README.md

# Welcome to DORiE!
(__D__UNE-__O__perated __Ri__chards equation solving __E__nvironment)

DORiE is a software suite for solving Richard's Equation. The core feature is a C++ PDE-solver powered by [DUNE](https://dune-project.org/) and the [DUNE-PDELab](https://dune-project.org/modules/dune-pdelab/) module. It implements a Discontinous Galerkin (DG) discretization scheme on structured rectangular / cubic and unstructured simplex grids in two and three spatial dimensions, and makes use of advanced features like adaptive grid refinement.

The suite encapsulates a documentation and various tools for program setup, program testing, and output analysis, which are mostly written in Python.

DORiE is developed and maintained by the [DORiE Developers](mailto:dorieteam@iup.uni-heidelberg.de) of the [TS-CCEES](http://ts.iup.uni-heidelberg.de/) research group at [IUP Heidelberg](http://www.iup.uni-heidelberg.de/), supervised by [Kurt Roth](http://ts.iup.uni-heidelberg.de/people/prof-dr-kurt-roth/), in collaboration with [Ole Klein](https://conan.iwr.uni-heidelberg.de/people/oklein/) and the [Scientific Computing Group](https://conan.iwr.uni-heidelberg.de/) of [IWR Heidelberg](https://typo.iwr.uni-heidelberg.de/home/).


# Installation Instructions

DORiE is a [DUNE](https://dune-project.org/) module and requires several other DUNE modules as well as third party software packages. Installation can be handled manually on your local machine, but for inexperienced users it is recommended to use the deployment software [Docker](https://www.docker.com/) instead.

No installation is necessary if you download DORiE as Docker image from
[Docker Hub](https://hub.docker.com/r/dorie/dorie/). You can then proceed
directly to the the instructions on [how to execute DORiE](#run-dorie-run).
The commands listed there are appended to the usual commands for running a
Docker container. See the description on Docker Hub for further details.

## Download Docker image

If you want to use any stable version of DORiE, or the most recent unstable
version, you can download the appropriate images from Docker Hub. To do so,
execute

    docker pull dorie/dorie[:<tag>]

Omitting the tag information downloads the image with tag `latest` which
refers to the latest stable version. You can download any tag by specifying
`<tag>`. The list of available tags can be found on Docker Hub and matches
the release tags list of the Git repository. The latest unstable version is
tagged as `devel`.

## Docker Installation

You can also compile any local version of DORiE into a new Docker image. To do
so, execute

    docker build -f docker/dorie.dockerfile -t <img> .

from within the top-level DORiE source directory. Docker will use a prepared
image from Dockerhub and install DORiE into it. The created image will be called
`<img>`.

## Manual Installation

Installing all packages manually can be quite an effort, but useful for developers who want to have easy access to the source files or users who prefer to run DORiE without the Docker overhead.

Whenever possible, dependencies should be installed using a package manager, like [APT](https://wiki.ubuntuusers.de/APT/) on Ubuntu or [Homebrew](http://brew.sh/) on Mac. Manual installation on a Windows environment is strongly discouraged!

DORiE is configured, built, and installed via the
[DUNE Buildsystem](https://dune-project.org/doc/installation/), using the
`dunecontrol` script to handle DUNE-internal dependencies.

### Dependencies
Depending on your system configuration, there will be more packages necessary to
install DORiE on your machine. See the step-by-step manual for further details.

The specified versions are the _supported_ ones, where compatibility is ensured
by CI tests.

| Software | Version/Branch | Comments |
| ---------| -------------- | -------- |
| CMake | 3.10.2 |
| GCC | 7.3 | Alternatively: LLVM Clang >=6
| git |
| pkg-config |
| HDF5 | 1.10 | with MPI support
| FFTW3 | 3.3.7 | with MPI support
| Python | 3.6 |
| pip | 3.6 |
| MPI | | Tested with OpenMPI 2.1.1
| SuperLU | 5.2 |
| [yaml-cpp](https://github.com/jbeder/yaml-cpp) | >= 5.2.0 |
| [dune-common](https://gitlab.dune-project.org/core/dune-common) | releases/2.6
| [dune-geometry](https://gitlab.dune-project.org/core/dune-geometry) | releases/2.6
| [dune-grid](https://gitlab.dune-project.org/core/dune-grid) | releases/2.6
| [dune-uggrid](https://gitlab.dune-project.org/staging/dune-uggrid) | releases/2.6
| [dune-istl](https://gitlab.dune-project.org/core/dune-istl) | releases/2.6
| [dune-localfunctions](https://gitlab.dune-project.org/core/dune-localfunctions) | releases/2.6
| [dune-functions](https://gitlab.dune-project.org/staging/dune-functions) | releases/2.6
| [dune-typetree](https://gitlab.dune-project.org/staging/dune-typetree) | releases/2.6
| [dune-pdelab](https://gitlab.dune-project.org/pdelab/dune-pdelab) | releases/2.6
| [dune-randomfield](https://gitlab.dune-project.org/oklein/dune-randomfield) | master


