Question

Three underscore separated elements make my strings :
– first (letters and digits)
– middle (letters, digits and underscore)
– last (letters and digits)

The last element is optional.

Note : I need to access my groups by their names, not their indices.

Examples :

String : abc_def
first : abc
middle : def
last : None

---

String : abc_def_xyz
first : abc
middle: def
last: xyz

---

String : abc_def_ghi_jkl_xyz
first : abc
middle : def_ghi_jkl
last : xyz

I can’t find the right regex…

I have two ideas so far :

Optional group

(?P<first>[a-z]+)_(?P<middle>\w+)(_(?P<last>[a-z]+))?

But the middle group matches until the end of the string :

String : abc_def_ghi_jkl_xyz
first : abc
middle : def_ghi_jkl_xyz
last : vide

Using the ‘|’

(?P<first>[a-z]+)_(?P<middle>\w+)_(?P<last>[a-z]+)|(?P<first>[a-z]+)_(?P<middle>\w+)

This expression is invalid : first and middle groups are declared two times. I though I could write an expression reusing the matched group from the first part of the expression :

(?P<first>[a-z]+)_(?P<middle>\w+)_(?P<last>[a-z]+)|(?P=first)_(?P=middle)

The expression is valid, however strings with just a first and a middle like abc_def are not matched.

Note

These strings are actually parts of a path I need to match. It could be paths like :

• /my/path/to/abc_def
• /my/path/to/abc_def/
• /my/path/to/abc_def/some/other/stuf
• /my/path/to/abc_def/some/other/stuf/
• /my/path/to/abc_def_ghi_jkl_xyz
• /my/path/to/abc_def_ghi_jkl_xyz/
• /my/path/to/abc_def_ghi_jkl_xyz/some/other/stuf
• /my/path/to/abc_def_ghi_jkl_xyz/some/other/stuf/
• …

Any idea to solve my problem solely with regular expressions ? Post-processing the matched groups is not an option.

Thank you very much !

Answer 1

Change the middle group to be non-greedy, and add beginning and end-of-string anchors:

^(?P<first>[a-z]+)_(?P<middle>\w+?)(_(?P<last>[a-z]+))?$

By default, the \w+will match as much as possible, which eats the rest of the string. Adding the ? tells it to match as little as possible.

Thanks to Tim Pietzcker for pointing out the anchor requirements.