Notes on Using Numerical Recipes in C++
with
Various Windows Compilers
USING THE NUMERICAL RECIPES IN C++ WITH THE BORLAND
C++ BUILDER VISUAL ENVIRONMENT
Let's assume that you have installed the C++ Recipes in
a directory that we will refer to symbolically as CPPDIR.
Typically CPPDIR will be something like "C:\numrec\c.210\cpp",
and this directory will contain the subdirectories "recipes",
"examples", "other" and "data". In the instructions
below, you will need to substitute you actual installation
directory path wherever we have written CPPDIR.
We consider the case in which you are going to write a
program that uses the Numerical Recipe named svdcmp.cpp.
In this example, we use the Numerical Recipes Example Routine
named xsvdcmp.cpp as our program, since it is a typical
program making use of svdcmp.cpp. This program may be found
in the "examples" subdirectory of the installation directory.
There are several ways in which to use your compiler to
prepare an executable program. We are just detailing one of
the possibilities, one that we have found convenient.
1. First, create a working directory named "xsvdcmp" as
a subdirectory of CPPDIR. (You may name the subdirectory
anything you like and locate it in some other place.
However, we have chosen the name xsvdcmp because
that is the name of the program we will prepare, and we have
chosen the CPPDIR location to keep the pathnames short in
our example.)
2. Into this working directory, copy the files that will
be necessary for the project. These files include:
CPPDIR\examples\xsvdcmp.cpp
CPPDIR\recipes\svdcmp.cpp
CPPDIR\recipes\pythag.cpp
CPPDIR\data\matrx3.dat
Routine xsvdcmp.cpp is the main program. It calls svdcmp(),
which, in turn, makes calls to pythag(). The data file
matrx3.dat is a file containing test matrices, and is
used by xsvdcmp.cpp as a source of input data.
(Note that it is not actually essential that the source
files xsvdcmp.cpp, svdcmp.cpp and pythag.cpp be copied into
a working directory. They could be compiled in their
original location. However, in some cases you will want
to modify the Numerical Recipes codes to suit your own
applications, and collecting the codes into a working
directory is a way of ensuring that the changes you make
do not affect any of your other projects.)
3. Start up the C++ Builder visual programming environment.
4. Make the menu selection "File/New.../Console Wizard".
Our notation here is meant to indicate that you click on the
"File" menu item, then select "New..." from the drop-down
list, and finally select "Console Wizard" from the panel of
options that appears.
In the "Console Wizard" setup box, select:
Window Type: Console
Execution Type: EXE
Then click on "Finish".
5. This will produce a project window with a generic program
name something like "Project1.cpp". We will next name our
project "p_xsvdcmp" and generate a program named "p_svdcmp.cpp"
by the following procedure:
6. Make the menu selection "File/Save Project As ..."
Use the file-selection dialogue box that comes up to navigate
to the directory CPPDIR\xsvdcmp\ and save the project as
"p_svdcmp.bpr". (Again, you could name it anything you like,
but stay with this for now.) The program window will now have
the title p_svdcmp.cpp, and will have contents something like
this:
#pragma hdrstop
#include
//---------------------------------------
#pragma argsused
int main(int argc, char* argv[])
{
return 0;
}
This bit of code framework is provided for you compliments of the
vendor. It can be turned into a workable C++ program by filling
in the routine "main()" with whatever programming instructions
you want to carry out. Presumably, one of these instructions
will involve a call to the routine svdcmp.cpp, since that is the
point of this exercise.
7. In the present case, for example, we could judiciously copy
the contents of the program xsvdcmp.cpp into this file, putting
the "include" statements near the top and the function "main()"
of xsvdcmp.cpp in place of the one-line "main()" routine that
has been automatically provided. This, however, leaves some
Borland-specific statements (particularly, the one about
"condefs.h") in the program, and makes it less portable. We
generally try to isolate vendor-specific code into a separate
file, and the following steps show how to do that.
