*** empty log message ***

[bbtk.git] / kernel / doc / bbtkUsersGuide / bbtkUsersGuide.tex

% ==========================================
\documentclass[11pt,final,a4paper]{article}
\input{config.tex}
\begin{document}
\title{The Black Box Toolkit\\User's Guide}
\date{\today}
\author{Laurent Guigues}
\maketitle
% ==========================================
\tableofcontents
% ==========================================


% ==========================================
%\section*{Abstract}
% ==========================================
\newpage
% ==========================================
\hrule
\section{What is {\bf bbtk} ?}
% ==========================================
\BBTK(\bbtkns) is a set of tools 
(\CPP libraries and executables) 
providing a \CPP framework for the definition 
of elementary processing \emph{units}, called {\bf black boxes}, 
and the definition and execution of processing \emph{chains} 
made up of these black boxes. 

% ==========================================
\subsection{The black box philosophy}
% ==========================================
\href{http://en.wikipedia.org/wiki/Black_box_\%28disambiguation\%29}{Wikipedia}
defines a {\bf black box} as 
\emph{``any component in a system in which only the input and output 
characteristics are of interest, without regard to its internal mechanism 
or structure''}. 
I would add something very important to this definition : 
not only the inputs and outputs are of interest but also 
\emph{what the box does} !
Hence, I would say that a black box is any \emph{\bf documented} 
component of a system, letting know the user 
\emph{\bf what} the box is supposed to do and 
\emph{\bf how to use it}
but not \emph{\bf how it does it}.

\BBTK provides a systematic framework 
to encapsulate (or ``wrap'') any 
existing \texttt{C} or \CPP processing code into an object 
(a black box) having a {\bf generic symbolic interface}, where 

\begin{itemize}
\item{\bf generic} means that the interface is \emph{the same} 
for all boxes. Hence one does not need to know which particular 
method allows, say, to set a particular input or 
get a particular output of the box. 
One can use a black box in a purely abstract way. 
\item{\bf symbolic} means that a particular 
input or output is referenced by a 'name', that is by a symbol 
which identifies the input or output. 
It also means that symbolic information (text!) is 
attached to a box: description of the box, author, 
description of its inputs and outputs, etc.
\end{itemize}

(in fact, genericity is achieved because the interface is symbolic. 
I let you think about this\dots)

Of course, symbolic data attached to box can be 
{\bf queried}: what are the inputs/outputs of the box ? 
what are their type ? description ? etc.
This allows {\bf automatic documentation} of boxes. 

The abstract definition of black boxes is the most basic 
aspect of \BBTK architecture. 
Another key aspect is the groupement of black boxes into 
so called {\bf packages}, 
which are \emph{dynamic libraries} which can also 
be queried, in particular about the boxes they provide. 
The package structure then offers a mechanism which 
is like a \emph{'plug-in'} mechanism.
\BBTK provides the methods to load a package at run-time, 
and create instances of the boxes it contains. 

These two mechanisms (black boxes and packages) 
then gives the way to:

\begin{itemize}
\item The definition of an {\bf interpreted language}, 
which allows to manipulate packages and boxes very easily in symbolic way. 
\BBTK provides one: \bbi (the Black Box Interpreter). 
\item {\bf Automatic documentation} of existing packages. 
\texttt{html} documentation of packages is proposed by 
\bbi.
\end{itemize}

Finally, these different components allow {\bf efficient}:

\begin{itemize}
\item {\bf capitalization and reuse} of existing processing units, 
including {\bf documentation} 
\item {\bf testing, prototyping} in a very simple script language
\item {\bf inter-operability} between atomic processings which 
have been written by different persons, using different libraries, etc. 
\end{itemize}

% ==========================================
\subsection{\bbtk components}
% ==========================================
\BBTK includes:
\begin{itemize}
\item A \CPP {\bf\emph{library}} - called \bbtk - which defines a framework 
(abstract classes) to develop black boxes and to store them into 
dynamic libraries, called black box \emph{packages}.
\item Different {\bf\emph{black box packages}}:
\begin{itemize}
\item {\bf\emph{std}}: the 'standard' package including basic useful boxes.
\item {\bf\emph{wx}}: basic graphical interface elements (widgets: sliders, buttons, etc. based on the \texttt{wxWidgets} library).  
\item {\bf\emph{itk}}: the basic image processing package, based on the \texttt{itk} library.
\item {\bf\emph{vtk}}: the basic image and surfaces processing package, based on the \texttt{vtk} library.
\item {\bf\emph{wxvtk}}: widget boxes based on the \texttt{vtk} library.
\item {\bf\emph{itkvtk}}: boxes to convert \texttt{itk} structures into \texttt{vtk} structures and conversally. 
\end{itemize}
\item An {\bf\emph{interpreter}}, called \bbi, which allows to 
load black box packages and to define and execute 
processing chains by connecting various black boxes of the already loaded packages.
\item {\bf\emph{Utilities}}:
\begin{itemize}
\item \bbfy generates the \CPP code of a black box from a 
description file written in \texttt{xml}.
%\item \bbdoc generates the html documentation of a black box package 
%(author, description, description of its black boxes: 
%author, description, inputs, outputs, and so on).
\item \bbCreatePackage allows to create the basic file architecture 
to start the development of a new black box package.
\end{itemize} 
\end{itemize}

The general architecture of \BBTK 
is shown in figure \ref{bb-architecture}.

\begin{figure}[!ht]
\caption{\label{bb-architecture}\BBTK architecture}
\begin{center}
\includegraphics[width=0.6\textwidth]{bb-architecture.png}
\end{center}
\end{figure}

% ==========================================
\subsection{Structure of this guide}
% ==========================================

This guide is divided into two parts. 

