SAS Event Stream Processing Python Interface

The ESPPy package enables you to create SAS Event Stream Processing (ESP) models programmatically in Python. Using ESPPy, you can connect to an ESP server and interact with projects and their components as Python objects. These objects include projects, continuous queries, windows, events, loggers, SAS Micro Analytic Service modules, routers, and analytical algorithms.

ESPPy has full integration with Jupyter notebooks including visualizing diagrams of your ESP projects, and support for streaming charts and images. This enables you to easily explore and prototype your ESP projects in a familiar notebook interface.

Installation

To install ESPPy, use pip. This installs ESPPy and the Python package dependencies.

pip install sas-esppy

Additional Requirements

In addition to the Python package dependencies, you also need the graphviz command-line tools to fully take advantage of ESPPy. Download them from http://www.graphviz.org/download/.

Performance Enhancement

ESPPy uses the ws4py websocket Python package. In some cases, you can improve performance greatly by installing the wsaccel package. This may not be available on all platforms though, and is left up to the user to install.

The Basics

To import the ESPPy package, use the same method as with any other Python package.

>>> import esppy

To connect to an ESP server, use the ESP class. In most cases, the only information that is needed is the hostname and port.

>>> esp = esppy.ESP('http://myesp.com:8777')

Getting Information about the Server

After you have connected to the server, you can get information about the server and projects.

>>> esp.server_info
{'analytics-license': True,
 'engine': 'esp',
 'http-admin': 8777,
 'pubsub': 8778,
 'version': 'X.X'}

# Currently no projects are loaded
>>> esp.get_projects()
{}

Loading a Project

To load a project, use the load_project method.

>>> esp.load_project('project.xml')

>>> esp.get_projects()
{'project': Project(name='project')}

To access continous queries and windows within projects, use the queries and windows attributes of the Project and ContinuousQuery objects, respectively.

>>> proj = esp.get_project('project')
>>> proj.queries
{'contquery': ContinuousQuery(name='contquery', project='project')}

>>> proj.queries['contquery'].windows
{'w_data': CopyWindow(name='w_data', continuous_query='contquery', project='project'),
 'w_request': SourceWindow(name='w_request', continuous_query='contquery', project='project'),
 'w_calculate': CalculateWindow(name='w_calculate', continuous_query='contquery', project='project')}

>>> dataw = proj.queries['contquery'].windows['w_data']

As a shortcut, you can drop the queries and windows attribute name. Projects and continuous queries act like dictionaries of those components.

>>> dataw = proj['contquery']['w_data']

Publishing Event Data

To publish events to a window, use the publish_events method. It accepts a file name, file-like object, DataFrame, or a string of CSV, XML, or JSON data.

>>> dataw.publish_events('data.csv')

Monitoring Events

You can subscribe to the events of any window in a project. By default, all event data are cached in the local window object.

>>> dataw.subscribe()
>>> dataw
       time        x        y        z
id                                    
6   0.15979 -2.30180  0.23155  10.6510
7   0.18982 -1.41650  1.18500  11.0730
8   0.22040 -0.27241  2.22010  11.9860
9   0.24976 -0.61292  2.22010  11.9860
10  0.27972  1.33480  4.24950  11.4140
11  0.31802  3.44590  7.58650  12.5990

To limit the number of cached events, use the limit parameter. For example, to only keep the last 20 events, enter the following line:

>>> dataw.subscribe(limit=20)

You can also limit the amount of time that events are collected using the horizon parameter. Use one of the following objects: datetime, date, time, or timedelta.

>>> dataw.subscribe(horizon=datetime.timedelta(hours=1))

You can also perform any DataFrame operation on your ESP windows.

>>> dataw.info()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 2108 entries, 6 to 2113
Data columns (total 4 columns):
time    2108 non-null float64
x       2108 non-null float64
y       2108 non-null float64
z       2108 non-null float64
dtypes: float64(4)
memory usage: 82.3 KB

>>> dataw.describe()
            time          x          y          z
count  20.000000  20.000000  20.000000  20.000000
mean   69.655050  -4.365320   8.589630  -1.675292
std     0.177469   1.832482   2.688911   2.108300
min    69.370000  -7.436700   4.862500  -5.175700
25%    69.512500  -5.911250   7.007675  -3.061150
50%    69.655000  -4.099700   7.722700  -1.702500
75%    69.797500  -2.945400   9.132350  -0.766110
max    69.940000  -1.566300  14.601000   3.214400

Using ESPPy Visualizations with JupyterLab

NOTE: These instructions assume you have Anaconda installed.

To use jupyterlab visualizations with ESPPy (available in version 6.2 or higher), perform the following steps:

1. Create a new Anaconda environment. For this example, the environment is called esp.

    $ conda create -n esp python=3.X

2. Activate the new environment.

$ conda activate esp

3. Install the following packages:

$ pip install jupyter
$ pip install jupyterlab
$ pip install matplotlib
$ pip install ipympl
$ pip install pandas
$ pip install requests
$ pip install image
$ pip install ws4py
$ pip install plotly
$ pip install ipyleaflet
$ pip install graphviz

4. Install the following Jupyterlab extensions:

$ jupyter labextension install @jupyter-widgets/jupyterlab-manager
$ jupyter labextension install plotlywidget
$ jupyter labextension install jupyter-leaflet

5. Install the following packages (WINDOWS ONLY):

$ conda install -c conda-forge python-graphviz

6. Create and change to a working directory.

$ cd $HOME
$ mkdir esppy
$ cd esppy

7. Install ESPPy.

pip install sas-esppy

8. Create a notebooks directory to store your notebooks.

$ mkdir notebooks

9. Start the Jupyterlab server. Select an available port. For this example, port 35000 was selected.

$ jupyter lab --port 35000

After you complete these steps, you can use the latest ESP graphics in your Jupyter notebooks.

Documentation

To view the full API documentation for ESPPy, see https://sassoftware.github.io/python-esppy/.