A great explanation of the numerous productivity methods and their benefits in Vim was written by Jim Dennis. For a number of years Jim Dennis was "The Answer Guy" columnist for the Linux Gazette. Also known as "curses curmudgeon" for using the Lynx browser many years after Netscape/Mozilla and IE became the dominant browsers, Dennis is a now a service engineer for Yahoo. Dennis' thorough explanation has now been posted here for the benefit of the DZone community.
The following explanations were prompted by a developer's question about productive shortcuts for Vim. The developer had two specific complaints about Vim. In Vim's search and replace, they were having to put "/" in front of each special character. They also had to use yy to copy one line, but usually they did not want to copy the whole line. The developer could put [number xx]yy to copy xx number of lines into the buffer, but they never knew if they were selecting exactly what they wanted.
Here is Jim Dennis' comprehensive response:
Your problem with Vim is that you don't grok vi.
You mention cutting with
yy and complain that you almost
never want to cut whole lines. In fact programmers, editing source
code, very often want to work on whole lines, ranges of lines and blocks
of code. However,
yy is only one of many way to yank text
into the anonymous copy buffer (or "register" as it's called in vi).
The "Zen" of vi is that you're speaking a language.
y is a verb. The statement
a simple statement which is, essentially, an abbreviation for
0go to the beginning of this line.
yyank from here (up to where?)
$up to the end of this line.
This can also be expressed as
the current line and paste a copy back into place; leaving a copy in the
anonymous register as a side effect). The
"verbs" take any movement as their "subject." Thus
"yank from here (the cursor) to the end of the current/next (big) word"
y'a is "yank from here to the line containing the mark
If you only understand basic up, down, left, and right cursor movements then vi will be no more productive than a copy of "notepad" for you. (Okay, you'll still have syntax highlighting and the ability to handle files larger than a piddling ~45KB or so; but work with me here).
vi has 26 "marks" and 26 "registers." A mark is set
to any cursor location using the
m command. Each mark is
designated by a single lower case letter. Thus
ma sets the
'a' mark to the current location, and
mz sets the
'z' mark. You can move to the line containing a mark using
' (single quote) command. Thus
to the beginning of the line containing the 'a' mark. You can
move to the precise location of any mark using the
(backquote) command. Thus
`z will move directly to the
exact location of the 'z' mark.
Because these are "movements" they can also be used as subjects for other "statements."
So, one way to cut an arbitrary selection of text would be to drop a
mark (I usually use 'a' as my "first" mark, 'z' as my
next mark, 'b' as another, and 'e' as yet another (I
don't recall ever having interactively used more than four marks in 15
years of using vi; one creates one's own conventions
regarding how marks and registers are used by macros that don't disturb
one's interactive context). Then we go the the other end of our desired
text; we can start at either end, it doesn't matter. Then we can
d`a to cut or
y`a to copy. Thus
the whole process has a 5 keystrokes overhead (six if we started in
"insert" mode and needed to Esc out command mode). Once
we've cut or copied then pasting in a copy is a single keystroke:
I say that this is one way to cut or copy text. However, it is only
one of many. Frequently we can more succinctly describe the range of
text without moving our cursor around and dropping a mark. For example
if I'm in a paragraph of text I can use
movements to the beginning or end of the paragraph respectively. So,
to move a paragraph of text I cut it using
(3 keystrokes). (If I happen to already be on the first or last line
of the paragraph I can then simply use
The notion of "paragraph" defaults to something which is usually intuitively reasonable. Thus it often works for code as well as prose.
Frequently we know some pattern (regular expression) that marks one
end or the other of the text in which we're interested. Searching
forwards or backwards are movements in vi. Thus they
can also be used as "subjects" in our "statements." So I can use
to cut from the current line to the next line containing the string
y?bar to copy from the current line to the most
recent (previous) line containing "bar." If I don't want whole lines I
can still use the search movements (as statements of their own), drop my
mark(s) and use the
`x commands as described previously.
In addition to "verbs" and "subjects" vi also has
"objects" (in the grammatical sense of the term). So far I've only
described the use of the anonymous register. However, I can use any of
the 26 "named" registers by prefixing the "object" reference
" (the double quote modifier). This if I use
I'm cutting the current line into the 'a' register and if I
"by/foo then I'm yanking a copy of the text from here
to the next line containing "foo" into the 'b' register. To
paste from a register I simply prefix the paste with the same modifier
"ap pastes a copy of the 'a' register's
contents into the text after the cursor and
"bP pastes a
copy from 'b' to before the current line.
This notion of "prefixes" also adds the analogs of grammatical
"adjectives" and "adverbs' to our text manipulation "language." Most
commands (verbs) and movement (verbs or objects, depending on context)
can also take numeric prefixes. Thus
3J means "join the
next three lines" and
d5} means "delete from the current
line through the end of the fifth paragraph down from here."
This is all intermediate level vi. None of it is Vim specific and there are far more advanced tricks in vi if you're ready to learn them. If you were to master just these intermediate concepts then you'd probably find that you rarely need to write any macros because the text manipulation language is sufficiently concise and expressive to do most things easily enough using the editor's "native" language.
A sampling of more advanced tricks:
There are a number of
: commands, most notably the
s/foo/bar/g global substitution technique. (That's not advanced
: commands can be). The whole
set of commands was historically inherited by vi's
previous incarnations as the ed (line editor) and later
the ex (extended line editor) utilities. In fact vi
is so named because it's the visual interface to ex.
