2.4.1 Simple Control Mode Commands

A full range of vi editing commands are available to you in control mode. The simplest of these move you around the command line and are summarized in Table 2.7. Vi-mode contains two "word" concepts. The simplest is any sequence of non-blank characters; we'll call this a non-blank word. The other is any sequence of only alphanumeric characters (letters and digits) or any sequence of only non-alphanumeric characters; we'll just call this a word. [10]

[10] Neither of these definitions is the same as the definition of a word in emacs-mode.

All of these commands except the last three can be preceded by a number that acts as a repeat count. The last two will be familiar to users of UNIX utilities (such as grep) that use regular expressions, as well as to vi users.

Time for a few examples. Let's say you type in this line and, before you hit RETURN, decide you want to change it:

$ fgrep -l Bob < ~pete/wk/names 

As shown, your cursor is beyond the last character of the line. First, type ESC to enter control mode; your cursor will move back one space so that it is on the s. Then if you type h, your cursor will move back to the e. If you type 3h from the e, you will end up at the n.

Now we will see the difference between the two "word" concepts. Go back to the end of the line by typing $. If you type b, the word in question is "names", and the cursor will end up on the n:

$ fgrep -l Bob < ~pete/wk/names

If you type b again, the next word is the slash (it's a "sequence" of non-alphanumeric characters), so the cursor ends up over it:

$ fgrep -l Bob < ~pete/wk/names

However, if you typed B instead of b, the non-blank word would be the entire pathname, and the cursor would end up at the beginning of it-that is, over the tilde:

$ fgrep -l Bob < ~pete/wk/names

You would have had to type b four times-or just 4b-to get the same effect, since there are four "words" in the part of the pathname to the left of /names: wk, slash, pete, and the leading tilde.

At this point, w and W do the opposite: typing w gets you over the p, since the tilde is a "word", while typing W brings you to the end of the line. But whereas w and W take you to the beginning of the next word, e and E take you to the end of the current word. Thus, if you type w with the cursor on the tilde, you get to:

$ fgrep -l Bob < ~pete/wk/names

Then typing e gets you to

$ fgrep -l Bob < ~pete/wk/names

And typing an additional w gets you to:

$ fgrep -l Bob < ~pete/wk/names

On the other hand, E gets you to the end of the current non-blank word-in this case, the end of the line. (If you find these commands non-mnemonic, you're right. The only way to assimilate them is through lots of practice.)

2.4.2 Entering and Changing Text

Now that you know how to enter control mode and move around on the command line, you need to know how to get back into input mode so you can make changes and type in additional commands. A number of commands take you from control mode into input mode; they are listed in Table 2.8 All of them enter input mode a bit differently.

Most likely, you will use either i or a consistently, and you may use R occasionally. I and A are abbreviations for 0i and $a respectively. To illustrate the difference between i, a, and R, say we start out with our example line:

$ fgrep -l Bob < ~pete/wk/names

If you type i followed by end, you will get:

$ fgrep -l Bob < ~pete/wkend/names

That is, the cursor will always appear to be under the / before names. But if you type a instead of i, you will notice the cursor move one space to the right. Then if you type nick, you will get:

$ fgrep -l Bob < ~pete/wk/nicknames

That is, the cursor will always be just after the last character you typed, until you type ESC to end your input. Finally, if you go back to the n in names, type R instead, and then type task, you will see:

$ fgrep -l Bob < ~pete/wk/tasks

In other words, you will be replacing (hence R) instead of inserting text.

Why capital R instead of lowercase r? The latter is a slightly different command, which replaces only one character and does not enter input mode. With r, the next single character overwrites the character under the cursor. So if we start with the original command line and type r followed by a semicolon, we get:

$ fgrep -l Bob < ~pete/wk;names

If you precede r with a number N, it will allow you to replace the next N existing characters on the line-but still not enter input mode. Lowercase r is effective for fixing erroneous option letters, I/O redirection characters, punctuation, etc.

2.4.3 Deletion Commands

Now that you know how to enter commands and move around the line, you need to know how to delete. The basic deletion command in vi-mode is d followed by one other letter. This letter determines what the unit and direction of deletion is, and it corresponds to a motion command, as listed previously in Table 2.7. Table 2.9 shows some commonly-used examples.

These commands have a few variations and abbreviations. If you use a c instead of d, you will enter input mode after it does the deletion. You can supply a numeric repeat count either before or after the d (or c). Table 2.10 lists the available abbreviations.

