Installation
There is an MSI installer for the DataExplore application if all you need is this software. This is recommended for anyone using windows not using the library directly as a widget.
For Linux users who just want the dataexplore app using snap might be easiest. This is supported in Ubuntu, Fedora, Solus, Arch and other distributions. Just install snapd with your package manager and then run:
sudo snap install dataexplore
For openSUSE Tumbleweed there are packages available from the openSUSE Build Service (OBS) for DataExplore. For instructions on how to install the Python 2 package see this download page, while the Python 3 package can be installed via following the prompts on this page.
On all operating systems installations of Python should include the pip tool. If not use your distributions package manager to install pip first. Then a simple call as follows should install all dependencies:
pip install pandastable
This might not work well in some cases because matplotlib has library dependencies that users might find confusing. Though it should work ok on recent versions of Ubuntu. Advice for each OS is given below.
- numpy
- pandas
- matplotlib
- numexpr
- statsmodels
- seaborn (requires scipy)
For all distributions the python tkinter library is needed. It may be present, but if not run:
sudo apt install python-tk (ubuntu)
sudo dnf install python-tkinter (fedora)
sudo eopkg install python-tkinter (solus)
Using pip should then work ok. matplotlib may require more packages such as python headers for compiling the extension. You might need the tk8.6-dev package to provide the tkagg backend. If you have problems using pip you can use the the package manager. In Ubuntu/Debian based distributions you can issue the command:
sudo apt install python-matplotlib
You should install with pip as a rule as it will provide the most recent versions of packages.
For python 3 installs
You need to use the command pip3 instead if python 2 is also on your system, like in Ubuntu. When installing packages with apt you likely need to specify python 3. e.g. python3-numpy instead of python-numpy.
For python 2.7 ONLY
You will also need the future package. Run pip install future to install them. Python 2.6 has NOT been tested and probably won't work.
It is much easier to install matplotlib in windows using the binary installer rather than using pip. You can download this here. Pick the appropriate file for your python version e.g. 'matplotlib-1.4.3.win32-py3.4.exe' for python 3.4.
pandas should install ok with the pip installer. In windows pip.exe is located under C:\Python34\Scripts. The future package is needed for python 2.7.
Note that the Python pydata stack can also be installed at once using miniconda. This includes a version of Python itself.
There are multiple packaged installers for scientific Python, the best of which is probably anaconda. Miniconda is a smaller version if you don't want all the packages. To use it download and run the Mac OS X installer from http://conda.pydata.org/miniconda.html. The installer will automatically configure your system to use the Anaconda Python. You can then use pip to install the package or you can use the conda package using:
conda install -c https://conda.anaconda.org/dmnfarrell pandastable
If using macports:
sudo port install py34-pip sudo pip-3.4 install matplotlib numpy pandas numexpr
For using homebrew see here.
You can download the latest tar.gz file here and do the following:
tar -xzvf pandastable.version.tar.gz
cd pandastable
sudo python3 setup.py install
Note that you still need to have installed the dependencies as above.
If you are using anaconda or miniconda you should install matplotlib with pip rather than conda. This is because the Tkagg backend seems broken with the conda package (confirmed in windows).