Why Use vim?, Part Two

Dig into some of vim’s programmable features: improvements in key mapping, a scripting language, and built-in and user-defined functions.

The last “Power Tools” column described two — extensive help and windows — of the many things that make vim better than (but stll compatible with) original vi. (You can get a list of more “most interesting” vim additions by typing :help vim-additions from within vim.) This month, let’s dig into some of vim’s programmable features: improvements in key mapping, a scripting language, and built-in and user-defined functions. Put down your mouse and get ready to roll!

Be Incompatible

By default, vim acts as much like vi as it can, modulo the worst vi “bugs.” However, some features require vim to act unlike vi. vim automatically “improves” itself if it finds a .vimrc file in your home directory. Otherwise, you can make vim act like vim by typing:

:set nocompatible

If you’re following along from your keyboard, please run that command now.

Key Mapping Plus

Original vi lets you make simple “programs” by mapping one or more keys to run a sequence of commands. For instance, the following command maps the two-key sequence \d to surround the current word with double quote (") characters:

:map \d i"^[Ea"^[

After defining this keymap, pressing \ followed quickly by d inserts a double quote to the left of the cursor (i"), returns to command mode (^[, which represents an ESC character), moves to the end of the current word (E), appends a double quote to the right side of the cursor (a"), and returns to command mode (^[).

Vim supports this syntax and improves upon it in two ways. Old vi made it tough to enter control characters, like Escape and Return — characters you use constantly in keymaps. You had to type Control-V before each control character that you wanted to store in the macro. Also, vi represents stored control characters as a two-character sequence: ^[ for Escape, for instance, and ^M for Return. But, if you actually typed ^[ or ^M on the keyboard, vi wouldn’t recognize those as Escape or Return; instead, it would treat the ^ as a “go to start of line” command and the [ as a, well, whatever a left bracket does. So, you generally can’t copy a keymap with your mouse, then paste it in later.

Vim accepts the old Control-V syntax but also understands new representations. Among others, <Esc> in a keymap stands for the Escape key, <F3> represents the F3 key, and <CR> stands for RETURN. As an example, here’s a keymap for the F3 key that adds three lines below the current line: a blank line, the line ^^^ NOTE ^^^, and another blank line:

:map <F3> o<CR>^^^ NOTE ^^^<CR><Esc>

(If you’re reading this article online, you can copy that keymap from the web browser window and paste it into your terminal window.)

As in vi, you can list currently-defined keymaps with the command :map. How do you tell the difference between literal characters and character representations in a keymap? Vim uses blue for character representations.

Figure One shows an example output from :map: an old vi keymap named K that breaks the current line before column 80, and the new F3 keymap that was just defined. Note that the <F3>, <CR>, and <Esc> are all in blue, which means each represents a single character.

Figure One: Displaying vim keymaps

Keymap Modes

In vim, the :map command actually defines the keymap in three modes: Normal, Visual, and Operator-pending. Normal mode is the same as in vi: keys you type invoke commands. Visual mode is like Normal mode, but movement commands extend a highlighted (selected) area. Operator-Pending mode comes when you’ve typed an operator such as d and vim is waiting for you to enter a motion command. (For more info, type :help vim-modes.)

So, for instance, the following keymap maps the Space key to move the cursor to the next whitespace character, either a Space or Tab. It does this by searching with the / command and a regular expression:

:map <Space> /[<Space><Tab>]<CR>

(You don’t need to type the character representations for Space and Tab inside the brackets: simply press the Space and Tab keys. But you do need to type <Space> for the keymap name, and similarly when you remove the keymap by typing :unmap <Space>. Typing a literal Space for the keymap name, instead of <Space>, causes an error.

If you’re used to using the Space key to move through a line of text instead of using the vi command l (lowercase “L”), this keymap will cause you problems. In that case, you might not want the keymap to work in Normal mode, but only in Operator-pending mode. That way, you can type d followed by a Space character to delete up to the next whitespace, or 10d and Space to delete to the tenth occurrence of whitespace, leaving Space by itself to move to the next character on the current line. You’d want to use the :omap (Operator-pending map) command instead of :map.

First, remove the Space keymap by typing :unmap<Space>. Now type (or copy and paste from this article’s browser window):

:omap <Space> /[<Space><Tab>]<CR>

Then try, say, 5d followed by Space. However, typing SPACE by itself should move your cursor along the current line because the omap won’t take effect here.

For more about this, type :help 40.1 to go to section 40.1 of vim ’s built-in help.

Introducing vim Scripting

Keymaps are great for simple programming, but they can’t make more than the simplest decisions. (For instance, a keymap aborts if any part of it fails, such as a text search not finding a match.) Keymaps can do recursion — repeatedly invoking themselves and/or other keymaps — but there’s little control. Other limitations apply, too.

