Why Emacs Keyboard Shorts are Painful. [was Good Lisp editor for Win]

Hi all,

I've done some research on emacs default shortcuts. Please see:

http://xahlee.org/emacs/emacs_kb_shortcuts_pain.html

Below is a text version

-----------------

Why Emacs's Keyboard Shortcuts Are Painful

Xah Lee, 2007-07

A important aspect in designing a keyboard shortcut set, for a
application that has intensive, repetitive, prolonged human-machine
interaction (such as coding and text editing), is to consider
ergonomic principles. Specifically: allocate keyboard shortcuts for
the most frequently used commands, and, the top most frequently used
commands should have most easily-pressed keystrokes. For example,
they should be on the home row.

This article shows why Emacs's keyboard shortcut set is the most
ergonomically bad.

The Swapping of Control and Meta Modifiers

Emacs's keyboard shortcuts is very inefficient. The primary cause is
because, emacs's keyboard shortcuts are designed with a keyboard that
practically has the Ctrl and Alt key positions swapped.

Space Cadet keyboard

above: The Space-cadet keyboard (Source↗, 2007-07) .

The common keyboard used around emacs era in the 1980s are those
keyboards from Lisp Machines↗. (see Space-cadet keyboard↗) The
keyboard on lisp machines have the Control key right besides the space
bar (similar to the position of Alt keys on PC keyboards), and Meta to
the left of Control. So, the Control key is the primary modifier, and
the Meta is secondary to Control. This is why, the shortcuts for the
most used commands in emacs involve the Control key instead of the
Meta key. (Example: The cursor movements: C-p, C-n, C-f, C-b, C-a, C-
e, the cut/paste/undo C-w, C-y, C-/, the kill-line C-k, the mark C-
SPC, the search C-s.) Lisp Machine's keyboards fell out of use alone
with Lisp Machines. Since the 1990s, the IBM PC keyboard↗ (and its
decedents) becomes the most popular and is used by some 98% of
personal computers today. The PC keyboard does not have Meta key but
have Alt instead. The Alt is placed right beside the space bar, while
Control is placed far to the corner.

Emacs did not change its keyboard shortcut bindings to adapt the PC
keyboard. Emacs simply remapped its Meta shortcuts to the Alt key by
default. (and kept on using the terminology Meta)

The tragedy of the Control/(Alt/Meta) swap made emacs keyboard
shortcuts very painful, and the frequent need to press the far-away
Control key creates the Emacs Pinky syndrome. (Many emacs-using
programer celebrities have injured their hands with emacs. (e.g.
Richard Stallman↗, Jamie Zawinski↗), and emacs's Ctrl and Meta
combinations are most cited as the major turn-off to potential users
among programers)

(For more photos of Lisp Machine's other keyboards (all have Control
as primary), see: Symbolics keyboard PN 364000↗ , Symbolics keyboard
PN 365407 Rev C↗ )

(The reason that Symbolics keyboards have Control as primary modifier,
is because in the early computing era, the use of Control Characters
as part of the non-printable chars from the ASCII↗ standard, is
important and frequent. This still can be seen in emacs today by the
standard use of “Form feed” character (ASCII 12, represented as ^L) as
a indication of “page break”/ section in much of Emacs Lisp's source
codes. (To go to the previous/next ^L, emacs uses the command forward-
page (C-x ]) and backward-page (C-x [)) At those times, there's not
much of the concept of “keyboard shortcuts”, but rather, modifier keys
are means to enter special data. Today, the Control key is primarily
used as a mechanism for keyboard shortcuts. Only in Telnet/ Terminal/
SSH applications, the Control key still remain of its original use.)

(The reason that modifier keys changed purpose, from special
characters data entry to being a shortcut to invoke commands for
software applications, is due to the change in computer use. In the
1970s, computer users are mathematicians and programers, and their
activity is writing programs for scientific research. Throughout the
decades, computers became cheaper, and PC was born and become a
household commodity. People use computers for its software
applications, beginning with application for professionals such as
word processor, spread sheets, image-editor in the 1990s, and today in
2007, most people use computer for email, web browsing, instant
messaging, gaming, music and video watching.)

