Table of Contents

Mastering IPython 4.0

Credits

About the Author

About the Reviewer

www.PacktPub.com

eBooks, discount offers, and more

Why subscribe?

Preface

What this book covers

What you need for this book

Who this book is for

Conventions

Reader feedback

Customer support

Downloading the example code

Downloading the color images of this book

Errata

Piracy

Questions

1. Using IPython for HPC

The need for speed

FORTRAN to the rescue – the problems FORTRAN addressed

Readability

Portability

Efficiency

The computing environment

Choosing between IPython and Fortran

Fortran

IPython

Object-orientation

Ease of adoption

Popularity – Fortran versus IPython

Useful libraries

The cost of building (and maintaining) software

Requirements and specification gathering

Development

Execution

Testing and maintenance

Alternatives

Cross-language development

Prototyping and exploratory development

An example case – Fast Fourier Transform

Fast Fourier Transform

Fortran

Python

Performance concerns

Software engineering concerns

Complexity-based metrics

Size-based metrics

Where we stand now

High Performance Computing

The HPC learning curve

Cloudy with a chance of parallelism (or Amazon's computer is bigger than yours)

HPC and parallelism

Clouds and HPC

Going parallel

Terminology

A parallel programming example

A serial program

A parallel equivalent

Discussion

Summary

2. Advanced Shell Topics

What is IPython?

Installing IPython

All-in-one distributions

Package management with conda

Canopy Package Manager

What happened to the Notebook?

Starting out with the terminal

IPython beyond Python

Shell integration

History

Magic commands

Creating custom magic commands

Cython

Configuring IPython

Debugging

Post-mortem debugging

Debugging at startup

Debugger commands

Read-Eval-Print Loop (REPL) and IPython architecture

Alternative development environments

Spyder

Canopy

PyDev

Others

Summary

3. Stepping Up to IPython for Parallel Computing

Serial processes

Program counters and address spaces

Batch systems

Multitasking and preemption

Time slicing

Threading

Threading in Python

Example

Limitations of threading

Global Interpreter Lock

What happens in an interpreter?

CPython

Multi-core machines

Kill GIL

Using multiple processors

The IPython parallel architecture

Overview

Components

The IPython Engine

The IPython Controller

The IPython Hub

The IPython Scheduler

Getting started with ipyparallel

ipcluster

Hello world

Using map_sync

Asynchronous calls

Synchronizing imports

Parallel magic commands

%px

%%px

%pxresult

%pxconfig

%autopx

Types of parallelism

SIMD

SPMD

ipcluster and mpiexec/mpirun

ipcluster and PBS

Starting the engines

Starting the controller

Using the scripts

MapReduce

Scatter and gather

A more sophisticated method

MIMD

MPMD

Task farming and load balancing

The @parallel function decorator

Data parallelism

No data dependence

External data dependence

Application steering

Debugging

First to the post

Graceful shutdown

Summary

4. Messaging with ZeroMQ and MPI

The storage hierarchy

Address spaces

Data locality

ZeroMQ

A sample ZeroMQ program

The server

The client

Messaging patterns in ZeroMQ

Pairwise

Server

Client

Discussion

Client/server

Server 1

Server 2

Client

Discussion

Publish/subscribe

Publisher

Subscriber

Discussion

Push/Pull

Ventilator

Worker

Sink

Discussion

Important ZeroMQ features

Issues using ZeroMQ

Startup and shutdown

Discovery

MPI

Hello World

Rank and role

Point-to-point communication

Broadcasting

Reduce

Discussion

Change the configuration

Divide the work

Parcel out the work

Process control

Master

Worker

ZeroMQ and IPython

ZeroMQ socket types

IPython components

Client

Engine(s)

