Explode UTF8 string regarding to uppercase or numeric characters - php

Question

As this question, I can split strings that includes upper cases like this:
function splitAtUpperCase($string){
return preg_replace('/([a-z0-9])?([A-Z])/','$1 $2',$string);
}
$string = 'setIfUnmodifiedSince';
echo splitAtUpperCase($string);
Output is "set If Unmodified Since"
But I need some modification:
That code snippet doesn't handle the cases, when these characters exist in string: ÇÖĞŞÜİ. I don't want to transliterate the characters. Then I lose meaning of word. I need to use some UTF characters. That code makes "HereÇonThen" to "HereÇon Then"
I also don't want to split uppercase abbreviations. If word is "IKnowYouWillComeASAPHere" I need it to be converted to "I Know You Will Come ASAP Here"
Don't explode if all letters are uppercase. Like "DONTCOMEHERE"
Explode also numeric values. "Before2013ends" to "Before 2013 ends"
Explode if first character is hash key (#).
cases and expected results
"comeHEREtomorrow" => "come HERE tomorrow"
"KissYouTODAY" => "kiss you TODAY"
"comeÜndeHere" => "come Ünde Here"
"NEVERSAYIT" => "NEVERSAYIT"
"2013willCome" => "2013 will Come"
"Before2013ends" => "Before 2013 ends"
"IKnowThat" => "I Know That"
"#whatiknow" => "# whatiknow"
For these cases I use subsequent str_replace operations. I look for a short solution that doesn't make too much for loops to check the words. It would be better to have it as preg_replace or etc. if possible.
Edit: Anyone can try his solution by changing convert function inside this PHP fiddle: http://ideone.com/9gajZ8

Answer 1

/([[:lower:][:digit:]])?([[:upper:]]+)/u should do it.
Here /u is used for Unicode characters. and ([[:upper:]]+) is used for Sequence of upper cased letters.
Note. Case of a letter depends on the character set you are using.

Answer 2

Some notes:
Use Unicode properties to search for upper-case & lower-case letters (and even title-case ones, f.ex. Dž Lj Nj Dz)
comeHEREtomorrow & IKnowThat won't work with one method, until you use some dictionaries to find exact words.
Because if you want to translate comeHEREtomorrow as come HERE tomorrow, IKnowThat will be IK now That (or even IK now T hat);
And if you want to translate IKnowThat as I Know That, comeHEREtomorrow will be come H E R E tomorrow
My solution: http://ideone.com/oALyTo (excludes non-letter & non-number charaters)

Answer 3

Well, I matched all of your test cases, but I still don't think it's a good solution. (One of the few flaws in test driven design).
I took a slightly different approach. Instead of trying to write a regular expression for what the place between a word should look like, I wrote a regular expression that looks for everything that apparently is a word, and then imploded.
function convert($keyword) {
$wResult = preg_match_all('/(^I|[[:upper:]]{2,}|[[:upper:]][[:lower:]]*|[[:lower:]]+|\d+|#)/u', $keyword, $matches);
return implode(' ',$matches[0]);
}
As you can see, this is what I decided qualified as a word:
^I A capital I at the beginning of the string. Break point: Icons.
[[:upper:]]{2,} Consecutive capitals. Break Point: WellIKnowThat
[[:upper:]][[:lower:]]* A single Capital followed by some lower case letters
[[:lower:]]+ A string of lower case letters
\d+ A string of digits
# A literal #
It's not perfect - there're still many breakpoints. You can continue to refine these word definitions, but frankly, there's always going to be an edge case you can't catch. Then you wind up slowly expanding this regular expression until it's totally unmanageable. You could try using a dictionary, but that breaks down eventually, too. What do you do with "whirlwind"? Or "ITan"? Is that "IT an", or "I Tan"? Case in point? Here it is after I tried to catch some of My errors. It's getting so huge, and it's still trivial to come up with strings it breaks on. This function is all about degrees - how much time is it worth spending to teach your algorithm all the funny points of all the world languages?
EDIT: After some work, And deciding that I could be separated out as its own word if and only if it was followed immediately by One Capital letter and one lower case letter, I've updated my attempt at an answer.
function convert($keyword, $debug = false) {
$wResult = preg_match_all('/I(?=[[:upper:]][[:lower:]])|[[:upper:]]{2,}|[[:upper:]][[:lower:]]*|[[:lower:]]+|\d+|#/u', $keyword, $matches);
if($debug){
var_dump($matches);
var_dump($matches[0]);
var_dump(implode(' ',$matches[0]));
}
return implode(' ',$matches[0]);
}
I also added some new test cases:
convert("Icons") = "Icons"
convert("WellIKnowThat") == "Well I Know That"
convert("ITan") == "I Tan"
convert("whirlwind") == "whirlwind"
I think this is about as good as it's going to get today. The final set of "Word Definitions" in order of preference, is:
Upper case I, provided it's followed by an upper case letter and a lower case letter:I(?=[[:upper:]][[:lower:]])
Two or more consecutive upper case letters: [[:upper:]]{2,}
A single uppercase Letter, followed by as many Lower case letters as possible: [[:upper:]][[:lower:]]*
one or more consecutive lower case letters: [[:lower:]]+
One or more consecutive digits: \d+
A literal pound symbol: #
I've added another word definition, a test case, and refined the testing fiddle. The new word definition matches the rule for I, but with A - the only other one letter word in the English Language.

Answer 4

you need Unicode Regex:
\p{Lu} for upercase and \p{Li} for lowercase
Hence, your usage will look like this:
/([\p{Ll}0-9])?([\p{Lu}])/