Welcome to the quick tour of Elvish. This tour works best if you have used another shell or programming language before.
If you are familiar with traditional shells like Bash, the sections basic shell language and shell scripting commands can help you "translate" your knowledge into Elvish.
If you are mainly interested in using Elvish interactively, jump directly to interactive features.
Many basic language features of Elvish are very familiar to traditional shells. A notable exception is control structures, covered below in the advanced language features section.
If you are familiar with bash, the following table shows some (rough) correspondence between Elvish and bash syntax:
Feature | Elvish | bash equivalent |
---|---|---|
Barewords | echo foo |
|
Single-quoted strings | echo 'foo' |
|
echo 'It''s good' |
echo 'It'\''s good' |
|
Double-quoted strings | echo "foo" |
|
echo "foo\nbar" |
echo $'foo\nbar' |
|
echo "foo: "$foo |
echo "foo: $foo" |
|
Comments | # comment |
|
Line continuation | echo foo ^ |
echo foo \ |
Brace expansion | echo {foo bar}.txt |
echo {foo,bar}.txt |
Wildcards | echo *.? |
|
echo **.go |
find . -name '*.go' |
|
echo *.?[set:ch] |
echo *.[ch] |
|
Tilde expansion | echo ~/foo |
|
Setting variables | var foo = bar |
foo=bar |
set foo = bar |
foo=bar |
|
foo=bar cmd |
||
Using variables | echo $foo |
|
echo $E:HOME |
echo $HOME |
|
Redirections | head -n10 < a.txt > b.txt |
|
Byte pipelines | head -n4 a.txt | grep x |
|
Output capture | ls -l (which elvish) |
ls -l $(which elvish) |
Background jobs | echo foo & |
|
Command sequence | a; b |
a && b |
Like traditional shells, unquoted words that don't contain special characters are treated as strings (such words are called barewords):
~> echo foobar
foobar
~> ls /
bin dev home lib64 mnt proc run srv tmp var
boot etc lib lost+found opt root sbin sys usr
~> vim a.c
This is one of the most distinctive syntactical features of shells; non-shell programming languages typically treat unquoted words as names of functions and variables.
Read the language reference on barewords to learn more.
Like traditional shells, single-quoted strings expand nothing; every character represents itself (except the single quote itself):
~> echo 'hello\world$'
hello\world$
Like Plan 9 rc, or zsh with
the RC_QUOTES
option turned on, the single quote itself can be written by
doubling it
~> echo 'it''s good'
it's good
Read the language reference on single-quoted strings to learn more.
Like many non-shell programming languages and $''
in bash, double-quoted
strings support C-like escape sequences (e.g. \n
for newline):
~> echo "foo\nbar"
foo
bar
Unlike traditional shells, Elvish does not support interpolation inside double-quoted strings. Instead, you can just write multiple words together, and they will be concatenated:
~> var x = foo
~> put 'x is '$x
▶ 'x is foo'
Read the language reference on double-quoted strings to learn more.
Comments start with #
and extend to the end of the line:
~> echo foo # this is a comment
foo
Line continuation in Elvish uses ^
instead of \
:
~> echo foo ^
bar
foo bar
Unlike traditional shells, line continuation is treated as whitespace. In
Elvish, the following code outputs foo bar
:
echo foo^
bar
However, in bash, the following code outputs foobar
:
echo foo\
bar
Brace expansions in Elvish work like in traditional shells, but use spaces instead of commas:
~> echo {foo bar}.txt
foo.txt bar.txt
The opening brace {
must not be followed by a whitespace, to disambiguate
from lambdas.
Note: commas might still work as a separator in Elvish's brace expansions, but it will eventually be deprecated and removed soon.
Read the language reference on braced lists to learn more.
The basic wildcard characters, *
and ?
, work like in traditional shells:
~> ls
bar.ch d1 d2 d3 foo.c foo.h lorem.go lorem.txt
~> echo *.?
foo.c foo.h
Elvish also supports **
, which matches multiple path components:
~> find . -name '*.go'
./d1/a.go
./d2/b.go
./lorem.go
./d3/d4/c.go
~> echo **.go
d1/a.go d2/b.go d3/d4/c.go lorem.go
Character classes are a bit more verbose in Elvish. First, a character set is
written like [set:ch]
, instead of just [ch]
. Second, they don't appear on
their own, but as a suffix to ?
. For example, to match files ending in either
.c
or .h
, use:
~> echo *.?[set:ch]
foo.c foo.h
A character set suffix can also be applied to *
. For example, to match files
who extension only contains c
and h
:
~> echo *.*[set:ch]
bar.ch foo.c foo.h
Read the language reference on wildcard expansion to learn more.