Controller

Hub

Scheduler

Connection diagram

Messaging use cases

Registration

Heartbeat

IOPub

Summary

5. Opening the Toolkit – The IPython API

Performance profiling

Using utils.timing

Using %%timeit

Using %%prun

The AsyncResult class

multiprocessing.pool.Pool

Blocking methods

Nonblocking methods

Obtaining results

An example program using various methods

mp.pool.AsyncResult

Getting results

An example program using various methods

AsyncResultSet metadata

Metadata keys

Other metadata

The Client class

Attributes

Methods

The View class

View attributes

Calling Python functions

Synchronous calls

Asynchronous calls

Configurable calls

Job control

DirectView

Data movement

Dictionary-style data access

Scatter and gather

Push and pull

Imports

Discussion

LoadBalancedView

Data movement

Imports

Summary

6. Works Well with Others – IPython and Third-Party Tools

R

The rpy2 module/extension

Installing rpy2

Using Rmagic

The %R magic

The %%R magic

Pulling and pushing

Graphics

Using rpy2.robjects

The basics

Interpreting a string as R

Octave

The oct2py module/extension

Installing oct2py

Using Octave magic

The %octave magic

Tricky issues

The %%octave magic

Pushing and pulling

Graphics

Using the Octave module

Pushing and pulling

Running Octave code

Hy

The hymagic module/extension

Installing hymagic

Using hymagic

The %hylang magic

The %%hylang magic

A quick introduction to Hy

Hello world!

Get used to parentheses

Arithmetic operations are in the wrong place

Function composition is everywhere

Control structures in Hy

Setting variable values

Defining functions

if statements

Conditionals

Loops

Calling Python

Summary

7. Seeing Is Believing– Visualization

Matplotlib

Starting matplotlib

An initial graph

Modifying the graph

Controlling interactivity

Bokeh

Starting Bokeh

An initial graph

Modifying the graph

Customizing graphs

Interactive plots

An example interactive plot

R

Installing ggplot2 and pandas

Using DataFrames

An initial graph

Modifying the graph

A different view

Python-nvd3

Starting Python-nvd3

An initial graph

Putting some tools together

A different type of plot

Summary

8. But It Worked in the Demo! – Testing

Unit testing

A quick introduction

Assertions

Environmental issues

Before it starts – setup

While it is running – mocks

After it finishes – teardown

Writing to be tested

unittest

Important concepts

A test using setUp and tearDown

One-time setUp and tearDown

Decorators

pytest

Installation

Back compatibility

Test discovery

Organizing test files

Assertions

A test using setUp and tearDown

Classic xUnit-style

Being verbose

Using fixtures

Skipping and failing

Monkeypatching

nose2

Installation

Back compatibility

Test discovery

Running individual tests

Assertions, setup, and teardown

Modified xUnit-style

Using decorators

Plugins

Generating XML with the junitxml plugin

Summary

9. Documentation

Inline comments

Using inline comments

Function annotations

Syntax

Parameters

Return values

Semantics

Type hints

Syntax

Semantics

Docstrings

Example

Inheriting docstrings

Recommended elements

One-line docstrings

Syntax

Semantics

Multiline docstrings

Syntax

Semantics

The API

Inputs

Functionality

Outputs

Error conditions

Relationship with other parts of the system

Example uses

Example

reStructuredText

History and goals

Customers

The solution

Overview

Paragraphs

Text styles

Literal blocks

Lists

Enumerated lists

Bulleted lists

Definition lists

Hyperlinks

Sections

Docutils

Installation

Usage

Documenting source files

Sphinx

Installation and startup

Specifying the source files

Summary

10. Visiting Jupyter

Installation and startup

The Dashboard

Creating a notebook

Interacting with Python scripts

Working with cells

Cell tricks

Cell scoping

Cell execution

Restart & Run All

Magics

Cell structure

Code cells

Markdown cells

Raw cells

Heading cells

Being graphic

Using matplotlib

Using Bokeh

Using R

Using Python-nvd3

Format conversion

Other formats

nbviewer

Summary

11. Into the Future

Some history

The Jupyter project

The Notebook

The console

Jupyter Client

The future of Jupyter

Official roadmap

Official subprojects

Direct creation

Incorporation

Incubation

External incubation

IPython

Current activity

The rise of parallelism

The megahertz wars end

The problem

A parallel with the past

The present

Problems are getting bigger and harder

Computers are becoming more parallel

Clouds are rolling in

There is no Big Idea

Pragmatic evolution of techniques

Better tools

The Next Big Idea

Growing professionalism

The NSF

Software Infrastructure for Sustained Innovation

Summary

Index

Mastering IPython 4.0

Mastering IPython 4.0

Copyright © 2016 Packt Publishing

All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, without the prior written permission of the publisher, except in the case of brief quotations embedded in critical articles or reviews.

Every effort has been made in the preparation of this book to ensure the accuracy of the information presented. However, the information contained in this book is sold without warranty, either express or implied. Neither the author, nor Packt Publishing, and its dealers and distributors will be held liable for any damages caused or alleged to be caused directly or indirectly by this book.

