sed: using part of the regular expression to parse the text

I have a line of text which looks like this:

/src/my_module_my_branch/my_module 

Example: /src/goodcode_dev/goodcode

and I need to convert it to:

/src/my_branch/my_module 

Example: /src/dev/goodcode

The problem both module and branch can include underscores. So I need first to identify the module from the third part and use it to extract out the branch name. Is there a way using sed to do such conversions?

If you have GNU sed, this should work. It assumes everything begins with /src/.

sed -r 's/^\/src\/([^/]+)_([^/]+)\/\1$/\/src\/\2\/\1/'

In English:

Invoke sed with extended regex (-r) so that + will work. Match beginning of line, /src/, group 1: one or more non-slashes, _, group 2: one or more non-slashes, /, group 1, end of line. Change to /src/, group 2, /, group 1.

First, find the last part, then use it to substitute

$ echo "/src/my_module_my_branch/my_module" | awk -F"/" '{n=$NF;gsub(n"_","",$3);print}' OFS="/"
/src/my_branch/my_module

If there isn't any difference between the "module" and "branch" parts, then I'd say no. That is, if my_module_my could be the module, and branch the branch, then how would sed know?