I'm working on an app which scrapes local websites to create a database of upcoming events, and I'm trying to use Regex to catch as many formats of dates as possible.
Consider the following sentence fragments:
I want to be able to scan these and catch as many dates as possible. At the moment I'm doing this in what is probably a flawed way (I'm not great at regex), going through several regex statements one after the other, like this
/([0-9]+?)(st|nd|rd|th) (of)? (Jan|Feb|Mar|etc)/i
/([0-9]+?)(st|nd|rd|th) (of)? (January|February|March|Etcetera)/i
/(Jan|Feb|Mar|etc) ([0-9]+?)(st|nd|rd|th)/i
/(January|February|March|Etcetera) ([0-9]+?)(st|nd|rd|th)/i
I could merge these all into one giant regex statement, but it seems like there must be a cleaner way of doing this in php, maybe a third-party library or something?
EDIT: The regex above may have errors - it's only meant as an example.
I wrote a function which extracts dates out of text by using strtotime():
function parse_date_tokens($tokens) {
# only try to extract a date if we have 2 or more tokens
if(!is_array($tokens) || count($tokens) < 2) return false;
return strtotime(implode(" ", $tokens));
}
function extract_dates($text) {
static $patterns = Array(
'/^[0-9]+(st|nd|rd|th|)?$/i', # day
'/^(Jan(uary)?|Feb(ruary)?|Mar(ch)?|etc)$/i', # month
'/^20[0-9]{2}$/', # year
'/^of$/' #words
);
# defines which of the above patterns aren't actually part of a date
static $drop_patterns = Array(
false,
false,
false,
true
);
$tokens = Array();
$result = Array();
$text = str_word_count($text, 1, '0123456789'); # get all words in text
# iterate words and search for matching patterns
foreach($text as $word) {
$found = false;
foreach($patterns as $key => $pattern) {
if(preg_match($pattern, $word)) {
if(!$drop_patterns[$key]) {
$tokens[] = $word;
}
$found = true;
break;
}
}
if(!$found) {
$result[] = parse_date_tokens($tokens);
$tokens = Array();
}
}
$result[] = parse_date_tokens($tokens);
return array_filter($result);
}
# test
$texts = Array(
"The focus of the seminar, on Saturday 2nd February 2013 will be [...]",
"Valentines Special @ The Radisson, Feb 14th",
"On Friday the 15th of February, a special Hollywood themed [...]",
"Symposium on Childhood Play on Friday, February 8th",
"Hosting a craft workshop March 9th - 11th in the old [...]"
);
$dates = extract_dates(implode(" ", $texts));
echo "Dates:
";
foreach($dates as $date) {
echo " " . date('d.m.Y H:i:s', $date) . "
";
}
This outputs:
Dates:
02.02.2013 00:00:00
14.02.2013 00:00:00
15.02.2013 00:00:00
08.02.2013 00:00:00
09.03.2013 00:00:00
This solution may not be perfect and certainly has its flaws but it's a quite simple solution for your problem.
For this kind of potentially complex regexes, I tend to break it down into simple pieces that can be individually unit-tested, maintained and evolved.
I use REL, a DSL (in Scala) that allows you to reassemble and reuse your regex pieces. This way, you can define your regex like these Date matchers and unit test on each part.
Also, your unit/spec tests can double as your doc for this bit of regex, indicating what is matched and what is not (which tends to be important with regexes).
In the upcoming version of REL (0.3), you will be directly able to export the Regex in, say, PCRE (thus, PHP) flavor to use it independently… For now only JavaScript and .NET translations are implemented in the github repository. Using the latest (not yet publicly committed) snapshot, the PCRE flavor of the English alphanumeric date regex is this:
/(?:(?:(?<!\d)(?<a_d1>(?>(?:(?:[23]?1)st|(?:2?2)nd|(?:2?3)rd|(?:[12]?[4-9]|[123]0)th)\b|0[1-9]|[12][0-9]|3[01]|[1-9]|[12][0-9]|3[01]))(?: ?+(?:of )?+))(?>(?<a_m1>jan(?>uary|\.)?|feb(?>ruary|r?\.?)?|mar(?>ch|\.)?|apr(?>il|\.)?|may|jun(?>e|\.)?|jul(?>y|\.)?|aug(?>ust|\.)?|sep(?>tember|t?\.?)?|oct(?>ober|\.)?|nov(?>ember|\.)?|dec(?>ember|\.)?))|(?:\b(?>(?<a_m2>jan(?>uary|\.)?|feb(?>ruary|r?\.?)?|mar(?>ch|\.)?|apr(?>il|\.)?|may|jun(?>e|\.)?|jul(?>y|\.)?|aug(?>ust|\.)?|sep(?>tember|t?\.?)?|oct(?>ober|\.)?|nov(?>ember|\.)?|dec(?>ember|\.)?)))(?:(?:(?: ?+)(?<a_d2>(?>(?:(?:[23]?1)st|(?:2?2)nd|(?:2?3)rd|(?:[12]?[4-9]|[123]0)th)\b|0[1-9]|[12][0-9]|3[01]|[1-9]|[12][0-9]|3[01]))(?!\d))?))(?:(?:,?+)(?:(?:(?: ?)(?<a_y>(?:1[7-9]|20)\d\d|'?+\d\d))(?!\d))|(?<=\b|\.))/i
Obtained via expressing fr.splayce.rel.matchers.en.Date.ALPHA using PCREFlavor (not yet in the GitHub repository). It will only match when there is a month, expressed in alphabetic form (feb, feb. or february), the ….Date.ALL regex also matching numerical forms like 2/21/2013 is more complex.
Also, this particular regex matches your examples but may still be a bit limited for your needs:
March 9th)2013, jan. 14th