[Developers] Data on locations

[Bert Bril]

Hi all,


Regularly we need to program things dependent on collecting data on 
locations throughout the survey. Sometimes it's just the locations you 
want, sometimes there are also values attached to each of the positions.

Conceptually, you can see this as a matrix of data: the columns are the 
different types of data, the rows are the 'samples'. For example:

Inl	Crl	Z	Ampl	AcImp
300	555	2.345	92456	3.2109e7
301	878	2.344	88888	1.2567e7

When we started the OpendTect project, we had no special objects for the 
situation. We simply made things like TypeSet<BinIDValue> and passed 
them (un)happily between objects.

If you try to find certain positions in such a set, you're in trouble. 
The search time is O(N), which gives typical O(N^2) bottlenecks. Of 
course this could be improved, so we made the BinIDValueSet (see 
General/binidvalset.h). This covers the data matrix with a BinID and any 
mumber of columns. The data is stored per inline (sorted) and for each 
inline per crossline (sorted). The cost of building the set is a bit 
higher than a set of BinID+Value, but when any search needs to be done, 
this is easily regained by the O(log(N)) search performance.

The BinIDValueSet does not bother about any description of the data 
columns, which is what the PosVecDataSet (General/posvecdataset.h) adds. 
It is essentially a BinIDValueSet + data column definitions. This 
enables not only easier sharing of data, but also more generalized 
store/retrieve.

A couple of years ago, we realized that this is not enough. The 
BinIDValueSet is not very good with 2D positioned data (trace numbers, 
(X,Y)), the positions may not be precise enough, there is no 
grouping/selection, the BinIDValueSet::Pos iterator is not very easy to 
use, and it is not a DataPack. That gave rise to the DataPointSet 
(General/datapointset.h). It is now the vehicle of choice for all kinds 
of location-based data sets. We made tools around it, most notably the 
uiDataPointSet enabling editing, cross-plotting, store/retrieve, and more.

Therefore: if you handle sets of data points with locations, you may 
want to take a look at the DataPointSet and its buddies. As long as the 
amount of points is not really huge, it will probably be a very useful 
and convenient object.


/Bert

-- 
-- Bert Bril / OpendTect developer at dGB
-- mailto:Bert.Bril at opendtect.org , http://opendtect.org