8. Erase the code that has been provided, and replace it with
the two statements:
#include
#define main
9. Add the necessary files to your project, using the menu
selection "Project/Add to Project" to navigate to and select
the each of the source-code files in the working directory:
CPPDIR\xsvdcmp\xsvdcmp.cpp
CPPDIR\xsvdcmp\svdcmp.cpp
CPPDIR\xsvdcmp\pythag.cpp
10. Your main project source code p_svdcmp.cpp will then read:
#include
USEUNIT("xsvdcmp.cpp");
USEUNIT("svdcmp.cpp");
USEUNIT("pythag.cpp");
#define main
11. There is one more issue. Program xsvdcmp.cpp has an
include-statement that will try to include "nr.h", which it will
look for in the working directory "CPPDIR\xsvdcmp". The actual
location of the "nr.h" file is the "other" subdirectory of the
Numerical Recipes installation directory. There are at least
two solutions to this problem:
a. Use the menu selection "Project/Options/
Directories/Conditionals", and add the text "..\other;"
to the beginning of the include path.
OR
b. Create a file named "nr.h" in the working directory,
and in this file place the single line:
#include "..\other\nr.h"
The virtue of either solution is that your installation will have
only one bona-fide "nr.h" file (the one in the "CPPDIR\other"
directory), rather than having copies in multiple locations.
Therefore, if you ever want to upgrade to a new version, or make
a correction to this file, you only need to edit a single file.
12. Now it is time to compile. Just choose the menu selection
"Project/Build p_svdcmp", wait for all the compilation and linking
to take place. If there are any error messages (other than
innocuous compiler warnings), you will need to do some debugging.
13. You may now run the program with the menu selection "Run/Run",
and witness the program output in the console window that appears.
14. One final note: The "xsvdcmp.cpp" program reads in
several matrices, one-by-one, and pauses between each case so
that you can see the output. Some of the Numerical Recipes
Example Routines, on the other hand, have the property that they
just do a calculation, print some output, and then terminate.
Unfortunately, under Borland C++ Builder, the completion of a
console application causes the console window to close
immediately, with the result that you get only a brief glimpse
at the data you have computed.
The solution to this problem is to create a file named nrexit.cpp
with the following contents:
#include
using namespace std;
// Function used to stall exit in Borland C++ Builder
void nrexit(void)
{
cin.get();
return;
}
#pragma exit nrexit
If you add this source file to your project along with the main
routine and recipes, then you will find that upon completing your
program, the system will leave the console window open until
you press a key on the keyboard.
USING THE NUMERICAL RECIPES IN C++ WITH BORLAND C++ BUILDER FROM
A DOS COMMAND LINE
It is also possible use the Borland C++ Builder compiler to
compile and link Numerical Recipes programs from a DOS command
line. In doing so, you give up the great advantages of the
integrated visual programming environment. However, you gain
the ability to embed your compilation instructions into scripts
or makefiles.
As above, we assume that the C++ Recipes are installed in a
directory that we shall refer to symbolically as CPPDIR. In
this directory there are subdirectories "recipes" containing
the Numerical Recipes, "examples" containing the Numerical
Recipes Example Routines, "other" containing the utility
routines, and "data" containing data files for testing. In
all of the instructions below, you will need to change the
symbol CPPDIR to the actual path of this installation directory.
We consider the case that you are going to write a
program that uses the Numerical Recipe named svdcmp.cpp.
In this example, we use the Numerical Recipes Example Routine
named xsvdcmp.cpp as our program, since it is a typical
program making use of svdcmp.cpp. This program may be found
in the "examples" subdirectory of CPPDIR.
1. First, create a working directory named "xsvdcmp" as a
subdirectory of CPPDIR
> mkdir xsvdcmp
2. Move into this subdirectory
> cd xsvdcmp
3. Move the necessary files into the working directory
> copy ..\examples\xsvdcmp.cpp
> copy ..\recipes\svdcmp.cpp
> copy ..\recipes\pythag.cpp
> copy ..\data\matrx3.dat
Routine xsvdcmp.cpp is the main program. It calls svdcmp(),
which, in turn, makes calls to pythag(). The data file matrx3.dat
is a file containing test matrices, and is used by xsvdcmp.cpp as
a source of input data.
4. The header file "nr.h" is also needed. You may copy this from
the directory "CPP\other", but we suggest instead that you create
a file named "nr.h" in the working directory, and in this file put
the single line
#include "..\other\nr.h"
Then when you include this "nr.h" file, it will in turn include
the "nr.h" file in the "CPPDIR\other" directory.
