An OpenModelica Python Interface and its use in PySimulator

Abstract

How can Python users be empowered with the robust simulation, compilation and scripting abilities of a non-proprietary object-oriented, equation based modeling language such as Modelica? The immediate objective of this work is to develop an application programming interface for the OpenModelica modeling and simulation environment that would bridge the gap between the two agile programming languages Python and Modelica.
The Python interface to OpenModelica – OMPython, is both a tool and a functional library that allows Python users to realize the full capabilities of OpenModelica's scripting and simulation environment requiring minimal setup actions. OMPython is designed to combine both the simulation and model building processes. Thus domain experts (people writing the models) and computational engineers (people writing the solver code) can work on one unified tool that is industrially viable for optimization of Modelica models, while offering a flexible platform for algorithm development and research.

Full text

An OpenModelica Python Interface and its use in PySimulator
Anand Kalaiarasi Ganeson1, Peter Fritzson1, Olena Rogovchenko1, Adeel Asghar1, Martin Sjölund1
Andreas Pfeiffer2
1PELAB – Programming Environment Lab, Dept. Computer Science
Linköping University, SE-581 83 Linköping, Sweden
2Institute of System Dynamics and Control, German Aerospace Center DLR, Oberpfaffenhofen
1ganan642@student.liu.se, {peter.fritzson, olena.rogovchenko, adeel.asghar, martin.sjolund}@liu.se
2Andreas.Pfeiffer@dlr.de
Abstract
How can Python users be empowered with the robust
simulation, compilation and scripting abilities of a non-
proprietary object-oriented, equation based modeling
language such as Modelica? The immediate objective
of this work is to develop an application programming
interface for the OpenModelica modeling and simula-
tion environment that would bridge the gap between the
two agile programming languages Python and Modeli-
ca.
The Python interface to OpenModelica – OMPy-
thon, is both a tool and a functional library that allows
Python users to realize the full capabilities of
OpenModelica's scripting and simulation environment
requiring minimal setup actions. OMPython is designed
to combine both the simulation and model building
processes. Thus domain experts (people writing the
models) and computational engineers (people writing
the solver code) can work on one unified tool that is
industrially viable for optimization of Modelica mod-
els, while offering a flexible platform for algorithm
development and research.
Keywords: Python, OpenModelica, OMPython, Python,
simulation, modeling, Modelica, Python simulator.
1 Introduction
Necessity is the mother of all inventions. Often in sci-
ence and engineering, the insufficiency of available
tools for researchers and developers creates difficulties
in exploring and investigating a certain subject. This
creates incentives to develop new infrastructures and
tools to fill the void. The goal behind the creation of the
Python interface to OpenModelica is to create a free,
open source, highly portable, Python based interactive
session handler for Modelica scripting and modeling,
thus catering to the needs of the Python user communi-
ty.
OMPython – the Python interface to OpenModelica
is developed in Python using tool communication based
on OmniORB and OmniORBpy - high performance
CORBA ORBs for Python. It provides seamless sup-
port to the Modelica Standard Library and the Modelica
Language Specification [3] supported by OpenModeli-
ca [2].
OMPython provides user-friendly features such as:
 Interactive session handling, parsing, interpretation
of commands and Modelica expressions for evalua-
tion, simulation, plotting, etc.
 Creating models, using pre-defined models, making
component interfaces and annotations.
 Interface to all OpenModelica API calls.
 Optimized result parser that gives access to every
element of the OpenModelica Compiler's (OMC)
output.
 Helper functions to allow manipulation of nested
dictionary data types.
 Easy access to the Modelica Standard library and
calling of OpenModelica commands.
 Provides an extensible, deployable and distributable
unit for developers.
Since OMPython is designed to function like a library,
it can be used from within any Python application that
requires the OpenModelica services. OMPython uses
the CORBA implementation of OmniORB and Om-
niORBpy to communicate with the OpenModelica
compiler.
2 Using OMPython
This section describes how to use OMPython and also
demonstrates its use in creating a simple Modelica
model callable from the Python interpreter. It also pre-
sents the two modes of operation specifically designed
for testing OpenModelica commands and using the
OMPython API as a Python library [1].
DOI Proceedings of the 9th International Modelica Conference 537
10.3384/ecp12076537 September 3-5, 2012, Munich, Germany












