 |  |
1.9. Special Characters and Quoting
The characters <, >, |, and & are four examples of special characters that have particular meanings to the shell. The wildcards we saw earlier in this chapter (*, ?, and [...]) are also special characters.
Table 1-6 gives indications of the meanings of all special characters within shell command lines only. Other characters have special meanings in specific situations, such as the regular expressions and string-handling operators we'll see in Chapter 3 and Chapter 4.
Table 1-6. Special characters
| Character | Meaning | See chapter |
|---|---|---|
| ~ | Home directory | 1 |
| ` | Command substitution (archaic) | 4 |
| # | Comment | 4 |
| $ | Variable expression | 3 |
| & | Background job | 1 |
| * | String wildcard | 1 |
| ( | Start subshell | 8 |
| ) | End subshell | 8 |
| \ | Quote next character | 1 |
| | | Pipe | 1 |
| [ | Start character-set wildcard | 1 |
| ] | End character-set wildcard | 1 |
| { | Start code block | 7 |
| } | End code block | 7 |
| ; | Shell command separator | 3 |
| ' | Strong quote | 1 |
| " | Weak quote | 1 |
| < | Input redirect | 1 |
| > | Output redirect | 1 |
| / | Pathname directory separator | 1 |
| ? | Single-character wildcard | 1 |
| % | Job name/number identifier | 8 |
1.9.1. Quoting
Sometimes you will want to use special characters literally, i.e., without their special meanings. This is called quoting. If you surround a string of characters with single quotes, you strip all characters within the quotes of any special meaning they might have.
The most obvious situation where you might need to quote a string is with the print command, which just takes its arguments and prints them to the standard output. What is the point of this? As you will see in later chapters, the shell does quite a bit of processing on command lines -- most of which involves some of the special characters listed in Table 1-6. print is a way of making the result of that processing available on the standard output.
But what if we wanted to print the string, 2 * 3 > 5 is a valid inequality? Suppose you typed this:
print 2 * 3 > 5 is a valid inequality.
You would get your shell prompt back, as if nothing happened! But then there would be a new file, with the name 5, containing "2", the names of all files in your current directory, and then the string 3 is a valid inequality. Make sure you understand why.[16]
[16] This should also teach you something about the flexibility of placing I/O redirectors anywhere on the command line -- even in places where they don't seem to make sense.
However, if you type:
print '2 * 3 > 5 is a valid inequality.'
the result is the string, taken literally. You needn't quote the entire line, just the portion containing special characters (or characters you think might be special, if you just want to be sure):
print '2 * 3 > 5' is a valid inequality.
This has exactly the same result.
Notice that Table 1-6 lists double quotes (") as weak quotes. A string in double quotes is subjected to some of the steps the shell takes to process command lines, but not all. (In other words, it treats only some special characters as special.) You'll see in later chapters why double quotes are sometimes preferable; Chapter 7 contains the most comprehensive explanation of the shell's rules for quoting and other aspects of command-line processing. For now, though, you should stick to single quotes.
1.9.2. Backslash-Escaping
Another way to change the meaning of a character is to precede it with a backslash (\). This is called backslash-escaping the character. In most cases, when you backslash-escape a character, you quote it. For example:
print 2 \* 3 \> 5 is a valid inequality.
produces the same results as if you surrounded the string with single quotes. To use a literal backslash, just surround it with quotes ('\') or, even better, backslash-escape it (\\).
Here is a more practical example of quoting special characters. A few Unix commands take arguments that often include wildcard characters, which need to be escaped so the shell doesn't process them first. The most common such command is find, which searches for files throughout entire directory trees.
To use find, you supply the root of the tree you want to search and arguments that describe the characteristics of the file(s) you want to find. For example, the command find . -name string -print searches the directory tree whose root is your current directory for files whose names match the string, and prints their names. (Other arguments allow you to search by the file's size, owner, permissions, date of last access, etc.)