Tilde expansion works likes in traditional shells. Assuming that the home
directory of the current user is /home/me
, and the home directory of elf
is
/home/elf
:
~> echo ~/foo
/home/me/foo
~> echo ~elf/foo
/home/elf/foo
Read the language reference on tilde expansion to learn more.
Variables are declared with the var
command, and set with the set
command:
~> var foo = bar
~> echo $foo
bar
~> set foo = quux
~> echo $foo
quux
The spaces around =
are mandatory.
Unlike traditional shells, variables must be declared before they can be set; setting an undeclared variable results in an error.
Like traditional shells, Elvish supports setting a variable temporarily for the
duration of a command, by prefixing the command with foo=bar
. For example:
~> var foo = original
~> fn f { echo $foo }
~> foo=temporary f
temporary
~> echo $foo
original
Read the language reference on the var
command,
the set
command and
temporary assignments to learn
more.
Like traditional shells, using the value of a variable requires the $
prefix.
~> var foo = bar
~> echo $foo
bar
Unlike traditional shells, variables must be declared before being used; if the
foo
variable wasn't declared with var
first, echo $foo
results in an
error.
Elvish does not perform $IFS
splitting on variables, so $foo
always
evaluates to one value, even if it contains whitespaces and newlines. You never
need to write echo "$foo"
again. (And in fact,
double-quoted strings do not
support interpolation).
Also unlike traditional shells, environment variables in Elvish live in a
separate E:
namespace:
~> echo $E:HOME
/home/elf
Read the language reference on variables, variable use and special namespaces to learn more.
Redirections in Elvish work like in traditional shells. For example, to save the
first 10 lines of a.txt
to a1.txt
:
~> head -n10 < a.txt > a1.txt
Read the language reference on redirections to learn more.
UNIX pipelines in Elvish (called byte pipelines, to distinguish from
value pipelines) work like in traditional shells. For
example, to find occurrences of x
in the first 4 lines of a.txt
:
~> cat a.txt
foo
barx
lorem
quux
lux
nox
~> head -n4 a.txt | grep x
barx
quux
Read the language reference on pipelines to learn more.
Output of commands can be captured and used as values with ()
. For example,
the following command shows details of the elvish
binary:
~> ls -l (which elvish)
-rwxr-xr-x 1 xiaq users 7813495 Mar 2 21:32 /home/xiaq/go/bin/elvish
Note: the same feature is usually known as command substitution in traditonal shells.
Unlike traditional shells, Elvish only splits the output on newlines, not any other whitespace characters.
Read the language reference on output capture to learn more.
Add &
to the end of a pipeline to make it run in the background, similar to
traditional shells:
~> echo foo &
foo
job echo foo & finished
Unlike traditional shells, the &
character does not serve to separate
commands. In bash you can write echo foo & echo bar
; in Elvish you still need
to terminate the first command with ;
or newline: echo foo &; echo bar
.
Read the language reference on background pipelines to learn more.
Join commands with a ;
or newline to run them sequentially (insert a newline
with Alt-Enter):
~> echo a; echo b
a
b
~> echo a
echo b
a
b
In Elvish, when a command fails (e.g. when an external command exits with a non-zero status), execution gets terminated.
~> echo before; false; echo after
before
Exception: false exited with 1
[tty 2], line 1: echo before; false; echo after
In this aspect, Elvish's behavior is similar to joining all commands with &&
or setting set -e
in traditional shells:
Building on a core of familiar shell-like syntax, the Elvish language incorporates many advanced features that make it a modern dynamic programming language.
Like in traditional shells, commands in Elvish can output bytes. The echo
command outputs bytes:
~> echo foo bar
foo bar
Additionally, commands can also output values. Values include not just
strings, but also lambdas, numbers,
lists and maps. The put
command outputs values:
~> put foo [foo] [&foo=bar] { put foo }
▶ foo
▶ [foo]
▶ [&foo=bar]
▶ <closure 0xc000347500>
Many builtin commands output values. For example, string functions in the str:
module outputs their results as values. This makes those functions work
seamlessly with strings that contain newlines or even NUL bytes:
~> use str
~> str:join ',' ["foo\nbar" "lorem\x00ipsum"]
▶ "foo\nbar,lorem\x00ipsum"
Unlike most programming languages, Elvish commands don't have return values. Instead, they use the value output to "return" their results.
Read the reference for builtin commands to learn which commands work with value inputs and outputs. Among them, here are some general-purpose primitives:
Command | Functionality |
---|---|
all
|
Passes value inputs to value outputs |
each
|
Applies a function to all values from value input |
put
|
Writes arguments as value outputs |
slurp
|
Convert byte input to a single string in value output |
Pipelines work with value outputs too. When forming pipelines, a command that
writes value outputs can be followed by a command that takes value inputs. For
example, the each
command takes value inputs, and applies a lambda to each one
of them:
~> put foo bar | each {|x| echo 'I got '$x }
I got foo
I got bar
Read the language reference on pipelines to learn more about pipelines in general.
