PMIP3 wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Working with Python ======

<note tip>This page will tell you how to select a python distribution and start (and exit!) the python interpreter

You can then read the [[jyp_steps|JYP's recommended steps for learning python]] for really working with python
</note>

===== Where to start =====

In order to start working with Python, you need to have a //Python distribution// installed on your local computer or on the remote Linux server(s) you work on. A //distribution// provides a //Python interpreter//, and Python extensions (aka Python //modules// or //packages//). You may have several distributions installed on your computer and you need to know how to initialize them, and which one you are using at a given time (type ''which python'' on Linux to determine where the python executable is located)

If you are using a Linux computer or a Mac, you should already have a **default python** installed. The following example shows where the python interpreter is installed on the //asterix// LSCE server (if it's in ''/usr/bin'', it's the default python) and which version it is (example below: version //2.6.6// compiled in May 2015)

<code># Which is the current python (e.g. where is it located)?
 > which python
/usr/bin/python

# Where is it coming from ('rpm' works on a RedHat-like Linux machine)?
 > rpm -qf /usr/bin/python
python-2.6.6-64.el6.x86_64

# You also get some information when you start the interpreter
 > python
Python 2.6.6 (r266:84292, May 22 2015, 08:34:51)
[GCC 4.4.7 20120313 (Red Hat 4.4.7-15)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>>
</code>

===== Python distributions available for LSCE users =====

This section will help you choose a distribution in the big Python ecosystem (many distributions, python version 2.7.* or 3.*, ...). You should use a distribution that is already available near your data (e.g. do the computation on a server near your data, do not move/duplicate the data!) and try to identify who is maintaining it, if you need help or extra packages.

Only install a distribution yourself if you need it on your local computer, or if you need to install some modules that can't be installed by the contacts listed below. A python distribution can require several Gb of disk space, so do not install it in a backed up //home// directory!

We suggest that you **use Python 2 rather than Python 3**. Or you can check the [[https://wiki.lsce.ipsl.fr/pmip3/doku.php/other:python:jyp_steps#python_27_vs_python_3|differences between both versions]] and try to write scripts that will work in both versions! 8-)

==== LSCE distribution ====

Contact: the LSCE system administrators ([[help-lsce@lsce.ipsl.fr]])

<code>
 > module avail
[...]
castem/12        ferret/6.9.5     hdf5/1.8.9       netcdf/4p        python/2.7.5     sun-java/7.0.45
[...]

 > module load python/2.7.5

 > which python
/usr/local/install/python-2.7.5/bin/python

 > python
Python 2.7.5 (default, Sep 18 2013, 15:47:43)
[GCC 4.4.7 20120313 (Red Hat 4.4.7-3)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>>
</code>

==== TGCC distribution ====

Contact: the TGCC hotline ([[hotline.tgcc@cea.fr]])

<code>
$ module avail
[...]
cmake/2.8.9(default)      hwloc/1.5                 netcdf-utils/4.3.3.1_hdf5 python/2.7.8              tix/8.4.3
cmake/3.2.2               hwloc/1.7.1(default)      octave/3.6.3              python/3.3.2              tk/8.5(default)
[...]

$ module load python/2.7.8
load module python/2.7.8 (Python)

$ which python
/usr/local/ccc_python/2.7.8_201409/bin/python

$ python
Python 2.7.8 (default, Aug 27 2014, 17:50:16)
[GCC 4.4.7 20120313 (Red Hat 4.4.7-4)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>>
</code>

==== UV-CDAT ====

Contact: Jean-Yves Peterschmitt (//JYP//) @ LSCE

=== What is UV-CDAT? ===

[[http://uv-cdat.llnl.gov/|UV-CDAT]] (//Ultrascale Visualization - Climate Data Analysis Tools//) is a python distribution developed specifically for the climate scientists, and is recommended by JYP.

UV-CDAT is **available for Mac and Linux**.

You can find out which **standard packages** are installed in the //Packages built by CDAT// column below. UV-CDAT also provides some **specific packages** that you will not find in other python distributions: **cdms2**, **cdutil**, **cdtime**, **genutil**, **vcs**, **cmor**, ...