You can use wildcards in the string, but you must quote them, so that the find command itself can match them against names of files in each directory it searches. The command find . -name '*.c' will match all files whose names end in .c anywhere in your current directory, subdirectories, sub-subdirectories, etc.
1.9.3. Quoting Quotation Marks
You can also use a backslash to include double quotes within a string. For example:
print \"2 \* 3 \> 5\" is a valid inequality.
produces the following output:
"2 * 3 > 5" is a valid inequality.
Within a double-quoted string, only the double quotes need to be escaped:
$ print "\"2 * 3 > 5\" is a valid inequality." "2 * 3 > 5" is a valid inequality.
However, this won't work with single quotes inside quoted expressions. For example, print 'Bob\'s hair is brown' will not give you Bob's hair is brown. You can get around this limitation in various ways. First, try eliminating the quotes:
print Bob\'s hair is brown
If no other characters are special (as is the case here), this works. Otherwise, you can use the following command:
print 'Bob'\''s hair is brown'
That is, '\'' (i.e., single quote, backslash, single quote, single quote) acts like a single quote within a quoted string. Why? The first ' in '\'' ends the quoted string we started with 'Bob, the \' inserts a literal single quote, and the next ' starts another quoted string that ends with the word "brown". If you understand this, you will have no trouble resolving the other bewildering issues that arise from the shell's often cryptic syntax.
A somewhat more legible mechanism, specific to ksh93, is available for cases where you need to quote single quotes. This is the shell's extended quoting mechanism: $'...'. This is known in ksh documentation as ANSI C quoting, since the rules closely resemble those of ANSI/ISO Standard C. The full details are provided in Chapter 7. Here is how to use ANSI C quoting for the previous example:
$ print $'Bob\'s hair is brown' Bob's hair is brown
1.9.4. Continuing Lines
A related issue is how to continue the text of a command beyond a single line on your terminal or workstation window. The answer is conceptually simple: just quote the ENTER key. After all, ENTER is really just another character.
You can do this in two ways: by ending a line with a backslash or by not closing a quote mark (i.e., by including ENTER in a quoted string). If you use the backslash, there must be nothing between it and the end of the line -- not even spaces or TABs.
Whether you use a backslash or a single quote, you are telling the shell to ignore the special meaning of the ENTER character. After you press ENTER, the shell understands that you haven't finished your command line (i.e., since you haven't typed a "real" ENTER), so it responds with a secondary prompt, which is > by default, and waits for you to finish the line. You can continue a line as many times as you wish.
For example, if you want the shell to print the first sentence of Thomas Hardy's The Return of the Native, you can type this:
$ print A Saturday afternoon in November was approaching the \ > time of twilight, and the vast tract of unenclosed wild known \ > as Egdon Heath embrowned itself moment by moment.
$ print 'A Saturday afternoon in November was approaching the > time of twilight, and the vast tract of unenclosed wild known > as Egdon Heath embrowned itself moment by moment.'
There is a difference between the two methods. The first prints the sentence as one long line. The second preserves the embedded newlines. Try both, and you'll see the difference.
1.9.5. Control Keys
Control keys -- those that you type by holding down the CONTROL (or CTRL) key and hitting another key -- are another type of special character. These normally don't print anything on your screen, but the operating system interprets a few of them as special commands. You already know one of them: ENTER is actually the same as CTRL-M (try it and see). You have probably also used the BACKSPACE or DEL key to erase typos on your command line.
Actually, many control keys have functions that don't really concern you -- yet you should know about them for future reference and in case you type them by accident.
Perhaps the most difficult thing about control keys is that they can differ from system to system. The usual arrangement is shown in Table 1-7, which lists the control keys that all major modern versions of Unix support. Note that CTRL-\ and CTRL-| (control-backslash and control-pipe) are the same character notated two different ways; the same is true of DEL and CTRL-?.
