How can I get a string that finishes with one of two characters and replace that last character conditionally by regexp?

Question

I'm looking for a regexp formula that finds strings starting with a dash and ending with a dash or a point, it order to manually evaluate cases where dashes must be replaced with em-dashes.

For example, the below text:

 -hi there. -hello-. It's nice -said while looking at the window- if you could come. 

Needs to be replaced with

 —hi there. —hello—. —good morning —he said. 

But this dashes must remain unchanged:

 1992-1994 MTS-O 

Since I dont think a fully automated solution is posible, i'm looking to speed up the manual review with a single regexp that replaces these two:

 –(.+?)– –(.+?)\\. 

With one that match a dash or a point at the end, and let me do a fast substitution that conditionally replace the en dash, when that is matched or keeps the point, if thats matched.

Answer 1

Maybe you can settle with a simple pattern as suggested. But that might cause problems with some edge-cases. It needs a little more to fulfill all your requirements.

..a regexp formula that finds a string starting with a dash and ending with a dash or a point ,

However, if you want to do it in one go you may need a PCRE pattern like this: Demo

(?=^-.*[.-]$)-|\G(?!^).*\K-

First, verify the whole string with a lookahead: (?=^-.*[.-]$) . If we've a match we are at position 1.

Then, we look for the first dash to replace it, followed by a \\G-continue alternative to match subsequent dashes that are not at the starting position (?!^) . We skip ahead to the next - with .* and use \\K to drop everything before it. Fun, right?

In general, I would suggest using two regexes. First to find/verify the pattern in question, and then do the replacement. But that is probably not an option in your environment.

Answer 2

My guess is that, maybe these simple expressions,

(?=-)-

or more accurately for ending with . :

(?=-.*\.$)-

with a simple replacement of — might work.

Demo