Possible bug in Mata regexr() function

[Bjarte Aagnes]

What looks like a bug in the Mata regexr() function is reported to tech-support:

. di regexr("A;",";","")
A
 
. mata:  regexr("A;",";","")
  A;

[Matthew J. Baker]

Bjarte Aagnes --

Might this have something to do with ; not being a regular character? I find the following works:

mata: ustrregexrf("A;",";","")
  A

Best,

Matthew J. Baker

[Matthew J. Baker]

Bjarte Aagnes --

The plot thickens. While the official Stata documentation says that the mata function regexr applies only to "plain" or "lower" Ascii, a semicolon is included in lower ascii. Another weird thing about this is the following experiment with a few different characters:

. mata: regexr("A:", ":", "")
  A:

. mata: regexr("A!", "!", "")
  A!

However, I find that things work correctly if I put a space at the end of the expression:

. mata: regexr("A; ", ";", "")
  A

I'm baffled!

@Matthew J. Baker

[Mauricio Caceres]

More generally, I find this is the case for any character:

mata:
for (k = 97; k <= 122; k++)  {
    regexr("A" + char(k), char(k), "")
}
end

Will print A followed by a-z. In other words, the last character is not being replaced. Consider:

mata:
for (k = 97; k <= 122; k++)  {
    regexr(char(k) + "A", char(k), "")
}
end

This just prints "A" all the way! However,

mata:
for (k = 97; k <= 122; k++)  {
    regexr("A" + char(k), "A" + char(k), "")
}
end

Works just fine (same for addming a space, as noted in the above posts). Sure seems like a bug. Very odd.

[Bjarte Aagnes]

This has been fixed. Update 20 feb 2019.