: commands normally operate over lines of text. ed
and ex were written in an era when terminal screens
were uncommon and many terminals were "teletype" (TTY) devices. So it
was common to work from printed copies of the text, using commands
through an extremely terse interface (common connection speeds were 110
baud, or, roughly, characters per second -- which is slower than a fast
typist; lags were common on multi-user interactive sessions;
additionally there was often some motivation to conserve paper).
So the syntax of most
: commands includes an address or
range of addresses (line number) followed by a command. Naturally one
could use literal line numbers:
:127,215 s/foo/bar to
change the first occurrence of "foo" into "bar" on each line between 127
and 215. One could also use some abbreviations such as
$ for current and last lines respectively. One could
also use relative prefixes
- to refer to
offsets after or before the curent line, respectively. Thus:
meaning "from the current line to the last line, join them all into one
:% is synonymous with
:1,$ (all the
:... g and
:... v commands bear some
explanation as they are incredibly powerful.
:... g is a
prefix for "globally" applying a subsequent command to all lines which
match a pattern (regular expression) while
:... v applies
such a command too all lines which do NOT match the given pattern. As
with other ex commands these can be prefixed by
addressing/range references. Thus
"delete any lines containing the string "foo" from the current one
through the next 21 lines" while
:.,$v/bar/d means "from
here to the end of the file, delete any lines which DON'T contain the
It's interesting that the common Unix command grep
was actually inspired by this ex command (and is named
after the way in which it was documented). The ex
:g/re/p (grep) was the way they documented how to
"globally" "print" lines containing a "regular expression" (re). When ed
and ex were used, the
:p command was one
of the first that anyone learned and often the first one used when
editing any file. It was how you printed the current contents (usually
just one page full at a time using
:.,+25p or some such).
: addresses can also refer to marks. Thus you can
:'a,'bg/foo/j to join any line containing the string
foo to its subsequent line, if it lies between the lines between the 'a'
and 'b' marks.
That's pretty obscure (I've only used something like that a few times in the last 15 years). However, I'll freely admit that I've often done things iteratively and interactively that could probably have been done more efficiently if I'd taken the time to think out the correct incantation.
Another very useful vi or ex
:r to read in the contents of another file.
:rfoo inserts the contents of the file named "foo" at
the current line.
More powerful is the
:r! command. This reads the
results of a command. It's the same as suspending the vi
session, running a command, redirecting its output to a temporary file,
resuming your vi session, and reading in the contents
from the temp. file.
Even more powerful are the
! (bang) and
(ex bang) commands. These also execute external
commands and read the results into the current text. However, they also
filter selections of our text through the command! This we can sort all
the lines in our file using
the vi "goto" command; it defaults to going to the last
line of the file, but can be prefixed by a line number, such as 1, the
first line). This is equivalent to the ex variant
Writers often use
! with the Unix fmt or
fold utilities for reformating or "word wrapping"
selections of text. A very common macro is
(reformat the current paragraph). Programmers sometimes use it to run
their code, or just portions of it, through indent or
other code reformatting tools.
! commands means that any
external utility or filter can be treated as an extension of our
editor. I have occasionally used these with scripts that pulled data
from a database, or with wget or lynx
commands that pulled data off a website, or ssh
commands that pulled data from remote systems.
Another useful ex command is
:source). This reads the contents of a file as a
series of commands. When you start vi it normally,
implicitly, performs a
file (and Vim usually does this on
naturally enough). The use of this is that you can change your editor
profile on the fly by simply sourcing in a new set of macros,
abbreviations, and editor settings. If you're sneaky you can even use
this as a trick for storing sequences of ex editing
commands to apply to files on demand.
For example I have a seven line file (36 characters) which runs a
file through wc, and inserts a C-style comment at the
top of the file containing that word count data. I can apply that
"macro" to a file by using a command like:
vim +'so mymacro.ex'
+ command line option to vi and Vim
is normally used to start the editing session at a given line number.
However it's a little known fact that one can follow the
by any valid ex command/expression, such as a "source"
command as I've done here).
Usually it's far easier to write such "macros" using Perl, AWK, sed (which is, in fact, like grep a utility inspired by the ed command).
@ command is probably the most obscure vi
command. In occasionally teaching advanced systems administration
courses for close to a decade I've met very few people who've ever used
@ executes the contents of a register as if it were a vi
or ex command.
Example: I often use:
:r!locate ... to find some file on my
system and read its name into my document. From there I delete any
extraneous hits, leaving only the full path to the file I'm interested
in. Rather than laboriously Tab-ing through each component
of the path (or worse, if I happen to be stuck on a machine without Tab
completion support in its copy of vi) I just use:
0i:r(to turn the current line into a valid :r command),
"cdd(to delete the line into the "c" register) and
@cexecute that command.
That's only 10 keystrokes (and the expression
is effectively a finger macro for me, so I can type it almost as
quickly as any common six letter word).
A sobering thought
I've only scratched to surface of vi's power and none of what I've described here isn't even part of the "improvements" for which vim is named! All of what I've described here should work on any old copy of vi from 20 or 30 years ago.
There are people who have used considerably more of vi's power than I ever will.