How to use preg_replace to remove excessive single spaces - php

Question

We are extracting text from PDF files, and there is a high frequency of results that contain malformed text. Specifically adding spaces between the characters of a word. e.g. SEATTLE is being returned as S E A T T L E.
Is there a RegEx expression for preg_replace that can remove any spaces in the case of n number of single character "words"? Specifically, remove spaces from any occurrence of a string that is more than 3 single alpha characters and is separated by spaces?
If googled this for awhile, but can't even imagine how to construct the expression. As expressed in a comment, I don't want ALL spaces removed, but only when there is an occurrence of >3 single alpha characters, e.g. Welcome to the Greater S E A T T L E area should become Welcome to the Greater SEATTLE area. The result is to be used in full text searching, so case sensitivity is not a concern.

Answer 1

You may use a simple approach with a preg_replace_callback. Match '~\b[A-Za-z](?: [A-Za-z]){2,}\b~' and str_replace spaces in the anonymous function:
$regex = '~\b[A-Za-z](?: [A-Za-z]){2,}\b~';
$result = preg_replace_callback($regex, function($m) {
return str_replace(" ", "", $m[0]);
}, $s);
See the regex demo.
To only match sequences of uppercase letters, remove a-z from the pattern:
$regex = '~\b[A-Z](?: [A-Z]){2,}\b~';
And another thing: there may be soft/hard spaces, tabs, other kind of whitespace. Then, use
$regex = '~\b[A-Za-z](?:\h[A-Za-z]){2,}\b~u';
^^ ^
Finally, to match any Unicode letter, use \p{L} (to only match uppercase ones, \p{Lu}) instead of [a-zA-Z]:
$regex = '~\b\p{L}(?:\h\p{L}){2,}\b~u';
NOTE: It will most probably fail to work in some cases, e.g. when there are one-letter words. You will have to handle those cases separately/manually. Anyway, there is no safe regex-only way to fix OCR issues.
Pattern details
\b - a word boundary
[A-Za-z] - a single letter
(?: [A-Za-z]){2,} - 2 or more occurrences of
- a space (\h matches any kind of horizontal whitespace)
[A-Za-z] - a single letter
\b - a word boundary
When usign u modifier, \h becomes Unicode-aware.

Answer 2

You could do this in one go:
(?i:(?<!\S)([a-z]) +((?1))|\G(?!\A) +((?1))\b)
See live demo here
Explanation:
(?i: # Start of non-capturing group with case-insensitive modifier on
(?<!\S) # Negative lookbehind to ensure there is no leading non-whitespace character
([a-z]) + # Capture one letter and at least one space
((?1)) # Capture one letter in 2nd capturing group
| # Or
\G(?!\A) + # Start match from where previous match ends
# with matching spaces
((?1))\b # Match a letter at word boundary
) # End of non-capturing group
PHP code:
$str = preg_replace('~(?i:(?<!\S)([a-z]) +((?1))|\G(?!\A) +((?1))\b)~', '$1$2$3', $str);

Answer 3

You may use this pure regex approach with lookarounds and \G:
$re = '~\b(?:(?=(?:\pL\h+){3}\pL\b)|(?<!^)\G)(\pL)\h+(?=\pL\b)~';
$repl = preg_replace($re, '$1', $str);
RegEx Demo
RegEx Details:
\b: Match word boundary
(?:: Start non-capture group
(?=(?:\pL\h+){3}\pL\b): Lookahead to assert we have 3+ single letters separated by 1+ spaces
|: OR
(?<!^)\G: \G asserts position at the end of the previous match. (?<!^) ensures we don't match start of the string for the first match
): End non-capture group
(\pL): Match a single letter and capture it
\h+: Followed by 1+ horizontal whitespace
(?=\pL\b): Assert that we only have a single letter ahead
In the replacement we use $1 which is the letter left of whitespace we capture