2.1 Installing OMPython
The two requirements for the operation of the API are
installations of OpenModelica 1.8.1 and Python 2.6.
Since OMPython is supplied together with the
OpenModelica installer, the standard source distribu-
tion of the API can be used to install it to the third party
libraries of the installed Python version. Building and
installing the module, for example in the Windows sys-
tems, is as simple as running one line of command
from the terminal.
python setup.py install
Now OMPython can be imported into any Python ap-
plication.
2.2 Executing OMPython
The API can be used in two modes, Test and Library,
each designed for a specific purpose.
2.2.1 Test
Like any new tool, it is important to give its users the
freedom to easily explore its capabilities, try its features
and possibly suggest new improvements.
For this purpose, the API can be executed in the test
mode by executing the run() method of the OMPython
module. This mode allows users to interactively send
OpenModelica commands to OMC via the CORBA
interface. The Python types of the OpenModelica out-
put are returned to the user. To illustrate this, in Figure
1 a few operations are presented from the Python ter-
minal.
Figure 1. OMPython executing OpenModelica commands
in the Test mode.
Creating new models in the text based Python terminal
is rather straightforward using OMPython. Figure 2
illustrates this and shows how a model can be saved
with a simple command.
Figure 2. Creating and saving a simple HelloWorld model
file using OMPython.
2.2.2 Library
Once modelers are familiar with the interface they
know what type of responses can be expected and can
use the module as a library to programmatically design,
simulate, plot, and do more with the models.
This can be done by executing the execute()
method of the OMPython module. The execute method
forms the essence of the OMPython API. It encapsu-
lates the OMC operations, CORBA functionalities,
parses the results to native Python data types and ex-
poses the API as a simple string processing method.
Each instance of the execute method returns a result
that the modeler can make use of. Additionally, com-
plicated data structures such as deeply nested dictionar-
ies are constructed, strictly typed, and are made availa-
ble to the user using this method.
The Code Listing 1 shown below provides a simple
Python script that uses OMPython as a library to per-
form a few tasks like loading Modelica libraries to
simulating pre-defined Modelica models. Figure 3 de-
picts the output of the program generated by OMPy-
thon on a standard Python terminal.
Code Listing 1
import OMPython
OMPython.execute("loadFile(\"c:/OpenModeli
ca1.8.1/testmodels/BouncingBall.mo\")")
result=OMPython.execute("simulate(Bouncing
Ball, stopTime=2, method=\'Euler\')")
print result
OMPython.execute("quit()")
An OpenModelica Python Interface and its use in PySimulator

538 Proceedings of the 9th International Modelica Conference DOI
September 3-5, 2012, Munich Germany 10.3384/ecp12076537


Figure 3. OMPython executing the Python script shown
above.
3 Deploying OMPython in
PySimulator
PySimulator is a Python-based Simulation and Analysis
tool that is developed by the German Aerospace Center
(DLR) in Germany. The tool uses plugins for simula-
tors based on Dymola [10], FMUs [11], and OpenMod-
elica [2]. It also provides analysis tools for some appli-
cations particularly in physics and engineering.
This section shows the integration of the new
OpenModelica simulator plugin for PySimulator using
OMPython.
3.1 The OpenModelica Plugin
The plugin for the OpenModelica simulator integrates
easily and well into the PySimulator package by using
the OMPython library. PySimulator's template for the
plugins provides convenient methods to implement
simulation routines, parameter settings, retrieve and use
simulation variables and more. Figure 4 shows a part of
the development package of PySimulator that includes
the OpenModelica plugin.
Figure 4. OpenModelica plugin using OMPython within
PySimulator.
The OpenModelica plugin defines and uses some fea-
tures of PySimulator for performing simulations, read-
ing result files, and displaying variables etc. The
plugins use PySimulator's plugin templates; this allows
other simulation packages to be integrated easily.
The deployment of the OpenModelica plugin within
the PySimulator project allows the project to benefit
from the full scripting capabilities of the latest
OpenModelica API.
3.2 Loading a Modelica Model
The integration of the OMPython module within the
OpenModelica plugin for PySimulator makes it possi-
ble for the modeler to quickly load Modelica files such
as models (.mo) or load a simulated model's executable
file.
The user can open these files from the menu bar by
selecting File > Open Model > OpenModelica.
In this introductory example we will use a pre-
defined model named Influenza to demonstrate the
use of OMPython in PySimulator. Figure 5 depicts the
graphical user interface of PySimulator when opening a
model file. Once the model file is selected, the model is
loaded into the variables browser and is ready to be
configured for simulations.
Session 5A: Simulation Tools
DOI Proceedings of the 9th International Modelica Conference 539
10.3384/ecp12076537 September 3-5, 2012, Munich, Germany

