ⌦s
kill-ring-save         copy
kill-region            ✂

Note: Due to a ideosyncracy of emacs lisp, the following commands: kill-word, backward-kill-word, kill-line, kill-sentence, are all counted as just kill-region. In symbols, the following “⌦w”, “⌫w”, “⌦l”, “⌦s”, all are just “✂”. See end of this report for detail.

Command Frequency Table

Total number of command calls: 203118

Percent of non-data-entry command calls: 67%

How The Report Is Compiled

Here we describe how the report is generated. Each person, uses this emacs code: command-frequency.el, which will accumulate a keyed-list of all command the user calls in emacs, with key being the command name and value being its count. After some time, the user calls command-frequency-write-file which will write the output into a file.

The raw files used in this report are here: command-frequency_xah.txt, command-frequency_marc.txt, command-frequency_marc2.txt, command-frequency_rgb.txt.

A python program command-frequency.py, is used to parse these raw command-frequency files. The program aggregate all user's raw data, then groups some commands together as described in this report, and replace some command names by a corresponding glyph, and print them in HTML table format. Commands used less than 0.1% of total non-data-entry commands are not displayed.

Caveats

For technical reasons, the command-frequency.el program, lumps commands backward-kill-word, kill-word, kill-line, kill-region, all into just kill-region. It would be better to modify command-frequency.el to work around this, because these commands are all very frequently used. It is important to know their exact frequency distribution.

The gist of command-frequency.el is just this line: “(add-hook 'post-command-hook 'command-frequency-record)”. The command-frequency-record simply add the value of the buildin var “this-command” to a keyed list. The reason that it lumps commands backward-kill-word, kill-word, kill-line, kill-region, all into just kill-region, is because the “this-command” actually record kill-region instead of any of the called commands. (thanks to Katsumi Yamaoka). Quote from lisp doc:

— Variable: this-command

    This variable records the name of the command now being executed
    by the editor command loop.

     ...

    Some commands set this variable during their execution, as a flag
    for whatever command runs next. In particular, the functions for
    killing text set this-command to kill-region so that any kill
    commands immediately following will know to append the killed text
    to the previous kill.

If you like to participate in this research, please download command-frequency.el and send me the output. I will incorporate your data into the report. Thank you.

A previous version of this report, based on Xah Lee only, is available at command-frequency2.html.

Thanks to Ryan Yeske for writing the elisp code to compile command stats. Thanks to Marc Shapiro and “rgb” for providing me their data.

See also:

• Emacs's Keyboard Shortcut Layout
• A Ergonomic Keyboard Shortcut Layout For Emacs
• Emacs Lisp Function Frequency

Page created: 2007-07.
© 2007 by Xah Lee.