Recorders

What is a Recorder?!

A Recorder saves values after a specific Driver execution, for data sharing and/or analysis.

A recorder is attached to a driver, and collects data from the driver’s owner system.

Introduction

Add a Recorder

Simply use the add_recorder of a Driver method passing a recorder object you want to use (see Available recorders section of this tutorial).

from cosapp.drivers import RunOnce
from cosapp.recorders import DataFrameRecorder
from cosapp.tests.library.systems import Multiply1

m = Multiply1('mult')
run = m.add_driver(RunOnce('run', verbose=True))
run.add_recorder(DataFrameRecorder())

<cosapp.recorders.dataframe_recorder.DataFrameRecorder at 0x7f7b187e3d50>

Default behaviour

By default, units are appended to variable names in headers. Special columns are also included:

• Section: User name reference

• Reference: Internal reference for the record

• Status: Status of the execution for the record

• Error code: Error code for the record

m.run_drivers()
run.recorder.export_data()

	Section	Status	Error code	Reference	K1	family_name	modelings	p_in.x	p_out.x
0			0	run	5.0	Multiply	[(Multiply1, 0.99, 0.1), (Multiply2, 1, 0.3), ...	1.0	5.0

Common options

All recorders share the following options at creation:

• includes: variable name or list of variable names to be recorded. It supports traditional * and ? wild cards characters, as well as evaluable expressions, such as cos(x). By default, this option is '*' (collect everything).

• excludes: variable name or list of variable names not to be recorded (None by default). Also supports * and ? wild cards. excludes shadow includes.

• section: section name of the recording (string).

• hold: Boolean, determining whether or not the recorder should hold previous data between computations (False by default).

• precision: integer specifying the number of digits used when writing floating point numbers (9 by default).

• raw_output: Boolean, False by default. If True, the recorder will not add units in variable headers, and will not add the special columns.

Attributes section, hold and precision can be modified after the recorder creation.

from cosapp.base import System

class Example(System):

    def setup(self):
        self.add_inward('one', 0.)
        self.add_inward('one_two', [0., 1.])
        self.add_inward('two_1', 'test')
        self.add_inward('two_2', 42.)

e = Example('test')
run = e.add_driver(RunOnce('run'))
run.add_recorder(DataFrameRecorder(includes=['one*', 'two_?'], excludes='one_*'))
e.run_drivers()

run.recorder.export_data()  # special method creating a pandas.DataFrame

	Section	Status	Error code	Reference	one	two_1	two_2
0			0	run	0.0	test	42.0

run.recorder.hold = True
e.one = 22.
e.run_drivers()
run.recorder.export_data()

	Section	Status	Error code	Reference	one	two_1	two_2
0			0	run	0.0	test	42.0
1			0	run	22.0	test	42.0

Available recorders

DataFrameRecorder

Records data in a memory buffer, which can be exported as a pandas.DataFrame object after execution, using method export_data().

DSVRecorder

Records data in a Delimited Separated Value (DSV) formatted file. The constructor requires the file path, as a string. It has two specific options:

• delimiter: delimiter symbol between columns (possibilities are ',', ';' or '\t').

• use_buffer: boolean specifiying if the data must be written during the execution (use_buffer=False), or afterwards (use_buffer=True). Default is False.

from cosapp.recorders import DSVRecorder
m = Multiply1('mult')
run = m.add_driver(RunOnce('run', verbose=True))
run.add_recorder(DSVRecorder('test.csv', delimiter=';', precision=2))
m.run_drivers()

%less test.csv