CSV Files¶
.csv
is currently the only supported file format for data input using trait2d.analysis.Track.from_file()
or trait2d.analysis.ListOfTracks.from_file()
. Other file formats will have to be converted to a pandas.DataFrame
first which can then be imported using trait2d.analysis.Track.from_dataframe()
.
Note
Rudimentary import form native Python dictionaries is supported through trait2d.analysis.Track.from_dict()
. It is the least flexible though, and thus not recommended.
Below, the expected strucure of the .csv
files is described. The same structure also applies to DataFrame
objects used for data import.
Data Structure¶
The .csv
file has to contain the following columns:
x
: x-component of the particle localisationsy
: y-component of the particle localisationst
: time component of the particle localisations
In case there is more than one track stored inside the .csv
file, the following column has to be present as well:
id
: unique ID of the track for each row
The default column names can also be changed using the col_name_*
keyword arguments. See trait2d.analysis.Track.from_file()
and trait2d.analysis.Track.from_dataframe()
for detailed information.
Note
trait2d.analysis.Track.from_dict()
does not allow custom column names.
Units¶
SI units are assumed, that is, metres for x
and y
and seconds for t
. However, units can also be specified using the unit_*
keyword arguments. See trait2d.analysis.Track.from_file()
and trait2d.analysis.Track.from_dataframe()
for detailed information.
Note
trait2d.analysis.Track.from_dict()
does not allow custom units.