Packt Publishing has endeavored to provide trademark information about all of the companies and products mentioned in this book by the appropriate use of capitals. However, Packt Publishing cannot guarantee the accuracy of this information.

First published: May 2016

Production reference: 1240516

Published by Packt Publishing Ltd.

Livery Place

35 Livery Street

Birmingham B3 2PB, UK.

ISBN 978-1-78588-841-0

www.packtpub.com

Credits

Author

Thomas Bitterman

Reviewers

James Davidheiser

Dipanjan Deb

Commissioning Editor

Veena Pagare

Acquisition Editor

Manish Nainani

Content Development Editor

Deepti Thore

Technical Editor

Tanmayee Patil

Copy Editor

Vikrant Phadke

Project Coordinator

Shweta H. Birwatkar

Proofreader

Safis Editing

Indexer

Monica Ajmera Mehta

Graphics

Disha Haria

Production Coordinator

Nilesh Mohite

Cover Work

Nilesh Mohite

About the Author

Thomas Bitterman has a PhD from Louisiana State University and is currently an assistant professor at Wittenberg University. He previously worked in the industry for many years, including a recent stint at the Ohio Supercomputer Center. Thomas has experience in such diverse areas as electronic commerce, enterprise messaging, wireless networking, supercomputing, and academia. He also likes to keep sharp, writing material for Packt Publishing and O'Reilly in his copious free time.

I would like to thank my girlfriend for putting up with the amount of time this writing has taken away.

The Ohio Supercomputer Center has been very generous with their resources. The AweSim infrastructure (https://awesim.org/en/) is truly years ahead of anything else in the field. The original architect must have been a genius.

And last (but by no means least), I would like to thank Deepti Thore, Manish Nainani, Tanmayee Patil and everyone else at Packt, without whose patience and expertise this project would have never come to fruition.

About the Reviewer

Dipanjan Deb is an experienced analytics professional with about 16 years of cumulative experience in machine/statistical learning, data mining, and predictive analytics across the healthcare, maritime, automotive, energy, CPG, and human resource domains. He is highly proficient in developing cutting-edge analytic solutions using open source and commercial packages to integrate multiple systems to provide massively parallelized and large-scale optimization.

He has extensive experience in building analytics teams of data scientists that deliver high-quality solutions. Dipanjan strategizes and collaborates with industry experts, technical experts, and data scientists to build analytic solutions that shorten the transition from a POC to commercial release.

He is well versed in overarching supervised, semi-supervised, and unsupervised learning algorithm implementations in R, Python, Vowpal Wabbit, Julia, and SAS; and distributed frameworks, including Hadoop and Spark, both in-premise and in cloud environments. He is a part-time Kaggler and IoT/IIoT enthusiast (Raspberry Pi and Arduino prototyping).

www.PacktPub.com

eBooks, discount offers, and more

Did you know that Packt offers eBook versions of every book published, with PDF and ePub files available? You can upgrade to the eBook version at www.PacktPub.com and as a print book customer, you are entitled to a discount on the eBook copy. Get in touch with us at for more details.

At www.PacktPub.com, you can also read a collection of free technical articles, sign up for a range of free newsletters and receive exclusive discounts and offers on Packt books and eBooks.

https://www2.packtpub.com/books/subscription/packtlib

Do you need instant solutions to your IT questions? PacktLib is Packt's online digital book library. Here, you can search, access, and read Packt's entire library of books.

Why subscribe?

Fully searchable across every book published by Packt

Copy and paste, print, and bookmark content

On demand and accessible via a web browser

Preface

Welcome to the world of IPython 4.0, which is used in high performance and parallel environments. Python itself has been gaining traction in these areas, and IPython builds on these strengths.

High-performance computing (HPC) has a number of characteristics that make it different from the majority of other computing fields. We will start with a brief overview of what makes HPC different and how IPython can be a game-changing technology.

We will then start on the IPython command line. Now that Jupyter has split from the IPython project, this is the primary means by which the developer will interface with the language. This is an important enough topic to devote two chapters to. In the first, we will concentrate on basic commands and gaining an understanding of how IPython carries them out. The second chapter will cover more advanced commands, leaving the reader with a solid grounding in what the command line has to offer.

After that, we will address some particulars of parallel programming. IPython parallel is a package that contains a great deal of functionality required for parallel computing in IPython. It supports a flexible set of parallel programming models and is critical if you want to harness the power of massively parallel architectures.

