I'm looking for a single pattern to match, free-standing collections of non-Unicode alpha numeric characters. I will eventually do a replace with a single space.

Prerequisite

• In regard to alpha characters, the Unicode category \p{L} is necessary
• In regard to numeric \d is adequate
• white space is included

Match Examples

_{'/' denotes any non-unicode alpha numeric character}

aàa 111 /
   ^   ^^

aàa / 111
   ^^^

aàa /// 111
   ^^^^^

aàa/// 111
   ^^^^

aàa ///111
   ^^^^

aàa *&^#* 111
   ^^^^^^^

)(*)* 111
^^^^^^

à- 1
 ^^

à -1
 ^^

Unmatched Examples

aàa///111

aàa-111

aà-/*&^*-a-1-1-1

What I have so far

• The pattern [^\p{L}\d] will match any non-alpha numeric pattern.
• Zero-width negative lookahead / lookbehind with word boundaries gets it closer e.g. (?<!\b)[^\p{L}\d](?!\b)

However, a pattern that solves all the above examples has been elusive

Note: my spidey senses tell me this is likely possible with a single pattern. Though, if this is more efficient or practical as 2 separate patterns, so be it.

CodePudding user response：

\b word boundaries are problematic because those match a boundary between \w and \W, but you're not using \w and \W.

It looks like you always want whitespace on one side or the other of a match so that needs to be worked in. Give this a try. It matches [^\p{L}\d\n]* either preceded or followed by [ \t] .

[ \t] [^\p{L}\d\n]*|[^\p{L}\d\n]*[ \t] 

Demo:

Visual Representation