Shell(5) - SerenityOS man pages

Name

The Shell Command Language

Introduction

The shell operates according to the following general steps:

Any text below is superseded by the formal grammar defined in the formal grammar section.

General Token Recognition

This section describes the general tokens the language accepts, it should be noted that due to nature of the language, some tokens are valid only in a specific context.

Bareword

String of characters that are not Special or Syntactic Elements

Glob

String of characters containing at least one of *? in bareword position

History Events

A designator starting with ! in bareword position that describes a word or a range of words from a previously entered command. Please look at the section named 'History Event Designators' for a more thorough explanation. Only allowed in interactive mode.

Single Quoted String

Any sequence of characters between two single quotes (')

Double Quoted String

Any sequence of Double Quoted String Part tokens:

Heredocs

Heredocs are made in two parts, the initiator and the contents, the initiator may be used in place of a string (i.e. wherever a string is allowed to be used), with the constraint that the contents must follow the sequence that the initiator is used in.

There are four different initiators:

Note that heredocs must be listed in the same order as they are used after a sequence that has been terminated with a newline.

Variable Reference

Any sequence of Identifier characters, or a Special Variable following a $. Variables may be followed by a Slice (see Slice)

Slice

Variables may be sliced into, which will allow the user to select a subset of entries in the contents of the variable. An expression of the form $identifier[slice-contents] can be used to slice into a variable, where slice-contents has semantics similar to Brace Expansions, but it may only evaluate to numeric values, that are used to index into the variable being sliced. Negative indices are allowed, and will index the contents from the end. It should be noted that the shell will always perform bounds-checking on the indices, and raise an error on out-of-bound accesses. Slices can slice into both lists and strings.

For example, $lst[1..-2] can be used to select a permutation of a 4-element list referred to by the variable lst, as the slice will evaluate to the list (1 0 -1 -2), which will select the indices 1, 0, 3, 2 (in that order).

Immediate Expressions

An expression of the form '${identifier expression...}', such expressions are expanded to other kinds of nodes before resolution, and are internal functions provided by the shell. Currently, the following functions are exposed:

Finds the length of the given expression. if either string or list is given, the shell will attempt to treat expression as that type, otherwise the type of expression will be inferred.

Finds the lengths of the entries in expression, this requires expression to be a list. If either string or list is given, the shell attempts to treat the elements of expression as that type, otherwise the types are individually inferred.

Evaluate expression

Any expression following a $ that is not a variable reference:

Lists

Any two expressions joined by the Join operator ( [whitespace]), or a variable reference referring to a list value

Comments

Any text following a # in bareword position, up to but not including a newline

Keywords

The following tokens:

Special characters

Any of the following:

Tilde

Any initial path segment starting with the character ~ in bareword position, Optionally followed by a bareword for the username

Redirections

The shell can create various redirections to file descriptors of a command before executing it, the general syntax for redirections is an optional file descriptor, followed by a redirection operator, followed by a destination.

There are four redirection operators corresponding to various file descriptor open modes: Read, Write, WriteAppend and ReadWrite, respectively <, >, >> and <>.

A special syntactic element &fd can reference a file descriptor as a destination.

Redirections take two main forms, Read/Write redirections, and fd closure redirections.

Read/Write
Close

Examples

# Redirect the standard error to a file, and close the standard input
$ 2> foo 1>&-

# Redirect a file as read-write into the standard input
$ 1<>foo

# Redirect the standard output to /dev/null
$ >/dev/null

Expansions

The shell performs various expansions, in different stages.

Brace Expansions

Brace expansions are of two kinds, normal brace expansions and range brace expansions. Normal brace expansions are sequences of optional expressions inside braces ({}), delimited by a comma (','); a missing expression is treated as an empty string literal. Such expressions are simply expanded to the expressions they enclose. Range brace expansions are of the form {start_expression..end_expression}, where start_expression and end_expression denote the bounds of an inclusive range, and can be one of two types:

Juxtapositions

Any two expressions joined without any operator are considered to be in a Juxtaposition, with the resulting value being the list product of two expressions. For instance, (1 2)(3 4) shall be evaluated to (13 14 23 24) by calculating the list product of the two expressions (1 2) and (3 4).

