Re: Regexp Provoking an Error Message Contains \y but the Error Message Displays \b

[Neil R. Ormos]

arnold@skeeve.com wrote:
> "Neil R. Ormos" wrote:

>> The regexp in this short example lacks a closing right parenthesis.

>>> ~/.local/bin/gawk-5.1.0   'BEGIN{ if (a ~ /( \y / ) print "test"}'
>> gawk-5.1.0: cmd. line:1: error: Unmatched ( or \(: /( \b /

>> I have read the manual's explanation of why Gawk requires \y and not \b
>> in regexps.

>> Is it intended that Gawk display "\b" in the error message?

> This is sort of amusing, at least if you're a geek like me. :-)

> You've hit a dark corner in the implementation.  When parsing a regexp
> and getting ready to compile the regexp, gawk turns \b into a literal
> backspace character, and it turns \y into the two characters \b so that
> the regexp routines will match a word boundary.  (It has worked this
> way for decades.)

> The error message comes from the regexp routines, which actually
> saw '\' and 'b'.

> There's not a lot I can do about this...

Fair enough.  Thank you for the explanation, which
confirmed my suspicion.

I was not aware of how much additional processing
of the error message was intended--e.g., whether
there was supposed to be a reciprocal translation
of \b back to the original \y, or perhaps
replacement of the "offending regexp" part of the
error message with a transcription of the original
regexp.

I am so used to seeing gawk's \y as equivalent as
equivalent to grep's \b that I probably would not
have noticed this at all had I not encountered it
with a more elaborate regexp and tried
unsuccessfully to find the offending line of code
by searching for what was quoted in the error
message.

Best regards,

--Neil