Possible bug with regex functions

[Weston Ley]

I believe there might be a bug with Stata's regex processor. In particular, the regex search pattern "\$" does not search for literal representations of the dollar sign but rather analyzes "\$" as an anchor. E.g. the command below returns the following

disp ustrregexra("\$ asdf", "\$", "dog")

$ asdfdog

When it should instead return

dog asdf

[Bjarte Aagnes]

disp ustrregexra("$ asdf", "[$]", "dog")

dog asdf

. disp ustrregexra("$ asdf", "\\$", "dog")
dog asdf

[Weston Ley]

disp ustrregexra("$ asdf", "[$]", "dog")

dog asdf

. disp ustrregexra("$ asdf", "\\$", "dog")
dog asdf

Thanks for the help. Is there a specific implementation of regex that Stata is modeled after (e.g. PCRE, PCRE2, Java)? I'm not familiar with any other implementation that requires two backslashes to literally search the dollar sign character.

[Nick Cox]

Part of an explanation perhaps starts from another place: the role of $ for Stata originally was to indicate a global macro name, and later adding some support for regular expressions had to respect that.

[Weston Ley]

Part of an explanation perhaps starts from another place: the role of $ for Stata originally was to indicate a global macro name, and later adding some support for regular expressions had to respect that.

That makes sense, I wasn't aware that Stata's regex commands also supported macros.

[Bjarte Aagnes]

• ref #4 https://www.stata.com/support/faqs/p...hes-and-macros/
• on Stata regex engines : https://www.statalist.org/forums/for...79#post1327779
• ICU regexp