Tildes

Any bareword starting with a tilde (~) and spanning up to the first path separator (/) - or EOL - is considered to be a tilde expansion with the text between the tilde and the separator being the username, which shall be expanded to a single string containing the home directory of the given username (or the current user if no username is provided).

Evaluate

Evaluate expressions take the general form of a dollar sign ($) followed by some expression, which is evaluated by the rules below.

Commands

A Command is a single simple command, containing arguments and redirections for a single program, or a compound command containing a shell control structure. The shell can evaluate a sequence of commands, a conditional relation between commands, or various semantic elements composed of commands and intrinsics.

Commands can be either calls to Shell builtins, or external programs.

Shell Semantic Elements

The commands can be composed into semantic elements, producing composite commands:

Sequences

A sequence of commands, executed serially independent of each other: Command ; Command ; Command ...

It should be noted that a newline (\\n) can be substituted for the semicolon (;).

Example

# Do one thing, then do another
echo foo; echo bar

Logical Relations

A sequence of commands whose execution depends somehow on the result of another

Command && Command && Command ... (AND)

Short-circuiting command evaluations, will cancel the entire chain should any command fails (have a non-zero exit code)

Command || Command || Command ... (OR)

Short-circuiting command evaluation, will continue down the chain if any command fails.

It should be noted that And chains bind more tightly than Or chains, so an expression of the form C1 && C2 || C3 is understood as "evaluate C1, if successful, evaluate C2, if not successful, evaluate C3".

Examples
# Create file if not found
test -f foo.txt || touch foo.txt

# Announce execution status of a command
rm test && echo "deleted!" || echo "failed with $?"

Control Structures

Conditionals

Conditionals can either be expressed with the Logical Relations, or via explicit if expressions. An if expression contains at least a condition command and a then clause, and optionally the else keyword followed by an else clause. An else clause may contain another if expression instead of a normal block.

The then clause must be surrounded by braces, but the else clause may also be another if expression.

An if expression evaluates either the then clause or (if available) the else clause, based on the exit code of the condition command; should the exit code be zero, the then clause will be executed, and if not, the else clause will.

Examples
# Remove a file if it exists, create it otherwise
if test -e the_file {
    rm the_file
} else {
    touch the_file
}

# Cond chain (if-elseif-else)
if A {
    echo A
} else if B {
    echo B
} else {
    echo C
}
For Loops

For Loops evaluate a sequence of commands once per element in a given list. The shell has two forms of for loops, one with an explicitly named iteration variable, and one with an implicitly named one. The general syntax follows the form for index index_name name in expr { sequence }, and allows omitting the index index_name name in part to implicitly name the variable it.

It should be noted that the index index_name section is optional, but if supplied, will require an explicit iteration variable as well. In other words, for index i in foo is not valid syntax.

A for-loop evaluates the sequence once per every element in the expr, setting the local variable name to the element being processed, and the local variable enum name to the enumeration index (if set).

The Shell shall cancel the for loop if two consecutive commands are interrupted via SIGINT (^C), and any other terminating signal aborts the loop entirely.

Examples
# Iterate over every non-hidden file in the current directory, and prepend '1-' to its name.
$ for * { mv $it 1-$it }

# Iterate over a sequence and write each element to a file
$ for i in $(seq 1 100) { echo $i >> foo }

# Iterate over some files and get their index
$ for index i x in * { echo file at index $i is named $x }
Infinite Loops

Infinite loops (as denoted by the keyword loop) can be used to repeat a block until the block runs break, or the loop terminates by external sources (interrupts, program exit, and terminating signals).

The behavior regarding SIGINT and other signals is the same as for loops (mentioned above).

Examples
# Keep deleting a file
loop {
    rm -f foo
}
Subshells

Subshells evaluate a given block in a new instance (fork) of the current shell process. to create a subshell, any valid shell code can be enclosed in braces.

