Installing the NMODL Framework

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Cloning Source

The NMODL Framework is maintained on github. The best way to get the sources is to simply clone the repository.

Note: This project uses git submodules which must be cloned along with the repository itself:

git clone --recursive https://github.com/BlueBrain/nmodl.git
cd nmodl

Prerequisites

To build the project from source, a modern C++ compiler with C++14 support is necessary. Make sure you have following packages available:

  • flex (>=2.6)

  • bison (>=3.0)

  • CMake (>=3.15)

  • Python (>=3.9)

  • Python packages : jinja2 (>=2.10), pyyaml (>=3.13), pytest (>=4.0.0), sympy (>=1.3), textwrap

On MacOS

Typically the versions of bison and flex provided by the system are outdated and not compatible with our requirements. To get recent version of all dependencies we recommend using homebrew:

brew install flex bison cmake python3

All of the Python dependencies (build, run, and development) can be installed using:

pip3 install --user -r requirements.txt

Make sure to have latest flex/bison in your PATH:

export PATH=/usr/local/opt/flex/bin:/usr/local/opt/bison/bin:/usr/local/bin/:$PATH

On Apple M1, corresponding brew paths are under /opt/homebrew/opt/:

export PATH=/opt/homebrew/opt/flex/bin:/opt/homebrew/opt/bison/bin:$PATH

On Ubuntu

On Ubuntu (>=18.04) flex/bison versions are recent enough and are installed along with the system toolchain:

apt-get install flex bison gcc python3 python3-pip

All of the Python dependencies (build, run, and development) can be installed using:

pip3 install --user -r requirements.txt

Build Project

Using CMake

Once all dependencies are in place, build project as:

mkdir -p nmodl/build
cd nmodl/build
cmake .. -DCMAKE_INSTALL_PREFIX=$HOME/nmodl
cmake --build . --parallel 4 --target install

Hint

By default NMODL is build with the CMake option NMODL_ENABLE_PYTHON_BINDINGS set to ON which increases a lot the compilation complexity and memory requirements. For that purpose it’s recommended to either disable this option if the Python bindings are not needed or restrict the number of parallel jobs running in parallel in the cmake command using cmake --parallel <number_of_parallel_jobs>. i.e. in a machine with 8 threads do cmake --parallel 4.

And set PYTHONPATH as:

export PYTHONPATH=$HOME/nmodl/lib:$PYTHONPATH

Flex / Bison Paths

If flex / bison are not in your default PATH, you can provide the path to cmake as:

cmake .. -DFLEX_EXECUTABLE=/usr/local/opt/flex/bin/flex \
         -DBISON_EXECUTABLE=/usr/local/opt/bison/bin/bison \
         -DCMAKE_INSTALL_PREFIX=$HOME/nmodl

Using the Python build system

If you are mainly interested in the NMODL Framework parsing and analysis tools and wish to use them from Python, we recommend building and installing using Python.

pip3 install --user .

This should build the NMODL framework and install it into your pip user site-packages folder such that it becomes available as a Python module.

Building a wheel

You can also build a wheel you can test and install in another environment using:

pip3 wheel . --no-deps [-C OPTION1=VALUE1 -C OPTION2=VALUE2...] [--wheel-dir DIRECTORY]

where the various OPTION values describe the build options (for a list of all available options, please consult the reference). Notably, due to a bug in CMake, on MacOS one should pass -C build-dir=DIRECTORY to the above.

When building without linking against libpython

NMODL uses an embedded python to symbolically evaluate differential equations. For this to work we would usually link against libpython, which is automatically taken care of by pybind11. In some cases, for instance when building a python wheel, we cannot link against libpython, because we cannot know where it will be at runtime. Instead, we load the python library (along with a wrapper library that manages calls to embedded python) at runtime. To disable linking against python and enabling dynamic loading of libpython at runtime we need to configure the build with the cmake option -DLINK_AGAINST_PYTHON=False.

