PHPExcel very slow - ways to improve?

Question

I am generating reports in .xlsx using PHPExcel. It was okay in the initial testing stages with small data sets (tens of rows, 3 sheets), but now when using it on a real pro

Answer 1

I ran into this issue as well. Thought I'd throw my two cents in since this question gets so many views.

Setting Cell Values

Instead of setting the value for each cell individually, use the fromArray() method. Taken and modified from the wiki.

$arrayData = array(
array(NULL, 2010, 2011, 2012),
array('Q1',   12,   15,   21),
array('Q2',   56,   73,   86),
array('Q3',   52,   61,   69),
array('Q4',   30,   32,    0),
);

$as = $objPHPExcel->getActiveSheet();

$as->fromArray(
    $arrayData,  // The data to set
    NULL,        // Array values with this value will not be set
    'C3'         // Top left coordinate of the worksheet range where
                 //    we want to set these values (default is A1)
);

Styling Cells

Static

It is also quicker to apply the styles for a range, than to set the style for each cell individually (noticing a pattern??).

$default_style = array(
    'font' => array(
        'name' => 'Verdana',
        'color' => array('rgb' => '000000'),
        'size' => 11
    ),
    'alignment' => array(
        'horizontal' => \PHPExcel_Style_Alignment::HORIZONTAL_CENTER,
        'vertical' => \PHPExcel_Style_Alignment::VERTICAL_CENTER
    ),
    'borders' => array(
        'allborders' => array(
            'style' => \PHPExcel_Style_Border::BORDER_THIN,
            'color' => array('rgb' => 'AAAAAA')
        )
    )
);

// Apply default style to whole sheet
$as->getDefaultStyle()->applyFromArray($default_style);

$titles = array(
    'Name',
    'Number',
    'Address',
    'Telephone'
);

$title_style = array(
    'font' => array(
        'bold' => true
    ),
    'fill' => array(
        'type' => \PHPExcel_Style_Fill::FILL_SOLID,
        'startcolor' => array('rgb' => '5CACEE')
    ),
    'alignment' => array(
        'wrap' => true
    )
);

$as->fromArray($titles, null, 'A1'); // Add titles

$last_col = $as->getHighestColumn(); // Get last column, as a letter

// Apply title style to titles
$as->getStyle('A1:'.$last_col.'1')->applyFromArray($title_style);

Dynamically

I use PHPExcel to check the data given in the spreadsheet with the current data in the database. Since each cell is checked individually, I put the styles in an array (null for no style), and used the loop below to get the range of cells to apply the style to.

/*
 * $row is previously set in a loop iterating through each 
 *     row from the DB, which is equal to a spreadsheet row.
 * $styles = array(0 => 'error', 1 => 'error', 2 => null, 3 => 'changed', ...);
 */
$start = $end = $style = null;
foreach ($styles as $col => $s) {
    if (!$style && !$s) continue;
    if ($style === $s) {
        $end = $col;
    } else {
        if ($style) {
            $array = null;
            switch ($style) {
                case 'changed':
                    $array = $this->changed_style;
                    break;
                case 'error':
                    $array = $this->error_style;
                    break;
                case 'ignored':
                    $array = $this->ignored_style;
                    break;
            }
            if ($array) { 
                $start = \PHPExcel_Cell::stringFromColumnIndex($start);
                $end = \PHPExcel_Cell::stringFromColumnIndex($end);
                $as->getStyle($start.$row.':'.$end.$row)->applyFromArray($array);
            }
        }
        $start = $end = $col;
        $style = $s;
    }
} 

Answer 2

I am in no means an expert in using PHPExcel, but the OfficeOpenXML format (the format of *.xlsx files) is itself a group of XML files packed in ZIP archive with *.xlsx extension. If you value your performance and know what kind of data you will be passing, maybe it is a better idea to build own XLSX generator, stripped down to the most important functions, maybe making some calculations on database layer etc. instead of parsing the whole document.

To do it, you can begin with analyzing files generated using smaller data sets (by changing extension from *.xlsx into *.zip, unpacking it and browsing through the contents of the single files). That way you could determine what you really need and generate it yourself (by creating appropriate XML files and packing them into ZIP archive, then renaming to have *.xlsx extension).