Figure 5. Loading Modelica models or model executables
in PySimulator
3.3 Using the OpenModelica plugin
The loaded Modelica model can be simulated from Py-
Simulator using the default simulation options or by
setting the simulation options before simulating from
the Integrator Control dialog box. The OpenModelica
plugin defines the simulation routine for the Modelica
models by using the execute method of the OMPython
API.
Figure 6 shows how the simulation options can be
set using PySimulator's Integrator control feature.
Figure 6. Preparing the simulation settings using the
Integrator Control.
3.4 Simulating the model
The initial simulation parameters and settings are pro-
vided as inputs to the OMC via the front-end of Py-
Simulator. The Run button of the Integrator control
triggers the simulate command of the OMC with the
supplied simulation options. The simulate command
has the following parameters,
 Simulation Interval
o Start Time
o Stop Time
 Algorithm
 Error Tolerance
 Step size
The user has the option to choose from a range of Nu-
merical integration algorithms from the Algorithm se-
lection box. The Integrator control dialog box also fil-
ters some parameters that are not available for some
integration solvers by disabling the field; avoiding error
and providing more accuracy in the results.
The Variables browser builds a tree structure of the
instance variables and highlights time-continuous vari-
ables in blue. The user can select these variables and
plot them in the Plot window by checking the check
box near the highlighted variables.
Figure 7 illustrates the Variables browser that al-
lows users to access the variables after the Influenza
model has been simulated with some simulation pa-
rameters set.
Figure 7. Variables browser of the simulated model.
3.5 Plotting variables from the simulated models
The Plot window of the PySimulator GUI provides ad-
ditional user interface controls for comparing different
An OpenModelica Python Interface and its use in PySimulator

540 Proceedings of the 9th International Modelica Conference DOI
September 3-5, 2012, Munich Germany 10.3384/ecp12076537

plots side-by-side, adding and removing plots and also
to save the plots.
Figure 8 shows the plotted variables in the plot win-
dow and the list of simulation variables in the Variables
browser along with the variables selected for plotting.
Figure 8. Plotted variables using PySimulator.
3.6 Using Simulated results
It is desirable to avoid simulating the model again eve-
ry time the user needs the simulation results. It is in-
stead preferable to use an existing simulation result file
for the calculations, this saves resources and time. Py-
Simulator supports opening the OpenModelica simula-
tion result files (.mat) and the model's executable file to
build the variable tree in the variables browser. The
user can then adjust some parameters from the variable
tree or the Integrator control to achieve the desired re-
sults.
4 The OMPython API
The Python interface to OpenModelica addresses its
functional requirements through the implementation of
two interrelated modules, OMPython and OMParser
[1]. This section introduces the two modules and
demonstrates their functionalities with some examples.
The following Figure 9 illustrates the functions of
the OMPython API with its components.
Figure 9. Functions of the OMPython API
4. 1 OMPython module
The OMPython module is the main interfacing
module of the OMPython API which is responsible
for providing the API as a tool and a Python library.
The following are its components:
4.1.1 Interactive Session handler
Each instance of the module creates an interactive
session between the user and the OMC. The session
handler uses the CORBA Interoperable Object Ref-
erence (IOR) file to maintain the user's session activ-
Session 5A: Simulation Tools
DOI Proceedings of the 9th International Modelica Conference 541
10.3384/ecp12076537 September 3-5, 2012, Munich, Germany