**Extra packages** are listed:
  * in the apropriate //Installation notes// pages for CDAT>=''2.8.0''
  * on the [[other:uvcdat:extra|Extra packages installed in UV-CDAT by JYP]] page for CDAT%%<=%%''2.10''

=== UV-CDAT versions maintained by JYP ===

The following versions are maintained by JYP, on the Linux servers where LSCE users have accounts

^  CDAT version  ^  JYP\\ version  ^  python\\ version  ^  Availability?  ^  Packages built by CDAT  ^  Installation notes  ^
|  **1.1.0**  |  08  |  2.7.3  |  LSCE, curie  |  n/a  |  [[other:uvcdat:installed#section110|1.1.0 notes]]  |
|  **1.5.1**  |  10  |  2.7.4  |  LSCE, ciclad  |  {{:other:python:cdat-1.5.1_jyp-10_build_info.txt}}  |  [[other:uvcdat:installed#section151|1.5.1 notes]]  |
|  **2.1.0**  |  13  |  2.7.8  |  LSCE, ciclad  |  {{:other:python:cdat-2.1.0_jyp-13_build_info.txt}}  |  [[other:uvcdat:installed#section210|2.1.0 notes]]  |
|  **2.4.0**  |  //Not installed//  |||||
|  **2.8.0**  |  14  |  2.7.12  |  LSCE  |  {{:other:uvcdat:uvcdat-2.8.0_install_dry_with-forge_160221.txt}}  |  [[:other:uvcdat:cdat_conda:cdat_2_8_0|2.8.0 notes]]  |

Notes:
  * For the conda based CDAT versions (>=''2.8.0''), you can get the list of currently available packages by typing: ''conda list''
  * For older CDAT versions (<=''2.1.0''), the ''cdat-VVV_jyp-NNN_build_info.txt'' files have been generated with\\ ''cd /my_build_directory_path; cat build_info.txt | sort > cdat-VVV_JYP-NNN_build_info.txt''

=== The basics of UV-CDAT initialization ===

The way you initialize UV-CDAT depends on:
  * which **server** you want to use it on
  * which **shell** you are using in your terminals: //tcsh// or other shells
    * Read [[other:newppl:starting#which_shell_are_you_using|Which shell are you using?]] if you are not sure
    * Warning: the shell in a (batch) script may be different from your interactive shell!
  * which **version** of UV-CDAT you want to use...
    * Note: You should use the most recent version of UV-CDAT available on a given server, unless you have a good reason to use an older version (e.g. if you need to use the //old-style vcs// of version 1.5.1).
Note: if you get an error when importing //cdms2// or //vcs// it means that either you have forgotten to initialize UV-CDAT, or that something went wrong during the initialization. In both cases, you are either still using the default python installed on your system, or another python distribution!

<code>$ python -c 'import cdms2, vcs'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
ImportError: No module named cdms2

# Am I using the correct python distribution?
$ which python
/usr/bin/python</code>

=== conda-based versions of UV-CDAT ===

<WRAP center round important 60%>
\\ The steps detailed in this section are for [[other:uvcdat:cdat_conda:|UV-CDAT 2.8.0 and later]]
</WRAP>

Note: if you need to maintain a **local** version on **your** machine, you can read [[other:uvcdat:conda_notes|Installing and maintaining UV-CDAT with conda]] and the [[other:uvcdat:cdat_conda:index#versions|versions specific notes]].

== Configuring .bashrc ==

You should **add** the following line at the end of your ''~/.bashrc'' configuration file (create the ''~/.bashrc'' file if it does not exist yet). This will basically make sure that your shell can find the ''conda'' executable it needs for initializing CDAT, and define a ''wp'' alias that you can use to find where the ''python'' executable is located

^  Server  ^  Line to add to ''.bashrc''  ^
|  **LSCE**  | ''source ~jypeter/.conda_jyp.sh''\\ \\ Note:{{ :other:python:conda_jyp.txt |conda_jyp}} content |
|  **ciclad**  | ''source ~jypmce/.conda_jyp.sh'' |


Note: when you add the specified line to your ''bashrc'' file, you will then always get the python supplied by conda, instead of the system's python, when you use ''bash''. If you don't like this permanent behavior, just type ''source ~jyp_login_on_this_server/.conda_jyp.sh'' in the bash shell where you want to use conda+CDAT, rather than adding it to the ''.bashrc'' file...