In order for NMODL binaries to know where to find libpython and our own libpywrapper two environment variables need to be present:

  • NMODL_PYLIB: This variable should point to the libpython shared-object (or dylib) file. On macos this could be for example:

export NMODL_PYLIB=/usr/local/Cellar/python/3.7.7/Frameworks/Python.framework/Versions/3.7/Python

Note: In order for all unit tests to function correctly when building without linking against libpython we must set NMODL_PYLIB before running cmake!

Testing the Installed Module

If you have installed the NMODL Framework using CMake, you can now run tests from the build directory as:

$ make test
Running tests...
Test project /Users/kumbhar/workarena/repos/bbp/incubator/nocmodl/cmake-build-debug
      Start  1: testmodtoken/NMODL Lexer returning valid ModToken object
 1/60 Test  #1: testmodtoken/NMODL Lexer returning valid ModToken object ...................................   Passed    0.01 sec
      Start  2: testlexer/NMODL Lexer returning valid token types
 2/60 Test  #2: testlexer/NMODL Lexer returning valid token types ..........................................   Passed    0.00 sec
      Start  3: testparser/Scenario: NMODL can define macros using DEFINE keyword
 3/60 Test  #3: testparser/Scenario: NMODL can define macros using DEFINE keyword ..........................   Passed    0.01 sec
      Start  4: testparser/Scenario: Macros can be used anywhere in the mod file
 4/60 Test  #4: testparser/Scenario: Macros can be used anywhere in the mod file ...........................   Passed    0.01 sec
      Start  5: testparser/Scenario: NMODL parser accepts empty unit specification
 5/60 Test  #5: testparser/Scenario: NMODL parser accepts empty unit specification .........................   Passed    0.01 sec
      Start  6: testparser/Scenario: NMODL parser running number of valid NMODL constructs
 6/60 Test  #6: testparser/Scenario: NMODL parser running number of valid NMODL constructs .................   Passed    0.04 sec
      Start  7: testparser/Scenario: NMODL parser running number of invalid NMODL constructs
 7/60 Test  #7: testparser/Scenario: NMODL parser running number of invalid NMODL constructs ...............   Passed    0.01 sec
      Start  8: testparser/Scenario: Legacy differential equation solver from NEURON solve number of ODE
 8/60 Test  #8: testparser/Scenario: Legacy differential equation solver from NEURON solve number of ODE ...   Passed    0.00 sec
 ...

To test the NMODL Framework python bindings, you can try a minimal example in your Python 3 interpeter as follows:

>>> import nmodl.dsl as nmodl
>>> driver = nmodl.NmodlDriver()
>>> modast = driver.parse_string("NEURON { SUFFIX hh }")
>>> print ('%s' % modast)
{"Program":[{"NeuronBlock":[{"StatementBlock":[{"Suffix":[{"Name":[{"String":[{"name":"SUFFIX"}]}]},{"Name":[{"String":[{"name":"hh"}]}]}]}]}]}]}
>>> print (nmodl.to_nmodl(modast))
NEURON {
    SUFFIX hh
}

You can also run all of the Python tests for a given wheel using:

bash packaging/test_wheel.bash PYTHON_EXECUTABLE WHEEL

where PYTHON_EXECUTABLE should be replaced by the path to the Python executable, and WHEEL should be replaced by the path to the wheel you wish to test.

NMODL is now setup correctly!

Generating Documentation

In order to build the documentation you must have additionally pandoc installed. Use your system’s package manager to do this (e.g. sudo apt-get install pandoc).

You can build the entire documentation simply by using the generate_docs.sh script:

bash docs/generate_docs.sh DIRECTORY [PYTHON_EXECUTABLE]

where DIRECTORY is where you want to put the output files. The HTML documentation will then be available in DIRECTORY/docs, and the temporary build will be stored in DIRECTORY/build. You can also specify the path to the Python executable if it is not picked up automatically.