If you google with "python environment construction", you will get about 200,000 hits, but the content is relatively old. The title says Data Scientist, but I recommend anaconda as well as Data Scientist.
―― 2.x or 3.x? 3.x has many libraries that don't work, so 2.x is recommended> There are libraries that don't work with 3.x. --Pip with easy_install, setuptools, but there is also a wheel ...> It's old. --virtualenv Required> That's not true. --In win, 64bit has many problems, so 32bit is recommended> old. --For win, download and install from Unofficial Binary> Thank you for your help, but I haven't used it recently.
The important thing is that you don't have to include ** official python **.
――It's better than before, but you need to use 2 system and 3 system properly depending on the purpose and library used. --Version control system required: pyenv, pythonz, etc. --There are cases where you want to use different development environments, such as some libraries that cannot coexist, or you want to try a new library for the dev version. --Environmental management system required: virtualenv, pyvenv (python 3.4 or later), etc. --Package management system makes beginners cry --Standard is pip (python package index) --Pip is not pre-installed up to python 3.3, so install it from the package management system called setuptools (currently distribute) ... ――The package management system needs package management. Isn't it Battery Included? ――I don't understand anymore. The page of here is detailed. --Linux and Mac come with python pre-installed, but most of them are python 2.x --When I try to upgrade the system python, it definitely fits. --In the end, you will have to put multiple versions of python with apt-get, yum, brew, but the environment variables will conflict. --I installed ipython with pip on python3 but can't use it> I installed it on another python orz ...
You don't have to worry about these in the definitive edition of the Environment Construction Law.
--One of the python distributions, you can install major libraries all-in-one. (numpy, scipy, pandas, ipython, jupyter, scikit-learn etc ...) --There is also a miniconda that has a minimal configuration. --Both python 2.x and 3.x are supported. (python 3.x since summer 2014) --Linux version, Mac version, Win version are available. --There are both 32bit and 64bit. --There are similar things like enthought and python (x, y), but only Anaconda supports all platforms & 2/3.
--You can use a package management system called conda. : ** Instead of pip **
--Currently, there are over 400 supported packages.
--Since pip is compiled on the client side, it may be moss depending on the environment.
--Since it can be used together with pip, libraries not included in conda can be installed with pip.
--conda can also version control. : ** Instead of pyenv **
--For example, if you include anaconda3-xx.xx, everything will be included based on python 3.5.
- conda create -n py2 python=2.7
--By typing the above command, you can build a virtual environment for python2.7.
--Use source activate py2 to enter the virtual environment of python2.7.
--conda can also be used for virtual environment management. : ** Instead of virtualenv / venv **
--You can install packages with conda or pip in a virtual environment.
--Since the virtual environment created by conda can absorb different versions of python, it can be said to be ** upward compatible with virtualenv **.
--In fact, when you try to use virtualenv with anaconda, you get a warning to create an environment with conda. (A virtualenv can be created.)
――Conda is the best.
――Three of the items raised in the issue of environment construction will be solved. (Because there is a fourth, linux / Mac should be combined with pyenv.)
windows is pre-installed and python is not included. However, since pyenv doesn't support windows at all, you should also consider version control. But don't worry. There is no problem if you add only anaconda.
-Download the installer for windows from Download Site. ――I think 3.5 is fine, but if you only use 2.7, please drop for 2.7. ――You don't have to worry so much because you can create a 2.7 environment later with conda. --Run the exe and follow the instructions to install. (You will be asked if you want to add it to the path on the way.)
Informal binaries don't resolve dependencies, so I'm sick of the complex dependencies I want. ..
The system comes pre-installed with python.
――Most of them are 2.x, so I would like to coexist with 3.x. ――It fits when you touch the system. --Even if you put it separately as ʻapt-get install python3` etc., you have to play with the environment variables yourself. ――There is actually a problem with anaconda alone. --Install curl, sqlite3, openssl, etc. together to keep the operation. --Recommendation is via pyenv. --The current pyenv is supports anaconda.
$ git clone https://github.com/yyuu/pyenv.git ~/.pyenv
$ echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bashrc
$ echo 'export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bashrc
$ echo 'eval "$(pyenv init -)"' >> ~/.bashrc
$ source ~/.bashrc
$ pyenv install -l | grep ana
#Check the latest version. anaconda3-2.5.0 (2 system is anaconda-2.5.0)
#If you like miniconda, please add miniconda.
$ pyenv install anaconda3-2.5.0
$ pyenv rehash
$ pyenv global anaconda3-2.5.0
#Set anaconda as the main python.
$ echo 'export PATH="$PYENV_ROOT/versions/anaconda3-2.5.0/bin/:$PATH"' >> ~/.bashrc
$ source ~/.bashrc
#Since activate batting with pyenv and anaconda, specify it in path.
$ conda update conda
#Update conda itself just in case.
(Corrected on 3/24)
$ echo 'export PATH="$PYENV_ROOT/versions/anaconda3-2.5.0/bin/:$PATH"' >> ~/.bashrc
Has side effects and will always use anaconda without accepting any pyenv global or pyenv local.
For those who use other than anaconda, or those who mainly want to operate with pyenv local, it is necessary to stop listing in Path and enter conda activate with the full path.
source ~ / .pyenv / versions / anaconda3-2.5.0 / bin / activate <environment name>
For those who do not switch the environment frequently and just use anaconda, it is better to specify it in the path.
(4/8) I created another article about activate collision problem when pyenv and anaconda coexist.
** I don't have a Mac, so it's hearsay. Please pardon. ** ** ↓ Since there is such a story, it seems better to put pyenv like Linux. The story that the Homebrew environment was blown away when Anaconda was added
I don't have it, so please refer to another person's article. I think it's the same except that pyenv becomes Homebrew instead of via github. Note from installing Homebrew to building an Anaconda environment for Python with pyenv
I haven't verified it myself, but as with linux, anaconda and pyenv activate will batting. I think it's easier to go through the path to anaconda.
