difficulty with gawk sub()

[Anton Treuenfels]

So I have a string like this:

"^(+|-|/|*|"

which is "automagically" generated during the initialization phase of an
expression parser. Ultimately it's to be used as a regular expression
pattern. As the form it's in is not a regular expression, I attempted to
remedy that by doing this:

sub( /|$/, ")", theVariableThatHoldsTheString )

with the goal of replacing that last "|" with a ")", ie., to look like this:

"^(+|-|/|*)"

But what I got was this:

")^(+|-|/|*|"

So...is there something wrong with the regular expression I used? I thought
it said "a '|' character followed by the end of the string". Is an anchored
match not a useful concept with sub() (and presumably gsub())? Something
else?

- Anton Treuenfels

[Luuk]

~/tmp> echo "^(+|-|/|*|" | awk '{ sub(/\|$/,")", $0); print $0 }'
^(+|-|/|*)

[Ed Morton]

`|` is an ERE metacharacter meaning `or` so your expression `/|$/` means
`nothing or end of string`. You need to escape it or put it in a bracket
expression to make it a literal: `/\|$/` or `/[|]$/`.

Ed.
>
> - Anton Treuenfels

[Anton Treuenfels]

"Ed Morton" <morto...@gmail.com> wrote in message
news:mibk4a$ksh$1...@dont-email.me...

Thanks - that eventually did occur to me (several hours later). Escaping, at
any rate, not making it a single member class.

I still wonder a bit that anything other than an error occurs when the
original expression is parsed. I did not know a complete lack of anything
had a defined meaning - though obviously it explains the result.

- Anton Treuenfels

[Janis Papanagnou]

Am 06.05.2015 um 05:16 schrieb Anton Treuenfels:
>
> "Ed Morton" <morto...@gmail.com> wrote in message
> news:mibk4a$ksh$1...@dont-email.me...
>>

>> `|` is an ERE metacharacter meaning `or` so your expression `/|$/`
>> means `nothing or end of string`. You need to escape it or put it in a
>> bracket expression to make it a literal: `/\|$/` or `/[|]$/`.
>>
>

> I still wonder a bit that anything other than an error occurs when the
> original expression is parsed. I did not know a complete lack of
> anything had a defined meaning - though obviously it explains the result.

As a simple example, matching "beaf" and "deadbeaf" can be expessed
with an "empty" subclause (avoiding repetitions of common parts).

Janis

>
> - Anton Treuenfels
>