Output capture works with value output too. Capturing value outputs always
recovers the exact values there were written. For example, the str:join
command joins a list of strings with a separator, and its output can be captured
and saved in a variable:
~> use str
~> var s = (str:join ',' ["foo\nbar" "lorem\x00ipsum"])
~> put $s
▶ "foo\nbar,lorem\x00ipsum"
Read the language reference on output capture to learn more.
Lists look like [a b c]
, and maps look like [&key1=value1 &key2=value2]
:
~> var li = [foo bar lorem ipsum]
~> put $li
▶ [foo bar lorem ipsum]
~> var map = [&k1=v2 &k2=v2]
~> put $map
▶ [&k1=v2 &k2=v2]
You can get elements of lists and maps by indexing them. Lists are zero-based and support slicing too:
~> put $li[0]
▶ foo
~> put $li[1..3]
▶ [bar lorem]
~> put $map[k1]
▶ v2
Read the language reference on lists and maps to learn more.
Elvish has a number type. There is no dedicated syntax for it; instead, it can
constructed using the num
builtin:
~> num 1
▶ (num 1)
~> num 1e2
▶ (num 100)
Most arithmetic commands in Elvish support both typed numbers and strings that can be converted to numbers. They usually output typed numbers:
~> + 1 2
▶ (num 3)
~> use math
~> math:pow (num 10) 3
▶ (num 1000)
Note: The set of number types will likely expand in future.
Read the language reference on numbers and the reference for the math module to learn more.
Elvish has two boolean values, $true
and $false
.
Read the language reference on booleans to learn more.
Many Elvish commands take options, which look like map pairs (&key=value
).
For example, the echo
command takes a sep
option that can be used to
override the default separator of space:
~> echo &sep=',' foo bar
foo,bar
~> echo &sep="\n" foo bar
foo
bar
Lambdas are first-class values in Elvish. They can be saved in variables, used as commands, passed to commands, and so on.
Lambdas can be written by enclosing its body with {
and }
:
~> var f = { echo "I'm a lambda" }
~> $f
I'm a lambda
~> put $f
▶ <closure 0xc000265bc0>
~> var g = (put $f)
~> $g
I'm a lambda
The opening brace {
must be followed by some whitespace, to disambiguate
from brace expansion.
Lambdas can take arguments and options, which can be written in a signature:
~> var f = {|a b &opt=default|
echo "a = "$a
echo "b = "$b
echo "opt = "$opt
}
~> $f foo bar
a = foo
b = bar
opt = default
~> $f foo bar &opt=option
a = foo
b = bar
opt = option
Read the language reference on functions to learn more about functions.
Control structures in Elvish look very different from traditional shells. For
example, this is how an if
command looks:
~> if (eq (uname) Linux) { echo "You're on Linux" }
You're on Linux
The if
command takes a conditional expression (an output capture in this
case), and the body to execute as a lambda. Since lambdas allow internal
newlines, you can also write it like this:
~> if (eq (uname) Linux) {
echo "You're on Linux"
}
You're on Linux
However, you must write the opening brace {
on the same line as if
. If you
write it on a separate line, Elvish would parse it as two separate commands.
The for
command looks like this:
~> for x [expressive versatile] {
echo "Elvish is "$x
}
Elvish is expressive
Elvish is versatile
Read the language reference on the if
command,
the for
command, and additionally
the while
command to learn more.
Elvish uses exceptions to signal errors. For example, calling a function with the wrong number of arguments throws an exception:
~> var f = { echo foo } # doesn't take arguments
~> $f a b
Exception: arity mismatch: arguments here must be 0 values, but is 2 values
[tty 2], line 1: $f a b
Moreover, non-zero exits from external commands are also turned into exceptions:
~> false
Exception: false exited with 1
[tty 3], line 1: false
Exceptions can be caught using the try
command:
~> try {
false
} except e {
echo 'got an exception'
}
got an exception
Read the language reference on
the exception value type and
the try
command to learn more.
The names of variables and functions can have namespaces prepended to their
names. Namespaces always end with :
.
The using variables section has already shown the E:
namespace. Other namespaces can be added by importing modules with use
. For
example, the str:
module provides string utilities:
~> use str
~> str:to-upper foo
▶ FOO
You can define your own modules by putting .elv
files in
~/.config/elvish/lib
(or ~\AppData\Roaming\elvish\lib
). For example, to
define a module called foo
, put the following in foo.elv
under the
aforementioned directory:
fn f {
echo 'in a function in foo'
}
This module can now be used like this:
~> use foo
~> foo:f
in a function in foo
Read the language reference on namespaces and modules to learn more.