The Choice Of Keys

The shortcut's key choices are primarily based on first letter of the
commands, not based on key position and finger strength or ease of
pressing the key. For example, the single char cursor moving shortcuts
(C-p previous-line ↑, C-n next-line ↓, C-b backward-char ←, C-f
forward-char →) are scattered around the keyboard with positions that
are most difficult to press. (these shortcuts all together accounts
for 43% of all commands executed by a keyboard shortcut) Of these, the
most frequently used is C-n (next-line), which accounts for 20% of all
shortcut calls, but is assigned to the letter n, positioned in the
middle of the keyboard, which is one of the most costly key to press.
Similarly, the second most used among these is the C-p (previous-
line), accounting for 16% of all shortcut command calls, is located in
a position above the right hand's pinky, also one of the most costly
key to press.

(Here we assumes the QWERTY keyboard layout. On the Dvorak layout, it
is about as bad.)

See also, a newsgroup post on “comp.emacs”. “Re: effective
emacs” (2008-06-01) by Daniel Weinreb↗. http://groups.google.com/group/comp.emacs/msg/
0342e0bc1aa05c0d.

«Emacs's default cursor moving shortcuts are “Ctrl+f”, “Ctrl+b”,
“Ctrl +n”, “Ctrl+p”. The keys f, b, n, p are scattered around the
keyboard and are not under the home row.»

That's true. At the time Guy Steele put together the Emacs
default key mappings, many people in the target user community (about
20 people at MIT!) were already using these key bindings. It would
have been hard to get the new Emacs bindings accepted by the community
if they differed for such basic commands. As you point out, anyone
using Emacs can very easily change this based on their own ergonomic
preferences.

Outdated Commands

A significant portion of emacs's major shortcuts (those with M-‹key›
or C-‹key›) are mapped to commands that are almost never used today.
Some of these occupies the most precious space (Home row with thumb:
For example: M-s (center-line), M-j (indent-new-comment-line), M-k
(kill-sentence)). Most programer who have used emacs for years never
use these commands. For example:

digit-argument, M-1 to M-9
negative-argument, M--

move-to-window-line, M-r
center-line, M-s
transpose-words, M-t
tab-to-tab-stop, M-i

M-g prefix, M-g
indent-new-comment-line, M-j
tmm-menubar, M-'

zap-to-char, M-z
back-to-indentation, M-m
tags-loop-continue, M-,
find-tag, M-.

(Note: Conversely, some commands that are used by every emacs user
every hour, such as Open (find-file; C-x C-f), Save (save-buffer; C-x
C-s), Close (kill-buffer; C-x k), Next Window/Tab (next-buffer C-x →)
all require multiple keystrokes with the difficult Control key.)

No Employment of the Shift Key

For historical reasons, emacs do not use any keybindings involving the
Shift with a letter. (Example: there's no “meta shift a”, or “control
shift a”) This is so because in early computing environment, such key
combination cannot be distinguished, due to a practical combination of
ASCII↗, Computer terminal↗, telnet↗.

Today, however, employing the Shift key as part of a shortcut with
other modifiers is common and convenient. For example, on Mac OS X,
Undo and Redo are Cmd+Z and Cmd+Shift+Z, Save and Save As are Cmd+S
and Cmd+Shift+S. On Mac and Windows, moving to next/previous field/
window/application often use the Shift key for reversing direction. In
text editing on both Mac and Windows, a modifier key with a arrow key
will move cursor by word/paragraph, and with Shift down will select
them while moving.

Using the Shift key as a reverse operation is very easy to remember,
and doesn't take another precious shortcut letter. By not using the
Shift key, commands with a logical reverse operation necessarily have
to find other key space, and overall making the shortcut set more
difficult to remember, or scattered, or more difficult to press.

