Tools Unix CS114 Fall 2002
Lecture 8 Friday, October 18, 2002
Shells
When you login, Unix presents you with a command line that allows you to execute programs. The core of Unix itself is actually not much more than a library of code. Programs make calls to routine inside Unix to display information on the terminal, to write to a le, to read from a le, to interact with the network, etc. In fact, the interaction with Unix that you perform on the command line is done through a special program that can read input from the command line, parse it, do I/O redirections, call appropriate commands with appropriate options, etc. It is called a command shell (or just shell for short). There are many avors of shells, split into two families, each family with a different feel, and each member of any given family with different features. The rst family is the Bourne shell family, with the following shells: sh, the very basic Bourne shell ksh the sh-compatible Korn shell bash the GNU Bourne-again shell, with features from both ksh and csh (see below) The second family is the C shell family, with shells: csh, the original C shell tcsh, an extended C shell with more functionality Typically, when we say that something works for the Bourne shell, it will work for all the shells in the Bourne shell family, and similarly for the C shell. When you login, the system starts up an initial shell, which is called the login shell. You can change which shell is you defauly shell when you login by invoking the chsh command at any point. Since a shell is just a program, you can always switch to a new shell by basically executing this shell. For example, you can start an instance of ksh by simply executing it: 1
babbage% ksh $ Since most shells will have different prompts (in my case, my csh has a prompt babbage%, while ksh has a prompt $), you can tell in this case that we are in a new shell. We can then issues commands interpreted by this new shell. We will see in the coming lectures some of the differences between the shells. One of them, for example, is I/O redirection, specically how to handle redirection to an existing le. The Bourne shells will typically overwrite the le, while the C shells will report an error. In a C shell, if you do want to overwrite a le at redirection, you can redirect using the &! redirection operator. This operator is not understood by Bourne shells. What are the roles of the shell? It turns out that the shell is what manages jobs and I/O redirection (cf. Lecture 6). The shells is also in charge of parsing the command line and performing substitutions. This is what we will focus on next.
Shell substitutions
When you enter a line at the shell prompt, before the shell executes the programs that you specify, it will make a pass over the line you entered to perform various kinds of substitutions. Here are the special characters that can be used to guide substitution: *, ?, [...] these are wildcard characters. are They replaced by the appropriate portion of lenames. For instance, the wildcard character *, by itself, is replaced by the list of all the les in the current directory. Hence, if you want to search for foo in all the les in the current directory, you can type fgrep foo *, which the shell will exand into fgrep foo file1 file2 file3 ... for all the les in the current directory. You can use wildcards to match part of le names. For instance, *.txt will match any lename ending with .txt, while a*b will match any lename starting with a and ending with b. Note that these are not regular expressions. They use wildcard characters. Another wildcard character is ?, which can stand for any letter. Hence, a?c will be expanded into all the lenames in the current directory that start with a, have one character following it and followed by a b. The [abc] construct can be replaced by any of the characters within the brackets. \ is used to escape the following character. Since some characters have special meaning to the shell, such as *, you need to escape them if you want to refer to the actual character. For example, if you have a le called * in your current directory, and you want only to search for foo in it, then you would need to write fgrep foo \*, escaping the special * character. $var where var is a shell variables (see next section). This is replaced by the value of the shell variable. 2
command this is replaced by whatever executing command sends to stdout. "..." indicates that within the double-quotes, only variable and command substitutions should occur (i.e., no wildcard expansion). ... indicates that within the single-quotes, no substitutions should occur (no wildcard, variable or command substitution). Avoiding substitution...