Combine regex patterns to match groups inside delimited strings

Question

Background

Looking to replace periods with dollar signs only within text that is delimited by dollar signs (never spanning lines). For example:

Names: $annie.bettie.cindy.dannie.ellie$. Only $a$ names. $a.b.c.d.e.f$.

Problem

The following regex almost works, but is too simple:

/([[:alnum:]])\.([[:alnum:]])/g

If a match exists outside of the delimiters ($), then too much will be replaced.

The following regex:

/\$.*?\$/g

Matches and groups the delimited strings:

Names: $annie.bettie.cindy.dannie.ellie$. Only $a$ names. $a.b.c.d.e.f$.

Question

How do I combine the two regular expressions so that the periods can be replaced with another string? For example:

Names: $annie.bettie.cindy.dannie.ellie$. Only $a$ names. $a.b.c.d.e.f$.

Ultimately will become:

Names: `r v$annie$bettie$cindy$dannie$ellie`. Only `r v$a` names. `r v$a$b$c$d$e$f`.

The trouble I'm having is matching the delimited dots.

The regular expression will be piped into sed from a terminal running bash.

Maybe [this SO thread](http://stackoverflow.com/questions/25241413/pass-the-matched-value-to-a-function-and-replace-with-the-return-value) can throw some light. — Wiktor Stribiżew, Sep 28 '16 at 06:58
can you explain your expected output? problem statement said you need to replace `.` with `$` within `$sometext$` pattern.. particularly I don't get the ```r v``` part as well as the missing `$` at end of pattern — Sundeep, Sep 28 '16 at 07:08
@Sundeep: The goal is to create a regex that converts simple [Pandoc](http://pandoc.org/) YAML variable references to the more powerful inline [knitr](http://yihui.name/knitr/) expressions. — Dave Jarvis, Sep 28 '16 at 15:51

score 1 · Accepted Answer · answered Sep 28 '16 at 09:52

This might work for you (GNU sed):

sed -r ':a;s/^(([^$]*\$[^$.]*\$)*[^$]*\$[^$.]*)\./\1\n/;ta;s/(\$[^$]*)\$/`r v\1`/g;y/\n/$/' file

Replace all periods within the groups by a newlines. Insert groups prefix and suffix literals and then translate the newlines to dollars.

score 1 · Answer 2 · answered Sep 28 '16 at 16:06

$ cat ip.txt 
Names: $annie.bettie.cindy.dannie.ellie$. Only $a$ names. $a.b.c.d.e.f$.

$ perl -pe '
BEGIN
{
    sub f
    {
        $a = $_[0] =~ tr/./$/r;
        $a =~ s/^/`r v/;
        $a =~ s/.$/`/;
        return $a;
    }
}
s/\$.*?\$/f($&)/ge
' ip.txt
Names: `r v$annie$bettie$cindy$dannie$ellie`. Only `r v$a` names. `r v$a$b$c$d$e$f`.

The subroutine f performs the necessary transformation for $sometext$ strings - first transliterate . to $, then add some string to beginning and finally remove last character to replace with required format
- The subroutine is put in a BEGIN block, which is executed before processing the input file line by line
s/\$.*?\$/f($&)/ge will extract the $sometext$ pattern and pass on to f subroutine. Perl knows to call it courtesy the e flag
-p switch means input line gets printed after all commands

Combine regex patterns to match groups inside delimited strings

Background

Problem

Question

2 Answers2