multiple eregi_replace / preg_replace : help

jayant

I have a array $strings
having data in form
$strings[0]="pattern/string 1";
$strings[1]="replace URL string 1";
$strings[2]="pattern/string 2";
$strings[3]="replace URL string 2";
$strings[4]="pattern/string 3";
$strings[5]="replace URL string 3";
$strings[6]="pattern/string 4";
$strings[7]="replace URL string 4";
...
...
...

I have text in variable $text.

What I want is that the perticular string in the text should get replace with a hyperlink.
Suppose we have text as:
qqw dskfh dflkhdsf lkhdf dlsrih ;.ldjf

I want "dskfh dflkhdsf" to point to a URL.

The problem that comes it that I might also want that "dskfh" should link to somewhere else.

If the text is "dskfh dflkhdsf" and I first replace "dskfh" then for "dskfh dflkhdsf" then dskfh will point but in this case I want that since the longer string matches, it should point to the URL i.e "dskfh dflkhdsf". So if I replace first "dskfh dflkhdsf" with its link which will be something as:
"<a href=#>dskfh dflkhdsf</a>"
After replacing "dskfh dflkhdsf" I replace "dskfh" the output will be "<a href=#><a href=#>dskfh</a> dflkhdsf</a>"

In short: if "abc" and "abc def" have to be linked with URL1 and URL2
then in the text:
"kjkhj abc lkjhsdf abc def ljh"
I want the output as
"kjkhj <a href=URL1>abc</a> lkjhsdf <a href=URL2>abc def</a> ljh"
instead of
"kjkhj <a href=URL1>abc</a> lkjhsdf <a href=URL2><a href=URL1>abc</a> def</a> ljh"
or
"kjkhj <a href=URL1>abc</a> lkjhsdf <a href=URL1>abc</a> def ljh"

How can I achieve this ?

Thanks in advance.
Prize: A BMW to the one who answers this correctly 😉

ednark

ok here is one solution... i don't have code but the explaination should suffice...

you want any pattern in your list, unless its already included in a hyperlink

(?!<a.?>).?pattern.*?(?!</a>)

you would need to wrap every pattern in this

only problem is you have to be careful about you order of applying replacements so the longest one goes first, but you should be able to sort your array of replacements

if there might already be hyperlinks, then you can use the same concept with two passes

pass 1)

every time you make a replcement surround it with some delim chars...

in the above case the delim chars are valid <a> tags

only match replcements not surrounded with your delim chars...

pass 2)

after all is said and done... remove all delim chars...

so maybe you should make the delim chars someting low on the asciitable

if this works.. do i really get a car?

jayant

Originally posted by ednark
ok here is one solution...
...................
...............
if this works.. do i really get a car?

The problem is a bit tough than this, I had thought of this one, but one major problem in this one was suppose I have "href" the text to be replaced with URL then it will also get replaced.

yes do you get the car. any BMW available in the market (please exclude vintage BMW's)

jayant

Originally posted by ednark

(?!<a.?>).?pattern.*?(?!</a>)

This might not work in all cases:

suppose we have already replaced pattern1 and pattern2, we are doing pattern3 now and the text looks like:
"dklhdf <a href=url1>pattern1</a> kljhdf sdfh pattern3 sdlf sdfhjb <a href=url2>pattern2</a> ljdf sdkfj "

Notice: when check for pattern3 with (?!<a.?>).?pattern.*?(?!</a>) , it will get satisfied and hence will not get linked. there will be some existing url's in the text also, hence: chances are nothing will get linked.

ednark

ok my first regex was slightly off...
this code works for my test... question is... did i miss a case anywhere

try this out

you might want to take the spaces out of the $preg variable and put each in the actual $strings array ' item ' vs 'item'

<?php

header('Content-type: text/plain');

$strings = array(
    'foo bar',
    'foo_with_bar.com',
    'foo',
    'foo_no_bar.com',
    'ednark',
    'i_want_a_bmw.com'
);

$input_text = <<<END_OF_STRING
    this is a long bit of text with some stuff in it that should be replaced and
    stuff like ednark_foo which should not.<br />
    i will say foo sometimes and foo bar other times.<br />
    if ednark wants a bmw he needs to figure this out.<br />
    sometimes there would already be a link <a href="already_linked_dont_replace.html"> foo </a>.
    also dont replace foo in this link <a href="dont_replace_me_either.html">foo bar </a>
END_OF_STRING;

