Computing and Statistical Data Analysis (PH4515, UofL PG Lectures) : Computing and Statistical Data Analysis (PH4515, UofL PG Lectures) Glen Cowan
RHUL Physics Computing and Statistical Data Analysis Glen Cowan
Physics Department
Royal Holloway, University of London
Egham, Surrey TW20 0EX
01784 443452
g.cowan@rhul.ac.uk
www.pp.rhul.ac.uk/~cowan/stat_course.html
Outline : Outline Glen Cowan
RHUL Physics Computing and Statistical Data Analysis 1st 4 weeks will be a crash course in C++
Quick overview of the important stuff
Use UNIX (Linux) environment
Intro to tools like ROOT, gmake, debugger
From around week 5, statistical data analysis
Probability, random variables, Monte Carlo methods
Statistical tests
Parameter estimation
Data analysis exercises will use C++ tools
Coursework, exams, etc. : Coursework, exams, etc. Glen Cowan
RHUL Physics Computing and Statistical Data Analysis For C++ part
Computer based exercises -- see course web site.
For data analysis part
More exercises, many computer based
For PH4515 students
Written exam at end of year (70% of mark),
no questions on C++, only statistical data analysis.
For PhD students
No material from this course in exam
C++ Outline : C++ Outline Glen Cowan
RHUL Physics Computing and Statistical Data Analysis Approximately by lecture for 1st 4 weeks:
1 Introduction to C++ and UNIX environment
2 Variables, types, expressions
3 Loops, type casting, functions
4 Files and streams
5 Arrays, strings, pointers
6 Classes, intro to Object Oriented Programming
7 Memory allocation, operator overloading, templates
8 Inheritance, STL, gmake, ddd
Some resources (computing part) : Some resources (computing part) Glen Cowan
RHUL Physics Computing and Statistical Data Analysis There are many web based resources, e.g.,
www.doc.ic.ac.uk/~wjk/C++Intro (Rob Miller, IC course)
www.cplusplus.com (online reference)
See links on course site or google for “C++ tutorial”, etc.
There are thousands of books – see e.g.
W. Savitch, Problem Solving with C++, 4th edition
(lots of detail – very thick).
B. Stroustrup, The C++ Programming Language
(the classic – even thicker).
Lippman, Lajoie (& Moo), C++ Primer, A-W, 1998.
Introduction to UNIX/Linux : We will learn C++ using the Linux operating system
Open source, quasi-free version of UNIX
UNIX and C developed ~1970 at Bell Labs
Short, cryptic commands: cd, ls, grep, …
Other operating systems in 1970s, 80s ‘better’, (e.g. VMS) but, fast ‘RISC processors’ in early 1990s needed a cheap solution → we got UNIX
In 1991, Linus Torvalds writes a free, open source version of UNIX called Linux.
We currently use the distribution from CERN Introduction to UNIX/Linux Glen Cowan
RHUL Physics Computing and Statistical Data Analysis
Basic UNIX : Basic UNIX Glen Cowan
RHUL Physics Computing and Statistical Data Analysis UNIX tasks divide neatly into:
interaction between operating system and computer (the kernel),
interaction between operating system and user (the shell).
Several shells (i.e. command sets) available: sh, csh, tcsh, bash, … Shell commands typed at a prompt, here [linappserv1]~>
often set to indicate name of computer: Command pwd to “print working directory”, i.e., show the directory
(folder) you’re sitting in.
UNIX file structure : UNIX file structure Glen Cowan
RHUL Physics Computing and Statistical Data Analysis Tree-like structure for files and directories (like folders): / ← the ‘root’ directory usr/ bin/ home/ sys/ tmp/ ... smith/ jones/ jackson/ ... WWW/ code/ thesis/ ...
Simple UNIX file tricks : Simple UNIX file tricks Glen Cowan
RHUL Physics Computing and Statistical Data Analysis A complete file name specifies the entire ‘path’
/home/jones/thesis/chapter1.tex
A tilde points to the home directory:
~/thesis/chapter1.tex ← the logged in user (e.g. jones)
~smith/analysis/result.dat ← a different user
Single dot points to current directory, two dots for the one above:
/home/jones/thesis ← current directory
../code ← same as /home/jones/code
A few UNIX commands (case sensitive!) : A few UNIX commands (case sensitive!) Glen Cowan
RHUL Physics Computing and Statistical Data Analysis pwd Show present working directory
ls List files in present working directory
ls -la List files of present working directory with details
man ls Show manual page for ls. Works for all commands.
man -k keyword Searches man pages for info on “keyword”.
cd Change present working directory to home directory.
mkdir foo Create subdirectory foo
cd foo Change to subdirectory foo (go down in tree)
cd .. Go up one directory in tree
rmdir foo Remove subdirectory foo (must be empty)
xemacs foo & Edit file foo with XEmacs (& to run in background)
more foo Display file foo (space for next page)
less foo Similar to more foo, but able to back up (q to quit)
rm foo Delete file foo
A few more UNIX commands : A few more UNIX commands Glen Cowan
RHUL Physics Computing and Statistical Data Analysis cp foo bar Copy file foo to file bar, e.g., cp ~smith/foo ./
copies Smith’s file foo to my current directory
mv foo bar Rename file foo to bar
lpr foo Print file foo. Use -P to specify print queue, e.g.,
lpr -Plj1 foo (site dependent).
ps Show existing processes
kill 345 Kill process 345 (kill -9 as last resort)
./foo Run executable program foo in current directory
ctrl-c Terminate currently executing program
chmod ug+x foo Change access mode so user and group have
privilege to execute foo (Check with ls -la) Better to read a book or online tutorial and use man pages
Introduction to C++ : Introduction to C++ Glen Cowan
RHUL Physics Computing and Statistical Data Analysis Language C developed (from B) ~ 1970 at Bell Labs
Used to create parts of UNIX
C++ derived from C in early 1980s by Bjarne Stroustrup
“C with classes”, i.e., user-defined data types that
allow “Object Oriented Programming”.