Most people tend to use D to delete to end of line, dd to delete an entire line, and x (as "backspace") to delete single characters. If you aren't a hardcore vi user, you may find it difficult to get some of the more esoteric deletion commands under your fingers.

Every good editor provides "un-delete" commands as well as delete commands, and vi-mode is no exception. Vi-mode maintains a delete buffer that stores all of the modifications to text on the current line only (note that this is different from the full vi editor). The command u undoes the last text modification command only, while U undoes all such commands on the current line. So if you make one change but want to undo it, type u; but if you make lots of changes and find that the original is closer to what you want, you can undo everything by typing U. A related command is . (dot), which redoes the last text modification command.

There is also a way to save text in the delete buffer without having deleted it in the first place: just type in a delete command but use y ("yank") instead of d. This does not modify anything, but it allows you to retrieve the yanked text as many times as you like later on. The command to retrieve yanked text is p, which inserts the text on the current line to the left of the cursor. The y and p commands are powerful but far better suited to "real vi" tasks like making global changes to documents or programs than to shell commands, so we doubt you'll use them very often.

2.4.4 Moving Around in the History File

The next group of vi control mode commands we will cover allows you to move around in and search your history file. This is the all-important functionality that lets you go back and fix an erroneous command without retyping the entire line. These commands are summarized in Table 2.11.

The first three can be preceded by repeat counts (e.g., 3k or 3- moves back three lines in the history file).

If you aren't familiar with vi and its cultural history, you may be wondering at the wisdom of choosing such seemingly poor mnemonics as h, j, k, and l for backward character, forward line, backward line, and forward character, respectively. Well, there actually is a rationale for the choices-other than that they are all together on the standard keyboard.

Bill Joy originally developed vi to run on Lear-Siegler ADM-3a terminals, which were the first popular models with addressable cursors (meaning that a program could send an ADM-3a a command to move the cursor to a specified location on the screen). The ADM-3a's h, j, k, and l keys had little arrows on them, so Joy decided to use those keys for appropriate commands in vi.

Another (partial) rationale for the command choices is that [CTRL-H] is the traditional backspace key, and [CTRL-J] denotes linefeed.

Perhaps + and - are better mnemonics than j and k, but the latter have the advantage of being more easily accessible to touch typists. In either case, these commands are the most basic ones for moving around the history file. To see how they work, let's take the same examples we used when discussing emacs-mode above.

You enter the example command (RETURN works in both input and control modes, as does LINEFEED or [CTRL-J]):

$ fgrep -l Bob < ~pete/wk/names

but you get an error message saying that your option letter was wrong. You want to change it to -s without having to retype the entire command. Assuming you are in control mode (you may have to type ESC to put yourself in control mode), you type k or - to get the command back. Your cursor will be at the beginning of the line:

$ fgrep -l Bob < ~pete/wk/names

Type w to get to the -, then l to get to the l. Now you can replace it by typing rs; press RETURN to run the command.

Now let's say you get another error message, and you finally decide to look at the manual page for the fgrep command. You remember having done this a while ago today, so rather than typing in the entire man(1) command, you search for the last one you used. To do this, type ESC to enter control mode (if you are already in control mode, this will have no effect), then type / followed by man or ma. To be on the safe side, you can also type ^ma; the ^ means match only lines that begin with ma. [11]

[11] Fans of vi and search utilities like grep should note that caret (^) for beginning-of-line is the only context operator vi-mode provides for search strings.

But typing /^ma doesn't give you what you want: instead, the shell gives you:

$ make myprogram

To search for "man" again, you can type n, which does another backward search using the last search string. Typing / again without an argument and hitting RETURN will accomplish the same thing.

The G command retrieves the command whose number is the same as the numeric prefix argument you supply. G depends on the command numbering scheme described in Chapter 3 in the section "Prompting Variables." Without a prefix argument, it goes to command number 1. This may be useful to former C shell users who still want to use command numbers.

2.4.5 Character-finding Commands

There are some additional motion commands in vi-mode, although they are less useful than the ones we saw earlier in the chapter. These commands allow you to move to the position of a particular character in the line. They are summarized in Table 2.12, in which x denotes any character.

All of these commands can be preceded by a repeat count.

