Why Sponsor Oils? | source | all docs for version 0.24.0 | all versions | oilshell.org
Oils Reference — Chapter Command Language
This chapter describes the command language for OSH, and some YSH extensions.
(in progress)
OSH:
print-files() {
for name in *.py; do
if test -x "$name"; then
echo "$name is executable"
fi
done
}
YSH:
proc print-files {
for name in *.py {
if test -x $name { # no quotes needed
echo "$name is executable"
}
}
}
Commands are composed of words. The first word may be the name of
proc or shell "function"hay defineExamples:
echo hi # a shell builtin doesn't start a process
ls /usr/bin ~/src # starts a new process
myproc "hello $name"
myshellfunc "hello $name"
myalias -l
Redirects are also allowed in any part of the command:
echo 'to stderr' >&2
echo >&2 'to stderr'
echo 'to file' > out.txt
echo > out.txt 'to file'
Bindings are allowed before a simple command:
PYTHONPATH=. mydir/myscript.py
These bindings set a variable and mark it exported. This binding is usually temporary, but when used with certain special builtins, it persists.
YSH prefix bindings look exactly like they do in shell:
PYTHONPATH=. mydir/myscript.py
However, they temporarily set ENV.PYTHONPATH, not $PYTHONPATH. This is
done by adding a new Dict to the prototype chain of the Obj.
The new ENV then becomes the environment of the child processes for the
command.
(In YSH, prefix bindings only mean one thing. They are temporary; they don't persist depending on whether the command is a special builtin.)
Run two commands in sequence like this:
echo one; echo two
or this:
echo one
echo two
Match a string against a series of glob patterns. Execute code in the section below the matching pattern.
path='foo.py'
case "$path" in
*.py)
echo 'python'
;;
*.sh)
echo 'shell'
;;
esac
For bash compatibility, the ;; terminator can be substituted with either:
;& - fall through to next arm, ignoring the condition;;& - fall through to next arm, respecting the conditionTest if a command exited with status zero (true). If so, execute the corresponding block of code.
Shell:
if test -d foo; then
echo 'foo is a directory'
elif test -f foo; then
echo 'foo is a file'
else
echo 'neither'
fi
YSH:
if test -d foo {
echo 'foo is a directory'
} elif test -f foo {
echo 'foo is a file'
} else {
echo 'neither'
}
Statically parsed boolean expressions, from bash and other shells:
x=42
if [[ $x -eq 42 ]]; then
echo yes
fi # => yes
Compare with the test builtin, which is dynamically parsed.
See bool-expr for the expression syntax.
Invert an exit code:
if ! test -d /tmp; then
echo "No temp directory
fi
mkdir -p /tmp && cp foo /tmp
ls || die "failed"
POSIX
POSIX
For loops iterate over words.
YSH style:
var mystr = 'one'
var myarray = :| two three |
for i in $mystr @myarray *.py {
echo $i
}
Shell style:
local mystr='one'
local myarray=(two three)
for i in "mystr" "${myarray[@]}" *.py; do
echo $i
done
Both fragments output 3 lines and then Python files on remaining lines.
A bash/ksh construct:
for (( i = 0; i < 5; ++i )); do
echo $i
done
These are keywords in Oils, not builtins!
Break out of a loop. (Not used for case statements!)
Continue to the next iteration of a loop.
Return from a function.
Exit the shell process with the given status:
exit 2
POSIX:
f() {
echo args "$@"
}
f 1 2 3
POSIX:
{ echo one; echo two; }
The trailing ; is necessary in OSH, but not YSH. In YSH, parse_brace makes
} is more of a special word.
( echo one; echo two )
In YSH, use forkwait instead of parentheses.
Pipelines are a traditional POSIX shell construct:
ls /tmp | grep ssh | sort
Related:
PIPESTATUS in OSH_pipeline_status in YSHStart a command as a background job. Don't wait for it to finish, and return control to the shell.
The PID of the job is recorded in the $! variable.
sleep 1 &
echo pid=$!
{ echo two; sleep 2 } &
wait
wait
In YSH, use the fork builtin.
The operators > and >> redirect the stdout of a process to a disk file.
The < operator redirects stdin from a disk file.
Examples of redirecting the stdout of a command:
echo foo > out.txt # overwrite out.txt
date >> stamp.txt # append to stamp.txt
Redirect to the stdin of a command:
cat < in.txt
Redirects are compatible with POSIX and bash, so they take descriptor numbers on the left:
make 2> stderr.txt # '2>' is valid, but '2 >' is not
Note that the word argument to file redirects is evaluated like bash, which is different than other arguments to other redirects:
tar -x -z < Python* # glob must expand to exactly 1 file
tar -x -z < $myvar # $myvar is split because it's unquoted
In other words, it's evaluated as a sequence of 1 word, which produces zero to N strings. But redirects are only valid when it produces exactly 1 string.
