calculateTextWidth throws an error for some fonts

[benlongstaff]

Undefined array key "Widths"
in vendor/smalot/pdfparser/src/Smalot/PdfParser/Font.php:279

Not all fonts have the widths data in the font header e.g. $font->getDetails() returns.

Array
(
    [Name] => WPJPNX+E13BP
    [Type] => Type0
    [Encoding] => Identity-H
    [BaseFont] => WPJPNX+E13BP
    [DescendantFonts] => Array
        (
            [0] => Array
                (
                    [Name] => WPJPNX+E13BP
                    [Type] => CIDFontType2
                    [Encoding] => Ansi
                    [BaseFont] => WPJPNX+E13BP
                    [CIDToGIDMap] => Identity
                    [DW] => 1000
                    [Subtype] => CIDFontType2
                )

        )

    [Subtype] => Type0
    [ToUnicode] => Array
        (
            [Filter] => FlateDecode
            [Length1] => 1024
            [Length] => 303
        )

)

vs the expected

Array
(
    [Name] => MyriadPro-Regular
    [Type] => Type1
    [Encoding] => WinAnsiEncoding
    [BaseFont] => MyriadPro-Regular
    [FirstChar] => 1
    [FontDescriptor] => Array
        (
            [Ascent] => 750
            [CapHeight] => 674
            [Descent] => -250
            [Flags] => 32
            [FontName] => MyriadPro-Regular
            [ItalicAngle] => 0
            [StemV] => 80
            [Type] => FontDescriptor
        )

    [LastChar] => 255
    [Subtype] => Type1
    [Widths] => Array
        (
            [0] => 0
            ...
            [254] => 471
        )

)

[k00ni]

Can you provide us the PDF?

[benlongstaff]

Unfortunately the files are Bank Statements, I will need to find a way to remove the elements with sensitive information.

Is there other information about the font I could provide in the meantime?

[k00ni]

The most helpful would be PHP exploit code which triggers the error. In the following (untested) a rough example. Please have a look.

/*
 * $elements must contain faulty data to trigger the error.
 * $header->getDetails() is used inside "calculateTextWidth".
 * If it doesnt return an array with key "Widths", the error occur.
 *
 * You can build $elements yourself or you place var_dump near
 * https://github.com/smalot/pdfparser/blob/master/src/Smalot/PdfParser/Font.php#L278
 * and use that.
 */
$elements = [
    'Name' => ''...',
    'Type' => '...',
    'Encoding' => '...',
    // 'Widths' => '...'       <=== must be missing to trigger the error
];
$header = new Header($elements);

$font = new Font(new Document(), $header);
$font->calculateTextWidth('', null); // call this to trigger error

[k00ni]

@benlongstaff Ignore my last comment. I realized it is more a basement for a unit test to first trigger the error and after fixing it, make sure it doesn't happen again.

We would need two things fix it:

1. a unit test which triggers the error (see my comment with example code above)
2. and a check inside the function to avoid the error in case no width is given.

Would you take the time and prepare a pull request? I will lead/assist you until its merged.

Does PDF specification allows no Widths on the font? If so a simple check should be fine (and/or setting a default value even). If it doesn't, its more an anomaly. In this case, what is the best way then? Stick with the check?

[GreyWyvern]

Fortunately, the PDF for issue #592 has this font-without-Width problem as well and we already have permission to use it. /samples/bugs/Issue592.pdf

The key thing is, what do we want PdfParser to do in this case? Return zero (0)? Something like (in Font.php):

    /**
     * Calculate text width with data from header 'Widths'. If width of character is not found then character is added to missing array.
     */
    public function calculateTextWidth(string $text, array &$missing = null): ?float
    {
        $index_map = array_flip($this->table);
        $details = $this->getDetails();

        // If 'Widths' is not defined for this font, return 0
        // See: https://github.com/smalot/pdfparser/issues/570
        if (!isset($details['Widths'])) return 0;

        $widths = $details['Widths'];
...

[k00ni]

The key thing is, what do we want PdfParser to do in this case? Return zero (0)?

I suggest -1 or null because it is an invalid width which is easy to check for. Whatever is returned in this case, the function header should be extended to reflect this behavior.

[GreyWyvern]

This function doesn't seem to be used by any other function in PdfParser after running a quick search, so I think returning null, -1 or even false would be okay.

[mbideau-atreal]

I also have the same issue : font with no Widths that generates a PHP Notice and fail to calculate text width.
The test PDF is the same as provided in the issue #629.

The following code, triggers the PHP Notice using the mentioned PDF sample.

<?php

require_once __DIR__.'/pdfparser/alt_autoload.php-dist';

$config = new \Smalot\PdfParser\Config();
$config->setDataTmFontInfoHasToBeIncluded(true);
$parser = new \Smalot\PdfParser\Parser(array(), $config);

$pdf = $parser->parseFile('/tmp/doc.pdf');

$pages = $pdf->getPages();
$lastpage = end($pages);
$data = $lastpage->getDataTm();

echo "Items:".PHP_EOL;
$current_text = null;
foreach($data as $item) {
    if(is_array($item)) {
        $text = $item[1];
        if ($text != $current_text) {
            echo "- '$text'".PHP_EOL;
            $font = $lastpage->getFont($item[2]);
            echo "  font: ".$font->getName()." (".$font->getType().")"." size: ".$item[3].PHP_EOL;
            $missing = array();
            echo "  text width: ".$font->calculateTextWidth($text, $missing)." (missing: ".implode(',', $missing).")".PHP_EOL;
            $current_text = $text;
        }
    }
}

PS: this code needs the fix of the issue #629 in order to detect the font properly

Is there something I can do when generating the PDF to fix this issue in the PDF ? I have (a little) control over the PDF generation.

I am mainly interested in making text width calculation works rather than preventing a PHP Notice.

Thank again for you software and contributors.
Best regards