alphabetic
characters and characters in 'iskeyword'), it is used as the search pattern,
with "\<" prepended (meaning: start of a word).  Otherwise "\<\k\k" is used
as search pattern (start of any keyword of at least two characters).

In Replace mode, the number of characters that are replaced depends on the
length of the matched string.  This works like typing the characters of the
matched string in Replace mode.

If there is not a valid keyword character before the cursor, any keyword of
at least two characters is matched.
	e.g., to get:
	    printf("(%g, %g, %g)", vector[0], vector[1], vector[2]);
	just type:
	    printf("(%g, %g, %g)", vector[0], ^P[1], ^P[2]);

The search wraps around the end of the file, the value of 'wrapscan' is not
used here.

Multiple repeats of the same completion are skipped; thus a different match
will be inserted at each CTRL-N and CTRL-P (unless there is only one
matching keyword).

Single character matches are never included, as they usually just get in
the way of what you were really after.
	e.g., to get:
		printf("name = %s\n", name);
	just type:
		printf("name = %s\n", n^P);
	or even:
		printf("name = %s\n", ^P);
The 'n' in '\n' is skipped.

After expanding a word, you can use CTRL-X CTRL-P or CTRL-X CTRL-N to get the
word following the expansion in other contexts.  These sequences search for
the text just expanded and further expand by getting an extra word.  This is
useful if you need to repeat a sequence of complicated words.  Although CTRL-P
and CTRL-N look just for strings of at least two characters, CTRL-X CTRL-P and
CTRL-X CTRL-N can be used to expand words of just one character.
	e.g., to get:
		M&eacute;xico
	you can type:
		M^N^P^X^P^X^P
CTRL-N starts the expansion and then CTRL-P takes back the single character
"M", the next two CTRL-X CTRL-P's get the words "&eacute" and ";xico".

If the previous expansion was split, because it got longer than 'textwidth',
then just the text in the current line will be used.

If the match found is at the end of a line, then the first word in the next
line will be inserted and the message "Word from other line" displayed, if
this word is accepted the next CTRL-X CTRL-P or CTRL-X CTRL-N will search
for those lines starting with this word.


Completing keywords in 'dictionary'			*compl-dictionary*

							*i_CTRL-X_CTRL-K*
CTRL-X CTRL-K		Search the files given with the 'dictionary' option
			for words that start with the keyword in front of the
			cursor.  This is like CTRL-N, but only the dictionary
			files are searched, not the current file.  The found
			keyword is inserted in front of the cursor.  This
			could potentially be pretty slow, since all matches
			are found before the first match is used.  By default,
			the 'dictionary' option is empty.
			For suggestions where to find a list of words, see the
			'dictionary' option.
			'ignorecase', 'smartcase' and 'infercase' apply.

	CTRL-K	or
	CTRL-N		Search forward for next matching keyword.  This
			keyword replaces the previous matching keyword.

	CTRL-P		Search backwards for next matching keyword.  This
			keyword replaces the previous matching keyword.


Completing words in 'thesaurus'				*compl-thesaurus*

							*i_CTRL-X_CTRL-T*
CTRL-X CTRL-T		Works as CTRL-X CTRL-K, but in a special way.  It uses
			the 'thesaurus' option instead of 'dictionary'.  If a
			match is found in the thesaurus file, all the
			remaining words on the same line are included as
			matches, even though they don't complete the word.
			Thus a word can be completely replaced.

	CTRL-T	or
	CTRL-N		Search forward for next matching keyword.  This
			keyword replaces the previous matching keyword.

	CTRL-P		Search backwards for next matching keyword.  This
			keyword replaces the previous matching keyword.

In the file used by the 'thesaurus' option each line in the file should
contain words with similar meaning, separated by non-keyword characters (white
space is preferred).  Maximum line length is 510 bytes.

For an example,