ities and log files for standard output and errors. The
session is closed down when the user issues the
quit() command to the OMC. This also removes
the temporary IOR file from the user's machine. The
log files facilitate the user with some straight for-
ward debugging and trace-backing purposes.
4.1.2 CORBA Communication
OMPython uses the Client-Server architecture of the
CORBA mechanism to interact with the OMC. OM-
Python implements the client side of the architec-
ture.
4.1.3 Modes of Operation
The module defines two modes of operation, each
designed for specific purposes.
 Test
 Library
The Test mode allows users to test OMPython while
the Library mode gives the user the ability to use the
results of OMPython.
4.1.4 Using the interface definition
The vital link between the client and the server pro-
cesses in this distributed implementation is the Inter-
face Definition Language (IDL) file. OMC defines
the omc_communication.idl file that it uses to
implement the Remote Procedure Calls (RPCs),
OMPython mirrors this IDL file to establish the RPC
from the client machine.
4.1.5 Get/Set helper functions
Due to the nature of the complicated string outputs
generated by the OMC such as Component Annota-
tions, the parser module of the OMPython module
generates nested dictionaries. Deeply nested diction-
aries in Python require cumbersome operations to
retrieve and set values inside dictionaries at various
levels. To simplify the multiple steps necessary to
perform a get or set operation within a dictionary,
OMPython defines the dot-notation get/set methods.
Figure 10 shows how the user can get and set the
values of any nested dictionary data type.
Figure 10. Get/Set helper function
4.1.6 Universal Typecaster
Since the variables in Python are dynamically typed,
the interpretation of the data types needs to be strict-
ly controlled during runtime. For this purpose, the
OMPython module defines a universal typecasting
function that typecasts the data to the correct types
before building the results.
4.1.7 Imports OMParser
Although the OMC outputs the results to the OMPy-
thon module via its CORBA interface, the results are
still in the String-to-String CORBA output format
which cannot be used intelligibly. So the OMPython
module uses its own built-in parser module the OM-
Parser to generate appropriate data structures for the
OMC retrieved results.
4.2 OMParser module
Since the results of the OMC are retrieved in a
String format over CORBA, some data treatment
must be done to ensure that the results are usable
correctly in Python.
The OMParser module is designed to do the fol-
lowing,
 Analyze the result string for categorical data.
 Group each category under a category name
 Typecast the data within these categories
 Build suitable data structure to hold these data so
that the results are easily accessible.
4.2.1 Understanding the Parsed output
Each command in OpenModelica produces a result
that can be categorized according to the statistics of
the pattern of data presented in the text. Grammar
based parsers were found to be tedious to use be-
An OpenModelica Python Interface and its use in PySimulator