You can use the stty(1) command to find out what your settings are and change them if you wish; see Chapter 8 for details. On modern Unix systems (including GNU/Linux), use stty -a to see your control-key settings:
$ stty -a speed 38400 baud; rows 24; columns 80; line = 0; intr = ^C; quit = ^\; erase = ^H; kill = ^U; eof = ^D; eol = <undef>; eol2 = <undef>; start = ^Q; stop = ^S; susp = ^Z; rprnt = ^R; werase = ^W; lnext = ^V; flush = ^O; min = 1; time = 0; ...
The ^X notation stands for CTRL-X.
Table 1-7. Control keys
| Control key | stty name | Function description |
|---|---|---|
| CTRL-C | intr | Stop current command. |
| CTRL-D | eof | End of input. |
| CTRL-\ or CTRL-| | quit | Stop current command, if CTRL-C doesn't work. |
| CTRL-S | stop | Halt output to screen. |
| CTRL-Q | start | Restart output to screen. |
| BACKSPACE or CTRL-H | erase | Erase last character. This is the most common setting. |
| DEL or CTRL-? | erase | Erase last character. This is a common alternative setting. for the erase character |
| CTRL-U | kill | Erase entire command line. |
| CTRL-Z | susp | Suspend current command (see Chapter 8). |
| CTRL-R | rprnt | Reprint the characters entered so far. |
The control key you will probably use most often is CTRL-C, sometimes called the interrupt key. This stops -- or tries to stop -- the command that is currently running. You will want to use this when you enter a command and find that it's taking too long, when you gave it the wrong arguments by mistake, when you change your mind about wanting to run it, and so on.
Sometimes CTRL-C doesn't work; in that case, if you really want to stop a job, try CTRL-\. But don't just type CTRL-\; always try CTRL-C first! Chapter 8 explains why in detail. For now, suffice it to say that CTRL-C gives the running job more of a chance to clean up before exiting, so that files and other resources are not left in funny states.
We've already seen an example of CTRL-D. When you are running a command that accepts standard input from your keyboard, CTRL-D (as the first character on the line) tells the process that your input is finished -- as if the process were reading a file and it reached the end of the file. mail is a utility in which this happens often. When you are typing in a message, you end by typing CTRL-D. This tells mail that your message is complete and ready to be sent. Most utilities that accept standard input understand CTRL-D as the end-of-input character, though many such programs accept commands like q, quit, exit, etc. The shell itself understands CTRL-D as the end-of-input character: as we saw earlier in this chapter, you can normally end a login session by typing CTRL-D at the shell prompt. You are just telling the shell that its command input is finished.
CTRL-S and CTRL-Q are called flow-control characters. They represent an antiquated way of stopping and restarting the flow of output from one device to another (e.g., from the computer to your terminal) that was useful when the speed of such output was low. They are rather obsolete in these days of high-speed local networks and dialup lines. In fact, under the latter conditions, CTRL-S and CTRL-Q are basically a nuisance. The only thing you really need to know about them is that if your screen output becomes "stuck," then you may have hit CTRL-S by accident. Type CTRL-Q to restart the output; any keys you may have hit in between will then take effect.
The final group of control characters gives you rudimentary ways to edit your command line. BACKSPACE or CTRL-H acts as a backspace key (in fact, some systems use the DEL or CTRL-? keys as "erase" instead of BACKSPACE); CTRL-U erases the entire line and lets you start over. Again, most of these are outmoded.[17] Instead of using these, go to the next chapter and read about the Korn shell's editing modes, which are among its most exciting features.
[17] Why are so many outmoded control keys still in use? They have nothing to do with the shell per se; instead, they are recognized by the tty driver, an old and hoary part of the operating system's lower depths that controls input and output to/from your terminal. It is, in fact, the tty driver that understands CTRL-D and signals end-of-input to programs reading from the terminal, not the programs themselves.
|  | |
| 1.8. Background Jobs |  | 2. Command-Line Editing |
Copyright © 2003 O'Reilly & Associates. All rights reserved.