$correct_output_text = <<<END_OF_STRING
    this is a long bit of text with some stuff in it that should be replaced and
    stuff like ednark_foo which should not.<br />
    i will say <a href="foo_no_bar.com">foo</a> sometimes and <a href="foo_with_bar.com">foo bar</a> other times.<br />
    if <a href="i_want_a_bmw.com">ednark</a> wants a bmw he needs to figure this out.<br />
    sometimes there would already be a link <a href="already_linked_dont_replace.html"> foo </a>.
    also dont replace <a href="foo_no_bar.com">foo</a> in this link <a href="dont_replace_me_either.html">foo bar </a>
END_OF_STRING;


$preg = '/(?!<a.*?>)\sX\s(?!<\/a>)/';

$corrected_text = $input_text;

while ( list( ,$search_for ) = each($strings) ) {
    list( ,$replace_with ) = each($strings);
    echo "Doing replace '$search_for' => '$replace_with'\n";
    $preg_search = str_replace( 'X', $search_for, $preg ); 
    $corrected_text = preg_replace( $preg_search, " <a href=\"$replace_with\">$search_for</a> ", $corrected_text );
}

echo "\n\nSUCCESS:\n\n    ". ($corrected_text==$correct_output_text);
echo "\n\nINPUT STRING:\n\n$input_text";
echo "\n\nTEST RESULT:\n\n$corrected_text";
echo "\n\nCORRECT RESULT:\n\n$correct_output_text";

?>

xblue

Hi Jayant,

quite tricky. This is how I'd do it:

sorting the array so that the longer of two similar keywords is always first (as ednark suggested).

then replacing, but only if the keywords is not already within tags, e.g. using the e modifier.

// dummy data
$str = "kjkhj abc lkjhsdf abc def ljh";
$keywords = array('abc def', 'abc');
$urls = array('page1', 'page2');

// version 1: executing preg_replace in a loop
for ($i=0; $i<count($keywords); $i++){
    $tmp = preg_replace("#(<a\\s[^>]+>.*?$keywords[$i].*?</a>)|($keywords[$i])#se", '"$0"=="$1" ? "$0" : "<a href=\"$urls[$i]\">$0</a>"', $str);
}
echo $tmp;
unset($tmp);
echo '<hr>';

// version 2: preparing pattern and replacement arguments in a loop
// and calling preg_replace only once, passing arrays as arguments
// personally I prefer this one
$patterns = array();
$replacements = array();
for ($i=0; $i<count($keywords); $i++){
    $patterns[] = "#(<a\\s[^>]+>.*?$keywords[$i].*?</a>)|($keywords[$i])#se";
    $replacements[] = '"$0"=="$1" ? "$0" : "<a href=\"'.$urls[$i].'\">$0</a>"';
}
echo preg_replace($patterns, $replacements, $str);

EDIT: just an additional note: if there are no other html tags contained in the string, it should be more efficient to use [^<] instead of .?
EDIT typo
EDIT: oh, and surely you could insert \b around the keywords to match at word boundaries only.

jayant

Originally posted by ednark

<a href="already_linked_dont_replace.html"> foo </a>.

This get linked up again, but if we have <a href="already_linked_dont_replace.html">foo </a>. it doesn't (notice no space between foo and >)
rest seems to work.

I am wondering what if X is already present, wont it create any problem?

jayant

Originally posted by xblue
quite tricky.

I worked on it for quite some time before asking for help. well I am also sure of sorting by length of strings in descending order.

To your code I ran the text of ednark, along with its strings and url's

in version 1 it only replaces ednark, and also replaces ednark in ednark_foo which shouldn't. other strings arent replaced

in version 2 it replaces other strings but replaces ednark of ednark_foo which it shouldnt while doesnt replace ednark which it should. foo also gets replaced in ednark_foo.

I wonder if I had done regex along with caffine probably I could have figured it out myself.

jayant

Well I have decifered one part after much thinking.
After sorting the list in desending order of size,
replace the string with a randomly generated string which possibly (99.99%) won't exist.

Now check whether it should be replaced to the URL or be changed back to original text.

This checking part is gonna be dicey too.

ednark

in version 2 it replaces other strings but replaces ednark of ednark_foo which it shouldnt while doesnt replace ednark which it should. foo also gets replaced in ednark_foo.

hmmm... this does not happen on my computer as i have wrapped the string in \s. my second example works perfect for me.

why don't you post up an example like i did... not fully working... but with a test string or two... and a correct result that you want after parsing... that would make things much easier