How can I remove all tokens with non-word characters in Perl?

Question

I am trying to come up with a regex for removing all words that contain non-word characters.

So if it contains a colon, comma, number, bracket etc then remove it from the line, not just the character but the word. I have this so far.

$wordline = s/\s.*\W.*?\s//g;

Does not have to be perfect so removing strings with dash and apostrophe is ok.

Answer 1

s/\w*([^\w\s]|\d)+\w* ?//g;

Answer 2

$wordline = join(" ", grep(/^\w+$/, split(/\s+/, $wordline)));

Answer 3

s/(?<!\S)(?![A-Za-z]+(?:\s|$))\S+(?!\S)//g

In regex-land, a "word character" is a letter, a digit, or an underscore ([A-Za-z0-9_]). It sounds like you're using it to mean just letters, so \w and \W won't do you any good. My regex matches:

• a bunch of non-whitespace characters: \S+

• not preceded: (?<!\S) or followed: (?!\S) by non-whitespace characters

• unless all the characters are letters: (?![A-Za-z]+(?:\s|$))

This will leave behind all the spaces surrounding the words that it deletes. Dealing with those correctly is a little trickier than you might expect; it's much easier to do in a separate step, e.g.:

s/^ +| +(?= |$)//g