Examples
# Run a block of code in the background, in a subshell, then detach it from the current shell
$ { for * { te $it } }&
$ disown
Functions

A function is a user-defined entity that can be used as a simple command to execute a compound command, optionally with some parameters. Such a function is defined via the syntax below:

function_name(explicitly_named_arguments...) { compound_command }

The function is named function_name, and has some explicitly named arguments explicitly_named_arguments..., which must be supplied by the caller, failure to do so will cause the command to exit with status 1.

The compound command shall be executed whenever the simple command function_name is executed. This execution shall be performed in a new local frame.

Additionally, should the simple command containing the function name be in a pipeline, or requested to be run in the background, this execution shall be moved to a subshell; naturally, in such a case any changes to the shell state (such as variables, aliases, etc) shall not be leaked to the parent shell process.

The passed arguments shall be stored in the special variables * and ARGV, and the explicitly named arguments shall be set, in order, from the first passed argument onwards.

The exit status of a function simple command shall be the exit status of the last command executed within the command, or 0 if the function has no commands. The declaration is not a command, and will not alter the exit status.

Examples
fn(a b c) {
    echo $a $b $c \( $* \)
}

$ fn 1 2 3 4
# 1 2 3 ( 1 2 3 4 )
Match Expressions

The pattern matching construct match shall choose from a sequence of patterns, and execute the corresponding action in a new frame. The choice is done by matching the result of the matched expression (after expansion) against the patterns (expanded down to either globs or literals). Multiple patterns can be attributed to a single given action by delimiting them with a pipe ('|') symbol. A pattern (or the series of) may be annotated with an extra as (...) clause, which allows globbed parts of the matching pattern to be named and used in the matching block.

The expanded matched expression can optionally be given a name using the as name clause after the matched expression, with which it may be accessible in the action clauses.

Examples
# Match the result of running 'make_some_value' (which is a list when captured by $(...))
match "$(make_some_value)" as value {
    (hello*) { echo "Hi!" }
    (say\ *) { echo "No, I will not $value" }
}

# Match the result of running 'make_some_value', cast to a string
# Note the `as (expr)` in the second pattern, which assigns whatever the `*` matches
# to the name `expr` inside the block.
match "$(make_some_value)" {
    hello* { echo "Hi!" }
    say\ * as (expr) { echo "No, I will not say $expr!" }
}

History Event Designators

History expansion may be utilized to reuse previously typed words or commands. Such expressions are of the general form !<event_designator>(:<word_designator>), where event_designator would select an entry in the shell history, and word_designator would select a word (or a range of words) from that entry.

Event designator effect
! Select the immediately preceding command
n Select the n'th entry in the history
-n Select the last n'th entry in the history
str Select the most recent entry starting with str
?str Select the most recent entry containing str
Word designator effect
n The n'th word, starting with 0 as the command
^ The first word (index 0)
$ The last word
x-y The range of words starting at x and ending at y (inclusive)

Formal Grammar

Shell Grammar

toplevel :: sequence?

sequence :: variable_decls? or_logical_sequence terminator sequence
          | variable_decls? or_logical_sequence '&' sequence
          | variable_decls? or_logical_sequence
          | variable_decls? function_decl (terminator sequence)?
          | variable_decls? terminator sequence

function_decl :: identifier '(' (ws* identifier)* ')' ws* '{' [!c] toplevel '}'

or_logical_sequence :: and_logical_sequence '|' '|' and_logical_sequence
                     | and_logical_sequence

and_logical_sequence :: pipe_sequence '&' '&' and_logical_sequence
                      | pipe_sequence

terminator :: ';'
            | '\n' [?!heredoc_stack.is_empty] heredoc_entries

heredoc_entries :: { .*? (heredoc_entry) '\n' } [each heredoc_entries]

variable_decls :: identifier '=' expression (' '+ variable_decls)? ' '*
                | identifier '=' '(' pipe_sequence ')' (' '+ variable_decls)? ' '*

