Why does Impala 2.0 regex .*? behavior differ from a typical implementation

[#1]

From the docs at https://docs.cloudera.com/documentation/enterprise/6/6.3/topics/impala_string_functions.html#string_functions__regexp_extract

"This example shows how a pattern string starting with .*? matches the shortest possible portion of the source string, returning the rightmost set of lowercase letters."

    
select regexp_extract('AbcdBCdefGHI','.*?([[:lower:]]+)',1);

returns def

Every other Regex impl I've worked with would return bcd

I can't make sense of the docs either - "shortest possible string.. returns the rightmost.." - the shortest possible string in a "search from the left" operation returns the leftmost, not the rightmost 

[#2]

@cjard ,

I believe this is a bug. Check this out: https://issues.apache.org/jira/browse/IMPALA-2917

I agree that the documentation needs to be fixed as well.

André

[#3]

I've advised the documentation team of the confusing wording for you. Thank you for letting us know. 

[#4]

Thanks all!

I'm curious - does it mean that the documentation will be heavily rewritten to detail how .*? implementation is atypical, or will the bug be fixed so the documentation can be tweaked to be correct? 

[#5]

That will be up to the documentation team. I alerted them to your concerns so they can look into it. 

[#6]

Fingers crossed it can push the reported bug up the review list a bit then! 🙂 Thanks for the excellent help and attention..