Re: [bug-gawk] syntax error when setting RS via string concatenation

[Ed Morton]

On 5/9/2016 8:32 AM, Andrew J. Schorr wrote:
> On Mon, May 09, 2016 at 08:16:17AM -0500, Ed Morton wrote:
> > I didn't expect to get this syntax error when setting RS via string 
> > concatenation:
> >
> >      $ awk 'BEGIN{RS="[x"; RS=RS"]"}'
> >      awk: cmd. line:1: fatal: Unmatched [, [^, [:, [., or [=: /[x/
> >
> > as opposed to setting it to a variable constructed by the same string 
> > concatenation
> >
> >      $ awk 'BEGIN{t="[x"; t=t"]"; RS=t}'
> >      $
> >
> > Is that a bug or a feature (and if so, what feature?).
> I think it's a feature. Consider this:
>
> bash-4.2$ ./gawk 'BEGIN {RS = "[x"}'
> gawk: cmd. line:1: fatal: Unmatched [, [^, [:, [., or [=: /[x/
>
> Why should setting RS to an invalid regexp succeed?
Because it's not being used at that point. It's not intuitively obvious that you 
can't build up the variable in a loop or some such before you use it to separate 
records which is what I was really doing:

     RS = "["
     for (char in charArray) {
          RS = RS char
          # do other stuff with char
     }
     RS = RS "]"

>   There is no way for gawk to
> know that it should ignore RS because you plan to fix it later.
Sure there is, just ignore its value until it's used. I assume testing its 
validity when evaluated rather than when populated would create a performance 
issue though.
> Your workaround using a temporary variable seems to be the
> correct solution.
That's fine, I can workaround it now I know it's an issue. Thanks for the quick 
response.

     Ed.
> Regards,
> Andy