PSENSE-COMMON

• Author brad.liang@percusense.com

Common Modules used by internal PercuSense python applications. This project follows semantic versioning (breaking.major.minor release).

Getting Started

General Instructions

Steps to try this out yourself:

1. Install the psense-common library:

From PyPi

$ pip install psense-common

Local Installation

$ git clone git@bitbucket.org:psense/psense-common.git
$ python setup.py install

2. Import the modules you need in your script.

from psense_common import PSenseParser)

3. Access module methods directly.

from psense_common import PSenseParser

filename = '/path/to/data/file'
parser = PSenseParser()
parser.identify_file_source(filename)

if parser.source:
    parser.load_rawfile(filename)
    print(parser.data) # this is of type pandas.DataFrame

AWS

psense_common/psense_aws_itfc.py introduces the class PSenseAWSInterface, which provides an interface to the Amazon DynamoDB datastore. Currently, data is stored on a per-record basis (NoSQL) for Experiment, Sensor, and Event data ("Experiments", "SensorData", and "Calibration" -- not fully supported, will be changed to "EventData" in migration from 0.x.x to 1.x.x).

The class contains query and put operations but does not provide provisions for deleting items from the datastore.

Credentials

At initialization, the class will create an AWS session that is persisted. The session object allows the user to authenticate just a single time (rather than for every communication).

In generating the session, bot3 will attempt to load a profile (if not provided, will try "PShield") that contains the necessary credentials for authentication/authorization to our dynamodb instances. If the profile doesn't exist or cannot be loaded, the class reverts to the environment default AWS credentials in the 'us-east-1' region.

config and credential files for aws are stored in the user's home directory: ~/.aws/. In Windows, the comparable location is C:\Users\[username]\.aws\

Usage

Get Sensor Data

from psense_common import PSenseAWSInterface
import pytz

aws = PSenseAWSInterface(debugmode=True)
aws.set_query_config(req_size=7200, query_count=2)

experiments = ['[my experiment id 1]',
               '[my experiment id 2]']

for expid in experiments:
    if not aws.verif_experiment(expid):
        print('invalid experiment id skipping {}'.format(expid))
        pass

    count, data = aws.get_sensor_data()
    data.index = data.index.tz_localize(pytz.utc).tz_convert(localtz)

    print(data)

Add Experiment and real-time Sensor data

from psense_common import (PSenseAWSInterface, PSenseParser)

# user variables
filename = '[vfp600 gamry file].txt'
expid = 'experimentid'

# initialize classes + helper func
parser = PSenseParser(debugmode=True)
aws = PSenseAWSInterface(debugmode=True)
aws.set_query_config(req_size=7200, query_count=2)

def tail(fin):
    "Listen for new lines added to file."
    while True:
        where = fin.tell()            
        line = fin.readline()
        if not line:
            time.sleep(SLEEP_INTERVAL)
            fin.seek(where)
        else:
            yield line

# identify the type of file we are parsing
if not parser.identify_file_source(filename):
    print('unknown file type. exiting')
    assert False


# confirm that the experiment id is valid
if not aws.verif_experiment(expid):
    print('invalid experiment id skipping {}'.format(expid))
    assert False


# add experiment to database
add_experiment_success, if_fail_reason = aws.add_experiment()
if if_fail_reason == 'error':
    print('Aborted. Error occurred in communication with AWS.')
    assert False
elif if_fail_reason == 'exists':
    print('Experiment already exists in database. Must delete existing experiment sensor data before uploading new values.')


# "tail" the file and send each record to ddb
with open(filename, 'r') as fin:
    # skip to the end of the file before beginning tail
    fin.seek(0, os.SEEK_END)

    for line in tail(fin):
        row = parser.parse_record(line.strip())
        aws.add_sensordata(*row)

Future notes: Sensor and Event data should remain as a blob-store -- we expect schema to change depending on the type of sensor (or event). For example, sensors with 3 working electrodes should contain more properties than sensors with a single signal.

Experiment ID Formatting

All sensor data is associated with a particular experiment ID. Experiment IDs should be unique and follow the PercuSense naming scheme.

Usage

Decoding an experiment:

from psense_common import (psense_format)

experiment = '[my experiment id]'
validate_name = psense_format.psense_exp_title()
validate_name.decode(experiment)

Generating an experiment id through console IO (must provide 3-digit device id):

from psense_common import (psense_format)

device_id = 'P01'
experiment = psense_format.setup_new_experiment(device_id)

Data Parsing

documentation pending

See above sections (General Instructions, AWS) for basic usage.

PercuSense Data Filter

documentation pending

Development

Tests

$ python setup.py test
$ coverage run --source=psense_common/ setup.py test
$ coverage report -m

Publishing

Bitbucket has been configured to run tests and publish directly to pypi after code is merged to master. See bitbucket-pipelines.yml for configuration details.

Manual publishing (not recommended):

$ rm -rf dist
$ python setup.py sdist
$ twine upload dist/*