Bug 88632 - For benchmark validation, it would be useful if RAND() optionally would be non-random
Summary: For benchmark validation, it would be useful if RAND() optionally would be no...
Status: RESOLVED FIXED
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: Spreadsheet (show other bugs)
Version: unspecified
Hardware: Other All
: medium normal
Assignee: Not Assigned
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-01-20 16:43 UTC by Tor Lillqvist
Modified: 2015-01-22 13:05 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:


Attachments

Description Tor Lillqvist 2015-01-20 16:43:07 UTC
In some special-purpose Calc use cases, it would be very useful if the random numbers produced by the RAND() function would be the same each time LibreOffice is run on a document. Ditto for RANDBETWEEN().
Comment 1 raal 2015-01-20 19:47:51 UTC
Could you specify this special use cases? I think it's not possible because rand() have to be random. There are not options in rand> http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part2.html#__RefHeading__1018560_715980110
Comment 2 Tor Lillqvist 2015-01-20 21:36:29 UTC
You don't understand.
Comment 3 Tor Lillqvist 2015-01-20 21:45:33 UTC
To be more verbose, of course I know that RAND() should be be (pseudo-)random in all normal use cases. I am talking about a special case that never would be entered by accident by a normal user. We already have the code for this is master, but I think this bug report is necessary to have a rationale for doing it in the libreoffice-4-4 branch, too. See https://gerrit.libreoffice.org/#/c/14053/ .
Comment 4 m.a.riosv 2015-01-20 23:19:57 UTC
Sometimes is difficult understand what is not explained.

Is the special case a secret?
Comment 5 Michael Meeks 2015-01-21 07:06:07 UTC
In benchmarking it is necessary to both be able to measure performance, but also correctness. Clearly if we have a benchmark that includes RAND() - it is then hard to check correctness - without being able to force deterministic behavior on RAND. Clearly that is not a common case - but it is vital for this corner-case. 

We're working on publishing some sensible documents that people can play with too - but that's still in progress.

HTH.
Comment 6 Tor Lillqvist 2015-01-21 08:35:34 UTC
Patch committed to master and 4-4.
Comment 7 m.a.riosv 2015-01-21 21:18:13 UTC
Thanks for detail and of course for the work. I think understandably now, may be it could be of interest for some users.

Truly appreciated to see some focus on performance, IMHO the area more needed of improvement in Calc.
Comment 8 Tor Lillqvist 2015-01-22 07:51:10 UTC
If you have a specific case of a document whose calculations or loading has slowed down from one version of LibreOffice to the next, or is significantly (an order of magnitude) slowed in LibreOffice than in a competing product, please file a detailed bug report with a minimal sample document. Just complaining in general, in a comment on a totally unrelated bug report, is not how to do it.
Comment 9 m.a.riosv 2015-01-22 13:05:28 UTC
(In reply to Tor Lillqvist from comment #8)
> please file a detailed bug report with a minimal sample document. Just
> complaining in general, in a comment on a totally unrelated bug report, is
> not how to do it.
Forgive me if I explain bad. I'm not complaining, just the contrary, thinking it could be a future enhancement.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.