We used to type hint against IWriter for writer part parent, but this
was not respected in writer part who often coded against the implementation
instead of the interface. With this change we make it a requirement for
writer parts to use their respective writer implementation instead of an
interface. This give legitimate access to all `BaseWriter` methods amongst
other things.
Having default values made it harder for end-user to figure out whether
it the arguement had to be supplied or not. Ommitting the argument may
lead to hard to debug issues, and is overall not a good idea.
Closes#110
In some exotic environnement (IBM AIX) it may happens that `//TRANSLIT` is not supported. In that case iconv would return false. We can easily identify the case and fallback on mbstring instead. That should give correct result.
Closes#135Closes#136
This is a hard limit of the BIFF format that cannot be worked around,
so it is best to throw exceptions than silently discard data or generate
corrupted files.
This drop a lot of non-core features code and delegate their maintainance
to third parties. Also it open the door to any missing implementation
out of the box, such as Redis for the moment.
Finally this consistently enforce a constraint where there can be one and
only one active cell at any point in time in code. This used to be true for
non-default implementation of cache, but it was not true for default
implementation where all cells were kept in-memory and thus were never
detached from their worksheet and thus were all kept functionnal at any
point in time.
This inconsistency of behavior between in-memory and off-memory could
lead to bugs when changing cache system if the end-user code was badly
written. Now end-user will never be able to write buggy code in the first
place, avoiding future headache when introducing caching.
Closes#3
We use to write the wrong property to file, leading to a warning
in PHP 7.1 since the `switch` never matched anything and the variable
was left as a string instead of as an int.
Closes https://github.com/PHPOffice/PHPExcel/issues/1110
Removed "unescape enclosure functionality", since the unescaping is already handled by fgetcsv,
and performing the unescaping again would actually result int the text from the cell being read wrong.
As an example try parsing the folowing CSV:
```
"<img alt="""" src=""http://example.com/image.jpg"" />"
```
With the additional unescaping it would have ended up as:
```
<img alt=" src="http://example.com/image.jpg" />
```
instead of the correct:
```
<img alt="" src="http://example.com/image.jpg" />
```
Fixes https://github.com/PHPOffice/PHPExcel/pull/1171
Settings: deleted libxml_disable_entity_loader() calls since they're not necessary.
Prevent setLibXmlLoaderOptions() and getLibXmlLoaderOptions() to call the libxml_disable_entity_loader() function which alter the global state of libxml.
See the discussion here https://github.com/PHPOffice/PhpSpreadsheet/issues/74
Those were deprecated a long time ago in 59932b0cac
but their replacement seems to have never been added. 5 years later the
best solution is to un-deprecate them and if someday their replacement are
actually added, then we can re-deprecate them properly.
Closes#22
Over the years PclZip became a maintenance problem. It's code base
is old and the official project seems to have died out. ZipArchive
is now more commonly available and offer an easier and more complete
API. So PclZip was dropped in order to focus our efforts on what matters.
Closes#18
We need to make note of two things:
- The old code was trying to clone an objectArray, and hence was performing a shallow copy of memoryDrawing objects (the __clone magic function was not getting invoked).
Instead, if one loops over the objectArray, and then clones the individual memory drawing objects, then the __clone function for the MemoryDrawing object is invoked.
- The __clone function for memory drawing was using the clone keyword which does not deal with circular references (Since memoryDrawing object had references to worksheet object, it was encountering an infinite loop). However, serializing and unserializing objects deals with circular references pretty well.
Fixes#92Closes#106
By design, UTF-8 allows any byte-oriented substring searching algorithm,
since the sequence of bytes for a character cannot occur anywhere else
([source](https://en.wikipedia.org/wiki/UTF-8#Advantages_3)).
So `str_replace()` also works for UTF-8-encoded strings, assuming that
the input strings are valid UTF-8 strings. The previous implementation
of mbStrReplace() did nothing to detect invalid strings.
Also, `str_replace()` does not support [Unicode equivalence](https://en.wikipedia.org/wiki/Unicode_equivalence),
but nor do the other `mb_string` functions, and nor does `=SUBSTITUTE()` in Excel
(tested on Excel for Mac version 15.19.1, Excel 2016 for Windows and LibreOffice 5.1).
Closes#109
MarkBaker