== Initializing UV-CDAT ==

<WRAP center round tip 60%>
When you initialize UV-CDAT, it will be available only in the **current** window/shell. In the other windows, you still get whatever your default python is. This should minimize potential side-effects
</WRAP>

  - If [[other:newppl:starting#which_shell_are_you_using|your shell]] is not //bash//, start a bash shell by typing ''bash''
    * You can later go back to your previous shell by typing ''exit'' or ''CTRL-D''
  - Activate the //conda environment// that uses the version of UV-CDAT you want to use
    * ''source activate conda_env_name''
    * Use the table below to determine which environments are available on which server
  - Use CDAT !

^  UV-CDAT version  ^  Server  ^  Environment name\\ and activation line  ^  Packages\\ (''conda list'')  ^
|  [[:other:uvcdat:cdat_conda:cdat_2_8_0|2.8.0]]  |  **LSCE**\\ **ciclad**  |**cdatm14** (''source activate cdatm14'')  |  {{ :other:uvcdat:cdat_conda:conda_list_lsce_cdatm14_170310.txt |cdatm14 installed packages}}  |
|  [[:other:uvcdat:cdat_conda:cdat_2_10|2.10]]  |  **LSCE**  |**cdatm15** (''source activate cdatm15'')  |  {{ :other:uvcdat:cdat_conda:uvcdat-2.10_list_170721.txt |cdatm14 installed packages}}  |
|  **cdatm** will always point to the latest stable version  ||||
|  //latest stable version//  |  **LSCE**  |**cdatm** (''source activate cdatm'')  |  ''conda list''  |

=== Older versions of UV-CDAT ===

<WRAP center round important 60%>
The steps detailed in this section are for **UV-CDAT ''2.1.0'' and earlier**
</WRAP>

== Initializing UV-CDAT for tcsh users ==

If you use **tcsh** as [[other:newppl:starting#which_shell_are_you_using|your login shell]], you can define a ''cdatm'' alias on all the IPSL servers on which UV-CDAT is installed (see details below), and move easily from one version of UV-CDAT to another:
  * type ''cdatm'' to initialize the default version of UV-CDAT
  * type ''cdatm %%--%%config //CONFIG_NAME//'' to initialize a specific version
    * You can find the allowed values of //CONFIG_NAME// (e.g. ''LSCE_10'', ''LSCE_13'', ...) in one of the tables below

<code> > which python
/usr/bin/python

 > cdatm
** CDAT 'uv-1.5.1' initialized!
** If you have problems with this installation,
** it is maintained by: Jean-Yves Peterschmitt - LSCE

 > which python
/home/share/unix_files/cdat/versions/cdat_install_uv-1.5.1_x86_64_gcc4_10/bin/python
</code>

You need to add the following line to the ''.cshrc'' file in your home directory, in order to define the ''cdatm'' alias. You can also define the simple ''wp'' alias, in order to always be able to determine quickly which python is initialized in the current terminal.

|  **LSCE**  | ''alias cdatm 'source ~jypeter/CDAT/Install/cdat_multi/cdat_multi.login ~jypeter/CDAT/Install/cdat_multi''' |
|  **ciclad**  | ''alias cdatm 'source ~jypmce/CDAT/Install/cdat_multi/cdat_multi.login ~jypmce/CDAT/Install/cdat_multi'''|
|  **curie**  | ''alias cdatm 'source ~p25jype/CDAT/Install/cdat_multi/cdat_multi.login ~p25jype/CDAT/Install/cdat_multi''' |
|  **All** servers  | ''alias wp 'which python'' |

== Initializing UV-CDAT in any kind of shell, or in a batch script ==

You first need to know where the ''python'' program provided by UV-CDAT is installed, something like ''/path/to/UV-CDAT/bin/python''

Then, all you need to do in order to initialize UV-CDAT is to //source// the ''setup_runtime'' file appropriate for [[other:newppl:starting#which_shell_are_you_using|your shell]]:
  * tcsh: type ''source /path/to/UV-CDAT/bin/setup_runtime**.csh**''
  * bash, sh/ksh: type ''source /path/to/UV-CDAT/bin/setup_runtime**.sh**''

The table below lists, for each server and version, the path where you can find the //setup_runtime// files