pipe_sequence :: command '|' pipe_sequence
               | command
               | control_structure '|' pipe_sequence
               | control_structure

control_structure[c] :: for_expr
                      | loop_expr
                      | if_expr
                      | subshell
                      | match_expr
                      | ?c: continuation_control

continuation_control :: 'break'
                      | 'continue'

for_expr :: 'for' ws+ (('index' ' '+ identifier ' '+)? identifier ' '+ 'in' ws*)? expression ws+ '{' [c] toplevel '}'

loop_expr :: 'loop' ws* '{' [c] toplevel '}'

if_expr :: 'if' ws+ or_logical_sequence ws+ '{' toplevel '}' else_clause?

else_clause :: else '{' toplevel '}'
             | else if_expr

subshell :: '{' toplevel '}'

match_expr :: 'match' ws+ expression ws* ('as' ws+ identifier)? '{' match_entry* '}'

match_entry :: match_pattern ws* (as identifier_list)? '{' toplevel '}'

identifier_list :: '(' (identifier ws*)* ')'

match_pattern :: expression (ws* '|' ws* expression)*

command :: redirection command
         | list_expression command?

redirection :: number? '>'{1,2} ' '* string_composite
             | number? '<' ' '* string_composite
             | number? '>' '&' number
             | number? '>' '&' '-'

list_expression :: ' '* expression (' '+ list_expression)?

expression :: evaluate expression?
            | string_composite expression?
            | comment expression?
            | immediate_expression expression?
            | history_designator expression?
            | '(' list_expression ')' expression?

evaluate :: '$' '(' pipe_sequence ')'
          | '$' expression          {eval / dynamic resolve}

string_composite :: string string_composite?
                  | variable string_composite?
                  | bareword string_composite?
                  | glob string_composite?
                  | brace_expansion string_composite?
                  | heredoc_initiator string_composite?    {append to heredoc_entries}

heredoc_initiator :: '<' '<' '-' bareword         {*bareword, interpolate, no deindent}
                   | '<' '<' '-' "'" [^']* "'"    {*string, no interpolate, no deindent}
                   | '<' '<' '~' bareword         {*bareword, interpolate, deindent}
                   | '<' '<' '~' "'" [^']* "'"    {*bareword, no interpolate, deindent}

string :: '"' dquoted_string_inner '"'
        | "'" [^']* "'"

dquoted_string_inner :: '\' . dquoted_string_inner?       {concat}
                      | variable dquoted_string_inner?    {compose}
                      | . dquoted_string_inner?
                      | '\' 'x' xdigit*2 dquoted_string_inner?
                      | '\' 'u' xdigit*8 dquoted_string_inner?
                      | '\' [abefrnt] dquoted_string_inner?

variable :: variable_ref slice?

variable_ref :: '$' identifier
          | '$' '$'
          | '$' '?'
          | '$' '*'
          | '$' '#'
          | ...

slice :: '[' brace_expansion_spec ']'

comment :: '#' [^\n]*

immediate_expression :: '$' '{' immediate_function expression* '}'

immediate_function :: identifier       { predetermined list of names, see Shell.h:ENUMERATE_SHELL_IMMEDIATE_FUNCTIONS }

history_designator :: '!' event_selector (':' word_selector_composite)?

event_selector :: '!'                  {== '-0'}
                | '?' bareword '?'
                | bareword             {number: index, otherwise: lookup}

word_selector_composite :: word_selector ('-' word_selector)?

word_selector :: number
               | '^'                   {== 0}
               | '$'                   {== end}

bareword :: [^"'*$&#|()[\]{} ?;<>] bareword?
          | '\' [^"'*$&#|()[\]{} ?;<>] bareword?

bareword_with_tilde_expansion :: '~' bareword?

glob :: [*?] bareword?
      | bareword [*?]

brace_expansion :: '{' brace_expansion_spec '}'

brace_expansion_spec :: expression? (',' expression?)*
                      | expression '..' expression

digit :: <native hex digit>
number :: <number in base 10>
identifier :: <string of word characters>