When original vi users need a more-complex edit, they can filter their buffer through a Linux utility like sort, cut, or perl. For instance, if you’re editing a file where the first column in each line has a number and the rest of the line is text, the following vi command uses awk to total the numbers in the first column and write the total under the last line:

:$r !awk '{sum += $1} END {print sum}' %

Or you can add a number before each line with:

:% !cat -n

Writing an awk script on the editor’s command line can be tedious, slow, and error-prone, though. Vim makes the job easier because you can edit its command line (for instance, use the left and right arrow keys while you type a command), and it also has command history (try the up and down arrow keys after typing a few commands). But vim also has a built-in scripting language that can handle many of the jobs that required external utilities in original vi.

You can write a vim script on its command line, or store a script in a file and read it with the :source command, as in :source ~/myvimscript. Third, you can store vim scripts in the .vimrc file in your home directory — a file that vim reads each time it starts.

Here’s a simple .vimrc file:

set syntax=on
echo “You’re in the” getcwd() “directory.”

The first line sets syntax highlighting. (To find out more, type :help 'syntax'.) The second is a simple vim script command that outputs a message as you start vim. The two quoted strings are output literally, and getcwd() is a function call that returns the name of your current directory.

Scripts and Functions

Let’s say that you have a file that describes the steps required to complex a task or procedure. You don’t want to number the steps until the procedure has been finished. Some of the steps are on a single line, and other steps fill several lines. Let’s write a keymap for \n that inserts a sequential number at the start of a line. Each time you type \n, vim will insert the next-higher number at the front of the line.

vim scripting supports both string and numeric values in a variable. A string is surrounded by double quotes (") and a number isn’t; an unquoted alphanumeric is a variable name. (For more information, type :help variables.) To set a variable, use the :let command. The syntax is: :let var=value.

Let’s use a variable named stepnum to hold the step number. Set the initial value from the command line by typing a colon, :, and then let stepnum=0:

:let stepnum = 0

Next, let’s write a user-defined function that increments stepnum by 1. vim’s function syntax is:

:function Name(var1, var2, ...)
:   body
.
:endfunction

The function name must start with a capital letter. The optional return statement returns a value from the function.

Here’s the function definition.

:function Nextstep()
:  let g:stepnum = g:stepnum + 1
:  return g:stepnum
:  endfunction

You can type it during an interactive vim session, if you’d like; vim automatically prompts with another colon and two spaces until you type endfunction.

All variables are local to a function unless you prefix them with g:, as was done with stepnum. Nextstep() returns the new value of stepnum.

Now for some pondering: If you simply run Nextstep() from the command line, it increments g:stepnum and does nothing else. Somehow Nextstep() must return a value as part of an expression that becomes an editor command.

There’s another challenge: you run a function from the command line (in Command-line mode) but want the function’s result to be used in Normal mode — as part of a command that inserts text in Insert mode.

We need two vim commands:

*normal executes Normal-mode commands that you type on the command line (after a colon prompt). Its argument is the normal-mode commands you want to execute. If you don’t terminate the normal-mode commands, normal will add an <Esc> or <C-C> (CTRL-C) for you.

For example, if you wanted to insert the text NOTE: at the start of the current line, you’d use vim’s I (insert at start of line) command, followed by the text, followed by Escape. You can do that from a command line, with :normal, like this:

:normal "INOTE:"

*execute evaluates its argument and runs the result as an command. The argument is a string, a vim expression, or a combination. For example, if the variable count contains a number, you could move the cursor ahead by count words with the following command:

:execute "normal " count . "w"

Vim’s . (dot) operator does concatenation. So, if count contains the number 8, this would execute the command 8w to move the cursor forward eight words. If this is moving a bit too quickly, try :help normal and :help execute, then experiment a bit.

Back to the step-numbering problem. So far, the variable stepnum is 0 and the function Nextstep() has been defined. Using :execute, you can run the following:

:execute "normal I" . Nextstep() . ". "

If stepnum is set to 3, the normal command emitted would be:

I3. <Esc>

That’s I concatenated with the return value of Nextstep(), concatenated with a dot and a space. As before, normal adds the Escape automatically. This numbers the current step 3..

Next, define a keymap named \n to run that :execute... command. Here’s the keymap:

:map \n :execute "normal I" . Nextstep() . ". "<CR>

Try it! Typing \n from Normal mode should insert a step number on the current line. Move down to the next step and type \n to insert the next number.

Much more...

There’s much more to vim functions and script writing. vim comes with many built-in functions. A function or a script can loop, test conditions and branch, open GUI confirmation boxes, tell you what text is under the cursor, test files, let you test and use vim ‘s internal buffers, and more.

Chapter 41 of the vim online help (as of this writing) covers scripts and functions. Typing :help usr_41 should take you there directly. Or, even better, start with the table of contents by typing :help and paging down a few screens to read more about the powerful new features that have made vim VI IMproved.

Jerry Peek is a freelance writer and instructor who has used Unix and Linux for 25 years. He’s happy to hear from readers; see http://www.jpeek.com/contact.html.

[Read previous article] [Read next article]
[Read Jerry’s other Linux Magazine articles]