C shell

The Unix system had been written almost exclusively in C, so the C shell's first objective was a command language that was more stylistically consistent with the rest of the system. The keywords, the use of parentheses, and the C shell's built-in expression grammar and support for arrays were all strongly influenced by C.

By today's standards, C shell may not seem particularly more C-like than many other popular scripting languages. But through the 80s and 90s, the difference was seen as striking, particularly when compared to Bourne shell (also known as sh), the then-dominant shell written by Stephen Bourne at Bell Labs. This example illustrates the C shell's more conventional expression operators and syntax.

#!/bin/sh
if [ $days -gt 365 ]
then
   echo This is over a year.
fi

#!/bin/csh
if ( $days > 365 ) then
   echo This is over a year.
endif

The Bourne sh lacked an expression grammar. The square bracketed condition had to be evaluated by the slower means of running the external test program. sh's if command took its argument words as a new command to be run as a child process. If the child exited with a zero return code, sh would look for a then clause (a separate statement, but often written joined on the same line with a semicolon) and run that nested block. Otherwise, it would run the else. Hard-linking the test program as both "test" and "[" gave the notational advantage of the square brackets and the appearance that the functionality of test was part of the sh language. sh's use of a reversed keyword to mark the end of a control block was a style borrowed from ALGOL 68.[12]

By contrast, csh could evaluate the expression directly, which made it faster. It also claimed better readability: Its expressions used a grammar and a set of operators mostly copied from C, none of its keywords were reversed and the overall style was also more like C.

Here is a second example, comparing scripts that calculate the first 10 powers of 2.

#!/bin/sh
i=2
j=1
while [ $j -le 10 ]
do
   echo '2 **' $j = $i
   i=`expr $i '*' 2`
   j=`expr $j + 1`
done

#!/bin/csh
set i = 2
set j = 1
while ( $j <= 10 )
   echo '2 **' $j = $i
   @ i *= 2
   @ j++
end

Again because of the lack of an expression grammar, the sh script uses command substitution and the expr command. (Modern POSIX shell does have such a grammar: the statement could be written i=$((i * 2)) or : "$((i *= 2))".)

Finally, here is a third example, showing the differing styles for a switch statement.

#!/bin/sh
for i in d*
do
   case $i in
      d?) echo $i is short ;;
      *) echo $i is long ;;
   esac
done

#!/bin/csh
foreach i ( d* )
   switch ( $i )
      case d?:
         echo $i is short
         breaksw
      default:
         echo $i is long
   endsw
end

In the sh script, ";;" marks the end of each case because sh disallows null statements otherwise.

The second objective was that the C shell should be better for interactive use. It introduced numerous new features that made it easier, faster and more friendly to use by typing commands at a terminal. Users could get things done with a lot fewer keystrokes and it ran faster. The most significant of these new features were the history and editing mechanisms, aliases, directory stacks, tilde notation, cdpath, job control, and path hashing. These new features proved very popular, and many of them have since been copied by other Unix shells.

History allows users to recall previous commands and rerun them by typing only a few quick keystrokes. For example, typing two exclamation marks ("!!")[13] as a command causes the immediately preceding command to be run. Other short keystroke combinations, e.g., "!$" (meaning "the final argument of the previous command"), allow bits and pieces of previous commands to be pasted together and edited to form a new command.

Editing can be done not only on the text of a previous command, but also on variable substitutions. Operators range from simple string search/replace to parsing a pathname to extract a specific segment.

Aliases allow the user to type the name of an alias and have the C shell expand it internally into whatever set of words the user has defined. For many simple situations, aliases run faster and are more convenient than scripts.

The directory stack allows the user to push or pop the current working directory, making it easier to jump back and forth between different places in the filesystem.

Tilde notation offers a shorthand way of specifying pathnames relative to the home directory using the "~" character.

The escape key can be used interactively to show possible completions of a filename at the end of the current command line.

Cdpath extends the notion of a search path to the cd (change directory) command: If the specified directory is not in the current directory, csh will try to find it in the cdpath directories.

