问题
I am currently trying to update my statsmodels package in Conda to the developer version statsmodels v0.11.0dev0.
As I am relatively new to Python, I am struggling heavily to understand different threads on how to update to the developer version.
On https://www.statsmodels.org/dev/install.html a short hint on how to install the developer version is given, nevertheless I cannot follow.
I have tried the pip install -e and python setup.py develop.
In order to specifically update the statsmodel package, what do I need to put in front? A stepwise example would be really helpful and highly appreciated.
回答1:
Normal install
(not applicable here to answer the question)
The documenation in https://www.statsmodels.org/dev/install.html clearly indicates two ways for installing statsmodels:
For the first option you need to install Anaconda to be able to run conda install -c conda-forge statsmodels.
And for the other way git is required. Then clone the github repository to be able to run pip install git+https://github.com/statsmodels/statsmodels and then python setup.py install.
Developer version
Run the following steps:
1) git clone git://github.com/statsmodels/statsmodels.git
2) pip install cython
3) pip install -r requirements.txt
4) python setup.py develop
which then results in proper compile of c files:
Compiling statsmodels/tsa/statespace/_initialization.pyx because it changed.
Compiling statsmodels/tsa/statespace/_representation.pyx because it changed.
Compiling statsmodels/tsa/statespace/_kalman_filter.pyx because it changed.
Compiling statsmodels/tsa/statespace/_filters/_inversions.pyx because it changed.
Compiling statsmodels/tsa/statespace/_filters/_univariate.pyx because it changed.
Compiling statsmodels/tsa/statespace/_filters/_univariate_diffuse.pyx because it changed.
Compiling statsmodels/tsa/statespace/_kalman_smoother.pyx because it changed.
Compiling statsmodels/tsa/statespace/_smoothers/_alternative.pyx because it changed.
Compiling statsmodels/tsa/statespace/_smoothers/_classical.pyx because it changed.
Compiling statsmodels/tsa/statespace/_smoothers/_conventional.pyx because it changed.
Compiling statsmodels/tsa/statespace/_smoothers/_univariate.pyx because it changed.
Compiling statsmodels/tsa/statespace/_smoothers/_univariate_diffuse.pyx because it changed.
Compiling statsmodels/tsa/statespace/_simulation_smoother.pyx because it changed.
Compiling statsmodels/tsa/statespace/_tools.pyx because it changed.
[ 1/14] Cythonizing statsmodels/tsa/statespace/_filters/_inversions.pyx
[ 2/14] Cythonizing statsmodels/tsa/statespace/_filters/_univariate.pyx
[ 3/14] Cythonizing statsmodels/tsa/statespace/_filters/_univariate_diffuse.pyx
[ 4/14] Cythonizing statsmodels/tsa/statespace/_initialization.pyx
[ 5/14] Cythonizing statsmodels/tsa/statespace/_kalman_filter.pyx
[ 6/14] Cythonizing statsmodels/tsa/statespace/_kalman_smoother.pyx
[ 7/14] Cythonizing statsmodels/tsa/statespace/_representation.pyx
[ 8/14] Cythonizing statsmodels/tsa/statespace/_simulation_smoother.pyx
[ 9/14] Cythonizing statsmodels/tsa/statespace/_smoothers/_alternative.pyx
[10/14] Cythonizing statsmodels/tsa/statespace/_smoothers/_classical.pyx
[11/14] Cythonizing statsmodels/tsa/statespace/_smoothers/_conventional.pyx
[12/14] Cythonizing statsmodels/tsa/statespace/_smoothers/_univariate.pyx
[13/14] Cythonizing statsmodels/tsa/statespace/_smoothers/_univariate_diffuse.pyx
warning: statsmodels/tsa/statespace/_smoothers/_univariate_diffuse.pyx:555:14: Unreachable code
warning: statsmodels/tsa/statespace/_smoothers/_univariate_diffuse.pyx:1130:14: Unreachable code
warning: statsmodels/tsa/statespace/_smoothers/_univariate_diffuse.pyx:1705:14: Unreachable code
warning: statsmodels/tsa/statespace/_smoothers/_univariate_diffuse.pyx:2280:14: Unreachable code
[14/14] Cythonizing statsmodels/tsa/statespace/_tools.pyx
[some compiler warnings are shown here]
5 warnings generated.
I have tested this on Mac and have gcc installed via Xcode as well, not sure if this plays a role here.
来源：https://stackoverflow.com/questions/57868744/how-to-update-to-the-developer-version-of-statsmodels-using-conda