<prefix>:<type>:<additional_info>
Before you can upload any of your (meta)data files to the platform, you need to tell the platform how your data is structured. For example, you need to tell the platform where the ID is stored, which properties you want to include, the data type of each property, etc. .
This can be done through the web UI, but also by uploading a data properties file which describes your data.
This data properties file is a .csv
file where each row specifies what kind of data a property in your data files contains.
There are 2 kind of data properties file:
A data properties file that describes the structure of index-based files like CSV. A property in such an index-based file is accessed through its index, for example the property at column 3.
A data properties file that describes the structure of name-based files like GeoJSON or Parquet. A property in such a name-based file is accessed through its name, for example the property with name 'speed'.
The syntax used in both kind of files is the same, but the structure is slightly different.
For each property, the data properties file needs to describe the following:
Property identifier: how/where to find the property. This will either be an index or a name.
Display Name (string): the display name used to provide a human-readable name for the property. This name will be shown in the UI.
Type (string with special syntax): the type of the data property. This is used to indicate which property contains the id, location, timestamp, … and to specify whether a property is a numeric property, a category (enum), or text. See the next section for the syntax.
Description (string): a human-understandable description of what the property represents. This is for example used in tooltips in the UI.
Include for analysis (boolean: true or false): only properties that are included for analysis can be used for filtering and styling. If you set this to false for the field that corresponds to the id of your records, your users will not be able to style by the id, filter by the id, or see an overview of the visible asset ids. This can be useful when privacy is a concern.
Accuracy value (number): the analytics engine rounds numeric properties down to the nearest multiple of the accuracy value. For example if the property represents the length of a ship in metres, using a value of 0.1 for the accuracy value will tell the analytics engine to work with values with 1 digit after the decimal point. For example, a length value 2.31 will then be rounded down to 2.3. The larger the accuracy value, the more compression the platform can perform and the better the performance. This is especially important for larger data sets. This setting is only used for numeric properties. It is also ignored for the mandatory properties (id, location and timestamp).
No data value (number, string or empty): some data sets use a special value to indicate that the actual value for a property is unknown or not set.
Minimum valid value (number or empty): when specified for a numeric property, all values below this value will be considered as unknown or not set. This setting is only used for numeric properties. It is also ignored for the mandatory properties (id, location and timestamp).
Maximum valid value (number or empty): when specified for a numeric property, all values above this value will be considered as unknown or not set. This setting is only used for numeric properties. It is also ignored for the mandatory properties (id, location and timestamp).
The data properties .csv
files have the following columns to describe the properties:
Display name |
Type |
Description |
Include |
Accuracy |
No data value |
Min valid value |
Max valid value |
Note that this doesn’t include the property identifier.
That one is implicit.
The first row in the data properties .csv
file describes the first property, the second row the second, etc. .
The following is an example of a data properties file for index-based files.
For readability, it is displayed as a table and not as a .csv
file (but you can download the .csv
version here):
MMSI |
id:long |
unique identifier |
TRUE |
0 |
|||
time |
time:yyyy-MM-dd’T’HH:mm:ss |
time string |
FALSE |
0 |
|||
latitude |
y:float |
y-coordinate |
FALSE |
0 |
|||
longitude |
x:float |
x-coordinate |
FALSE |
0 |
|||
Speed over Ground |
float |
Speed over Ground |
TRUE |
1 |
102.3 |
||
Course over Ground |
float |
Course over Ground |
TRUE |
1 |
360.0 |
||
Heading |
heading:float |
Heading |
TRUE |
1 |
511 |
0 |
360 |
Vessel name |
string |
Name of the vessel |
FALSE |
0 |
|||
IMO |
string |
IMO |
FALSE |
0 |
|||
Callsign |
string |
Callsign |
FALSE |
0 |
|||
Vessel type |
enum |
VesselType |
TRUE |
0 |
|||
Status |
enum |
Status |
TRUE |
0 |
|||
Length |
length:float |
length |
TRUE |
0.1 |
0 |
||
Width |
width:float |
width |
TRUE |
0.1 |
0 |
||
Draft |
float |
Draft |
TRUE |
0.1 |
0 |
||
Cargo |
float |
Cargo |
TRUE |
1 |
In the above example, the line
Heading |
heading:float |
Heading |
TRUE |
1 |
511 |
0 |
360 |
indicates that the corresponding property (e.g. a column in a CSV file)
Should use Heading as display name in the xyzt.ai platform
Contains numbers that are stored as floats and that represent a heading
The description to use for that column is also Heading
The column should be included in the analytics
The accuracy value for this numeric property is 1, all heading values are rounded down to the nearest integer
When the value of the property is 511, it should be interpreted as "no data"
Only values in the interval [0,360] are valid values
and since the property is specified on row 7, the property identifier (=the index) is 7.
The data properties .csv
files have the following columns to describe the properties:
Property identifier |
Display name |
Type |
Description |
Accuracy |
No data value |
Min valid value |
Max valid value |
Note that this doesn’t include the include for analysis flag. That one is implicit. You only specify the properties that you want to include, and leave out all the others.
The following is an example of a data properties file for index-based files.
