LINUX CLASSES - DATA MANIPULATION

How Do I Select Certain Records From a File?

The grep command selects and prints lines from a file (or a bunch of files) that match a pattern. Let's say your friend Bill sent you an email recently with his phone number, and you want to call him ASAP to order some books. Instead of launching your email program and sifting through all the messages, you can scan your in-box file, like this:

grep 'number' /var/mail/hermie
call No Starch Press at this number: 800/420-7240.
noted that recently, an alarming number of alien spacecrafts
among colleagues at a number of different organizations

Here, grep has pulled out just the lines that contain the word number. The first line is obviously what you were after, while the others just happened to match the pattern. The general form of the grep command is this:

grep <flags> <pattern> <files>

The most useful grep flags are shown here:

-i Ignore uppercase and lowercase when comparing.
-v Print only lines that do not match the pattern.
-c Print only a count of the matching lines.
-n Display the line number before each matching line.

When grep performs its pattern matching, it expects you to provide a regular expression for the pattern. Regular expressions can be very simple or quite complex, so we won't get into a lot of details here. Here are the most common types of regular expressions:

abc Match lines containing the string "abc" anywhere.
^abc Match lines starting with "abc."
abc$ Match lines ending with "abc."
a..c Match lines containing "a" and "c" separated by any two characters (the dot matches any single character).
a.*c Match lines containing "a" and "c" separated by any number of characters (the dot- asterisk means match zero or more characters).

Regular expressions also come into play when using vi, sed, awk, and other Unix commands. If you want to master Unix, take time to understand regular expressions. Here is a sample poem.txt file and some grep commands to demonstrate regular-expression pattern matching:

Mary had a little lamb
Mary fried a lot of spam
Jack ate a Spam sandwich
Jill had a lamb spamwich

To print all lines containing spam (respecting uppercase and lowercase), enter

grep 'spam' poem.txt
Mary fried a lot of spam
Jill had a lamb spamwich

To print all lines containing spam (ignoring uppercase and lowercase), enter

grep -i 'spam' poem.txt
Mary fried a lot of spam
Jack ate a Spam sandwich
Jill had a lamb spamwich

To print just the number of lines containing the word spam (ignoring uppercase and lowercase), enter

grep -ic 'spam' poem.txt
3

To print all lines not containing spam (ignoring uppercase and lowercase), enter

grep -i -v 'spam' poem.txt
Mary had a little lamb

To print all lines starting with Mary, enter

grep '^Mary' poem.txt
Mary had a little lamb
Mary fried a lot of spam

To print all lines ending with ich, enter

grep 'ich$' poem.txt
Jack ate a Spam sandwich
Jill had a lamb spamwich

To print all lines containing had followed by lamb, enter

grep 'had.*lamb' poem.txt
Mary had a little lamb
Jill had a lamb spamwich

If you want to learn more about regular expressions, start with the man regexp command. There's also a good book called Mastering Regular Expressions, by Jeffrey Friedl, published by O'Reilly & Associates.

For more information on the grep command, see the grep manual.

Previous Lesson: Selecting Columns
Next Lesson: Search & Replace

[ RETURN TO INDEX ]

Comments (most recent first)

Parash (11 Aug 2010, 10:14)

Hi Bob, I have a requirement which is a bit tricky and I am unable to resolve. I have list of table names in a directory.Some of the tables in the list contain "$" in their names(system tables).Also there are some other files in the same directory and some of the names of these files contain the names as a part of it present in the table list that I have. Now when I try using an egrep(as I am using some other check conditions as well like files beginning with ^table|^p_) with the names from the table list in iteration in the directory, I get an error:"egrep: $ anchor not at end of pattern." for those table names which have $ as a part of their name.I am unable to escape it.I am also typing the directory structure and the logic I am applying.Please understand that my requirement does not include replacing the "$" in the file name.Looking forward for your help.Thanks.

My table list
$ cat table.list
ACTIVE_LOGINS
AB$_BC$_MEM_MC_G
ATTACHMENT

also in the same directory there are files such as:
$
$ ls -1 *AB*
index_AB$_TABLE1_I.sql.TEMPLATE
table_AB$_BC$_MEM_MC_G.sql.TEMPLATE
view_AB$BC$_MEM_MC_S.sql

My Requirement logic:
for table in `cat table.list`
do
echo table=$table
ORIG_TABLE_FILE=`ls | egrep "(^p_|^table_)"${table}".sql.TEMPLATE`
echo ORIG_TABLE_FILE=$ORIG_TABLE_FILE
for file in $(egrep "(ON ${table} \()|(TABLE ${table}\$)" `ls . | grep -v ^table_ | grep -v ^p_` | cut -f1 -d: | uniq)
do
echo file=$file
done
done

sapna (05 Jul 2010, 03:23)

How can I select a few files with similar name (except for number at the end) and put the selected ones in another folder

umar ayaz (16 Jun 2010, 05:20)

Good for basic understanding

Tom (12 May 2010, 00:00)

Hi all....I have a doubt...Can we give the directory name istead of filename in grep command

Bob Rankin (20 Apr 2010, 17:58)

@agh - Fixed now, thanks!

agh (16 Apr 2010, 07:22)

Thank you for the tutorial. I was wondering why the first line in the example got taken by grep. I guess that maybe it is a small mistke.

grep 'number' /var/mail/hermie
can call No Starch Press at 800/420-7240. Office hours are

No number on this line :).

therealenki (09 Apr 2010, 07:15)

Hi Bob,
A useful addition to this page is recursive use of grep. Lots of folks search and ask about this and it's use is NOT obvious cos of shell wildcard expansion.
Example:
Search all 'c' source files in this and sub directories for the text 'link(something) list' starting at the current directory.

Don't use: grep -r 'link.* list' *.c (doesn't work - file not found)

Use: grep -r 'link.* list' . --include "*.c"
(of course this presumes that ur copy of grep actually has the recurse functionality built in otherwise u will need to combine find & grep)

Hope this helps others.

Arlino (22 Feb 2010, 07:11)

Hi Bob, thanks for the hint. Although the grep help command seens to be clear, I was wondering what means -i, --ignore-case ignore case distinction (extracted from #grep --help). I´m not english native but reading you post cleared everything. It seems silly I know. Again, thanks a lot.


<Send This Link to a Friend> <Bookmark This Page>

	* Linux Classes *