Recording incoming keywords from google searches.

[erbio]

I am trying to record incoming keywords from google searches.

So far I have this code (Im new)

<?php
echo $_SERVER['HTTP_REFERER'];
?>

For example, this is where I would grab the keyword from -

http://search.yahoo.com/search?p=KEY...=UTF-8&fr=moz2

Can someone provide me with the code to do this?

[kows]

I'm sure you could do this with regular expressions, but I'm guessing you have no idea how they work. instead, you can do something simple like this to ignore the un-needed URL stuffs:

<pre>
<?php
  /*
   * grab our referer and see if we need to parse it
   * (we don't want to waste time parsing it if it's empty)
   */ 
  $r = $_SERVER['HTTP_REFERER'];

  if($r != ""){
    $p = parse_url($r);
    /*
     * make all parameter values lowercase for easy comparing
     */
    foreach($p as $key => $value)
      $p[$key] = strtolower($value);

    /*
     * see if they're using a search engine we recognize
     */
    if($p['host'] == "www.google.com" && $p['path'] == "/search" ||
       $p['host'] == "search.yahoo.com" && substr($p['path'], 0, 7) == "/search"){

      /*
       * determine which host we have, so that we can
       * grab the correct parameter of the query string
       */
      switch($p['host']){
        case "www.google.com": $k = "q"; break;
        case "search.yahoo.com": $k = "p"; break;
      }
      /*
       * loop through each parameter in the query string
       * to find the one we want (defined as $k above)
       */
      $params = explode('&', $p['query']);
      $i = 0;
      while($i < count($params)){
        $v = split('=', $params[$i]);
        if($v[0] == $k){
          /*
           * this is the parameter we wanted, let's decode it
           * and break out of this loop
           */
          $keywords = urldecode($v[1]);
          break;
        }
        $i++;
      }
    }
  }
  if(isset($keywords)){
    echo "We've determined you've reached this page via a search engine!\n";
    echo 'The keywords you were searching for were: "<strong>' . $keywords . '</strong>"';
  }else
    echo "You didn't reach this page via a search engine!";
?>
</pre>

I made referers from yahoo and google both work in that example. if you want to add any more, you'll have to add them yourself.

edit: revised code just a little.

[dclamp]

I'm sure you could do this with regular expressions, but I'm guessing you have no idea how they work. instead, you can do something simple like this to ignore the un-needed URL stuffs:

<pre>
<?php
  /*
   * grab our referer and see if we need to parse it
   * (we don't want to waste time parsing it if it's empty)
   */ 
  $r = $_SERVER['HTTP_REFERER'];

  if($r != ""){
    $p = parse_url($r);
    /*
     * make all parameter values lowercase for easy comparing
     */
    foreach($p as $key => $value)
      $p[$key] = strtolower($value);

    /*
     * see if they're using a search engine we recognize
     */
    if($p['host'] == "www.google.com" && $p['path'] == "/search" ||
       $p['host'] == "search.yahoo.com" && substr($p['path'], 0, 7) == "/search"){

      /*
       * determine which host we have, so that we can
       * grab the correct parameter of the query string
       */
      switch($p['host']){
        case "www.google.com": $k = "q"; break;
        case "search.yahoo.com": $k = "p"; break;
      }
      /*
       * loop through each parameter in the query string
       * to find the one we want (defined as $k above)
       */
      $params = explode('&', $p['query']);
      $i = 0;
      while($i < count($params)){
        $v = split('=', $params[$i]);
        if($v[0] == $k){
          /*
           * this is the parameter we wanted, let's decode it
           * and break out of this loop
           */
          $keywords = urldecode($v[1]);
          break;
        }
        $i++;
      }
    }
  }
  if(isset($keywords)){
    echo "We've determined you've reached this page via a search engine!\n";
    echo 'The keywords you were searching for were: "<strong>' . $keywords . '</strong>"';
//RIGHT HERE 
  }else {
    echo "You didn't reach this page via a search engine!";
?>
</pre>

I made referers from yahoo and google both work in that example. if you want to add any more, you'll have to add them yourself.

edit: revised code just a little.

you forgot a open bracket on the last else. i added it in the quote

[erbio]

Nice! Now if I could just record the keywords to either a flat file or database and display them within a page that would be so awesome.

[dclamp]

$sql = "INSERT INTO `searches` SET date='".date()."', keywords='".$keywords."'";
$query = mysql_query($sql) or DIE(mysql_error()); 

That will work with Kows' code.

