Python for Science

Shane Grigsby

What is python? Why python?

• Interpreted, object oriented language• Free and open source

• Focus is on readability• Fast to write and debug code• Large community– Lots of documentation– Lots of packages…

• General purpose language

The python scientific stack:

Python: Fast to write, slow to run?

• Depends on how you use it– if something is slow there is probably a faster way to do it!

Are you using a numeric library?Are you slicing through arrays, or looping lists?Is your code vectorized?

Numpy calls fortran code to do array operationsMany other libraries call C for operations… …or have functions written in both C and python

e.g., scipy.spatial.kdtree vs. scipy.spatial.cKDTree

How is python different from MATLAB?

• Indexing starts at 0• Space delimited• Default behavior is element-by-element when dealing with arrays

• Functions use ()’s, indexes use []’s, tuples and dictionaries use {}’s• You don’t have to use a ‘;’ on every command• Object oriented

See also: http://mathesaurus.sourceforge.net/matlab-python-xref.pdf

Today’s tutorial• Intro to the scientific stack• Importing modules• Intro to the development environments:

– Spyder– iPython

• Indexing• Defining functions• Plotting and graphics• Intro to data structures• Basic looping (maybe…)• Additional pandas

– data import from clipboard– time series (on your own)

Notebooks:• SARP python tutorial• Prism Data• Regression loopsFiles:• Leaf_Angles.csv• good_12_leafangle.h5• monthly.nc # Optional• *_spec.txt

Terminals and Prompts

• We’ll use python and python tools from three different ‘prompts’:– The system prompt (i.e., cmd)– From spyder– From the iPython Notebook

• Note that these will all run the same version of python, but with slightly different behaviors

Notebooks:• SARP python tutorial• Prism Data• Regression loopsFiles:• Leaf_Angles.csv• good_12_leafangle.h5• monthly.nc # Optional

Imports

Basic python is sparse……but we can import!

import tables

import numpy as np

from pylab import *

From scipy import ones, array

Example time: Using the iPython notebook

• Notes:– ‘%’s are specific to the iPython NB’s; they won’t

work in your terminal– We’ll use: %pylab inline• This doesn’t work for 3D or interactive plots (yet)• Use spyder or ipython (without the notebook) to access

interactive graphics.

Imports

• Pull from the python install directory first– i.e., lib/python2.7/site-packages

• Pull from the current directory second• Conflicting imports are replaced with the last

import

Defining a function• Multiline comment• Keyword arguments

– can use default values• Definition syntax

– return is optional• Constants defined at

the top of the script

• Top line brings in physical constants, so we don’t have to define them ourselves…

from scipy.constants import *

def Xwave(wavelength, temp, unit=1): X_wave = (h*c)/(k*(wavelength*unit)*temp) return X_wave def Lwave(wavelength, temp, unit=1):

"""Calculates L given wavelength and Temp To get M, multiply by pi Note that units are: W * m**-2 * sr**-1 * m**-1 I.e, units are given in meter of spectrum multiply by nm to get: W * m**-2 *sr**-1 nm**-1""”

X_funct= Xwave(wavelength, temp, unit) L=(2*h*(c**2))/(((wavelength*unit)**5)*(exp(X_funct)-1)) return L

ang = 1E-10nm = 1E-9um = 1E-6Cm = 1E-2hH = 1.0kH = 1E3mH = 1E6gH = 1E9tH = 1E12

Defining Functions

• Functions are defined using ‘def’, function name, ‘()’’s with parameters, and an ending ‘:’

• The function body is demarcated using white space (as in for loops)

• Functions are called using the function name, ‘()’’s, and input parameters– Note that the input parameters don’t have to

match the names that requested…

Looping in pythonfor item in list: print item • ‘item’ is a variable name; it is not declared in advance; it is arbitrary

(i.e., could be ‘i’, ‘point’, or some other label). There could also be more them one variable here—see next slide…

• ‘for’ and ‘in’ are syntactically required; they bracket our variables.• ‘list’ could be another data structure (dictionary, array, etc.), or could be

a function in advanced use.• Note: else and elseif are not required, but can be used• Note: the white space is required—either a tab or four spaces• Note: we don’t need to know how many items are in the data structure

for i in range(len(list)): # NO… print list[i] # executes, but is wrong

A more advanced loop:from liblas import fileimport scipy

f = file.File('/Users/grigsbye/Downloads/Alameda_park_trees_pts.las',mode='r')treeData = scipy.ones((len(f),3))for i, p in enumerate(f): treeData[i,0], treeData[i,1], treeData[i,2] = p.x,p.y,p.z

• First line imports a special module to read .las files• Third line reads a .las file into a python object we can loop over• Fourth line creates a scipy/numpy array to hold our data• ‘Enumerate’ returns an index number for each point that we are looping over• The last line assigns the x, y, and z values to our array, incrementing the index by one with

each loop • For a more complete guide to looping, see:http://nedbatchelder.com/text/iter.html