The virtue of this method is that your installation will have only
one bona-fide include file (the one in the "other" directory),
rather than having copies in multiple locations. Therefore, if
you ever want to upgrade to new versions, or to make corrections,
you only need to modify the master file.
5. Next, compile each of the source files, producing a
corresponding object file .obj for each one:
> bcc32 -c -w- -Od xsvdcmp.cpp
> bcc32 -c -w- -Od svdcmp.cpp
> bcc32 -c -w- -Od pythag.cpp
The first of these commands, for example, compiles the source
file xsvdcmp.cpp and produces the object file xsvdcmp.obj.
6. Then link all of the object files to produce an executable
named "xsvdcmp.exe".
> ILINK32 /ap /w- /x /C xsvdcmp.obj svdcmp.obj pythag.obj
c0x32.obj cw32.lib import32.lib
(This command belongs on a single line.)
The first three object files on this line are object files
produced in the previous step. The following object file and
libraries are compiler-related and should just be
added routinely to your link commands.
7. Following the linking operation, you may run the executable:
> xsvdcmp
USING THE NUMERICAL RECIPES IN C++ WITH THE MICROSOFT
VISUAL C++ DEVELOPMENT ENVIRONMENT
Let's assume that the C++ Recipes have been installed in a
directory that we shall refer to symbolically as CPPDIR. In
this directory there are subdirectories "recipes" containing
the Numerical Recipes, "examples" containing the Numerical
Recipes Example Routines, "other" containing the utilities,
and "data" containing data files for testing. You will need
to change any references to the symbol CPPDIR below to the
actual name of the directory in which these subdirectories
are located.
We shall consider the case that you are going to write a
program that uses the Numerical Recipe named svdcmp.cpp.
In this example, we use the Numerical Recipes Example Routine
named xsvdcmp.cpp as our program, since it is a typical
program making use of svdcmp.cpp. This program may be found
in the "examples" subdirectory of CPPDIR.
There are several ways in which to use your compiler to
prepare the executable program. We are just detailing one of
the possibilities, one that we have found convenient.
1. First, create a working directory named "xsvdcmp" as
a subdirectory of CPPDIR. (You may name the subdirectory
anything you like and locate it in some other place,
of course. However, we have chosen the name xsvdcmp because
that is the name of the program we will prepare, and we have
chosen the CPPDIR location to keep the pathnames short in
our example.)
2. Into this working directory, copy the files that will
be necessary for the project. These files include:
CPPDIR\examples\xsvdcmp.cpp
CPPDIR\recipes\svdcmp.cpp
CPPDIR\recipes\pythag.cpp
CPPDIR\data\matrx3.dat
Routine xsvdcmp.cpp is the main program. It calls svdcmp(),
which, in turn, makes calls to pythag(). The data file
matrx3.dat is a file containing test matrices, and is
used by xsvdcmp.cpp as a source of input data.
(Note that it is not actually essential that the source
files xsvdcmp.cpp, svdcmp.cpp and pythag.cpp be copied into
a working directory. They could be compiled in their
original location. However, in some cases you will want
to modify the Numerical Recipes codes to suit your own
applications, and collecting the codes into a working
directory is a way of ensuring that the changes you make
do not affect any of your other projects.)
3. Start up the Visual C++ programming environment.
4. Make the menu selection "File/New/Win32 Console Application"
Our notation here is meant to indicate that you click on the
"File" menu item, then select "New..." from the drop-down
list, and finally select "Win32 Console Application" from the
panel of options that appears. You will also need to fill in
this information in the panel:
Project Name: xsvdcmp
Location: CPPDIR\xsvdcmp
"Create new workspace" should be checked
In the "Location:" entry, you should replace the symbol CPPDIR
by the actual path name of the directory in which the xsvdcmp
subdirectory has been created.
Click OK and you will then be asked "What kind of Console
Application?". The answer is "Empty Project"
You will then be shown a summary of your choices. Click on OK.