[kows]

you forgot a open bracket on the last else. i added it in the quote

umm, no -- I didn't miss anything. the change you made in that quote will actually cause a fatal error because I don't use any curly brackets on that else. that echo line terminates the else statement, because it's the first statement after the "else." if I were to put multiple lines after that else, though, it will always run those lines unless the statements were enclosed in curly brackets.

for example, this works fine (of course, the lines can't all just be comments):

if(statement){
    //multiple lines
    //multiple lines
    //multiple lines
    //multiple lines
  }else
    //one line 

while what you changed wouldn't:

if(statement){
    //multiple lines
    //multiple lines
    //multiple lines
    //multiple lines
  }else{
    //one line 

but, if you did add an extra curly bracket at the very bottom, it would run fine.. but it's a waste of 2 characters if you're not doing anything else with that statement.

[dclamp]

umm, no -- I didn't miss anything. the change you made in that quote will actually cause a fatal error because I don't use any curly brackets on that else. that echo line terminates the else statement, because it's the first statement after the "else." if I were to put multiple lines after that else, though, it will always run those lines unless the statements were enclosed in curly brackets.

for example, this works fine (of course, the lines can't all just be comments):

if(statement){
    //multiple lines
    //multiple lines
    //multiple lines
    //multiple lines
  }else
    //one line 

while what you changed wouldn't:

if(statement){
    //multiple lines
    //multiple lines
    //multiple lines
    //multiple lines
  }else{
    //one line 

but, if you did add an extra curly bracket at the very bottom, it would run fine.. but it's a waste of 2 characters if you're not doing anything else with that statement.

so you only put a { if you plan to put a } at the end?

if(statement){
    //multiple lines
    //multiple lines
    //multiple lines
    //multiple lines
  }else{
    //one line
  } 

[kows]

you only need to use curly brackets on a statement if it has multiple lines.

examples:

//this will run and do what is expected
while($array = mysql_fetch_array($query))
  print_r($array);

//this will also run, but print_r($array) will only be fired once.
//this will only execute print_r on the LAST value found from $query
//because it is NOT a part of the while() statement.
//$i will be equal to the number of results found
while($array = mysql_fetch_array($query))
  $i++
  print_r($array);

//instead of the above, you would need to use the following for
//expected results:
while($array = mysql_fetch_array($query)){
  $i++;
  print_r($array);
}

//you can also nest statements like so:
if($variable == true)
  foreach($array as $key => $value)
    echo $key . ' => ' . $value . "\n";

//or, like this:
if($variable == false)
  foreach($array as $key => $value){
    $i++;
    echo $key . ' => ' . $value;
  } 

when you exclude the curly brackets, all you need to worry about is a terminating line (the terminating line is a valid-syntax statement that ends with a terminator (semi-colon -> (";"))). in the case of if() statements, you need to have a terminating line for each statement, like so (and, you can also add multi-lined statements at will by using curly brackets, also shown below):

if($variable == true)
  echo "variable is true";
elseif($variable == false && $variable2 == false)
  echo "variable two is true, but variable is false";
elseif($variable == false && $variable2 == true){
  //this is a multi-line statement
  echo "variable two is true, but variable and variable three are false";
  $variable3 = false;
}else
  echo "blah blah"; 

[visualAd]

While it is acceptable to omit curly brackets for a single statement construct, it is always a good idea to put them in as it improves the readability of the code and decreases the chances of introducing parse errors should another line be added.

As for as PHP is concerned it is not a waste as prior to parsing all white space is stripped from the source code.

[CornedBee]

As for as PHP is concerned it is not a waste as prior to parsing all white space is stripped from the source code.

Conceptually true, if technically incorrect.

[visualAd]

Conceptually true, if technically incorrect.

What happens? Is it just ignored? The interpreter changes the source code into a type of byte code. This is what is executed.

[CornedBee]

Reading source code is typically a two-stage process. First comes the scanner, which converts the text into a series of tokens. A token is the smallest syntactic piece of a language. It can be annotated. For example, this code:

if($i == 10) { echo "Hello"; }

would get scanned into this token stream (single characters are often represented by the character code):
IF >> '(' >> VAR("i") >> EQ >> NUM(10) >> ')' >> '{' >> ECHO >> STRING("Hello") >> ';' >> '}'[/code]
The second step is the parser, which knows the syntactic tree structure of the language and fits the tokens it gets into the tree. The result is the parse tree. It would perhaps parse the tokens into this:

     block_if
   expr         block
  eq_expr  '{'   statements   '}'
expr EQ expr        statement
VAR     NUM        echo_statement
                ECHO    expr     ';'
                        STRING

The next step would be to convert this to an abstract syntax tree, but that's already beside the point. What is important is that the scanner simply doesn't include whitespace in the tokens it sends. So ignoring whitespace is part of the process, and not a preprocessing step. (That would be extremely inefficient.)

PHP uses the T_ prefix for its token names, by the way. So if you ever encounter another "Unexpected T_STRING.", you know where it comes from. It's the parser parsing, for example, its echo_statement, and after the expression that is output, it expects a semicolon, but the scanner gives it a T_STRING.

[visualAd]

Thinking of it logically that is what would happen.

Point being - use as much white space as you like the only effect will be visually and the file size.