$ echo 'export PATH="$PYENV_ROOT/versions/anaconda3-2.5.0/bin/:$PATH"' >> ~/.bashrc
$ source ~/.bashrc
You can buy a MacBook Air-like one for about one-third the price (^^
See here Create ubuntu environment with crouton It is possible to put it, but it is a little troublesome.
--Building a virtual environment
** conda create -n
conda create -n py2 python=2.7 numpy scipy pandas jupyter
#It is also possible to put them together as anaconda.
conda create -n anaconda2 python=2.7 anaconda
--Check virtual environment
conda env list
#It also appears here.
conda info -e
--Entering and exiting the virtual environment
#Enter the virtual environment
source activate py2
#activate py2 on windows
#Get out of the virtual environment
source deactivate
#deactivate on windows
--Delete virtual environment
conda remove -n py2 --all
--Package installation & uninstallation
conda install numpy scipy #Same as pip, multiple OK with space delimiter
conda install numpy=1.10.4 #Version can be specified
conda install -n py2 numpy scipy # -You can also specify the environment name with n
conda update numpy # update
pip install numpy #You can also use pip. Use this when not in conda
source activate py2;pip install numpy #Must be activated when entering the virtual environment
conda uninstall -n py2 numpy #Uninstall
--Check the package
conda list
#List of currently included packages
conda list -n py2
# -You can also select under virtual environment with n
conda list --export > env.txt
conda create -n py2_copy --file env.txt
#You can export it and reuse it,
#Packages put in with pip cannot be exported, so you need to output them separately with pip freeze.
Someone may have posted a package on anaconda cloud (anaconda.org) that is not provided by anaconda.
anaconda search -t conda ggplot
#Come out in various ways
# ...
#bokeh/ggplot | 0.6.8 | conda | linux-64, win-32, win-64, linux-32, osx-64
# : ggplot for python
#...
anaconda show bokeh/ggplot # <USER/PACKAGE>You can see the details by specifying with
#Using Anaconda Cloud api site https://api.anaconda.org
#Name: ggplot
#Summary: ggplot for python
#Access: public
#Package Types: conda
#Versions:
# + 0.6.5
# + 0.6.8
conda install -c bokeh ggplot #USER name-Install the package by specifying with c
It's hard to see and it seems that you can't search by OS, so it may be faster to search from https://anaconda.org/.
――I think there are people who say that data science is R! --You can also install R with conda.
conda create -n r -c r r-irkernel
With this, you can even create an environment for using R in the now popular Jupyter.
――You can use R normally, or you can use it via Rstudio. --Jupyter usage of R is easy because it depends on the version of R and rzmq.
Great, Anaconda.
It is a little off the definitive edition. (3/24 postscript)
pyenv It can't be used on windows, but there are some good things to do if you bite pyenv on Mac / Linux.
--Easy to rebuild the environment.
――When you can't help it, you can assume that it wasn't pyenv uninstall.
--Alternatively, when a new version of anaconda comes out, you can start using the new anaconda.
--You can use other than Cpython such as pypy, jython, stackless.
--pyenv local can be used.
pyenv local
Of these, pyenv local is very useful.
For example, if you set the following, anaconda2 will start up instead of anaconda3 only when you move to py2.
mkdir py2
cd py2
pyenv local anaconda2-2.5.0
In fact, you can also specify the virtual environment of ** anaconda locally. ** **
conda create -n py2 python=2.7
pyenv local anaconda3-2.5.0/envs/py2
python
#Python 2.7.11 |Continuum Analytics, Inc.| (default, Dec 6 2015, 18:08:32)
#[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)] on linux2
However, [Linux version environment construction](http://qiita.com/y__sama/items/5b62d31cb7e6ed50f02c#%E7%92%B0%E5%A2%83%E6%A7%8B%E7%AF%89%E6 As described in% 96% B9% E6% B3% 95-1), $ echo'export PATH =" $ PYENV_ROOT / versions / anaconda3-2.5.0 / bin /: $ PATH "'>> ~ / .bashrc If is set, pyenv local will be ignored and anaconda3-2.5.0 will be read.
Whether to cut the environment in the directory for each project and operate with pyenv local or explicitly hitsource activate <virtual environment name>when switching is necessary depends on your preference and usage content. I think so please choose the one you like.
pyenv-virtualenv
Very confusing is the pyenv plugin called pyenv-virtualenv. This is different from python's virtual environment management system virtualenv, and you can switch between virtual environments created with virtualenv or venv (python3.4 or later). Moreover, it supports even the environment created by conda. Therefore, it solves the batting problem of pyenv local and source activate.
Installation is easy, clone from git and place it in the plugin folder under the pyenv installation folder.
git clone git://github.com/yyuu/pyenv-virtualenv.git ~/.pyenv/plugins/pyenv-virtualenv
souce ~/.bashrc
#Check the environment
pyenv virtualenvs
#>>>anaconda3-2.5.0 (created from /home/vagrant/.pyenv/versions/anaconda3-2.5.0)
#>>>anaconda3-2.5.0/envs/py2 (created from/home/vagrant/.pyenv/versions/anaconda32.5.0/envs/py2)
#Activate the environment
pyenv activate anaconda3-2.5.0/envs/py2
python
#>>>Python 2.7.11 |Continuum Analytics, Inc.| (default, Dec 6 2015, 18:08:32)
#Deactivate the environment
pyenv deactivate
You can build an environment with conda and switch the environment with pyenv activate and pyenv local.
For those who often switch virtual environments on Mac / Linux, it may be more productive to include pyenv-virtualenv.
Recommended Posts