正则表达式匹配一个或多个组太多

Question

I have a series of filenames of varying complexity. Basically, they are always split up by [_]{ASSET}_[OPTIONAL_DESCRIPTION]_v{#####}.{EXT}. ([]s are optional, in this case). Within that format though, each piece can be arbitrarily complex. (leading _s are arbitrary)

character_thing_v001.md
character_Description_v001.md
character_Some_Long_Description_v001.md
character_thing_with_additional_info_v001.md
character_thing_with_additional_info_Description_v001.md
character_thing_with_additional_info_More_Description_Info_v001.md
character_with_additional_info_Complete234ly_arbitrary_Description_v001.md
_character_thing_v001.md
___character_Description_v001.md
____character_Some_Long_Description_v001.md
__character_thing_with_additional_info_v001.md
__character_thing_with_additional_info_Description_v001.md
___character_thing_with_additional_info_More_Description_Info_v001.md

I made a lookahead assertion to separate ASSET and DESCRIPTION and everything worked fine until just recently, when my boss threw a wrench in the system. Now I have to support assets whose convention could be "some_undercase" OR "CAPS_###". I modified to allow AZ and made descriptionText match anything. That's where the mess started.

     (?:[_]+)?
     (?P<assetText>[a-zA-Z0-9]+
       (?=_[a-zA-Z0-9]+)?  # lookahead and optionally assert _Capital
         (?:(?:_[a-zA-Z0-9]+)+)?  # match next group if it exists
     )  # get full match
     (?:[_]+)?
     \_(?P<descriptionText>.+)?
     \_v(?P<versionIncrement>\d+)
     \.(?:\.)?
       (?P<extension>(?:md|some|other|extension|options)) 

This gets me part of the way there but it has problems that you can view, here

Now that the ASSET can have capitals, the lookahead matches too much for ASSET and starts going into the DESCRIPTION. This pattern is one of several that gets automatically generated so I'm looking for a way to solve the root of the problem, rather than write around it. Any guidance would be really appreciated, thank you.

Answer 1

I can't really follow the logic of some of the parts of your regex that seem unnecessary.

Doesn't this simplified regex do the same job?

_*
(?P<assetText>[a-zA-Z0-9]+(_[a-z_0-9]+)?)
(_  (?P<descriptionText>[a-zA-Z0-9_]+)  )?
_v(?P<versionIncrement>[0-9]+)
(?P<extension>\.[A-Za-z0-9]+)

Perhaps the (natural-language) rules for what constitutes an asset and what constitutes an optional description need to be clarified:

• Can an "asset" contain an underscore (I'm assuming not, from the template in your first sentence)?
  • If yes: what's the rule for where "asset" ends and "description" begins? Is it that the description always starts with an upper-case letter?
    • If yes: what are the rules for where upper-case letters can and cannot appear with the "asset"? If there are no restrictions, then the split between asset and description is truly ill-defined.