As shown in examples above, Elvish supports calling external commands directly by writing their name. If an external command exits with a non-zero code, it throws an exception.
Unfortunately, many of the advanced language features are only available for internal commands and functions. For example:
They can only write byte output, not value output.
They only take string arguments; non-string arguments are implicitly coerced to strings.
They don't take options.
Read the language reference on ordinary commands to learn more about when Elvish decides that a command is an external command.
Read the API of the interactive editor to learn more about UI customization options.
Press Tab to start completion. For example, after typing vim
and
Space, press Tab to complete filenames:
@ttyshot learn/tour/completion
Basic operations should be quite intuitive:
To navigate the candidate list, use arrow keys ▲ ▼ ◀ ▶ or Tab and Shift-Tab.
To accept the selected candidate, press Enter.
To cancel, press Escape.
As indicated by the horizontal scrollbar, you can scroll to the right to find additional results that don't fit in the terminal.
You may have noticed that the cursor has moved to the right of "COMPLETING
argument". This indicates that you can continue typing to filter candidates. For
example, after typing .md
, the UI looks like this:
@ttyshot learn/tour/completion-filter
Read the reference on completion API to learn how to program and customize tab completion.
Elvish has several UI features for working with command history.
Press ▲ to fetch the last command. This is called history walking mode:
@ttyshot learn/tour/history-walk
Press ▲ to go further back, ▼ to go forward, or Escape to cancel.
To restrict to commands that start with a prefix, simply type the prefix before
pressing ▲. For example, to walk through commands starting with
echo
, type echo
before pressing ▲:
@ttyshot learn/tour/history-walk-prefix
Press Ctrl-R to list the full command history:
@ttyshot learn/tour/history-list
Like in completion mode, type to filter the list, press ▲ and ▼ to navigate the list, Enter to insert the selected entry, or Escape to cancel.
Finally, Elvish has a last command mode dedicated to inserting parts of the last command. Press Alt-, to trigger it:
@ttyshot learn/tour/lastcmd
Elvish remembers which directories you have visited. Press Ctrl-L to list visited directories. Use ▲ and ▼ to navigate the list, Enter to change to that directory, or Escape to cancel.
@ttyshot learn/tour/location
Type to filter:
@ttyshot learn/tour/location-filter
Press Ctrl-N to start the builtin filesystem navigator.
@ttyshot learn/tour/navigation
Unlike other modes, the cursor stays in the main buffer in navigation mode. This allows you to continue typing commands; while doing that, you can press Enter to insert the selected filename. You can also press Alt-Enter to insert the filename without exiting navigation mode; this is useful when you want to insert multiple filenames.
Elvish's interactive startup script is rc.elv
.
Non-interactive Elvish sessions do not have a startup script.
Elvish doesn't support POSIX aliases, but you can get a similar experience simply by defining functions:
fn ls {|@a| e:ls --color $@a }
The e:
prefix (for "external") ensures that the external command named ls
will be called. Otherwise this definition will result in infinite recursion.
The left and right prompts can be customized by assigning functions to
edit:prompt
and
edit:rprompt
. The following example defines
prompts similar to the default, but uses fancy Unicode.
@ttyshot learn/tour/unicode-prompts
The tilde-abbr
command abbreviates home
directory to a tilde. The constantly
command
returns a function that always writes the same value(s) to the value output. The
styled
command writes styled output.
Another common task in the interactive startup script is to set the search path.
You can do set the environment variable directly (all environment variables have
a E:
prefix):
set E:PATH = /opts/bin:/bin:/usr/bin
But it is usually nicer to set the $paths
instead:
set paths = [/opts/bin /bin /usr/bin]
Elvish has its own set of builtin commands. This section helps you find commands that correspond to commands in traditional shells.
To force Elvish to treat a command as an external command, prefix it with
e:
.
In Elvish, environment variables live in the
E:
namespace. There is no concept
of exporting a variable to the environment; environment variables are always
"exported" to child processes, and non-environment variables never are.
To build reusable libraries, use Elvish's module mechanism.
To execute a dynamic piece of code for side effect, use
eval
. If the code lives in a file, write
eval (slurp < /path/to/file)
.
Due to Elvish's scoping rules, files executed using either of the mechanism
above can't create new variables in the current namespace. For example,
eval 'var foo = bar'; echo $foo
won't work. However, the REPL's namespace
can be manipulated with edit:add-var
.
To test files, use commands in the path module.
To compare numbers, use number comparison commands.
To compare strings, use string comparison commands.
To perform boolean operations, use
and
,
or
or
not
. Note: and
and or
are part of the
language rather than the builtin module, since they perform
short-circuit evaluation
and don't always evaluate all the arguments.
To check if an external command exists, use has-external.
To query the path of an external command, use search-external.