Well into the 1980s, most users only had simple character-mode terminals that precluded multiple windows, so they could only work on one task at a time. The C shell's job control allowed the user to suspend the current activity and create a new instance of the C shell, called a job, by typing ^Z. The user could then switch back and forth between jobs using the fg command. The active job was said to be in the foreground. Other jobs were said to be either suspended (stopped) or running in the background.

Path hashing speeds up the C shell's search for executable files. Rather than performing a filesystem call in each path directory, one at a time, until it either finds the file or runs out of possibilities, the C shell consults an internal hash table built by scanning the path directories. That table can usually tell the C shell where to find the file (if it exists) without having to search and can be refreshed with the rehash command.

A basic statement is one that simply runs a command. The first word is taken as name of the command to be run and may be either an internal command, e.g., echo, or an external command. The rest of the words are passed as arguments to the command.

At the basic statement level, here are some of the features of the grammar:

The C shell, like all Unix shells, treats any command-line argument that contains wildcard characters as a pattern and replaces it with the list of all the filenames that match (see globbing).

• * matches any number of characters.
• ? matches any single character.
• [...] matches any of the characters inside the square brackets. Ranges are allowed, using the hyphen.
• [^...] matches any character not in the set.

The C shell also introduced several notational conveniences (sometimes known as extended globbing), since copied by other Unix shells.

• abc{def,ghi} is alternation (aka brace expansion) and expands to abcdef abcghi.
• ~ means the current user's home directory.
• ~user means user's home directory.

Multiple directory-level wildcards, e.g., "*/*.c", are supported.

Since version 6.17.01, recursive wildcarding à la zsh (e.g. "**/*.c" or "***/*.html") is also supported with the globstar option.

Giving the shell the responsibility for interpreting wildcards was an important decision on Unix. It meant that wildcards would work with every command, and always in the same way. However, the decision relied on Unix's ability to pass long argument lists efficiently through the exec system call that csh uses to execute commands. By contrast, on Windows, wildcard interpretation is conventionally performed by each application. This is a legacy of MS-DOS, which only allowed a 128-byte command line to be passed to an application, making wildcarding by the DOS command prompt impractical. Although modern Windows can pass command lines of up to roughly 32K Unicode characters, the burden for wildcard interpretation remains with the application.

By default, when csh runs a command, the command inherits the csh's stdio file handles for stdin, stdout and stderr, which normally all point to the console window where the C shell is running. The i/o redirection operators allow the command to use a file instead for input or output.

• > file means stdout will be written to file, overwriting it if it exists, and creating it if it doesn't. Errors still come to the shell window.
• >& file means both stdout and stderr will be written to file, overwriting it if it exists, and creating it if it doesn't.
• >> file means stdout will be appended at the end of file.
• >>& file means both stdout and stderr will be appended at the end of file.
• < file means stdin will be read from file.
• << string is a here document. Stdin will read the following lines up to the one that matches string.

Redirecting stderr alone isn't possible without the aid of a sub-shell.

mkfifo ~/filter
setenv filter "~/filter"
cat $filter & ( ( ls /root/ || echo No access. ) > $filter ) >& /dev/null

Commands can be joined on the same line.

• ; means run the first command and then the next.
• && means run the first command and, if it succeeds with a 0 return code, run the next.
• || means run the first command and, if it fails with a non-zero return code, run the next.

Commands can be connected using a pipe, which causes the output of one command to be fed into the input of the next. Both commands run concurrently.

• | means connect stdout to stdin of the next command. Errors still come to the shell window.
• |& means connect both stdout and stderr to stdin of the next command.

Running concurrently means "in parallel". In a multi-core (multiple processor) system, the piped commands may literally be executing at the same time, otherwise the scheduler in the operating system time-slices between them.

Given a command, e.g., "a | b", the shell creates a pipe, then starts both a and b with stdio for the two commands redirected so that a writes its stdout into the input of the pipe while b reads stdin from the output of the pipe. Pipes are implemented by the operating system with a certain amount of buffering so that a can write for a while before the pipe fills but once the pipe fills any new write will block inside the OS until b reads enough to unblock new writes. If b tries to read more data than is available, it will block until a has written more data or until the pipe closes, e.g., if a exits.

