http://search.cpan.org/~jonatayah/Captcha-AreYouAHuman-0.00003/lib/Captcha/AreYouAHuman.pm

Search “CAPTCHA”.

Let’s say you need to write out some tables to a PDF that you are dynamically building.  You grab the multicell command, everything looks good, and then you get an automatic page break.  Suddenly, your PDF looks like puke on a plate: the cells overwrite each other, rows are on top of rows, rows stretch awkwardly across the page break.

Here’s a way to make MultiCell page breaks look good.  I tried this in TCPDF 5 — the use of the current page function was not working for me.  (You’re not writing in iTextSharp anymore!)

protected function writeMulticells($cell_array) {
    $this->pdf->setAutoPageBreak(false);
    $this->pdf->startTransaction();
    foreach ($cell_array as $row) {
        $this->pdf->MultiCell($row[0], $row[1],
                                    $row[2], $row[3],
                                    $row[4], $row[5],
                                    $row[6]);
    }
    $this->pdf->ln(); 

    if ($this->pdf->getY() > $this->pdf->getPageHeight() - 30) {
        $this->pdf->rollbackTransaction(true);
        $this->pdf->AddPage();
        foreach ($cell_array as $row) {
            $this->pdf->MultiCell($row[0], $row[1], $row[2],
               $row[3], $row[4], $row[5], $row[6]);
        }
        $this->pdf->ln();
    }
    $this->pdf->commitTransaction();
    $this->pdf->setAutoPageBreak(true, 30);
}

(I just happened to have this inside a CodeIgniter model, hence the protected function.)

What we do is the following:

1. Disable auto page breaks.
2. Start a transaction.
3. Write the row.
4. If you would have crossed into the footer, roll back, make a new page, and write the row again.
5. Commit the transaction.
6. Re-enable auto page breaks.

All of the examples I could find suggested using the getPage function, but that didn’t work all of the time in my particular application. Your mileage may vary, of course, but here’s an alternative option.

• Using Curl, I was submitting by POST with an array.  This caused it to use multipart/form-data encoding, which sure looks like it doesn’t work.  To get it to use www-form-urlencode, I had to urlencode the arguments myself and pass the argument string, not the array.
• The ampersand (&) is not a field separator — it’s a stop character.  The spec calls for all key/value pairs to end in an ampersand.  If you pull out any old library that will do the previous one — or, let’s say you had one homebrewed… — you might not have put an ampersand on the end.  Without the ampersand, you’ll get that dreaded (92) error.

Do that, and, like magic, it works.