542 Proceedings of the 9th International Modelica Conference DOI
September 3-5, 2012, Munich Germany 10.3384/ecp12076537

cause of the complexity of the patterns of data. This
is also the case because the OpenModelica imple-
mentation has two types of APIs. One is typed,
which could use grammar and the other is untyped,
which cannot.
OMParser follows a few simple rules to parse the
OMC output:
 Result strings that do not contain a pair of curly
braces "{}" are simply typecasted to their respec-
tive types.
For example:
>>getVectorizationLimit()
20
>>getNthInheritedClass(Modelica.Electr
ical.Analog.Basic.Resistor,1)
Modelica.Electrical.Analog.Interfaces.
OnePort
 Result strings that include one or more pairs of
curly braces "{}" are categorized for making dic-
tionary types.
For example:
>>getClassNames()
{'SET1':{'Set1': ['ModelicaServices',
'Modelica']}}
 Data contained within double quotes " " are for-
matted to string types; removing the escape se-
quences in-order to keep the semantics.
For example:
>>getModelicaPath()
"C:/OpenModelica1.8.0/lib/omlibrary"
4.2.2 The Dictionary data type in Python
Dictionaries are useful as they allow to group data
with different data types under one root dictionary
name. Dictionaries in Python are indexed by keys
unlike sequences, which are indexed by a range of
numbers.
It is best to think of dictionaries as an unordered
set of key:value pairs, with the requirement that the
keys are always unique. The common operation on
dictionaries is to store a value associate with a key
and retrieve the value using the key. This provides
us the flexibility of creating keys at runtime and ac-
cessing these values using their keys later. All data
within the dictionary are stored in a named diction-
ary. An empty dictionary is represented by a pair of
braces {}.
In the result returned by the OMC, the compli-
cated result strings are usually the ones found within
the curly braces. In order to make a meaningful cat-
egorization of the data within these brackets and to
avoid the potential complexities linked to creating
dynamic variables, we introduce the following nota-
tions that are used within the dictionaries to catego-
rize the OMC results,
 SET
 Set
 Subset
 Element
 Results
 Values
In this section, to explain these categories, we use
the parsed output of OMPython obtained using the
Test mode.
4.2.3 SET
A SET (note the capital letters) is used to group data
that belong to the first set of balanced curly brackets.
According to the needed semantics of the results, a
SET can contain Sets, Subsets, Elements, Values and
Results.
A SET can also be empty, denoted by {}. The
SETs are named with an increasing index starting
from 1 (one). This feature was planned to eliminate
the need for dynamic variable creation and having
duplicate Keys. The SET belongs within the diction-
ary called "result".
For example:
>>strtok("abcbdef","b")
{'SET1': {'Values': ['"a","c","def"']}}
The command strtok tokenizes the string
"abcbdef" at every occurrence of b and produces a
SET with values "a", "c", "def". Each value of
the SET is then usable in Python.
4.2.4 Set
A set is used to group all data within a SET that is
enclosed within a pair of balanced {}s. A Set can
contain only Values and Elements. A set can also be
empty, it can be depicted as {{}}, the outer brackets
compose the SET, the inner brackets are the Set
within the SET.
4.2.5 Subset
A Subset is a two-level deep set that is found within
a SET. A subset can contain multiple Sets within its
enclosure.
For example:
{SET1 {Subset1{Set1},{Set2},{Set3}}}
Session 5A: Simulation Tools
DOI Proceedings of the 9th International Modelica Conference 543
10.3384/ecp12076537 September 3-5, 2012, Munich, Germany

4.2.6 Element
Elements are the data which are grouped within a
pair of Parentheses (). As observed from the OMC
result strings, elements have an element name that
describes the data within them, so elements can be
grouped by their names.
In some cases such as when using the untyped
OpenModelica API calls, element structures do not
have a name, in these cases the data contained with-
in the parenthesis is parsed into outputs generated by
the typed API calls, such as set, values, etc. Also, in
some cases many elements have the same names, so
they are indexed by increasing numbers starting
from 1 (one). Elements have the special property of
having one or more Sets and Subsets within them.
However, they are still enclosed within the SET.
For example:
>>getClassAttributes(test.mymodel)
{'SET1': {'Elements': {'rec1':
{'Properties': {'Results': {'comment':
None, 'restriction': 'MODEL',
'startLine': 1, 'partial': False,
'name': '"mymodel"', 'encapsulated':
False, 'startColumn': 14, 'readonly':
'"writable"', 'endColumn': 69,
'file': '"<interactive>"', 'endLine': 1,
'final': False}}}}}}
In this example,the result contains a SET with an
Element named rec1 which has Properties which
are Results (see section 4.2.7) of the element.
4.2.7 Results
Data that is related by the assignment operator "=",
within the SETs are denoted as Results. These as-
signments cannot be assigned to their actual values
unless they are related by a Name = Value relation-
ship. So, they form the sub-dictionary called Results
within the Element (for example). These values can
then be related and stored using the key:value pair
relationship.
For example:
>>getClassAttributes(test.mymodel)
{'SET1':{'Elements':{'rec1':
{'Properties': {'Results':{'comment':
None, 'restriction': 'MODEL',
'startLine': 1, 'partial': False,
'name': '"mymodel"', 'encapsulated':
False, 'startColumn':14, 'readonly':
'"writable"', 'endColumn': 69, 'file':
'"<interactive>"', 'endLine': 1,
'final': False}}}}}}
4.2.8 Values
Data within any or all of SETs, Sets, Elements and
Subsets that are not assignments and separated by
commas are grouped together into a list called "Val-
ues". The Values list may also contain empty dic-
tionaries, due to Python's representation of a null
string "" as {} - an empty dictionary. Although a
null string is still a null value, sometimes it is possi-
ble to observe data grouped into Values to look like
Sets within the Values list.
For example:
>>getNthConnection(Modelica.Electrical.A
nalog.Examples.ChuaCircuit,2)
{'SET1': {'Set1': ['G.n', 'Nr.p', {}]}}
4.2.9 The Simulation results
The simulate() command produces output that has
no SET or Set data in it. Instead, for the sake of sim-
plicity, the result contains two dictionaries namely,
SimulationResults and SimulationOptions.
For example:
>>simulate(BouncingBall)
{'SimulationOptions': {'options': "''",
'storeInTemp': False, 'cflags': "''",
'simflags': "''", 'variableFilter':
"'.*'", 'noClean': False,
'outputFormat': "'mat'", 'method':
"'dassl'",'measureTime':False,
'stopTime':1.0, 'startTime': 0.0,
'numberOfIntervals': 500, 'tolerance':
1e-
06,'fileNamePrefix':"'BouncingBall'"},'S
imulationResults':{'timeCompile':4.75231
650258347,'timeBackend':0.01602630977192
6,
'messages':None,'timeFrontend':1.4200466
8806536,'timeSimulation':0.1197039958177
84,'timeTemplates':0.0230460728977474,'t
imeSimCode':0.0139967955849597,'timeTota
l':6.3452533928534,'resultFile':'"C:/Use
rs/ganan642/BouncingBall_res.mat"'}}
4.2.10 The Record types
Some commands produce result strings with Record
constructs, these data are categorized for making
dictionaries too. To keep the uniformity and simplic-
ity, the data of Record types are grouped into the
dictionary RecordResults.
For example:
>>checkSettings()
{'RecordResults': {'RTLIBS': '" -static-
libgcc -luuid -lole32 -lws2_32"', 'OMC_F
An OpenModelica Python Interface and its use in PySimulator