Java syntax based largely on C++ (head start if you know java)
C++ is case sensitive (a not same as A).
Currently most widely used programming language in High
Energy Physics and many other science/engineering fields.
Recent switch after four decades of FORTRAN.
Compiling and running a simple C++ program : Compiling and running a simple C++ program Glen Cowan
RHUL Physics Computing and Statistical Data Analysis // My first C++ program
#include
using namespace std;
int main(){
cout << "Hello World!" << endl;
return 0;
} Using,e.g., xemacs, create a file HelloWorld.cc containing: We now need to compile the file (creates machine-readable code):
g++ -o HelloWorld HelloWorld.cc
Invokes compiler (gcc) name of output file source code Run the program: ./HelloWorld ← you type this
Hello World! ← computer shows this
Notes on compiling/linking : Notes on compiling/linking Glen Cowan
RHUL Physics Computing and Statistical Data Analysis g++ -o HelloWorld HelloWorld.cc
is an abbreviated way of saying first
g++ -c HelloWorld.cc
Compiler (-c) produces HelloWorld.o. (‘object files’)
Then ‘link’ the object file(s) with
g++ -o HelloWorld HelloWorld.o If the program contains more than one source file, list with
spaces; use \ to continue to a new line: g++ -o HelloWorld HelloWorld.cc Bonjour.cc \
GruessGott.cc YoDude.cc
Writing programs in the Real World : Writing programs in the Real World Glen Cowan
RHUL Physics Computing and Statistical Data Analysis Usually create a new directory for each new program.
For trivial programs, type compile commands by hand.
For less trivial but still small projects, create a file (a ‘script’) to contain the commands needed to build the program: #!/bin/sh
# File build.sh to build HelloWorld
g++ -o HelloWorld HelloWorld.cc Bonjour.cc \
GruessGott.cc YoDude.cc To use, must first have ‘execute access’ for the file:
chmod ug+x build.sh ← do this only once
./build.sh ← executes the script
A closer look at HelloWorld.cc : A closer look at HelloWorld.cc Glen Cowan
RHUL Physics Computing and Statistical Data Analysis // My first C++ program is a comment (preferred style) The older ‘C style’ comments are also allowed (cannot be nested): /*
These lines
here are comments
*/
/* and so are these */ You should include enough comments in your code to make it
understandable by someone else (or by yourself, later).
Each file should start with comments indicating author’s name,
main purpose of the code, required input, etc.
More HelloWorld.cc − include statements : More HelloWorld.cc − include statements Glen Cowan
RHUL Physics Computing and Statistical Data Analysis #include is a compiler directive. Compiler directives start with #. These statements are not executed
at run time but rather provide information to the compiler.
#include tells the compiler that the code will use
library routines whose definitions can be found in a file called
iostream, usually located somewhere under /usr/include
Old style was #include
iostream contains functions that perform i/o operations to
communicate with keyboard and monitor.
In this case, we are using the iostream object cout to send text
to the monitor. We will include it in almost all programs.
More HelloWorld.cc : More HelloWorld.cc Glen Cowan
RHUL Physics Computing and Statistical Data Analysis using namespace std; More later. For now, just do it. A C++ program is made up of functions. Every program contains
exactly one function called main:
int main(){
// body of program goes here
return 0;
}
Functions “return” a value of a given type; main returns int (integer).
The () are for arguments. Here main takes no arguments.
The body of a function is enclosed in curly braces: { }
return 0; means main returns a value of 0.
Finishing up HelloWorld.cc : Finishing up HelloWorld.cc Glen Cowan
RHUL Physics Computing and Statistical Data Analysis The ‘meat’ of HelloWorld is contained in the line
cout << "Hello World!" << endl;
Like all statements, it ends with a semi-colon.
cout is an “output stream object”.
You send strings (sequences of characters) to cout with <<
We will see it also works for numerical quantities (automatic
conversion to strings), e.g., cout << "x = " << x << endl;
Sending endl to cout indicates a new line. (Try omitting this.)
Old style was "Hello World!\n"
Wrapping up lecture 1 : Wrapping up lecture 1 Glen Cowan
RHUL Physics Computing and Statistical Data Analysis We have seen just enough UNIX to get started. Try out the
commands from the lecture and have a look at the online tutorials.
We have seen how to compile and run the simplest possible C++
program. Log in, enter the code into a file and get it to run.
If you can’t get it to work, shout for help.
Try entering the compile commands into a short script and build
the program in this way.
Later we will see a more elegant (read: cryptic) way of
building larger programs with a utility called gmake.
Next lecture: variables, types, expressions