Dear Arian,

Thanks for your interest in FastBit work.  At this time, 64-bit 
integers are internally handled as doubles in many places!  As you 
have realized already, doubles only give you 48-bit precision while 
64-bit integers give you 64-bit precision.  Clearly there is a 
problem.  However, this is not an issue that we can fix easily without 
major rewrite of the software.  If it is possible, I suggest that you 
break 64-bit integers into two 32-bit ones.

 From our side, we need to look into redesigning the software to get 
rid of this issue in a reasonable way..

John

PS: Your set of tests might also be revealing some other errors in the 
current version of FastBit code, I have not had the time to fully 
explore the issues yet.  Thanks.


On 3/5/2010 5:31 AM, Arian Bär wrote:
> Hi John,
>
> I am currently evaluating Fastbit 1.1.6 for the use in our research
> project in the topic of network monitoring. I encountered a problem a
> problem with the processing of unsigned 64-bit integers. It seams that
> ibis is not supporting them, if the full value range is used. Here is
> a example output where rara can produce a hit and ibis not, given the
> same query. It would be very kind if u can help me with that.
>
> commandline output:
>
> $ ../ardea -d test -m id:i,foo:ul -r 0,18446744073709551615
> /home/abaer/programs/fastbit-ibis1.1.6/examples/.libs/lt-ardea: verbose level > 0
> Will attempt to parse 1 row
>   with the following column names and types
>       id:i,foo:ul
>
>
> $ ../rara test id=0 foo
> SELECT foo FROM test WHERE id=0
> -->  foo (with counts)
> 18446744073709551615, 1
>
> $ ../ibis -d test -q "select foo where id=0"
> doQuery:: estimate(id=0) took 0 CPU seconds, 0.000777245 elapsed
> seconds.  The number of hits is 0
> $ ../ibis -d test -q "select foo where id=0" -v
>
> /home/abaer/programs/fastbit-ibis1.1.6/examples/.libs/lt-ibis: batch
> mode, log level 1
>
> Completed construction of an ibis::part named test
>    1 row and 2 columns
> query[9_n8CJiG-mB----0]::setWhereClause -- where "id=0"
> query[9_n8CJiG-mB----0]::estimate -- time to compute the bounds: 0
> sec(CPU), 0.000341892 sec(elapsed).
> query[9_n8CJiG-mB----0]::estimate -- # of hits for query "id=0" is 0
> doQuery:: estimate(id=0) took 0 CPU seconds, 0.000902891 elapsed
> seconds.  The number of hits is 0
> /home/abaer/programs/fastbit-ibis1.1.6/examples/.libs/lt-ibis -- total
> CPU time 0.01 s, total elapsed time 0.00174713 s
> $ ../ibis -d test -q "select foo where id=0" -v2
>
> /home/abaer/programs/fastbit-ibis1.1.6/examples/.libs/lt-ibis: batch
> mode, log level 1
>
> Completed construction of an ibis::part named test
>    1 row and 2 columns
> query[9_n8CJiG-qJ----0]::setWhereClause -- where "id=0"
> query[9_n8CJiG-qJ----0]::estimate -- time to compute the bounds: 0
> sec(CPU), 0.000295162 sec(elapsed).
> query[9_n8CJiG-qJ----0]::estimate -- # of hits for query "id=0" is 0
> doQuery:: estimate(id=0) took 0 CPU seconds, 0.000799179 elapsed
> seconds.  The number of hits is 0
> /home/abaer/programs/fastbit-ibis1.1.6/examples/.libs/lt-ibis -- total
> CPU time 0 s, total elapsed time 0.00158167 s
>
>
> It would be very kind if u can help me with this.
>
> With kind regards,
> Arian Bär
_______________________________________________
FastBit-users mailing list
[email protected]
https://hpcrdm.lbl.gov/cgi-bin/mailman/listinfo/fastbit-users

Reply via email to