544 Proceedings of the 9th International Modelica Conference DOI
September 3-5, 2012, Munich Germany 10.3384/ecp12076537

OUND': True, 'MODELICAUSERCFLAGS': None,
'C_COMPILER_RESPONDING': False, 'OPENMO
DELICAHOME': '"C:/OpenModelica1.8.1/"',
'CREATE_FILE_WORKS': False, 'SYSTEM_INFO
':None, 'CONFIGURE_CMDLINE': '"Manually
created Makefiles for OMDev',
'RecordName':
'OpenModelica.Scripting.CheckSettingsRes
ult','OMC_PATH':'"C:/OpenModelica1.8.1//
bin/omc.exe"','WORKING_DIRECTORY':'"C:/U
sers/ganan642"', 'REMOVE_FILE_WORKS':
True, 'OS':
'"Windows_NT"','OPENMODELICALIBRARY':'"C
:/OpenModelica1.8.1/lib/omlibrary"','C_C
OMPILER': '"gcc"'}}
5 OMPython Implementation
The implementation of the OMPython API relies on
the Client–Server architecture of CORBA to com-
municate to the OMC [2]. OMPython acts as the
client that requests the services of OMC and OMC
behaves like the server and replies to the Python
module using the OmniORB and OmniORBpy –
Object Request Brokers (ORBs) of CORBA as the
communication platform.
This section briefly describes how the API uses
CORBA and its other features to achieve its re-
quirements.
5.1 The OMC CORBA interface
The OpenModelica Complier – OMC can be in-
voked using two methods:
 Executed at the operating system level, like a
program.
 Invoked as a server from a client application us-
ing a CORBA client-server interface.
OMPython uses the second method to start OMC
since this allows the API to interactively query the
compiler/interpreter for its services.
5.2 OMC Client Server architecture
Figure 11 gives an overview of the OpenModelica
client server architecture. OMPython plays the role
of the client in this architecture. It sends queries and
receives replies from the OMC via the CORBA in-
terface. The messages and expressions from the
CORBA interface are processed in two groups. The
first group consists of the commands which are
evaluated by the Ceval module and the second
group contains the expressions that are handled by
the Interactive module.
Figure 11. Client-Server of OpenModelica with some
interactive tool interfaces
Messages in the CORBA interface are classified into
two groups. The first group consists of the user
commands or expressions; these are evaluated by the
Ceval module. The second group contains the decla-
ration of variables, classes, assignments, etc. The
client-server API calls are processed by the Inter-
active module.
5.3 Using OMC through CORBA
The OMC process can be invoked from CORBA by
executing the OMC executable file using special
parameters passed to it. The default location of the
OMC executable file is in the $OPENMODELICA-
HOME/bin directory. OMPython invokes OMC with
some special flags +d=interactiveCorba
+c=random_string which instructs OMC to start
and enable the interactive CORBA communication
and also use a timestamp to name the CORBA In-
teroperable Object Reference (IOR) file that will be
created. The timestamp is needed to differentiate the
different instances of OMC that have been started by
different client processes simultaneously.
The default location where the IOR file is created
is in the temp directory. Normally, when OMC is
started with the +d=interactiveCorba flag, it will
create a file named openmodelica.objid. On
Windows (for example), if the +c flag was given,
the file name is suffixed with the random string to
avoid name conflicts between the simultaneously
running OMC server processes. This file contains
the CORBA IOR.
5.4 Using the CORBA IOR file
The IOR file contains the CORBA object reference
in string format. The CORBA object is created by
reading the strings written inside the IOR file.
Session 5A: Simulation Tools
DOI Proceedings of the 9th International Modelica Conference 545
10.3384/ecp12076537 September 3-5, 2012, Munich, Germany