If a word contains a dollar sign, "$", the following characters are taken as the name of a variable and the reference is replaced by the value of that variable. Various editing operators, typed as suffixes to the reference, allow pathname editing (e.g., ":e" to extract just the extension) and other operations.

Quoting mechanisms allow otherwise special characters, such as whitespace, wildcards, parentheses, and dollar signs, to be taken as literal text.

• \ means take the next character as an ordinary literal character.
• "string" is a weak quote. Enclosed whitespace and wildcards are taken as literals, but variable and command substitutions are still performed.
• 'string' is a strong quote. The entire enclosed string is taken as a literal.

Double quotes inside double quotes should be escaped with "\"". The same applies to the dollar symbol, to prevent variable expansion "\$". For backticks, to prevent command substitution nesting, single quotes are required "'\`'".

Command substitution allows the output of one command to be used as arguments to another.

• `command` means take the output of command, parse it into words and paste them back into the command line.

Nesting must be done within double quotes, with "\`".

Normally, when the C shell starts a command, it waits for the command to finish before giving the user another prompt signaling that a new command can be typed.

• command & means start command in the background and prompt immediately for a new command.

A subshell is a separate child copy of the shell that inherits the current state but can then make changes, e.g., to the current directory, without affecting the parent.

• ( commands ) means run commands in a subshell.

The C shell provides control structures for both condition-testing and iteration. The condition-testing control structures are the if and switch statements. The iteration control structures are the while, foreach and repeat statements.

There are two forms of the if statement. The short form is typed on a single line but can specify only a single command if the expression is true.

if ( expression ) command

The long form uses then, else and endif keywords to allow for blocks of commands to be nested inside the condition.

if ( expression1 ) then
    commands
else if ( expression2 ) then
    commands
    ...
else
    commands
endif

If the else and if keywords appear on the same line, csh chains, rather than nests them; the block is terminated with a single endif.

The switch statement compares a string against a list of patterns, which may contain wildcard characters. If nothing matches, the default action, if there is one, is taken.

switch ( string )
    case pattern1:
        commands
        breaksw
    case pattern2:
        commands
        breaksw
        ...
    default:
        commands
        breaksw
endsw

The while statement evaluates an expression. If it is true, the shell runs the nested commands and then repeats for as long as the expression remains true.

while ( expression )
    commands
end

The foreach statement takes a list of values, usually a list of filenames produced by wildcarding, and then for each, sets the loop variable to that value and runs the nested commands.

foreach loop-variable ( list-of-values )
    commands
end

The repeat statement repeats a single command an integral number of times.

repeat integer command

The C shell implements both shell and environment variables.[14] Environment variables, created using the setenv statement, are always simple strings, passed to any child processes, which retrieve these variables via the envp[] argument to main().

Shell variables, created using the set or @ statements, are internal to C shell. They are not passed to child processes. Shell variables can be either simple strings or arrays of strings. Some of the shell variables are predefined and used to control various internal C shell options, e.g., what should happen if a wildcard fails to match anything.

In current versions of csh, strings can be of arbitrary length, well into millions of characters.

The C shell implements a 32-bit integer expression grammar with operators borrowed from C but with a few additional operators for string comparisons and filesystem tests, e.g., testing for the existence of a file. Operators must be separated by whitespace from their operands. Variables are referenced as $name.

Operator precedence is also borrowed from C, but with different operator associativity rules to resolve the ambiguity of what comes first in a sequence of equal precedence operators. In C, the associativity is left-to-right for most operators; in C shell, it is right-to-left. For example,

// C groups from the left
int i = 10 / 5 * 2;
printf( "%d\n", i ); // prints 4
i = 7 - 4 + 2;
printf( "%d\n", i ); // prints 5
i = 2 >> 1 << 4;
printf( "%d\n", i ); // prints 16

# C shell groups from the right
@ i = 10 / 5 * 2
echo $i # prints 1
@ i = 7 - 4 + 2
echo $i # prints 1
@ i = ( 2 >> 1 << 4 )
echo $i # prints 0

The parentheses in the C shell example are to avoid having the bit-shifting operators confused as I/O redirection operators. In either language, parentheses can always be used to explicitly specify the desired order of evaluation, even if only for clarity.