[reportlab-users] Patch to support tables with oversize cells

[Robin Becker]

On 24/05/2022 18:53, Lennart Regebro wrote:
> On Tue, May 24, 2022 at 11:28 AM Robin Becker <robin at reportlab.com> wrote:
> 
>> Hi Lennart,
>>
>> thanks for the new patch. I assume everyone will recognize you as genius
>> if it all works :)
>>
> 
> If that was true I would not have to make spreadsheets to figure out the
> logic when I can't get all the cases to work at the same time. :-)
> 
> [image: image.png]
> 
Hi Lennart, I guess genius status is deferred a bit. Whenn I apply the patch I find that we are getting an error related 
to deepcopy of the cell values in the script docs/genAll.py which creates reportlab-userguide.pdf

> (.py310) robin at minikat:~/devel/reportlab/REPOS/reportlab/docs
> $ python genAll.py 
> "/home/robin/devel/reportlab/.py310/bin/python" genuserguide.py 
> Built story contains 1842 flowables...
 > .........
>   File "/home/robin/devel/reportlab/reportlab/platypus/tables.py", line 2032, in split
>     result = self._splitRows(availHeight, doInRowSplit=not self.splitByRow)
>   File "/home/robin/devel/reportlab/reportlab/platypus/tables.py", line 1589, in _splitRows
>     data = deepcopy(self._cellvalues)
 > .......>   File "/home/robin/devel/reportlab/.py310/lib/python3.10/copy.py", line 161, in deepcopy
>     rv = reductor(4)
> TypeError: cannot pickle '_io.BufferedReader' object



I did a quick debug and find that the problem is caused by one of the cell values being/containing an image eg


>> /home/robin/devel/reportlab/.py310/lib/python3.10/copy.py(165)deepcopy()
> -> raise
> (Pdb) p reductor
> <built-in method __reduce_ex__ of _io.BufferedReader object at 0x7fb33f846350>
> (Pdb) p x
> <_io.BufferedReader name='../images/replogo.gif'>

the code that fails is I think a bit like this

I = Image('../images/replogo.gif')
P0 = Paragraph(.......
P = Paragraph(......
data=  [['A',	'B', 'C',	  P0, 'D'],
	['00', '01', '02', [I,P], '04'],
	['10', '11', '12', [P,I], '14'],
	['20', '21', '22',	'23', '24'],
	['30', '31', '32',	'33', '34']]
t=Table(data,style=[......

I suppose we have either have to make ImageReader support a deepcopy or defer / change the deepcopy. In this case I 
don't think the splitInRow is in operation so previously we didn't create a deepcopy. Deferring to the special 
splitInRow case should save us time in the normal case, but obviously we would like to support the special cases.

First off it seems overkill to deepcopy the whole data. Previously the rows went either into one table or its successor 
no copying needed. Now I assume the problem is that we are splitting cells and want to have two separate versions of the 
same cells.

It seems reasonable to me that the cells / styles that get split must be created in two parts one for the top part of 
the table and the second  for the lower table. Since these must be new I don't see where the copy comes in.

I will have to think about the ImageReader problem as it's the most likely problem causer; in addition there are far too 
many options which attempt to save storage reduce open files etc etc etc. I think years ago we worried more about 
storage so defrred reading the whole image into memory. The deferral caused problems with too many open files etc etc. 
It all needs cleaning up.

Of course the real issue here is that we don't require any of reportlab content to be immutable. A lot of things are 
made harder because of that.
-- 
Robin Becker