>
>>// almost 10x slower than XSSF!
>>There is something to work on.
>
> I ran the SSPerformaceTest and found the following values:
> HSSF   default  10000x100  5 sec
> SXSSF  default   10000x100  7sec
>
> I did not do the XSSF test because my VMWare doen't have that much heap
> space. But never mind, the HSSF value is the same as in your test and the
> SXSSF value is slightly above, which is 5 times faster then XSSF so that I
> can't reproduce your 302 seconds. Did you use the original SSPerformanceTest
> or a modified version?
>

I will double check. It might be that I ran my benchmarks from my IDE
(IntelliJ Idea).

>>Alex, would you like to continue working on this feature and submit more
> patches?
>
> Absolutely, but keep in mind that for the moment my Excel knowlege is
> minimal. I will look into the two issues Boolean and Formulas and I am
> confident to having that work as expected. Is there a unit test for the Cell
> class? Besides of issues writing the XML, I suspect strongly that the
> current implementation has a number of bugs in the getters/setters. As an
> example consider setting a cell type to string and then setting a numeric
> formua (Formulas are typed in Excel, no?). Is that an error or should the
> type change to numeric then? My feeling is, that it would be more productive
> if you took care of that part of the code. I think that I can manage to get
> the XML right without too much trouble. What do you think?

When you set a formula then the cell's type becomes CELL_TYPE_FORMULA
and Cell.getCachedFormulaResultType() returns CELL_TYPE_NUMERIC or
CELL_TYPE_STRING depending on the formula.

This functionality is well covered in by our unit tests, for example,
look at BaseTestCell in
http://svn.apache.org/repos/asf/poi/trunk/src/testcases/org/apache/poi/ss/usermodel/

I would be good to run SXSSF against the whole test suite. I plan to
do so when I have time.

Yegor

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to