Introduction au Python pour des scientifiques Module1 ...dkhvoros/ptraining/Journee-python-2014... · Introduction au Python pour des scientifiques Module1: Initiation ... 10h00 –

Introduction au Python pour des scientifiques

Module1: Initiation

Sophie Cloché, Marc-Antoine Drouin, Dmitry Khvorostyanov,Estelle Lorant, Vincent Noël, Jean-Yves Peterschmitt

avec des contributions précieuses de

Ludmila Klenov, Yohann Morille

Institut Pierre et Simon Laplace

Journée de formation IPSL – 2 décembre 2014 Ecole Polytechnique, Palaiseau, France

Programme de la journéeMatin (maintenant) ! 9h30 : accueil café ! 10h00 – 12h45: Introduction générale à Python (amphi Becquerel) !!Midi !12h45 – 14h00 : repas (salle Aquarium) !!Après-midi 14h – 18h00: Travaux Pratiques (salles informatiques 32 et 33 de l'X)

Presentation outline

● Introduction to Python (Dmitry) ● Language syntax introduction (Dmitry & Marc-Antoine) ● Coffee break/Questions ● Scientific Python libraries (Marc-Antoine) ● Hands-on session overview (Dmitry)

Python: Brief History

• February 1991, the 1st published version (0.9.0): classes, exception handling, functions, core datatypes (list, dict, str...), module system

• January 1994, v1.0: functional programming paradigm tools lambda, map, filter and reduce.

• October 1996, v1.4: keyword arguments [ex.: c = jump(high = True)], complex numbers

• October, 2000, v2.0: list comprehensions (we'll see one today!), a garbage collection system (makes life easier with memory management)

• December, 2001, v2.2: new-style classes (unification of Python's types (written in C) and classes (written in Python) into one hierarchy; generators

• December, 2008, v3.0: "Reduce feature duplication by removing old ways of doing things" => no more backward compatibility!

• June, 2009, v3.1, July, 2010, v2.7. Python 2.7 was released to coincide with Python 3.1, included some features 3.1, as well as warnings for old features. The 2.7 was the last 2.x release.

Van Rossum at OSCON 2006 Photo taken from WikiPedia

Guido Van Rossum: "Benevolent Dictator For Life"

!Python creator, he oversees the

development process, making decisions

Why Python?● High-level flexible programming language

● Development speed ● Code readability ● Collaborative development, program component (re-)usability

combined with

● Rich variety of libraries for wide range of applications:

● Scientific computing − Data analysis, statistics, input/output − Data visualization − Image/signal processing, data mining, machine learning....

● Shell programming (a readable bash alternative!) ● Web / Databases ● Human-computer interaction (graphical user interfaces, virtual reality...) ● Video-games, networking, sound/text processing, linguistics, neurosciences... ● Embedded systems

● Free, open-source, permissive license (PSFL)

● Large, friendly, dynamically growing community

- AFPY: rencontres, forums, formations, documents éducatifs !! - "Planets" et "User Groups" de Python: forums, blogs, groupes thématiques... Stack Overflow: réponses à presque toutes vos questions! ! - EuroSciPy, PyCon, EuroPython, PyCon UK, OSCON, ... - échanges de compétences, tutoriaux, événements sociaux !! - Google, BitTorrent, SpamBayes, Plone CMS, MailMan, Zope, Django - Python-powered! !Des initiatives humanitaires dans des pays en développement: - Argentine Py Group: CDPedia - One Laptop Per Child: Sugar - Bungeni: système d'information pour les parlements africains + ouverts et accessibles - Python African Tour: Formations Maroc, Sénégal, Zambie, Nigéria...