^  Server  ^  CDAT\\ version  ^  JYP\\ version  ^  Path  ^
|  LSCE  |  1.1.0  |  LSCE_08  |  ''/home/share/unix_files/cdat/versions/cdat_install_uv-1.1.0_x86_64_gcc4_08/bin''  |
|  LSCE  |  1.5.1  |  LSCE_10  |  ''/home/share/unix_files/cdat/versions/cdat_install_uv-1.5.1_x86_64_gcc4_10/bin''  |
|  LSCE  |  2.1.0  |  LSCE_13  |  ''/home/share/unix_files/cdat/versions/cdat_install_uv-2.1.0_x86_64_gcc4_13/bin''  |
|  ciclad  |  1.5.1  |  ciclad_10-ng  |  ''/data/jypmce/cdat/versions/cdat_install_uv-1.5.1_x86_64_gcc4_10-ng/bin''  |
|  ciclad  |  2.1.0  |  ciclad_13  |  ''/data/jypmce/cdat/versions/cdat_install_uv-2.1.0_x86_64_gcc4_13/bin''  |
|  curie  |  1.1.0  |  curie_08  |  ''/ccc/work/cont003/dsm/p25jype/cdat/versions/cdat_install_uv-1.1.0_x86_64_gcc4_curie_08/bin''\\ **On curie**, you need to use **setup_cdat**.sh or **setup_cdat**.csh to initialize CDAT,\\ and also to load modules providing access to old librairies\\ ''module load gnu/4.6.3'' and ''module load qt/4.8.6''  |

Example: if you want to use UV-CDAT 2.1.0 at LSCE in a bash shell, you need to do the following (you only need the //source// line. The other lines are optional, just for checking that the initialization was OK)

<code>bash-4.1$ source /home/share/unix_files/cdat/versions/cdat_install_uv-2.1.0_x86_64_gcc4_13/bin/setup_runtime.sh

bash-4.1$ which python
/home/share/unix_files/cdat/versions/cdat_install_uv-2.1.0_x86_64_gcc4_13/bin/python

bash-4.1$ python -c 'import cdms2, vcs'
bash-4.1$</code>

You probably don't want to have to determine what you should type each time you want to use UV-CDAT in a window, so you should define a simple shortcut alias that you can type each time you need to initialize UV-CDAT. If, for example, you want to use UV-CDAT 2.1.0 at LSCE:
  * tcsh: add the following alias definitions to ''~/.cshrc''\\ ''alias cdat13 'source /home/share/unix_files/cdat/versions/cdat_install_uv-2.1.0_x86_64_gcc4_13/bin/setup_runtime.csh'''\\  ''alias wp 'which python'''
  * bash: add the following alias definition to ''~/.bashrc'', ''~/.profile'' or ''~/.bash_profile'' (whichever works for you, or read ''man bash''...)\\ ''alias cdat13='source /home/share/unix_files/cdat/versions/cdat_install_uv-2.1.0_x86_64_gcc4_13/bin/setup_runtime.sh'''\\ ''alias wp='which python'''

If you are going to use a UV-CDAT python script on a cluster or supercomputer, using some kind of batch system using a bash or ksh script, use the following lines before you run the python script
<code>source /home/share/unix_files/cdat/versions/cdat_install_uv-2.1.0_x86_64_gcc4_13/bin/setup_runtime.sh

# Check which python we will be using
mypython=`which python`
echo
echo -e "This script will use the following python: $mypython\n\n"</code>

==== Canopy ====

[[https://www.enthought.com/products/canopy/|Enthought Canopy]] is //a Scientific and Analytic Python Deployment with Integrated Analysis Environment// provided by [[https://www.enthought.com/|Enthought]]. It used to be called EPD (//Enthought Python Distribution//)

Canopy is **available for Windows, Mac and Linux**. You can download the free //Canopy Express// that will already provide many extensions. If you are entitled to use //Canopy Academic//, login from inside Canopy Express and download the extra modules you need


==== Anaconda ====

[[https://www.continuum.io/why-anaconda|Anaconda]] is a distribution similar to [[#canopy]] provided by [[https://www.continuum.io/|Continuum Analytics]].