6 Measurements
In this section, we present some performance meas-
urements for the OMPython API.
The measurements shown are based on the re-
sponse time of the Python interpreter/compiler that
performs the various functions of establishing the
CORBA communication, sending commands to the
OMC, receiving CORBA outputs, parsing the
CORBA outputs and finally displaying the results to
the user.
Figure 12 illustrates a simple script that simulates
a Modelica model and plots a variable using the Plot
generated by OpenModelica. It also shows the re-
ceived response times of each command that was
executed to perform the simulation. Table 1 and Ta-
ble 2 show the time statistics collected from five
unique runs of two simple scripts using the OMPy-
thon API. The time is measured in Seconds. Figure
13 and Figure 14 illustrate the overhead between the
average output and the unparsed output's response
times.
These measurements aim to give an idea about the
overhead of the OMPython API in addition to the
CORBA overhead that is needed for OMC commu-
nication.
Figure 12. Measuring response times of simulations
for the BouncingBall model.
Command
Average re-
sponse time
(s)
Average un-
parsed re-
sponse time
(s)
load-
File("c:/Ope
nModeli-
ca1.8.1/mode
ls/BouncingB
all.mo")
0.09223065
0.0421344389
simu-
late(Bouncin
gBall)
2.60921512
1.8922307169
plot(h)
0.03251472
0.0183359414
Table 1. Response time comparisons for loading,
simulating and plotting variables using OMPython.
Figure 13. Measuring response time for Simulations in
OMPython
Command
Average re-
sponse time
(s)
Average un-
parsed re-
sponse time (s)
getVersion()
0.0680588293
0.0590995445
loadMod-
el(Modelica)
5.971103887
4.4708573210
getElemen-
tsInfo(Modeli
ca.Electrical
.Analog.Basic
.Resistor)
0.0264064349
0.0190346404
getClass-
Names()
0.3907942649
0.2707218157
getTempDirec-
toryPath()
0.0244359882
0.0193691690
getSettings()
0.0327650196
0.0234227783
Table 2. Measuring response times of some
OpenModelica commands in OMPython
An OpenModelica Python Interface and its use in PySimulator

