Third party PDF libraries must now be installed via composer and naturally
via composer autoloading mechanism. Because of that it is not necessary
to specify their path on disk. The usage is simplified and it allows us
to include them in our unit tests.
This also means that from now on PhpSpreadsheet must use composer autoloader
mechanism. The internal autoloading implementation was dropped.
The OOXML spec defines the font sizes (the `sz` element) as a double, so the value needs to be wrapped in `PHPExcel_Shared_String::FormatNumber()` to avoid breaking in non-English locales.
`RichText` tries to extract container cell's font via
`Cell->getParent()->getStyle(...)`, which crashes because
`getParent()` does not return a worksheet. Changing it to
`getWorksheet()` fixes the problem.
If $a or $b are not strings, you can get an error that fmod needs param 1 ($a) to be of type double.
MS Excel does not fall over when you insert an empty string in MOD, so I'm guessing PHPExcel should do the same?
For (still) unknown reasons, PHPExcel detects a wrong field type.
If there is no BSE Index, we will fail here and other fields are not read.
So if the BSE Index < 1, continue with the next field.
TODO: Why is there no BSE Index?
Is this a new Office Version?
Password protected field?
More likely : a uncompatible picture
PHP Notice: Undefined offset: -1 in Classes/PHPExcel/Reader/Excel5.php on line 1063
PHP Fatal error: Call to a member function getBlipType() on a non-object in Classes/PHPExcel/Reader/Excel5.php on line 1064
removeColumn doesn't remove all cells, because CacheBase::deleteCacheData only unsets cell data if they are objects.
But cached cell data could be arrays.
I moved the unset from inside the if block (line 150 ff) to after the if block.
We wrapped lines to make easier the source easier to read, and to have more manageable diffs.
This was done with something like:
```sh
pandoc --wrap=auto --atx-headers -f markdown -t markdown-fenced_code_attributes+pipe_tables+raw_html+intraword_underscores`
```
We aim for long pages instead of several small one, to make
`CTRL+F`ing easier and URL less prone to changes in the future.
There is still way to improve it though...
