Regex to "normalize" usage of SPACE after . , : chars (and some exceptions)

Question

I need to normalize some texts (product descriptions) in regard to the correct usage of .,,,: symbols (no space before and one space after)

The regex I've come up with is this:

$variation['DESCRIPTION'] = preg_replace('#\s*([:,.])\s*(?! )#', '$1 ', $variation['DESCRIPTION']);

The problem is that this matches four cases it shouldn't touch:

Any decimal number, like 5.5
Any thousand separator, like 4,500
A "fixed" phrase in Greek, ό,τι
The ellipsis symbol, ... - Basically ellipsis is a totally special case, that I'm thinking should be taken care of in a separate preg_replace maybe? I mean, the three dots should be treated as one thing, meaning that some text ... should indeed be matched and converted to some text... but not to some text. . .

Especially for the numeric exception, I know it can be achieved with some negative lookahead/lookbehind but unfortunately I can't combine them in my current pattern.

This is a fiddle for you to check (the cases that shouldn't be matched are in lines 2, 3, 4).

EDIT: Both of the solutions posted below work fine, but end up adding a space after the last fullstop of the description. This is not much of a problem, as earlier in my code, I was taking care of the s and spaces at the beginning and end of the description, so I moved this preg_replace before that one...

So, the final code I ended up using is this:

$variation['DESCRIPTION'] = preg_replace('#\s*([:,.])(?!(?<=\d.)\d)(?!(?<=ό,)τι)\s*#ui', '$1 ', $variation['DESCRIPTION']);
$variation['DESCRIPTION'] = preg_replace('#^\s*(
)*\s*|\s*(
)*\s*$#', '', $variation['DESCRIPTION']);

So the only thing that's left to achieve is alter this code so that it treats the ellipsis the way I describe above.

Any help with this last requirement will be very much appreciated! TIA

user_3832970 · Answer

You can add two lookaheads containing lookbehinds:

\s*(\.{2,}|[:,.](?!(?<=ό,)τι)(?!(?<=\d.)\d))(?!\s*)\s*

See the regex demo. Note that I also added \s* to the last lookahead and swapped it with the consuming \s* to fail the match if there is after any zero or more whitespaces after the :, , or ..

Details

\s* - zero or more whitespaces
(\.{2,}|[:,.]) - Group 1: two or more dots, or a :, , or .
(?!(?<=ό,)τι) - fail the match if the next two chars are τι preceded with ό,
(?!(?<=\d.)\d) - fail the match if the next char is a digit preceded with a digit and any char (note that a . is enough since the [:,.] already match the char allowed/required, here, we just need to "jump" over that matched char)
(?!\s*) - a negative lookahead that fails the match if there are zero or more whitespaces, , zero or more whitespaces, /> immediately to the right of the current location.


\s* - zero or more whitespaces.

user_2943403 · Answer

If Wiktor's lookaround-heavy pattern is too difficult for you to conceptualize/maintain/adapt, then perhap a match&ignore technique will be easier for you. Admittedly, Wiktor's pattern is optimized for performance.

Pattern:

~                        #starting pattern delimiter 
\s*                      #zero or more whitespaces
(?:                      #start non-capturing group #1
  (?:                    #start non-capturing group #2
    \.\d+                #match float expression not requiring leading digits
    |                    #or
    \d{1,3}(?:,\d{3})+   #match number containing thousands separators
    |                    #or
    ό,τι                 #match literal greek phrase
    |                    #or
                 #match html break tag
  )                      #end non-capturing group #2
  (*SKIP)(*FAIL)         #discard anything matched by group #2
  |                      #or
  (                      #start capture group #1
    \.{3}                #match three dots as ellipsis
    |                    #or
    [:,.]                #match literal colon, comma, or dot
  )                      #end capture group #1
)                        #end non-capturing group #1
\s*                      #zero or more whitespaces
~                        #ending pattern delimiter

As you wish to extend your pattern to include more disqualifying rules, just add another pipe and add a subpattern to match the unwanted substring.

To ensure that three qualifying dots are matched as an ellipsis, match this before checking for individual characters.

Code: (Demo)

$text = <<- Softfoam floor
- Binding with laces

Specs:
• Something

• Something else
• One more

Children's Form Champion

Children's set that will give a comfortable feeling for endless hours of play.
It consists of a cardigan and trousers ,made of soft fabric and have rib cuffs and legs for a better fit.

• Normal fit

• Cardigan  :Rib cuffs, zippers throughout length, high neck, Champion logo 
 
• Pants: Elastic waist with drawstring, ribbed legs, Champion logo. Don't worry,there'll be ...more!
TEXT;

echo preg_replace(
         '~\s*(?:(?:\.\d+|\d{1,3}(?:,\d{3})+|ό,τι|)(*SKIP)(*FAIL)|(\.{3}|[:,.]))\s*~',
         '$1 ',
         $text
     );

Output:

Composition: 80% Polyamide, 15% Elastane, 5% Wool. Side length 50.5 cm <---- THIS SHOULDN'T BE MATCHED
Value 4,500 <---- THIS SHOULDN'T BE MATCHED EITHER

What about $1,234,567.89?
Or... .1mm one tenth of a millimeter?

ό,τι<---- THIS IS A FIXED PHRASE IN GREEK AND THEREFORE SHOULDN'T BE MATCHED
Comfort and timeless design characterize the Puma Smash V2 made of suede leather. They can be worn all the time, being a unique choice for those who want to stand out. Made of rubber. 
- Softfoam floor
- Binding with laces

Specs: 
• Something

• Something else
• One more

Children's Form Champion

Children's set that will give a comfortable feeling for endless hours of play. 
It consists of a cardigan and trousers, made of soft fabric and have rib cuffs and legs for a better fit. 

• Normal fit

• Cardigan: Rib cuffs, zippers throughout length, high neck, Champion logo 
 
• Pants: Elastic waist with drawstring, ribbed legs, Champion logo. Don't worry, there'll be ...more!

Regex to "normalize" usage of SPACE after . , : chars (and some exceptions)

2 Answers