Trajectory Databases - Mobility Data - page 57

3.4.1 Loading Data

First, we show how the raw trajectory data from a CSV text file Traj.csv can
be imported to a SECONDO database. 6 We assume the text file has the schema

(Id: int, Line: int, Up: bool, Time: instant, PosX: real,
PosY: real)

whichwould correspond to rawdata observed for the Trains relation introduced
in Section 3.2 .

let TrainsRaw = [const rel(tuple([Id: int, Line: int, Up: bool,
Time: instant, PosX: real, PosY: real])) value ()]
csvimport['Traj.csv', 0, "", ","]
projectextend[Id, Line, Up, Time; Pos: makepoint(.PosX, .PosY)]
consume;

This creates a relation

TrainsRaw(Id: int, Line: int, Up: bool, Time: instant, Pos: point)

where attribute Pos contains the position data as a point (easting, northing). In
the following, we briefly investigate two different ways to represent trajectories
more effectively according to the data model of Section 3.2 within SECONDO:
the compact representation and the unit representation .

3.4.2 Compact Representation

In TrainsRaw the information on a vehicle is distributed among many tuples.
Using the model of spatio-temporal data types (Section 3.2 ), we now express the
same data in a relation with only a single tuple per vehicle. The data type mpoint
is used to capture the temporal development of attribute Pos . We achieve this
by grouping TrainsRaw by Id and applying the approximate operator to each
group. Using Time as the least significant sorting criterion prior to grouping
guarantees that the positions for each train enter the approximate operator in
increasing temporal order:

let Trains = TrainsRaw feed
sortby[Id, Line, Up, Time]
groupby[Id, Line, Up; Trip: group feed approximate[Time, Pos] ]
consume;

The result is the relation Trains with the schema shown earlier. This is what
we call the compact representation of moving object data. It is easy to apply

6

It is also possible to import NMEA recordings using an operator mneaimport.