A Flaw in Keybinding Policy

Any major software, maintains a guide for the developers about the
choices of keyboard shortcuts, so that the shortcuts will be
consistent. Emacs has this in its Emacs Lisp manual: Elisp Manual: Key-
Binding-Conventions.

This guide, indicates that the only key space reserved for users to
define, are the function keys F5 to F9, and key stroke sequence
starting with Ctrl+c followed by a single letter key.

This is a severe restraint to the utility of customized shortcuts. F5
to F9 are only 6 keys. The key sequence starting with C-c followed by
a letter, is a difficult sequence to execute, and there are only 26
spaces there.

The function keys, F1 to F12, are very good candidates for user
defined shortcut space, similarly for the digit key shortcuts, 0 to 9.
These keys can be used with any combination of Control, Meta, Shift.
For example, a user might define them to insert various templates,
headers/footers, a system of customized HTML/XML tags. Or, she might
assign them to various special emacs modes such as dired, shell, ftp,
email, calendar, calc, *scratch*, make-frame-command (Open a new
window), insert signature.

It seems too drastic a policy, to limit user defined keys to only F5
to F9, and key sequence of Control+c followed by a single letter key.

------------------------------------------------------------------------------

Epilogue: Failure to Change

Today, most commonly used keyboard shortcuts have been somewhat
informally standardized. For example, C/X/V is for Copy/Cut/Paste. O
is for Open. S is for Save, Shift-S is for Save As. P is for Print. F
is for Find/Search. Tab is for next, Shift tab for previous. These are
common conventions today in every application across Microsoft Windows
and Macintosh (and in Linux too in general).

These shortcut conventions are primarily brought about by Apple
Computer Inc's
Human interface guidelines↗ and IBM's Common User Access↗ in the
1990s.

In the early 1990s, DOS era software, each application has its own
scheme of
shortcuts. The following is a excerpt from the Wikipedia article on
Common
User Access↗:

CUA was a detailed specification and set strict rules about how
applications should look and function. Its aim was in part to bring
about harmony between MS-DOS applications, which until then had
implemented totally different user interfaces.

Examples:

+ In WordPerfect, the command to open a file was [F7], [3].
+ In Lotus 1-2-3, a file was opened with [/] (to open the
menus), [W]
(for Workspace), [R] (for Retrieve).
+ In Microsoft Word, a file was opened with [Esc] (to open the
menus),
[T] (for Transfer), [L] (for Load).
+ In WordStar, it was [Ctrl]+[K]+[O].
+ In Emacs, a file was opened with [Ctrl]+[x] followed by [Ctrl]+
[f]
(for find-file).

Some programs used [Esc] to cancel an action, some used it to
complete one; WordPerfect used it to repeat a character. Some programs
used [End] to go to the end of a line, some used it to complete
filling in a form. [F1] was often help but in WordPerfect that was
[F3]. [Ins] sometimes toggled between overtype and inserting
characters, but some programs used it for “paste”.

Thus, every program had to be learned individually and its
complete user interface memorized. It was a sign of expertise to have
learned the UIs of dozens of applications, since a novice user facing
a new program would find their existing knowledge of a similar
application absolutely no use whatsoever.

Commercial software have updated themselves with time (or went
extinct), but emacs has not.

If we take a survey of the market share of text editors (including
IDEs) among professional programers (as defined by those who make a
living by computer programing), then, it is my guess, that emacs from
mid 1980s to early 1990s, has more than 50% of market share, but
gradually declined. Today, perhaps less that 5% of professional
programers use emacs (possibly even below 1%). I think, part of the
reason being that emacs has not modernized (not in the sense of being
fashionable, but in the sense of keeping with hardware and software
changes in the IT industry). The other major reason, is because emacs
itself is not a IDE in a modern sense, and most programing development
using compiled languages such as Pascal, C, C++, Java, C#, have moved
on with IDE platforms integrated with these languages's compiler
application.

See also: The Modernization of Emacs.

Xah
∑ http://xahlee.org/

☄
.

Loading