Grep doesn't seem to find all matches in a file I expect

Question

I'm not sure if I don't fully understand grep or if regexes are the source of my problem, so I have two questions. I have a simple test file named test.txt with the following contents:

$ cat test.txt Settings.xml blah Settings_1.xml blah Settings_2.xml

When I run grep in a directory containing only the above test file with the following command, it returns with no matches:

$ grep -ir "Settings*xml"

1) Why is the wildcard * not catching the period?

And when I run grep as such:

$ grep -ir "Settings*.xml"

the only difference being the period after the wildcard, the results are:

test.txt:Settings.xml

2) Why is grep not finding the other two matches?

Solution 1

The reason is that * is a special character in regular expressions and means zero or more preceding characters. You have to escape * to mean a literal * character with \. So in your examples:

grep -ir "Settings*xml"

would search for a string that starts with Setting, and then has zero or more s characters and xml at the end. There is no such string in your file because xml is always preceded with .. And this:

grep -ir "Settings*.xml"

would search for a string that starts with Setting, and then has zero or more s and .xml after zero or more s letters.

Your first regex would match something like this:

Settingssxml

Solution 2

This other answer explains what happened, it answers your explicit questions. My answer is intended to introduce a broader context.

I guess you expected * to match zero or more characters (any characters) and . to literally mean .. This works with shell globbing, i.e. if you had files like this:

$ ls -1
Settings.xml
blah
Settings_1.xml
Settings_2.xml

then (say, in bash) you could do:

$ echo Settings*.xml
Settings.xml Settings_1.xml Settings_2.xml

You didn't get what you expected because grep uses regex syntax where:

• . matches (almost) any character,
• * means "zero or more preceding characters",
• \ forces the next character to be interpreted literally.

That's why instead of "Settings*.xml" you should have used "Settings.*\.xml". In this case:

• .* does what you thought * would do,
• \. does what you thought . would do.

Comments

[Ben Sandeen]

It looks like the * wasn't doing what I thought (see the answer from @ArkadiuszDrabczyk) A solution that does return what I want is: $ grep -ir "Settings[[:alnum:][:punct:]]*.xml"

[slhck]

You mean * in your first sentences.