Programs running in parallel but on separate processors often need to exchange information despite having separate address spaces. They do so by sending messages. We will cover two messaging systems, ZeroMQ and MPI, and in relation to both how they are used in already existing programs and how they interact with IPython.

We will then take a deeper look at libraries that can enhance your productivity, whether included in IPython itself or provided by third-parties. There are far too many tools to cover in this book, and more are being written all the time, but a few will be particularly applicable to parallel and HPC projects.

An important feature of IPython is its support for visualization of datasets and results. We will cover some of IPython's extensive capabilities, whether built-in to the language or through external tools.

Rounding off the book will be material on testing and documentation. These oft-neglected topics separate truly professional code from also-rans, and we will look at IPython's support for these phases of development. Finally, we will discuss where the field is going. Part of the fun of programming is that everything changes every other year (if not sooner), and we will speculate on what the future might hold.

What this book covers

Chapter 1, Using IPython for HPC, discusses the distinctive features of parallel and HPC computing and how IPython fits in (and how it does not).

Chapter 2, Advanced Shell Topics, introduces the basics of working with the command line including debugging, shell commands, and embedding, and describes the architecture that underlies it.

Chapter 3, Stepping Up to IPython for Parallel Computing, explores the features of IPython that relate directly to parallel computing. Different parallel architectures will be introduced and IPython's support for them will be described.

Chapter 4, Messaging with ZeroMQ and MPI, covers these messaging systems and how they can be used with IPython and parallel architectures.

Chapter 5, Opening the Toolkit – The IPython API, introduces some of the more useful libraries included with IPython, including performance profiling, AsyncResult, and View.

Chapter 6, Works Well with Others – IPython and Third-Party Tools, describes tools created by third-parties, including R, Octave, and Hy. The appropriate magics are introduced, passing data between the languages is demonstrated, and sample programs are examined.

Chapter 7, Seeing Is Believing – Visualization, provides an overview of various tools that can be used to produce visual representations of data and results. Matplotlib, bokeh, R, and Python-nvd3 are covered.

Chapter 8, But It Worked in the Demo! – Testing, covers issues related to unit testing programs and the tools IPython provides to support this process. Frameworks discussed include unittest, pytest, and nose2.

Chapter 9, Documentation, discusses the different audience for documentation and their requirements. The use of docstrings with reStructuredText, docutils, and Sphinx is demonstrated in the context of good documentation standards.

Chapter 10, Visiting Jupyter, introduces the Jupyter notebook and describes its use as a laboratory notebook combining data and calculations.

Chapter 11, Into the Future, reflects on the current rapid rate of change and speculates on what the future may hold, both in terms of the recent split between IPython and the Jupyter project and relative to some emerging trends in scientific computing in general.

What you need for this book

This book was written using the IPython 4.0 and 4.0.1 (stable) releases from August 2015 through March 2016; all examples and functions should work in these versions. When third-party libraries are required, the version used will be noted at that time. Given the rate of change of the IPython 4 implementation, the various third-party libraries, and the field in general, it is an unfortunate fact that getting every example in this book to run on every reader's machine is doubtful. Add to that the differences in machine architecture and configuration and the problem only worsens. Despite efforts to write straightforward, portable code, the reader should not be surprised if some work is required to make the odd example work on their system.

Who this book is for

This book is for IPython developers who want to make the most of IPython and perform advanced scientific computing with IPython, utilizing the ease of interactive computing.

It is ideal for users who wish to learn about the interactive and parallel computing properties of IPython 4.0, along with its integration with third-party tools and concepts such as testing and documenting results.

Conventions

In this book, you will find a number of text styles that distinguish between different kinds of information. Here are some examples of these styles and an explanation of their meaning.

Code words in text, database table names, folder names, filenames, file extensions, pathnames, dummy URLs, user input, and Twitter handles are shown as follows: These methods must be named setUpClass and tearDownClass and must be implemented as class methods.

A block of code is set as follows:

    def setUp(self):

        print(Doing setUp)

        self.vals = np.zeros((10), dtype=np.int32)

        self.randGen = myrand.MyRand( )

Any command-line input or output is written as follows:

pip install pytest

New terms and important words are shown in bold. Words that you see on the screen, for example, in menus or dialog boxes, appear in the text like this: At a finer level of detail are the bugs listed under the Issues tag and the new features under the Pulls tag.

Note

Warnings or important notes appear in a box like this.