Starting with the previous example: let's say you want to change Bob to Rob. Make sure that you're at the end of the line (or, in any case, to the left of the B in Bob); then, if you type FB, your cursor will move to the B:

$ fgrep -l Bob < ~pete/wk/names

At this point, you could type r to replace the B with R. But let's say you wanted to change Bob to Blob. You would need to move one space to the right of the B. Of course, you could just type l. But, given that you're somewhere to the right of Bob, the fastest way to move to the o would be to type TB instead of FB followed by l.

As an example of how the repeat count can be used with character-finding commands, let's say you want to change the filename from names to namfile. In this case, assuming your cursor is still on the B, you need to get to the third e to the right, so you can type 3te, followed by l to put the cursor back on the e in names.

The character-finding commands also have associated delete commands. Read the command definitions in the previous table and mentally substitute "delete" for move. You'll get what happens when you precede the given character-finding command with a d. The deletion includes the character given as argument. For example, assume that your cursor is under the n in names:

$ fgrep -l Bob < ~pete/wk/names

If you want to change names to aides, one possibility is to type dfm. This means "delete right to next occurrence of m," i.e., delete "nam." Then you can type i (to enter input mode) and then "aid" to complete the change.

One final command rounds out the vi control mode commands for getting around on the current line: you can use the pipe character (|) for moving to a specific column, whose number is given by a numeric prefix argument. Column counts start at 1; count only your input, not the space taken up by the prompt string. The default repeat count is 1, of course, which means that typing | by itself is equivalent to 0 (see Table 2.7).

2.4.6 Filename Completion

Although the character-finding commands and | are not particularly useful, vi-mode provides one additional feature that we think you will use quite often: filename completion. This feature is not part of the real vi editor, and it was undoubtedly inspired by similar features in emacs and, originally, in the TOPS-20 operating system for DEC mainframes.

The rationale behind filename completion is simple: you should have to type only as much of a filename as is necessary to distinguish it from other filenames in the same directory. Backslash (\) is the command that tells the Korn shell to do filename completion in vi-mode. If you type in a word, type ESC to enter control mode, and then type \, one of four things will happen; they are the same as for ESC ESC in emacs-mode:

1. If there is no file whose name begins with the word, the shell will beep and nothing further will happen.

2. If there is exactly one way to complete the filename and the file is a regular file, the shell will type the rest of the filename, followed by a space in case you want to type in more command arguments.

3. If there is exactly one way to complete the filename and the file is a directory, the shell will complete the filename, followed by a slash.

4. If there is more than one way to complete the filename, the shell will complete out to the longest common prefix among the available choices.

A related command is *, which is the same as ESC * in emacs-mode as described earlier in this chapter. [12] It behaves similarly to ESC \, but if there is more than one completion possibility (number four in the list above), it lists all of them and allows you to type further. Thus, it resembles the * shell wildcard character.

[12] If you count the ESC needed to get out of input mode, the vi-mode command is identical to emacs-mode.

Less useful is the command =, which does the same kind of filename expansion as the * shell wildcard, but in a different way. Instead of expanding the filenames onto the command line, it prints them in a numbered list with one filename on each line. Then it gives you your shell prompt back and retypes whatever was on your command line before you typed =. For example, if the files in your directory include program.c and problem.c, and you type pro followed by ESC and then =, you will see this:

$ cc pro
1) problem.c
2) program.c

2.4.7 Miscellaneous Commands

Several miscellaneous commands round out vi-mode; some of them are quite esoteric. They are listed in Table 2.13.

[13] The line is also "executed" by the shell. However, # is the shell's comment character, so the shell ignores it.

The first of these can be preceded by a repeat count. A repeat count of n preceding the ~ changes the case of the next n characters. [14] The cursor will advance accordingly.

[14] This, in our opinion, is a design flaw in the vi editor that the Korn shell authors might have corrected. Letting the user append a motion command to ~ and having it behave analogously to d or y would have been much more useful; that way, a word could be case-twiddled with only two keystrokes.

A repeat count preceding _ causes the n-th word in the previous command to be inserted in the current line; without the count, the last word is used. Omitting the repeat count is useful because a filename is usually the last thing on a UNIX command line, and because users often run several commands in a row on the same file. With this feature, you can type all of the commands (except the first) followed by ESC _, and the shell will insert the filename.

Finally, the command @ allows you to create keyboard shortcuts by interacting with the shell's alias facility (see Chapter 3). If you create an alias called _x, where x is a letter, then the shell will expand the alias on the current line (but not run it) if you type @ followed by x. As with the similar facility in emacs-mode, we don't find this particularly useful.