### Optional Packages
| Software | Version/Branch | Comments |
| ---------| -------------- | -------- |
| [dune-testtools](https://gitlab.dune-project.org/quality/dune-testtools) | releases/2.6 | Handles system tests
| [doxygen](http://www.stack.nl/~dimitri/doxygen/) | 1.8.13 | Builds documentation
| [METIS](http://glaros.dtc.umn.edu/gkhome/views/metis) | 5 | For parallel runs
| [ParMETIS](http://glaros.dtc.umn.edu/gkhome/views/metis) | 4 | For parallel runs


### Step-by-step Instructions
These instructions are suitable for a clean **Ubuntu** or **macOS** setup. The main difference between the two systems is the package manager. Debian-based systems have the APT manager already built in. On Mac, we recommend installing [Homebrew](http://brew.sh/). If you prefer to use [MacPorts](https://www.macports.org/), you need to check if some of the packages require different installation options than displayed here.

If you installed [Anaconda](https://conda.io/docs/user-guide/install/download.html) on your machine, you don't need to install Python or Pip. Simply skip these packages when using the package managers for installing the software. However, notice the warnings when compiling DORiE below!

1. **macOS** users need to start by installing the Apple Command Line Tools by executing

        xcode-select --install

2. Install third party packages:
    
    **Ubuntu:**
    
        apt update 
        apt install cmake doxygen gcc g++ gfortran \
                    git libatlas-base-dev libfftw3-dev libfftw3-mpi-dev \
                    libfreetype6-dev libhdf5-mpi-dev libopenmpi-dev libpng-dev \
                    libsuperlu-dev libyaml-cpp-dev libxft-dev \
                    python3-dev python3-pip

    **macOS:**

        brew update
        brew install cmake doxygen gcc libpng open-mpi \
                     pkg-config python3 superlu yaml-cpp
        brew install hdf5 --with-mpi
        brew install fftw --with-mpi

3. The parallel linear solver of DORiE can make use of the ParMETIS package. If you want to run DORiE in parallel on multiple processes, additionally install METIS and ParMETIS:

    **Ubuntu:**

        apt install libmetis-dev libparmetis-dev

    **macOS:** _Support is dropped because ParMETIS is currently unavailable from Homebrew._

    **Parallel runs without these two packages are possible but not supported!**

4. Clone the DUNE modules and DORiE into a suitable folder on your machine. Use `git checkout` to switch to the correct branches (see above). Enter the parent folder, and call

        CMAKE_FLAGS="-DDUNE_PYTHON_VIRTUALENV_SETUP=True -DDUNE_PYTHON_ALLOW_GET_PIP=True" ./dune-common/bin/dunecontrol all

    to build all DUNE modules. Additionally, you can add `MAKE_FLAGS="-j X"` before the call to `dunecontrol` to compile on `X` processes in parallel.

    If you installed software into paths not appended to your `PATH` variable, you will have to add `CMAKE_FLAGS` to the call to make sure that CMake finds all packages. Alternatively, you can add a custom options file. See the [DUNE Installation Instructions](https://dune-project.org/doc/installation/) for details. CMake will throw an error if required packages are not found.

    **Warning:** Anacoda supplies its own version of HDF5 which is typically found first by CMake. If you have Anaconda installed on your machine, add

        HDF5_ROOT=/usr/local

    to the `CMAKE_FLAGS` in the call to `dunecontrol` above.
 
5. After a successful build you can call the `dorie` wrapper located in `dorie/build-cmake/bin` from anywhere on your machine, but it is more handy to add this location to your search path. You can add the directory to your `PATH` variable by calling

        PATH=<path/to/>dorie/build-cmake/bin:$PATH

    However, this will only last for your current terminal session. Another option is to add this path permanently in your shell configuration file.

    **Installing DUNE and DORiE via `make install` into your usual work environment is strongly discouraged!**

### Experimental Features
The local operator implementing Richards equation's discretization supports
multiple scheme settings. Setting these via the config file is disabled by
default. You can enable this feature by reconfiguring DORiE with the CMake flag
`-DEXPERIMENTAL_DG_FEATURES=On`, and rebuilding it.

The configuration settings in the section `[dg.experimental]` will then override
the default settings.


## Recommended Third-Party Software

The following software packages are cross-platform, so you should be able to find a release that fits your operating system:

* [Gmsh](http://gmsh.info/): An open-source CAD that can be used to create the `.msh` files used by DORiE to define triangular meshes.
* [ParaView](http://www.paraview.org/): A powerful post-processing tool for VTK files. Offers both visualization and data analysis tools.

# Usage

## Documentation

The documentation of the latest release branch can be found [online](https://dorie-doc.netlify.com/).

