Welcome to the Emacs Shell
[Note: This README is also available in Japanese; see README.ja]
Eshell is a command shell implemented entirely in Emacs Lisp. It invokes no
external processes beyond those requested by the user. It is intended to be a
functional replacement for command shells such as bash, zsh, rc, 4dos; since
Emacs itself is capable of handling most of the tasks accomplished by such
tools.
Despite the sheer fact that running an Emacs shell can be fun, here are a few
of the unique features offered by Eshell. More are documented in greater
detail under EshellFeatures.
# Integration with the Emacs Lisp programming environment
The ability to have the same shell on every system Emacs has been ported
to. Since Eshell imposes no external requirements, and relies upon only the
Lisp functions exposed by Emacs, it is quite operating system
independent. Several of the common UNIX commands, such as ls, mv, rm, ln,
etc., have been implemented in Lisp in order to provide a more consistent work
environment.
For those who might be using an older version of Eshell, version 2.1
represents an entirely new, module-based architecture. It supports most of the
features offered by modern shells. Here is a brief list of some of its more
visible features:
- Command argument completion (tcsh, zsh)
- Input history management (bash)
- Intelligent output scrolling
- Psuedo-devices (such as "/dev/clip" for copying to the clipboard)
- Extended globbing (zsh)
- Argument and globbing predication (zsh)
- I/O redirection to buffers, files, symbols, processes, etc.
- Many niceties otherwise seen only in 4DOS
- Alias functions, both Lisp and Eshell-syntax
- Piping, sequenced commands, background jobs, etc...
Eshell was designed to run on Emacs 20.4 or higher, and XEmacs 21.1 or
higher. It was tested on GNU/Linux, Windows NT 4.0, and Windows 2000.
If you try to run Eshell on other versions of Emacs, here are some common
problems you may run into, and how to overcome them:
Emacs 20.3: Pcomplete fails to byte compile. Instead of typing "make install"
in the Pcomplete directory, type: "make pcmpl-auto.el install_el". Then, copy
the file "pcmpl-auto.el" into the directory that pcomplete was just installed
into. Why this is failing is a mystery to me. It appears that byte-compiling
pcomplete from a running Emacs session works just fine.
To start using Eshell, add the following to your .emacs file:
(load "eshell-auto")
This will define all of the necessary autoloads.
Now type `M-x eshell`. See the INSTALL file for full installation
instructions.
# Philosophy
A shell is a layer which metaphorically surrounds the kernel, or heart of an
operating system. This kernel can be seen as an engine of pure functionality,
waiting to serve, while the user programs take advantage of that functionality
to accomplish their purpose.
The shell's role is to make that functionality accessible to the user in an
unformed state. Very roughly, it associates kernel functionality with textual
commands, allowing the user to interact with the operating system via
linguistic constructs. Process invocation is perhaps the most significant form
this takes, using the kernel's fork and exec functions.
Other programs also interact with the functionality of the kernel, but these
user applications typically offer a specific range of functionality, and thus
are not classed as "shells" proper. (What they lose in quiddity, they gain in
rigidity).
Emacs is also a user application, but it does make the functionality of the
kernel accessible through an interpreted language -- namely, Lisp. For that
reason, there is little preventing Emacs from serving the same role as a
modern shell. It too can manipulate the kernel in an unpredetermined way to
cause system changes. All it's missing is the shell-ish linguistic model.
Enter Eshell. Eshell translates "shell-like" syntax into Lisp in order to
exercise the kernel in the same manner as typical system shells. There is a
fundamental difference here, however, although it may seem subtle at first...
Shells like csh and Bourne shell were written several decades ago, in
different times, under more restrictive circumstances. This confined
perspective shows itself in the paradigm used by nearly all command-line
shells since. They are linear in conception, byte stream-based, sequential,
and confined to movement within a single host machine.
Emacs, on the other hand, is more than just a limited translator that can
invoke subprocesses and redirect file handles. It also manages character
buffers, windowing frames, network connections, registers, bookmarks,
processes, etc. In other words, it's a very multi-dimensional environment,
within which eshell emulates a highly linear methodology.
Taking a moment, let's look at how this could affect the future of a shell
allowed to develop in such a wider field of play:
There is no reason why directory movement should be linear, and confined to a
single file-system. Emacs, through w3 and ange-ftp, has access to the entire
Web. Why not allow a user to cd to multiple directories simultaneously, for
example? It might make some tasks easier, such as diff'ing files separated by
very long pathnames. Data sources are available from anywhere Emacs can
derive information from: not just from files or the output of other processes.
Multiple shell invocations all share the same environment -- even the same
process list! It would be possible to have "process views", so that one buffer
is watching standard output, another standard error, and another the result of
standard output grep'd through a regular expression... It is not necessary to
"leave" the shell, losing all input and output history, environment variables,
directory stack, etc. Emacs could save the contents of your eshell
environment, and restore all of it (or at least as much as possible) each time
you restart. This could occur automatically, without requiring complex
initialization scripts. Typos occur all of the time; many of them are repeats
of common errors, such as dri for dir. Since executing non-existent programs
is rarely the intention of the user, eshell could prompt for the replacement
string, and then record that in a database of known misspellings. (Note: The
typo at the beginning of this paragraph wasn't discovered until two months
after I wrote the text; it was not intentional). Emacs' register and
bookmarking facilities can be used for remembering where you've been, and what
you've seen -- to varying levels of persistence. They could perhaps even be
tied to specific "moments" during eshell execution, which would include the
environment at that time, as well as other variables. Although this would
require functionality orthogonal to Emacs' own bookmarking facilities, the
interface used could be made to operate very similarly. This presents a brief
idea of what the fuller dimensionality of an Emacs shell could offer. It's not
just the language of a shell that determines how it's used, but also the
Weltanschauung underlying its design -- and which is felt behind even the
smallest feature. I would hope the freedom provided by using Emacs as a parent
environment will invite rich ideas from others. It certainly feels as though
all I've done so far is to tie down the horse, so to speak, so that he will
run at a man's pace.
The author of Eshell has been a long-time user of the following shells, all of
which contributed to Eshell's design:
- rc
- bash
- zsh
- sh
- 4nt
- csh