How big is the problem workbook? Can you tell us its size in MB, the
number of sheets and the size of data matrix on each sheet. Every little
bit of information helps.
Can you obfuscate sensitive data? Without a sample file we can't do much
to help you.
I suspect the slowness may be related to the support for array formulas
which was introduced right after 3.6-FINAL. Does your workbook contain
array formulas? They look something like SUM(C11:C12*D11:D12).
Evaluation of array formulas is not yet fully implemented and it
explains why you got incorrect formula results.
Indirect evaluation of defined names is unlikely to affect performance.
It was an isolated fix in the implementation of INDIRECT.
Try to save the file as .xls and run the same code against HSSF. How
much is it faster?
Regards,
Yegor
On 9/28/2010 10:27 PM, Ryan Skow wrote:
Unfortunately the 3.7b3 is no faster. I can make a Bugzilla entry,
but can't attach the spreadsheet in question since it is
confidential. I've been trying to make test spreadsheets to try to
narrow down what exactly it is having trouble with, but have so far
been unsuccessful. It may have something to do with 'indirect
evaluation of defined names' as 3.6 does not support them hence the
reason I am trying to use the beta. I get the following error under 3.6:
"Indirect evaluation of defined names not supported yet"
I'm not even sure what 'indirect evaluation' is, but 3.7b2 seemed to
at least remove that error message. Although, the values being
calculated in the cells is incorrect which is an entirely different
problem I will look at if I can get past the 60 second load time.
Anyway, just wanted to give a status update and mention the 'indirect
evaluation' issue in case that means something to someone and it sheds
some light on the problem. I'm going to continue trying to make a
test spreadsheet that exhibits the same slowdown from 3.6 to 3.7b3.
Any ideas on what to try would be much appreciated.
Thanks!
On 9/23/2010 4:28 PM, David Fisher wrote:
You might give trunk a try, or wait until tomorrow when it is likely
that POI 3.7beta3 will be released. (Voting is in process, no one has
objected and enough positive votes so far. The link is on the thread
- http://permalink.gmane.org/gmane.comp.jakarta.poi.devel/19169)
If that fails to solve your issue then please submit a bugzilla
entry. If possible a clean file attachment that demonstrates the
trouble would really help.
Dave
On Sep 23, 2010, at 2:16 PM, Ryan Skow wrote:
Using 'new XSSFWorkbook' within POI 3.7 beta 2 is very slow compared
to doing the same with the stable POI 3.6. The xlsx file in
question has many formulas which reference other cells which also
contain formulas. In 3.6, creating a new XSSFWorkbook takes about 10
seconds. In 3.7b2 it takes over 10 minutes to get past the
constructor. The relative time between the two versions to open the
same spreadsheet leads me to believe something is amiss with 3.7b2.
The only question related to 'slowness' in the FAQ is around
logging. As far as I know logging is not enabled with either
configuration I have been using and am specifically specifying the
NullLogger.
Please let me know what additional information would be useful. I
mainly am just curious if anyone else is experiencing the slowdown
or if it is something specific to my spreadsheet or if there is a
quick fix.
Thanks!
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