For readability, it is displayed as a table and not as a .csv
file (but you can download the .csv
version here):
mmsi |
MMSI |
id:long |
unique identifier |
0 |
|||
t_recording |
time |
time:yyyy-MM-dd’T’HH:mm:ss |
time string |
0 |
|||
lat |
latitude |
y:float |
y-coordinate |
0 |
|||
lon |
longitude |
x:float |
x-coordinate |
0 |
|||
speed_over_ground |
Speed over Ground |
float |
Speed over Ground |
1 |
102.3 |
||
course_over_ground |
Course over Ground |
float |
Course over Ground |
1 |
360.0 |
||
heading |
Heading |
heading:float |
Heading |
1 |
511 |
0 |
360 |
v_type |
Vessel type |
enum |
VesselType |
0 |
|||
status |
Status |
enum |
Status |
0 |
|||
length |
Length |
length:float |
length |
0.1 |
0 |
||
width |
Width |
width:float |
width |
0.1 |
0 |
||
draft |
Draft |
float |
Draft |
0.1 |
0 |
||
cargo |
Cargo |
float |
Cargo |
1 |
In the above example, the line
heading |
Heading |
heading:float |
Heading |
1 |
511 |
0 |
360 |
indicates that the corresponding property
Is identified by heading
in your (meta)data files
Should use Heading as display name in the xyzt.ai platform
Contains numbers that are stored as floats and that represent a heading
The description to use for that property is also Heading
The accuracy value for this numeric property is 1, all heading values are rounded down to the nearest integer
When the value of the property is 511, it should be interpreted as "no data"
Only values in the interval [0,360] are valid values
And since the property is specified, it must be included for analytics.
The type property definition consists out of 3 parts:
<prefix>:<type>:<additional_info>
The following prefixes can be used for all types of data sets.
id:
use this prefix for the property representing the identifier of the record.
There can be only one property defined as the id.
Note that point events data sets don’t use an id, so you cannot use this prefix for point events data sets.
The following prefixes can only be used for movement data sets.
x:
use this prefix for the property representing the WGS84 longitude coordinate in degrees form (e.g., 179.0523).
In case your data uses a different coordinate reference system, you can specify this as additional info. There can be only one property defined as the x coordinate.
y:
use this prefix for the property representing the WGS84 latitude coordinate in degrees form (e.g., 50.0234).
In case your data uses a different coordinate reference system, you can specify this as additional info. There can be only one property defined as the y coordinate.
z:
use this prefix for an elevation expressed as meters above sea level.
It is used for 3D coordinates, for example when tracking air tracks.
There can be only one property defined as the z coordinate.
width:
, length:
: use the prefixes for the properties representing the width and length of the asset.
The visual analytics page uses these properties to scale the icons that represent the current position of the asset.
The platform assumes these properties are expressed in meters.
toFront:
, toBack:
, toLeft:
, toRight:
use the prefixes if the relative position of the sensor recording the location of the asset is known.
Sometimes the location-recording sensor is located off-center. For large assets like container ships, this additional offset can introduce an error of several meters on the location of the asset. When the data exposes this offset, the platform can compensate for it.
The following image illustrates how the platform interprets the different options:
The platform assumes these properties are expressed in meters.
heading:
use this prefix for a heading expressed in degrees clockwise from the north.
The prefixes for points data sets are x:
, y:
and z:
, with the same semantics as described
in the movement data sets section.
The following prefixes can only be used for time series data sets.
displayname:
use this prefix for the property that contains a human-readable name of the area or measurement device.
The value of this property will be shown in the UI.
For example, if you have a GeoJSON file defining sensors where each feature or sensor has properties like:
{
"properties": {
"sensor_id": "DGWBETLJGDIUTEDSG",
"sensor_name": "Kitchen sensor"
}
}
While the sensor_id
property is used as id
(and also used in your measurement .csv
files to indicate which sensor took the measurement), you don’t want that "DGWBETLJGDIUTEDSG"
to show in the UI when referring to that sensor.
Instead, you’ll want to use the value of the sensor_name
property, which is done by marking it as the displayname:
.
There can be at most one property defined as the display name. When none of the properties are marked as such, the id property will be used as display name.
The type is one of the following: time
, string
, float
, double
, short
, int
, long
or enum
.
time
: indicates that this property contains the time stamp of the records.
There can only one property defined as the timestamp.
string
: indicates that this property contains text.
float
: indicates that this property contains a single precision (32 bit) floating point number.
double
: indicates that this property contains a double precision (64 bit) floating point number.
short:
: indicates that this property contains a half precision (16 bit) integer number in the range [-32768, 32767].
int
: indicates that this property contains a single precision (32 bit) integer number in the
range [-2147483648, 2147483647].
long
: indicates that this property contains a single precision (64 bit) integer number.
enum
: indicates that this property contains a values that are categories or enumerations.For example
different brands of a car, or types of a vessel.
You can use the following flowchart to assist you in selecting the correct type:
Some properties allow specifying additional info:
When the x:
and y:
coordinates are expressed in a different coordinate reference system, you specify the EPSG code as additional info.
For example x:float:EPSG:4269
and y:float:EPSG:4269
The platform tries to auto-detect the time format for the time
property.
However, if the format is known, it can be specified by defining the pattern as additional info.
For example time:yyyy-MM-dd’T’HH:mm:ss
.
The syntax for the patterns is documented here.