(Related: YSH uses shopt --set simple_word_eval, which means that globs that
match nothing evaluate to zero strings, not themselves.)
Redirect to a file descriptor:
echo 'to stderr' >&2
Here documents let you write the stdin of a process in the shell program.
Specify a delimiter word (like EOF) after the redir operator (like <<).
If it's unquoted, then $ expansion happens, like a double-quoted string:
cat <<EOF
here doc with $double ${quoted} substitution
EOF
If the delimiter is quoted, then $ expansion does not happen, like a
single-quoted string:
cat <<'EOF'
price is $3.99
EOF
Leading tabs can be stripped with the <<- operator:
myfunc() {
cat <<-EOF
here doc with one tab leading tab stripped
EOF
}
The <<< operator means that the argument is a stdin string, not a
chosen delimiter.
cat <<< 'here string'
The string plus a newline is the stdin value, which is consistent with
GNU bash.
You can also use YSH multi-line strings as "here strings". For example:
Double-quoted:
cat <<< """
double
quoted = $x
"""
Single-quoted:
cat <<< '''
price is
$3.99
'''
J8-style with escapes:
cat <<< u'''
j8 style string price is
mu = \u{3bc}
'''
In these cases, a trailing newline is not added. For example, the first example is equivalent to:
write --end '' -- """
double
quoted = $x
"""
time [-p] pipeline
Measures the time taken by a command / pipeline. It uses the getrusage()
function from libc.
Note that time is a KEYWORD, not a builtin!
Internal commands (procs and builtins) accept typed arguments in parentheses:
json write (myobj)
Redirects can also appear after the typed args:
json write (myobj) >out.txt
Expressions in brackets like this:
assert [42 === x]
Are syntactic sugar for:
assert (^[42 === x])
That is, it's single arg of type value.Expr.
Redirects can also appear after the lazy typed args:
assert [42 === x] >out.txt
Blocks can be passed to simple commands, either literally:
cd /tmp {
echo $PWD # prints /tmp
}
echo $PWD
Or as an expression:
var block = ^(echo $PWD)
cd /tmp (; ; block)
Note that cd has no typed or named arguments, so the two semicolons are
preceded by nothing.
When passed to procs, blocks capture the enclosing stack frame:
var x = 42
myproc {
# lexical scope is respected
echo "x = $x" # x = 42
}
Redirects can appear after the block arg:
cd /tmp {
echo $PWD # prints /tmp
} >out.txt
Like the shell case statement, the Ysh case statement has string/glob patterns.
var s = 'README.md'
case (s) {
*.py { echo 'Python' }
*.cc | *.h { echo 'C++' }
* { echo 'Other' }
}
# => Other
We also generated it to typed data within ():
var x = 43
case (x) {
(30 + 12) { echo 'the integer 42' }
(else) { echo 'neither' }
}
# => neither
The else is a special keyword that matches any value.
case (s) {
/ dot* '.md' / { echo 'Markdown' }
(else) { echo 'neither' }
}
# => Markdown
Like shell, you can use a command:
if test --file $x {
echo "$x is a file"
}
You can also use an expression:
if (x > 0) {
echo 'positive'
}
This is a shell-style loop over "words":
for name in README.md *.py {
echo $name
}
# => README.md
# => foo.py
You can also ask for the index:
for i, name in README.md *.py {
echo "$i $name"
}
# => 0 README.md
# => 1 foo.py
stdinHere's how to iterate over the lines of stdin:
for line in (io.stdin) {
echo $line
}
Likewise, you can ask for the index with for i, line in (io.stdin) { ....
You can use an expression as the condition:
var x = 5
while (x < 0) {
setvar x -= 1
}
You or a command:
while test -f myfile {
echo 'myfile'
sleep 1
}
Expressions are enclosed in ().
Iterating over a List or Range is like iterating over words or lines:
var mylist = [42, 43]
for item in (mylist) {
echo $item
}
# => 42
# => 43
var n = 5
for i in (3 .. n) {
echo $i
}
# => 3
# => 4
However, there are three ways of iterating over a Dict:
for key in (mydict) {
echo $key
}
for key, value in (mydict) {
echo "$key $value"
}
for i, key, value in (mydict) {
echo "$i $key $value"
}
That is, if you ask for two things, you'll get the key and value. If you ask for three, you'll also get the index.