5. Next add the necessary source code files to your project.
Select "Project/Add to Project/Files"
Use file selector to select xsvdcmp.cpp, svdcmp.cpp
and pythag.cpp
6. Program xsvdcmp.cpp has an include-statement that will try
to include "nr.h", which it will look for in the working
directory "CPPDIR\xsvdcmp". The actual location of the "nr.h"
file is the "other" subdirectory of the Numerical Recipes
installation directory.
To deal with this, create a file named "nr.h" in the
working directory, and in this file place the single line:
#include "..\other\nr.h"
The virtue of this step is that your installation will have
only one bona-fide "nr.h" file (the one in the "CPPDIR\other"
directory), rather than having copies in multiple locations.
Therefore, if you ever want to upgrade to a new version, or
make a correction to this file, you only need to edit a single
file.
7. Then compile and link the program with the menu selection
"Build/Build xsvdcmp.exe". If there are any errors reported
(other than innocuous compiler warnings) you will need to do
some debugging.
8. Run the program by selecting "Build/Execute xsvdcmp.exe" from
the menu.
USING THE NUMERICAL RECIPES IN C++ WITH MICROSOFT VISUAL C++
FROM A DOS COMMAND LINE
It is also possible use the Microsoft Visual C++ compiler to
compile and link Numerical Recipes programs from a DOS command
line. In doing so, you give up the great advantages of the
integrated visual programming environment. However,
you gain the ability to embed your compilation instructions into
scripts or makefiles.
As above, we assume that the C++ Recipes are installed in a
directory that we shall refer to symbolically as CPPDIR. In
this directory there are subdirectories "recipes" containing
the Numerical Recipes, "examples" containing the Numerical
Recipes Example Routines, "other" containing the utility
routines, and "data" containing data files for testing. In
all of the instructions below, you will need to change the
symbol CPPDIR to the actual path of this installation directory.
We shall consider the case that you are going to write a
program that uses the Numerical Recipe named svdcmp.cpp.
In this example, we use the Numerical Recipes Example Routine
named xsvdcmp.cpp as our program, since it is a typical
program making use of svdcmp.cpp. This program may be found
in the "examples" subdirectory in the Numerical Recipes
installation.
1. First, create a working directory named "xsvdcmp" as
a subdirectory of CPPDIR
> mkdir xsvdcmp
2. Move into that subdirectory
> cd xsvdcmp
3. Move the necessary files into your working directory
> copy ..\examples\xsvdcmp.cpp
> copy ..\recipes\svdcmp.cpp
> copy ..\recipes\pythag.cpp
> copy ..\data\matrx3.dat
Routine xsvdcmp.cpp is the main program. It calls svdcmp,
which, in turn, makes a call to pythag. The data file matrx3.dat
is a file containing test matrices, and is used by xsvdcmp.cpp
as a source of input data.
4. The header files "nr.h" is also needed. Create
a file named "nr.h" and in this file put the single line
#include "..\other\nr.h"
Then when you include this "nr.h" file, it will in turn include
the "nr.h" file in the "CPPDIR\other" directory.
The virtue of this method is that your installation will have
only one bona-fide include file (the one in the "other" directory),
rather than having copies in multiple locations. Therefore, if you
ever want to upgrade to new versions, or to make corrections, you
only need to modify the master file.
5. Next, compile each of the source files, producing a
corresponding object file .obj for each one.
> CL /c /w xsvdcmp.cpp
> CL /c /w svdcmp.cpp
> CL /c /w pythag.cpp
The compile command used here assumes that the environment
variables PATH and INCLUDE have been properly set, so
that the compiler, include files and libraries can be found.
(When you install Visual C++, the setup creates a batch file,
VCVARS32.BAT, containing commands to modify the PATH, LIB, and
INCLUDE environment variables. If these variables haven't been
set properly, run VCVARS32.BAT before you compile at the
command prompt. VCVARS32.BAT is located in the \bin
subdirectory of the C++ installation.
6. Link all of the object files to produce an executable
named "xsvdcmp.exe".
> LINK xsvdcmp.obj svdcmp.obj pythag.obj
(This command assumes that the LIB environment variable is
properly defined. See the note above about VCVARS32.BAT.)
The three object files on this line are object files
produced in the previous step. If there are any error
messages issued (other than innocuous compiler warnings),
then you will need to do some debugging before continuing.
7. Following the linking operation, you may run the
executable:
> xsvdcmp