All,

I'm new to fastbit and investigating if I can use it for a windows 7 64bit
desktop application. For now I'm using ardea to convert the data to the
required binary format. Most of my data are string values which could be
represented/stored by a unique integer value with use of some dictionary. I
understand from the documentation that the fastbit library provides this
functionality for category columns. However I can't get ardea to create the
dictionary files in the output directory for category columns.

- Can ardea be configured to create the dictionary files for category
columns?
- Are category columns represented with strings or integer values in the
binary format (if a dictionary file is not created)?
- What happens if a category value doesn't occur in the user supplied
dictionary or the user supplied dictionary file doesn't exist at all.

- Is it be possible to query the category data using a range of integer
values (that represent the category values)?
The reason that I ask this is that for this application it makes sense to
group the category values together (on many levels for most columns) and I
hope to use an ordered integer representation to improve the performance of
querying those category values (see example below) while not storing
additional grouping information for all category values for all records.
The grouping/levels logic would then be implemented in the application
logic (preferably by using the fastbit dictionary.

Best regards,

Matthieu


Example of grouping of category values:
IntegerRepresentation,Value, Level0, Level1,Level2,etc
1,Jimmy'sBoa,BoaConstrictor,Snake,Reptile
2,Kaa,BoaConstrictor,Snake,Reptile
3,Nagini,Python,Snake,Reptile
4,Winnie-the-Pooh,BrownBear,Bear,Mammal.
5,Baloo,BrownBear,Bear,Mammal

Example of query:
All Snakes would translate to Value>0 and Value <4
instead of Value==Jimmy'sBoa or Value==Kaa or Value==Nagini
_______________________________________________
FastBit-users mailing list
[email protected]
https://hpcrdm.lbl.gov/cgi-bin/mailman/listinfo/fastbit-users

Reply via email to