Anaconda is **available for Windows, Mac and Linux**

Note: Anaconda provides and uses ''conda'' for its installation. Since UV-CDAT is now also installed and maintained with ''conda'', you can read the [[other:uvcdat:conda_notes|Installing and maintaining UV-CDAT with conda]] page for more information, even if you are not going to use UV-CDAT

===== Launching Python =====

Once you have initialized the [[#some_python_distributions|python distribution]] you want to use, follow the instructions below to //start the interpreter from a shell// (on a Windows computer, you will have to start python from the //Start// menu or by clicking a shortcut on your desktop).

Remember that on Linux/Mac you can check which python you are using by typing ''which python''!

==== Useful keyboard shortcuts ====

^  Key  ^  Effect  ^
|  **CTRL-D**  |  Exit the interpreter  |
|  **↑ and ↓**  |  Go to previous/next line(s)  |
|  **CTRL-A**  |  Go to the beginning of the line  |
|  **CTRL-E**  |  Go to the end of the line  |
|  **CTRL-K**  |  Erase from the cursor to the end of the line  |
|  **CTRL-U**  |  Erase from the beginning of the line to the cursor  |
|  **TAB** x 2  |  Do some TAB-completion (context dependent)\\ e.g. ''a = np.cum**TABTAB**''  |
|  **CTRL-C**  |  Interrupt a running script  |
|  **CTRL-Z**  |  Suspend the interpreter and go back to the shell\\ **Do not forget** to go back to the interpreter with ''fg''\\ or to kill it (with ''jobs'' and ''kill %NN'')  |

==== Stand-alone script ====

A python script is just like any other shell script. You don't have to explicitly call the python interpreter, if the interpreter is specified in a ''#!'' [[https://en.wikipedia.org/wiki/Shebang_%28Unix%29|shebang comment]] on the first line of the script with, and the script has its execution bit set (''chmod +x my_script.py''). 

<code>
jypeter@asterix1 - ...jypeter - 54 >cat basic_script.py
#!/usr/bin/env python

import sys

script_name = sys.argv[0]

print('Hello world, I am the ' + script_name + ' script')

# The end

jypeter@asterix1 - ...jypeter - 55 >chmod +x basic_script.py

ypeter@asterix1 - ...jypeter - 56 ># Type here what is required to initialize the distribution you want

jypeter@asterix1 - ...jypeter - 57 >./basic_script.py
Hello world, I am the ./basic_script.py script
</code>

==== Standard interpreter ====

  * ''python'': start the interpreter
  * ''python script.py'': execute //script.py// and **exit**
  * ''python -i script'': execute //script.py// and **stay in the interpreter** (''-i'' = //**I**nteractive//)

Type ''man python'' if you want to see what other command line options are available

==== ipython interpreter ====

The //ipython// interpreter provides more options and commands than the standard python interpreter, but takes more time to load. If you are going to develop by starting and exiting the interpreter many times, it's faster to use the standard interpreter

Starting ipython: ''ipython''

  * [[http://ipython.readthedocs.org/en/stable/interactive/tutorial.html|Using IPython for interactive work]]
  * [[https://damontallen.github.io/IPython-quick-ref-sheets/|Ipython-quick-ref-sheets]]

==== ipython notebook ====

The ipython notebook is a way to interact with python (and other supported interpreted languages) inside a web browser. You can mix //cells// with python commands, cells with the output of the python commands (possibly graphics generated by the commands), and text (using some wiki-like rich text format). This interactive web page, aka notebook, can be saved in a ''my_notebook.ipynb'' file and re-used later (e.g. the notebook provided in the [[other:python:jyp_steps#part_1|introduction to Python, part 1]]).

Starting the notebook server: ''ipython notebook''

<note warning>**Warning!** When you start the notebook server, you will start a python process and a web browser, and a new python process will be started each time you open a new notebook.

Please make sure that you shutdown cleanly each notebook and the server when you are finished, in order not to clutter the local computer or remote server with lots of python processes (especially a server shared by many users!).

You may want to [[other:newppl:starting#determining_the_load_of_a_linux_server|use the 'top' command]] in order to monitor what is happening</note>

/* standard page footer */

\\ \\ \\ 
----
[ [[pmip3:|PMIP3 Wiki Home]] ] -
[ [[pmip3:wiki_help|Help!]] ] -
[ [[wiki:syntax|Wiki syntax]] ]