Skip to content

Latest commit

 

History

History
245 lines (183 loc) · 6.57 KB

README.md

File metadata and controls

245 lines (183 loc) · 6.57 KB

Data Binner

The Data Binner is a generic framework that will take input data, then generate unique key names that represent a bin name at a specific level. It is often helpful to sort data into differnt bins in order to build things like metrics or charts with your data.

Data Flow

Data is provided to a Binner along with a bin name and optional other configuration. The Binner then extracts the value of the provided data you would like to bin and generates a number of bin names that can be used for your own purposes. You could for example use the bin names to increment counters with those names.

Converting Data

The Data Binner currently supports json data as an input, but provides an interface to be able to handle other types of input data. See the DataExtractor interface and the JsonDataExtractor implmentation for more information.

Json Data

Our main supported format of input data at the moment is Json data. The JsonDataExtractor supports the ability to extract data from nested properties by specifying the field name in dot form. So if you had data like:

{
    "car": {
        "make": "Subaru",
        "model": "Impreza WRX"
    }
}

And you wanted to create bins for the make of the cars, you would need to specify a dataFieldName of: car.make

More on the dataFieldName property is contained in the Binners section.

Binners

The Data Binner provides a few Binners out of the box. With each Binner, you must provide a binName. If the name of the field you are trying to count in your data is different than the binName then you may provide an optional dataFieldName. The simpliest example would be a Literal Binner like so.

Using a bin name the same as the field name:

{
    "action": "Punch"
}

With the Binner configured like:

Binner binner = new LiteralBinner("action");
List<String> bins = binner.generateBinNames(jsonString);

If you want your bin name to differ from your data's field name:

Binner binner = new LiteralBinner("myOtherName", "action");
List<String> bins = binner.generateBinNames(jsonString);

Literal Binner

The Literal Binner looks at literal (or string) values within your data. For instance, if you had a field named "action" and it had values like "Punch", "Kick", "Run" then the Literal Binner would generate bin names for those literal values. An example:

If you had input data:

{
    "action": "Punch"
}

The Literal Binner configured like:

Binner binner = new LiteralBinner("action");
List<String> bins = binner.generateBinNames(jsonString);

Would generate Bins like:

action.All
action.Punch

Date Binner

The Date Binner takes provided Date values and will break it out into multiple Date Bins at different Date Granularities. An example:

Input data:

{
    "myDate": "2012-04-23T18:25:43.511Z"
}

The Date Binner would be configured like:

Binner binner = new DateBinner("date", "myDate", DateGranularity.MIN);
List<String> bins = binner.generateBinNames(jsonString);

Would generate Bins like:

date.All
date.2012
date.201225
date.20120423
date.2012042306
date.201204230625

Numeric Binner

The Numeric Binner takes provided numeric values and will bin it at different bin levels in powers of 10. Example:

Input data:

{
    "airSpeed": 456.5
}

The Binner would be configured like:

Binner binner = new NumericBinner("airSpeed", 10);
List<String> bins = binner.generateBinNames(jsonString);

Would generate Bins like:

airSpeed.All
airSpeed.456-457
airSpeed.450-460
airSpeed.400-500
airSpeed.0-1000
airSpeed.0-10000
airSpeed.0-100000
airSpeed.0-1000000
airSpeed.0-10000000
airSpeed.0-100000000
airSpeed.0-1000000000

Note the Numeric Binner takes an optional "maxLevel" as a parameter to the constructor. The Bins will grow by powers of 10 until it reaches your specified maxLevel (10^maxLevel).

GeoTile Binner

The Geo Tile Binner will take provided location data and bin it into Geographic Tiles, like those used in a Web Mercator Map Tiling Service. It will generate Bin names with a "zoomLevel"-"xCoord"-"yCoord" system. An example:

Input data:

{
    "point": {
        "x": 88.2,
        "y": -99.3
    }
}

The Binner would be configured like:

Binner binner = new GeoTileBinner("geo", "point", 10, "x", "y");
List<String> bins = binner.generateBinNames(jsonString);

Would generate Bins like:

geo.All
geo.0-0-0
geo.1-0-0
geo.2-0-0
geo.3-1-0
geo.4-3-0
geo.5-7-0
geo.6-14-0
geo.7-28-0
geo.8-57-0
geo.9-114-0

Note the GeoTileBinner takes a number of optional parameters:

  • A maxLevel which is the max zoom level to generate bins down to
  • The name of the Latitude parameter within your data if your data is provided as a map/object like the example above. If you do not provide a name, we assume the name of the property is called lat
  • The name of the Longitude parameter within your data if your data is provided as a map/object like the example above. If you do not provide a name, we assume the name of the property is called lon

The GeoTileBinner will also handle data in the form of an array (or List) with lat, lon like so:

{
    "point": [45.0, 45.0]
}

And a Binner like:

Binner binner = new GeoTileBinner("geo", "point");
List<String> bins = binner.generateBinNames(jsonString);

Merged Binner

The Merged Binner is a special Binner that takes the output of other Binners and generates merged Bin names based on those generated Bin names. This is helpful if you want to be able to know answer a question like: how many Subaru WRXs were made in Dec of 2016. You might look up a bin name of date.201612.model.Impreza WRX to get this count. The Merged Binner took the Date and Literal Bins and combined them for you to create this Bin name.

Input data:

{
    "myDate": "2012-04-23T18:25:43.511Z",
    "car": {
        "make": "Subaru",
        "model": "Impreza WRX"
    }
}

The Binners would be configured like:

Binner dateBinner = new DateBinner("date", "myDate", DateGranularity.MIN);
Binner modelBinner = new LiteralBinner("car.model");

Binner mergedBinner = new MergedBinner(Arrays.asList(dateBinner, modelBinner));
        
List<String> bins = mergedBinner.generateBinNames(data);

Would generate Bins like:

date.All.car.model.All
date.2012.car.model.All
date.201225.car.model.All
date.20120423.car.model.All
date.2012042306.car.model.All
date.201204230625.car.model.All
date.All.car.model.Impreza WRX
date.2012.car.model.Impreza WRX
date.201225.car.model.Impreza WRX
date.20120423.car.model.Impreza WRX
date.2012042306.car.model.Impreza WRX
date.201204230625.car.model.Impreza WRX