546 Proceedings of the 9th International Modelica Conference DOI
September 3-5, 2012, Munich Germany 10.3384/ecp12076537

Figure 14. Measuring response times of some
OpenModelica commands in OMPython
7 Related Work
Some Simulation packages are available for Python
but these packages do not implement an equation-
based solving system. Also, they do not provide a
Modelica based modeling and simulation environ-
ment, but rather present their custom model types.
 PySCeS – The Python Simulator for Cellular
Systems. It uses the model description language
to define its models. Supports solvers like LSO-
DA, sections for non-linear root finding algo-
rithms, Metabolic control analysis, Mat-
plotlib/Gnuplot plotting interfaces, etc. It is re-
leased under a new BSD style license and is open
source software [4].
 SimPy – Simulation in Python, is an object-
oriented, process-based discrete-event simulation
language for Python. It is released under the
GNU Lesser GPL (LGPL) license version 2.1. It
features data collection capabilities, GUI and
plotting packages. It provides the modeler with
the active and passive components of a simula-
tion model and monitor variables for gathering
statistics [5]
 JModelica.org [12], MWORKS [13], and Amesim
[14] are other Modelica tools providing a Python
scripting API.
8 Conclusion
OMPython is a versatile Python library for
OpenModelica that can be used by engineers, scien-
tists, researchers and interested architects to explore
and develop Modelica based modeling and simula-
tion efforts. It is free, open source and is distributed
with the OpenModelica installation which gives the
user the potential to use the full collection of Model-
ica libraries that can assist in performing complex
simulations and analyses.
The OMPython API places minimal require-
ments on the user while offering an industry viable
standard modeling and simulation environment.
We suggest some future work that can be done to
enrich the usage of the OMPython API. The API can
be expanded to provide access to the GUI based fea-
tures of other OpenModelica tools such as OMEdit.
User interfaces can be easily built on top of OMPy-
thon to implement additional graphic features. Fur-
ther interesting efforts can be made if the OpenMod-
elica API can be designed to expose its commands
as interface definitions in the
omc_communication.idl file.
9 Acknowledgments
This work has been supported by Serc, by the Swe-
dish Strategic Research Foundation in the EDOp and
HIPo projects and Vinnova in the RTSIM and
ITEA2 OPENPROD projects. The Open Source
Modelica Consortium supports the OpenModelica
work.
References
[1] Anand Kalaiarasi Ganeson. Design and Im-
plementation of a User Friendly OpenModeli-
ca – Python interface, Master thesis LIU-
IDA/LITH-EX-A12/037SE, Linköping Uni-
versity, Sweden, 2012
[2] Open Source Modelica Consortium.
OpenModelica System Documentation Version
1.8.1, April 2012.
http://www.openmodelica.org
[3] Modelica Association. The Modelica Lan-
guage Specification Version 3.2, March 24th
2010. http://www.modelica.org. Modelica As-
sociation. Modelica Standard Library 3.1.
Aug. 2009. http://www.modelica.org.
[4] PySCeS.http://pysces.sourceforge.net/index.ht
ml
[5] SimPy. http://simpy.sourceforge.net/
[6] Mark Lutz. Programming Python. ISBN
9781449302856, O'Reilly, 2011.
[7] omniORB 4.1.6 and omniORBpy 3.6. The
omni-ORB version 4.1 User's guide, the om-
niORBpy version 3 User's guide.
[8] http://omniorb.sourceforge.net/
[9] Andreas Pfeiffer, M. Hellerer, S. Hartweg,
Martin Otter, and M. Reiner. PySimulator – A
Simulation and Analysis Environment in Py-
Session 5A: Simulation Tools
DOI Proceedings of the 9th International Modelica Conference 547
10.3384/ecp12076537 September 3-5, 2012, Munich, Germany

thon with Plugin Infrastructure. Submitted to
the 9th International Modelica Conference,
Munich, Germany, September. 2012.
[10] Dassault Systèmes AB:
la, www.dymola.com.
[11] MODELISAR consortium: Functional Mock-
up Interface for Model Exchange, Version 1.0,
2010. www.functional-mockup-interface.org
[12] JModelica.org. http://JModelica.org. Accessed
May 20, 2012.
[13] MWORKS. http://en.tongyuan.cc/. Accessed
May 20, 2012.
[14] LMS Inc. Amesim tool suite.
http://www.lmsintl.com/imagine-amesim-
suite. Accessed May 20, 2012.
An OpenModelica Python Interface and its use in PySimulator

548 Proceedings of the 9th International Modelica Conference DOI
September 3-5, 2012, Munich Germany 10.3384/ecp12076537