[reportlab-users] Tons of table questions

Robin Becker robin at reportlab.com
Thu Feb 26 07:14:59 EST 2009


Paul Barrass wrote:

> Robin Becker wrote:

>> Dinu there was a long discussion about table times a few years ago. In

>> particular there was a longTable speedup proposed and contributed by

>> Henning von Bargen. That was about 5 years ago. In rl_config.py

>> there's a variable to turn this optimisation on and off for the

>> default Table class. If you set it to one in your local_rl_config then

>> you should get rid of the n**2 behaviour.

> In my experience with long tables, I've found the non-linear behaviour

> both with and without LongTable() - I think the LongTable optimisation

> optimises the calculation of /where/ to split (which can improve

> performance where cell widths/heights are variable) but the non-linear

> behaviour is occurring during the actual splitting, where LongTable is

> identical to Table.

>

......
yes there's bound to be an element of n**2 since the split has to create the new
table from the old by copying.

We currently don't have the concept of a view into the table data.

The original problem which the longTable optimisation addressed was that we
scanned the whole table each time again rather than splitting early.
--
Robin Becker


More information about the reportlab-users mailing list