如何使用 sed 替换正则表达式捕获组？

Question

I have a large file with many scattered file paths that look like

lolsed_bulsh.png

I want to prepend these file names with an extended path like:

/full/path/lolsed_bullsh.png

I'm having a hard time matching and capturing these. currently i'm trying variations of:

cat myfile.txt| sed s/\(.+\)\.png/\/full\/path\/\1/g | ack /full/path

I think sed has some regex or capture group behavior I'm not understanding

Answer 1

In your regex change + with * :

sed -E "s/(.*)\.png/\/full\/path\/\1/g" <<< "lolsed_bulsh.png"

It prints:

/full/path/lolsed_bulsh

NOTE: The non standard -E option is to avoid escaping ( and )

Answer 2

Save yourself some escaping by choosing a different separator (and -E option), for example:

cat myfile.txt | sed -E "s|(..*)\.png|/full/path/\1|g" | ack /full/path

Note that where supported, the -E option ensures ( and ) don't need escaping.

Answer 3

sed uses POSIX BRE, and BRE doesn't support one or more quantifier + . The quantifier + is only supported in POSIX ERE . However, POSIX sed uses BRE and has no option to switch to ERE .

Use ..* to simulate .+ if you want to maintain portability.

Or if you can assume that the code is always run on GNU sed, you can use GNU extension \+ . Alternatively, you can also use the GNU extension -r flag to switch to POSIX ERE. The -E flag in higuaro's answer has been tagged for inclusion in POSIX.1 Issue 8 , and exists in POSIX.1-202x Draft 1 (June 2020) .