1) typo in new Id: iMput.

2) wrong fix for 2nd "Error" call in Substitute(): you disabled Error call but didn't disable Exit, you need to skip Exit there too, so Substitute will return non empty result when user replaces regex "\b" to e.g. "__".

AlexeyT wrote

1) typo in new Id: iMput.

Indeed, I looked right over that. I fixed that.

I read your comment, but I am not sure that this is desirable behaviour.

For example sed also does not consider the empty string a word boundary:

home:~> echo '' | sed s/\\b/x/g

home:~> echo 'a' | sed s/\\b/x/g
xax

For empty string, it does not change anything. I think this is correct, because there is no word, so there also cannot be a word boundary.

So I preferred not to change this behaviour.

Michael.

My app does mass replace with regex. in big text, user enters regex "\b" and replaces it to eg "__". app calls Substitute() with each found match, match is zero length so your code will NOT replace it to "__". but it's needed to replace it to "__".

AlexeyT wrote

My app does mass replace with regex. in big text, user enters regex "\b" and replaces it to eg "__". app calls Substitute() with each found match, match is zero length so your code will NOT replace it to "__". but it's needed to replace it to "__".

That can well be, but maybe in that case you need to change your app for this particular use case. It's very much a corner case.

The alternative is to introduce a new boolean option for TRegexp which controls the behaviour of substitute.

Since there are already 9 boolean properties, it may be better to introduce a Options set property and mark the others deprecated.

Michael.

Yes to this idea!

---- On Sat, 15 Feb 2020 10:31:45 -0500 Michael Van Canneyt wrote ----

Since there are already 9 boolean properties, it may be better to introduce a Options set property and mark the others deprecated.