The documentation can be built after DORiE has been properly configured (i.e., by calling `dunecontrol`). Note that you might have to re-configure DORiE once after installing it, because some dependencies are installed at configure time (e.g. by `dunecontrol --only=dorie configure`). To build the documentation, just run

    dunecontrol --only=dorie make doc

or navigate to the `dorie/build-cmake` directory and call

    make doc

You will then find the index page of the documentation at
`dorie/build-cmake/doc/html/index.html`.

## Run, DORiE, Run!
DORiE provides a command line interface (CLI) for all its functions.
The required Python modules and all their dependencies are readily installed
into a virtual environment (`venv`), which has to be activated within a
shell session. You can do so by activating it in your current session
(Manual Installation only) or by running the Docker application.

### Run the `venv` using the Docker application
If you did not install DORiE locally, you can use the Docker
application to boot up the virtual environment in a mounted directory of choice.

Start up the Docker application by calling

    docker run -it -v <dir>:/mnt <img>

where you replace `<dir>` with a local directory for storing input and output
data, and `<img>` with `dorie/dorie[:<tag>]`.
If you built a local Docker image, replace `<img>` appropriately.
You can also insert `$PWD` as `<dir>` if you want to mount your
current working directory.

The command boots up a (`bash`) shell inside a Docker container and mounts
the directory `<dir>` and all its subdirectories into the directory `/mnt`
inside the container. Your shell session starts in this directory with the
virtual environment activated.

Notice, that you can only use **local file paths** in all configuration settings
due to the directory mount.

### Activate the `venv` locally
To activate the virtual environment within your current shell session, execute

    source <path/to/>dorie/build-cmake/activate

where you replace `<path/to/>` with the path to the appropriate directory.

Your shell will now display the prefix `(dune-env)` to indicate that it is
configured appropriately. You can exit the environent at any time by simply
executing

    deactivate

Notice that any virtual environment only applies to, and lasts for your current
terminal session!

_With the virtual environment activated,_ you can now navigate to any directory
that you would like to contain your simulation input and/or output data.

### Execute the application
Any command to the DORiE application has the signature

    dorie <cmd> [<opts>] [<args>]

Using the `-h` or `--help` option, you can find all available commands and
further help.

To start your first simulation run, create a new directory and enter it.
Place some exemplary configuration and boundary condition data files in there
by calling

    dorie create

DORiE encapsulates two main routines, the Parameter Field Generator (PFG) and
the main program routine. Each takes a single `.ini` configuration file as
argument.

Tweak the parameters of `parfield.ini` to your liking and then call

    dorie pfg parfield.ini

to create a parameter field file. Do the same with the `config.ini`, while
ensuring that you use the parameter field file you just created. Then call

    dorie run config.ini

to execute the main program.

## Troubleshooting
CMake heavily caches the results of its configuration process. In case you encounter errors or strange behavior, especially after an update, you should delete the DORiE build folder (called `build-cmake` by default) and re-install DORiE using `dunecontrol`.

If the problem persists, take a look at the [List of Known Bugs](https://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/issues), or create an issue yourself. For problems related to the installation, refer to the sections below.

### Debugging
DORiE can be built with debugging flags via CMake. To do so, run

    CMAKE_FLAGS="-DCMAKE_BUILD_TYPE=Debug" dunecontrol --only=dorie all

After building, a debugger can hook into the executables.

**Note:** If no `CMAKE_BUILD_TYPE` is specified during re-configuration, the last configuration build type is used. If no CMake files exist, it defaults to `Release`. You will find the actual value displayed in the final output of CMake.

To re-create a release build, configure DORiE with the release build type by executing

    CMAKE_FLAGS="-DCMAKE_BUILD_TYPE=Release" dunecontrol --only=dorie all

#### Debugging inside Docker
A debugger needs special security privileges usually not provided by the Docker engine. To enable debugging inside Docker, these privileges have to be granted when calling `docker run` by adding the options

    --security-opt seccomp=unconfined

The debugger `gdb` is pre-installed in the Docker container.

### Running System Tests
DORiE includes a testing system for comparing its results the ones of ODE solvers
or former versions of itself. This ensures that DORiE is running correctly and
producing the expected results. We distinguish _unit tests_ for testing certain
features of the code, and _system tests_ for verifying the results of the
final application. As system tests require executing the DUNE solvers,
it is recommended to build them in a `Release` environment.

| Test category | Build tests | Execute tests | Recommended build type |
| ------------- | ----------- | ------------- | ---------------------- |
| Unit tests | `make build_unit_tests` | `make_unit_tests` | `Debug`
| System tests | `make build_system_tests` | `make_system_tests` | `Release`

The `make` commands are to be executed from within the `build-cmake` directory.

The unit tests automatically include compiler flags for code coverage reports.
After building them with `Debug` flags and executing them, you can
retrieve code coverage information using the
[`gcovr`](https://gcovr.com/index.html) utility.

### Further Help
[Open an issue](https://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/issues/new), or write to the [DORiE developer mailing list](mailto:dorieteam@iup.uni-heidelberg.de).