perl非贪婪的正则表达式匹配太多了

Question

I have a file with something like

<post href="http://example.com/" description="Example website" tag="more text"/>

What I want to get is Example website . Doing:

cat file | perl -pe 's/.*description=".*?"//'

works as expected, and I get tag="more text"/> , but when trying:

cat file | perl -pe 's/.*description="(.*)?"/\1/'

I get Example website" tag="more text/> , while I was expecting to get Example website . So it seems there's something with the capturing and the backreference that is not working as intended, and although I think I might understand why, I'm not sure how to solve it.

I could always do:

cat file | perl -pe 's/.*description="//;s/".*//'

but I really want to understand how to solve it with the regular expression, instead of doing two substitutions.

Answer 1

You aren't using non-greedy, you have greediness within an optional capture group as the question mark is right after the group's end parenthesis:

Change:

description="(.*)?"

to:

description="(.*?)"

and you should have your expected results.

Answer 2

The ? metacharacter has two meanings in regular expressions.

When it follows a character like * or + which allows an expression to be matched a variable number of times, it is the "non-greedy" modifier.

.*?
a+?
(foo){3,}?               # actually, I'm not sure about this one

it In other contexts, it means "match 0 or 1 times"

abc?d                    # matches "abcd" or "abd"

By putting the ? outside the capture group, you have changed it to the second meaning. Put it inside the capture group, like @smerny said.

(.*?)