Plugin Structure

• Create datasource_XXX.py file under system/plugin, where XXX is the name of the datasource.
• Derive a plugin class DataSource_XXX from DataSource in datasource.py.
• Implement:
  • __init__()
  • get_channels()
  • get_timeseries()
  • get_object()
  • get_blob() (usually empty)
  • get_dataframe() (optional)
• Datasource class provides methods that can be used by plugin:
  • resample()

The minimal / empty class will be:

from datasource import DataSource

class Datasource_XXX(Datasource):
  def __init__(self, project_config, config):
    super.__init__(project_config, config)

  def get_channels(self):
    return []
    
  def get_timeseries(self, channels, length, to, resampling=None, reducer='last'):
    return {}

  def get_object(self, channels, length, to):
    return {}
    
  def get_blob(self, channel, params, output):
    return None

Method Definitions

User Function __init___

  def __init__(self, project_config, config):
    super.__init__(project_config, config)
    ...

User Function get_channels()

• return a list of channels

  def get_channels(self):
    result = []
    ...
    return result

• Return in JSON or string: [ {"name": NAME, "type": TYPE, ...}, ... ]
• type is one of: timeseries, histogram, table, and tree

Optional User Function get_timeseries()

• return time-series data

  def get_timeseries(self, channels, length, to, resampling=None, reducer='last'):
    result = {}
    ...
    return result

• Return in JSON or string, format as defined in the DataModel.md document
• If resampling is not None, apply resampling.
  • If resampling <= 0, just align data points among channels with inferring the intervals from data.
  • reducer can be:
    • first: value of the first non-NaN data point
    • last: value of the last non-NaN data point
    • mean: mean of non-NaN data point values
    • median: median of non-NaN data point values
    • sum: sum of non-NaN data point values
    • count: number of non-NaN data point values
    • std: standard deviation of non-NaN data point values
    • min: minimum data point value
    • max: maximum data point value
  • If resampling is not supported by the data source, use self.resample().

Optional User Function get_object()

• return object time-series, or single object

  def get_object(self, channels, length, to):
    result = {}
    ...
    return result

• Return in JSON or string, format as defined in the DataModel.md document

Optional User Function get_blob()

• Fill blob content

  def get_blob(self, channel, params, output):
    ...
    return mime_type

• Fill the blob content into “output” and return the mime_type

Optional User Function get_dataframe()

• return time-aligned time-series data table
• implement this only if datasource can do this efficiently

def get_dataframe(self, channels, length, to, resampling=None, reducer='last', timezone='local'):
    result = []
    ...
    return result

• The parameters are same as get_timeseries()
  
• Return in JSON or CSV text

The JSON format is

[
  ["DateTime", "TimeStamp", CH0, CH1, ... ],
  RAW[0],
  RAW[1],
  ...
]

where RAW[k] is:

[ DATETIME[k], TIMESTAMP[k], X0[k], X1[k], ... ]

Utility Function for Users resample()

def resample(self, set_of_timeseries, length, to, interval, reducer):
    return RESAMPLED_TIME_SERIES

This will be used in user’s get_timeseries(), if the data source does not efficiently support resampling, typically like:

class DataSource_XXX(DataSource):
    def get_timeseries(self, channels, length, to, resampling=None, reducer='last'):
        result = {}
        ...

        if resampling is None:
            return result
            
        return self.resample(result, length, to, resampling, reducer)