Large, friendly, dynamically growing community

map(lambda x: round(x**2, 1), [0.1, 0.4, 5]) >> [0.0, 0.2, 25.0]!filter(lambda x: x % 3 == 0, [1,2,6,5,7]) >> [6]!

reduce(lambda x, y: x * y, [1,2,3,4]) >> 24

● Functional

class ncVar:! def __init__(self, file): ! self.var = []! self.nc = !! ! netCDF4.Dataset(file)! def read(self, varname):! self.var = self.nc.variables!! ! [varname][:]! !V = ncVar('O3.nc')!V.read('tem2')!print(V.var.shape)

● Object-oriented

def filter_div(v,N) : ! y = []! for x in v:! if (x % N == 0):! y.append(x)! return y!!filter_div([1,2,6,5,7],3) >> [6]

● Imperative

High-level and flexible programming language

dates = array([datetime(*map(int,row[:5])) for row in data])

● Multiple ways to do the same thing, often with very compact syntax:

Three programming paradigms

=> see TP2 for three other ways to do this !

●Can get very quickly what you need with a simple straightforward code:

import cartopy.crs as ccrs!import netCDF4!!nc = netCDF4.Dataset('sinwave.nc')!lons = nc.variables['lon'][:]!lats = nc.variables['lat'][:]!data = nc.variables['data'][:]!!ax = plt.axes(projection=ccrs.Mollweide())!ax.contourf(lons, lats, data,! transform=ccrs.PlateCarree(),cmap='spectral') !ax.coastlines()!ax.set_global()!show()

High-level and flexible language + libraries

import cartopy.crs as ccrs!import matplotlib.pyplot as plt!ax = plt.axes(projection=ccrs.Mollweide())!ax.stock_img()!plt.show()

...And create complex software architectures : !Twisted : ~ 208,000 lines of code (network framework) Zope : ~ 404,000 (web application server & framework) Plone : ~ 573,000 (content managment system)

High-level and flexible language + libraries

Python est utilisé dans de nombreux domaines: !programmation système, réseaux, automatisation, ingénierie, serveurs

web, bases de données, images de synthèse, manipulation de texte, jeux vidéos, data mining, applications commerciales...

!... et en sciences ( Ouah ! ) !!!Python s’adapte à plusieurs types de programmation : !• Séquentielle / parallèle • Procédurale/ fonctionnele / orientée objet • Test-driven • Bête/intuitive (populaire dans les labos)

High-level and flexible language: résumé

● Available with Linux (standard distribution included), Mac OS X (standard included), Windows, Android, etc.

● Interactive mode (python, ipython, notebook) or automatic (scripts, libraries, frameworks)

● Interpreters and compilers (CPython, PyPy, Cython, py2exe) ● Interfaces with other languages : Fortran (f2py), C/C++, Java (Jython, JPype,

Jepp)....

High-level and flexible language + framework

Est-ce que Python va dominer le monde? [ du travail ]

Percentage of jobs in web job listings since 2005

(from indeed.com)

% growth% job postings

Distributions Python

• python.org = python + librairies standard o numpy, etc à installer soi-meme : complexe

• distributions "intégrées" python scientifique o python + librairie standard + numpy + lecture/écriture + etc. o plus facile, à privilégier si possible o vérifier la license... (parfois payant)

• CANOPY (Enthought Python Distribution, conseillé) http://www.enthought.com

• Anaconda (http://www.continuum.io/downloads)

• UV-CDAT (http://uvcdat.llnl.gov)

• SAGE (calcul formel) http://www.sagemath.org • python(x,y) http://www.pythonxy.com

•Restez pour l'après-midi •Google : Python, science,

numpy, matplotlib, tutorials...

•Bouquins (souvent en anglais,mais pas toujours)

Pour aller plus loin

TPs de cet après-midi

login machine python2... python40 !!Mot de pass: sera communiqué discrètement !Actualiser les PATHs: !$ export PATH=$HOME/../python1/anaconda/bin:$PATH !Données, énoncés, solutions (seront communiqué au fur et à mésure de votre progréssion): cp $HOME/../python1/TPx . ou x = [1, 2, 3, 4, 5]

Merci pour votre attentionQuestions ?