There is also specification of OfficeOpenXML, which is large (a couple thousands of pages), thus I do not propose reading it unless you really want to. Creating files to match the way they were generated by PHPExcel should be enough.

The solution mentioned above does not include any PHPExcel-related tips, because I am not an expert in it. I have been previously interested in OOXML standarization process however, and would be happy if knowledge about this standard would help you solve your problem.

Answer 3

Is it populating the worksheet? or saving? that you find too slow?

How are you populating the spreadsheet with the data?

• Using the fromArray() method is more efficient than populating each individual cell, especially if you use the Advanced Value Binder to set cell datatypes automatically.

• If you're setting values for every individual cell in a sheet using

  $objPHPExcel->getActiveSheet()->setCellValue('A1',$x);
  $objPHPExcel->getActiveSheet()->setCellValue('B1',$y);
  

  use

  $sheet = $objPHPExcel->getActiveSheet();
  $sheet->setCellValue('A1',$x);
  $sheet->setCellValue('B1',$y);
  

  so that you're only accessing the getActiveSheet() method once; or take advantage of the fluent interface to set multiple cells with only a single call to $objPHPExcel->getActiveSheet()

  $objPHPExcel->getActiveSheet()->setCellValue('A1',$x)
                                ->setCellValue('B1',$y);
  

You've commented on applying styles to ranges of cells:

• You also have the option to use applyFromArray() to set a whole variety of style settings in one go.
• It's a lot more efficient if you can apply styles to a column or a row rather than simply to a range

If you're using formulae in your workbook, when saving:

• Use

  $objWriter->setPreCalculateFormulas(false)
  

  to disable calculating the formulae within PHPExcel itself.

Those are just a few hints to help boost performance, and there's plenty more suggested in the forum threads. They won't all necessarily help, too much depends on your specific workbook to give any absolutes, but you should be able to improve that slow speed. Even the little notebook that I use for development can write a 3 worksheet, 20 column, 2,000 row Excel 2007 file faster than your production server.

EDIT

If it was possible to simply improve the speed of PHPExcel itself, I'd have done so long ago. As it is, I'm constantly performance testing to see how its speed can be improved. If you want faster speeds than PHPExcel itself can give, then there's a list of alternative libraries here.

Answer 4

In my case I increased performance by changing cache storage method to in memory gzip cache_in_memory_gzip

$cm = \PHPExcel_CachedObjectStorageFactory::cache_in_memory_gzip;
\PHPExcel_Settings::setCacheStorageMethod($cm);

Answer 5

I had the exact same issue. Got a 5000 row, 32 column CSV file that took forever to process. It turns out almost all the time spent "processing" is actually the character encoding which is set to encode everything to UTF8 by default. So if you go into your config\excel.php file and scroll down to encoding, just set it as:

/*
|--------------------------------------------------------------------------
| Import encoding
|--------------------------------------------------------------------------
*/
    'encoding' => array(

        'input'  => '',
        'output' => ''

    ),

With this alone - the above mentioned file takes around 8sec to process. You might want to warn your client to save the CSV correctly though.

Answer 6

One performance tip I hadn't seen before relates to adding worksheets – or more specifically, setting the title of a worksheet. If you add many worksheets, the order of operations can have a huge impact. For the following tests, I used a spreadsheet with 120 populated worksheets and timed how long it took to create another 120 empty worksheets.

First, using the steps shown in the docs:

for ($i = 0; $i < 120; $i++) {
    $sheet = $spreadsheet->createSheet();
    $sheet->setTitle('Sheet Title' . $i);
}
// Time: 12.5605s

Second, alternative method from the docs:

for ($i = 0; $i < 120; $i++) {
    $sheet = new Worksheet($spreadsheet, 'Sheet Title' . $i);
    $spreadsheet->addSheet($sheet);
}
// Time: 0.0266s

Most of the performance gap between the above two methods can be closed by using the second parameter of setTitle (if it is safe to do so in your case; see the docs):

for ($i = 0; $i < 120; $i++) {
    $sheet = $spreadsheet->createSheet();
    $sheet->setTitle('Sheet Title' . $i, false);
}
// Time: 0.5793s