CODEWITHSUNDEEP
phpregexpreg-match-alltext-extractionword-boundaries
Dave
Dave

Reputation: 63

Match all space-delimited "words" containing at least 1 letter and 1 number and may contain slashes and hyphens

I have the following string:

SEDCVBNT S800BG09 7GFHFGD6H 324235346 RHGF7U S8-00BG/09 7687678

and the following regex:

preg_match_all('/\b(?=.+[0-9])(?=.+[A-Z])[A-Z0-9-\/]{4,20}/i', $string, $matches)

What I'm trying to achieve is to return all of the whole "words" that:

Unfortunately, the above regex returns purely alphabetical and purely numeric words as well:

Array (
  [0] => Array (
      [0] => SEDCVBNT
      [1] => S800BG09
      [2] => 7GFHFGD6H
      [3] => 324235346
      [4] => RHGF7U
      [5] => S8-00BG/09
  )
)

I don't want SEDCVBNT or 324235346 to be returned.

Upvotes: 2

Views: 185

Answers (3)

mickmackusa
mickmackusa

Reputation: 47874

Word boundary markers (\b) cannot be relied upon for identifying the edges of a "word" for this task because, for one example, a word ending in slash followed by a space will not satisfy a word boundary. A word boundary is only appropriate when determining the zero-width position between a \w and \W (and vice versa).

Code: (Demo)

$string = 'SEDCVBNT S800BG09 7GFHFGD6H 324235346 RHGF7U S8-00BG/09 7687678';
preg_match_all(
    '~
      (?:^|\s)      #match start of string or whitespace
      \K            #release previously matched characters
      (?=\S*[a-z])  #lookahead for zero or more visible characters followed by letter
      (?=\S*\d)     #lookahead for zero or more visible characters followed by number
      [a-z\d/-]+    #match one or more consecutive whitelisted characters
      (?=\s|$)      #lookahead for a whitespace or the end of string
     ~xi',          #ignore literal whitespaces in pattern, use case-insensitivity with letters
    $string,
    $m
);
var_export($m);

Upvotes: 0

Jon
Jon

Reputation: 437336

You need slightly advanced regex syntax for this one.

The regex I came up with is

(?<=\s|^)(?=[\w/-]*\d[\w/-]*)(?=[\w/-]*[A-Za-z][\w/-]*)([\w/-])+(?=\s|$)

Let's explain it:

  • The syntax [\w/-] comes up a lot; this means "any word character (which includes letters, digits, accented letters etc) or a slash or a dash" -- effectively, all characters that you consider to be part of a valid token.
  • The regex uses positive lookahead to make sure that, at the place where a match is attempted, the following text does satisfy certain criteria. Positive lookahead looks like this: (?=[\w/-]*\d[\w/-]*).
  • It also uses positive (the one at the end: (?=\s|$)) and negative (at the beginning: (?<=\s|^)) lookahead to make sure that a match is only made if the whole text token begins after a whitespace character or is at the beginning of the input string (\s|^) and is followed by with a whitespace character or terminates the input string (\s|$).
  • Since the two interior lookahead patterns are almost identical to the capture group pattern ([\w/-])+, in effect I 'm using them to only match text that matches multiple patterns: both of the lookaheads and the capture group pattern at the end.
  • The first lookahead ensures that the next token includes at least one digit (\d).
  • The second lookahead ensures that the next token includes at least one letter (A-Za-z).
  • The capture group matches one or more word characters and/or / and -.

Therefore, for the capture group to match, the text being examined must:

  1. Be preceded either by whitespace or the beginning of the input string (this prevents partial-word matches starting after a disallowed character)
  2. Include at least one digit in the next stretch of allowed characters (first positive lookahead)
  3. Include at least one letter in the next stretch of allowed characters (second positive lookahead)
  4. Be comprised only of word characters, / and - (capturing group).
  5. Be followed either by whitespace or the end of the input string (this prevents partial-word matches ending at a disallowed character).

Which is exactly what you require. :)

Note: refiddle.com seems to not play well with negative lookbehind, so the regexp after the link does not include the initial (?<=\s|^) part. This means that it will erroneously match the DEF456 in ABC123$DEF456.

Upvotes: 2

agent-j
agent-j

Reputation: 27913

Here is the raw regex: \b(?=\S*?\d)(?=\S*?[a-z])\S+?(?=$|\s)

preg_match_all('/\b(?=\S*?\d)(?=\S*?[a-z])\S+?(?=$|\s)/i', $string, $matches)

Upvotes: -1

Related Questions