Write step

Write step can be used to write data back into datapoint and timeseries. Write step should be used carefully as they are modifying data within the time-series database. Therefore they are by default de-activated on experiment mode.

The config of a write step is a datapoints mapping dictionary between :

• “df.column” : "datapoint.hardwareId”
• “df.column” : “template.property”

When using a template, the write block will use the template property and registration to map the datapoint. Write step will using the index timestamp from your data frame, add a transformation block in combination with relative date variables to manipulate the writing time. (e.g. for forecasting)

Here’s an example of write step config using a mapping based on template properties :

{
  "type": "write",
  "config": {
    "datapoints" : {
      "sum_all_columns" : "bearing_output"
    }
  },
  "inputs": [
    { "dataframe" : "sum_dataframe" }
  ]
}

Alternatively, the mapping between your data frame columns and datapoint hardwareId from the platform can be done dynamically based on your logic. For this, the config should contains the key map_property_name pointing to a property name. The property itself must contains the dict mapping between as key (data frame columns) and as values (hardwareId).

Here’s an example of write step config using a dynamic mapping:

{
  "type": "write",
  "config": {
    "map_property_name" : "your_property_name"
  },
  "inputs": [
    { "dataframe" : "sum_dataframe" }
  ]
}

Don’t forget to pass, by a script step or directly when calling your pipeline, a property with the name your_property_name and as value a dict with columns name/hardware id key value pairs.

As a reminder properties are passed as dict between all pipeline steps. It’s initialized at pipeline execution and is accessible with context.properties[key].