Automancer docs


Version: 1.0


This module adds support for data recording.


Data recording can be performed using the record: state attribute. Multiple fields, corresponding to variables that you wish to record, can be defined. The resulting data is stored to a data object.

    - name: Temperature
      value: %{{ devices.TempSensor.readout }}
      unit: 5 degC
    - name: Pressure
      value: %{{ devices.PressureSensor.readout }}
      dtype: <f4
      unit: hPa
  save: data.xlsx
  - ...
  - ...

The following options are available for each field:

  • fields (list, required) – The list of fields to record:
    • name (string) – The field's name.
    • value (dynamic expression, required) – The field's value, to be evaluated every time one of the dependency variables in the expression changes.
    • unit (quantity) – The unit in which the value should be printed. For example, the value 1 GB with unit 200 MB would be recorded as 5. Only supported if value is scalar. Defaults to the unit of value.
    • dtype (string) – The field's data type, as defined in numpy. For scalar values, the byte order must be explicitly defined and the native byte order (described as =) may not be used. Defaults to the smallest data type which can cover the resulting value (e.g. adding an i4 an u8 will result in an i8), encoded as little-endian.
    • disconnected_value (value) – The value to write when one of the dependency variables is unavailable. Defaults to np.nan for floating-point values, 0 for other scalar values and False for boolean values.
    • frequency (quantity) – A higher bound for the frequency at which to record this field.
  • format (string) – The format used to save records (see below). Required if their is no output file path or if its extension is ambiguous.
  • save (file object, required) – The file object in which to save records.
  • when_paused (boolean) – Whether to record when paused. Defaults to false.

A warning will be emitted if any of the recorded values is overflowing and cannot be stored with the provided data type. The written value is undefined behavior in this case.

Choosing an output format

The following output formats are supported.

npy.npyNumpy array file, can be loaded with np.load()
npz.npzDeflateZIP file containing a Numpy array file called arr_0.npy, can be loaded with np.load()
xlsx.xlsxExcel file
csv.csvCSV file
json.jsonJSON file
arraynumpy.ndarray object
dataframepandas.DataFrame object

Certain formats support streaming, that is, the possibility for data to be stored gradually rather than all at once. There are mainly two benefits to this approach:

  • Storing data gradually requires less memory as newly-obtained chunks of data are stored immediately rather than kept in memory. Without streaming, the program can be killed by the OS if it uses too much memory.
  • If the program crashes, you will be able to recover most of the data as it already written in persistent storage. Without streaming, all of the data is lost.

Recording the status of multiple devices

When using floating-point number data types, the disconnection of a device can be indicated by storing NaN instead of the recorded value. For other data types, there is no sentinel value to accomplish this goal. We can use the disconnected_value attribute to create our own sentinel value, such as the largest possible value, e.g. 232 - 1 for a 4 bytes unsigned integer, written as u4. When this is not an option, the only solution left is to use a boolean field in addition to the value field:

  - name: Temperature
    value: %{{ devices.TempSensor.readout }}
  - name: Temperature connected
    value: %{{ devices.TempSensor.readout.connected }}

The <node>.connected value is a boolean which become false when a device is disconnected, indicating that the first field has invalid data during that time. When dealing with more than one field, we can group their respective <node>.connected value into one or more bytes.

  - name: Temperature 1
    value: %{{ devices.TempSensor1.readout }}
  - name: Temperature 2
    value: %{{ devices.TempSensor2.readout }}
  - name: Temperatures connected
    dtype: <u1
    value: %{{ sum((node << index for index, node in enumerate([devices.TempSensor1.readout.connected, devices.TempSensor2.readout.connected]))) }}

Bit 0 of the third field now corresponds to the connection status of TempSensor1 and bit 1 to that of TempSensor2.