The first part (\ref{bbi}) 
is devoted to the use of the \emph{black box interpreter} \bbi. 
This is the highest level of use of the toolkit, which 
allows to create and execute processing chains by connecting 
black boxes of existing packages. 

The second part (\ref{cpp}) explains how to 
use the black box toolkit framework in \CPP code, 
typically to develop large applications which 
involve complex graphical interfaces.
 

% ==========================================
% ==========================================
% ==========================================
% ==========================================
% ==========================================
% ==========================================
\vspace{0.5cm}\hrule
\section{The black box interpreter (bbi)}
\label{bbi}
% ==========================================

% ==========================================
%\subsection{Structure of this part}
% ==========================================

Using the black box interpreter is very simple. 
Everything is done with only a few commands.
The philosophy of this part is also very simple: 
it introduces the \bbi commands using examples, 
starting with the most simple commands.  
The first section of this part 
(\ref{bbi-getting-started})
is designed like a tutorial, 
which progressively introduces all the concepts of \bbi. 
We suggest you run \bbi and follow the examples, 
to see how it works in practice. 
At the end of this section, 
you will be able to use \bbi and write 
own black box processing scripts. 

After this tutorial, 
the section \ref{bbi-more-on} 
(called \emph{more on...}) 
goes deeper into various issues of \bbi. 
Read it at your convenience, 
either linearly to learn more about \bbi, 
or in random order to get an answer 
to a particular question. 

Finally, the section \ref{bbi-reference}
summarizes all the commands of \bbi, 
their parameters and effect. 
Use it as a reference.

% ==========================================
\subsection{Getting started}
\label{bbi-getting-started}
% ==========================================
\subsubsection{Creating and executing black boxes}
% ==========================================

To run the black box interpreter, 
open a console and type \texttt{bbi} 
or double click on the application icon.
You get a window which looks like the one in figure 
\ref{bbi-fig-bbi-gui} 
(the exact appearance of \bbi is system and \bbtk version dependent)
\footnote{If you compiled \bbtk without \wx then \bbi does not have a 
graphical interface but a simple prompt).

\begin{figure}[!ht]
\caption{\label{bbi-fig-bbi-gui}The black box interpreter interface}
\begin{center}
\includegraphics[width=0.7\textwidth]{bbi-gui.png}
\end{center}
\end{figure}

The 'Command' tab is subdivided into two parts : 
one single line zone at the bottom in which you can enter your commands and 
one multiple line zone in which \bbi prints out the result of your commands.


Try typing in the input zone (in this manual, 
the commands entered by the user will be preceded by a prompt '>') :
\begin{verbatim}
> help 
\end{verbatim}

you get the list of the commands of the interpreter:
\begin{verbatim}
Available commands:
 author
 category
 config
 connect
 define
 delete
 description
 endefine
 endpackage
 exec
 graph
 help
 include
 index
 input
 load
 message
 new
 output
 package
 print
 quit
 reset
 set
 unload
\end{verbatim}

To get help on a particular command type \texttt{help <command-name>},
for example:
\begin{verbatim}
> help quit
\end{verbatim}

gives:
\begin{verbatim}
 quit: 
   usage: quit
    Quits the program (during script execution it stops the complete execution)
\end{verbatim}

The \texttt{help} command has multiple usages. 
It is used to get help on almost anything in \bbi !
Type \texttt{'help help'} to get help on the \texttt{help} command itself:
\begin{verbatim}
> help help
 usage: 
         (1) help 
         (2) help <command name> 
         (3) help packages [all]
         (4) help <package name> [all]
         (5) help <black box type> 
         (6) help <black box name>
  Effect:
         (1) Lists all available commands;
         (2) Prints help on a particular command; 
         (3) Lists the packages loaded and their black boxes.
             Add 'all' to list adaptors; 
         (4) Prints short help on the black boxes of a package.
             Add 'all' to include adaptors; 
         (5) Prints full help on a black box type; 
         (6) Prints information on the inputs, outputs and connections of a black box instance.
\end{verbatim}

At start \bbi does not know any black box. 
If you type \texttt{'help packages'}, which is 
the third form of the \texttt{help} command, you get:
\begin{verbatim}
> help packages
user
  workspace
\end{verbatim}

which means that \bbi only knows one package 
(library of black boxes) called \texttt{user}
and which contains a black box called \texttt{workspace}.
The \texttt{user} package is an internal package to \bbi, 
which stores user-defined black box types. 
At start, it already contains 
one box, called \texttt{workspace}.
\texttt{workspace} is a special type of black box, 
called complex black box, whose purpose is 
to store other black boxes. 
Any black box you create in \bbi is stored 
in \texttt{workspace}  
(this will be explained in details in sections 
\ref{bbi-writing-scripts} and 
\ref{bbi-more-on-complex-black-boxes}). 

If you type \texttt{'help workspace'}, you get:
\begin{verbatim}
> help workspace
Complex Black Box <user::workspace>
 User's workspace
 By: bbi (internal)
 Category(s) : complex box;
 * No inputs
 * No outputs
 * No boxes
\end{verbatim}

In the text displayed, 
the \texttt{user::} prepended to the name \texttt{workspace} 
means that the box \texttt{workspace} 
belongs to the \texttt{user} package. 
Then comes a description and three lines which 
tell that \texttt{workspace} does not have any input 
nor output nor boxes yet.

In order to let \bbi know of some black boxes, 
you must load another package. 
The \texttt{std} package is the ``standard'' package, 
which contains basic useful black boxes. 

To load it, type:
\begin{verbatim}
> include std
\end{verbatim}

Then if you type:
\begin{verbatim}
> help packages
\end{verbatim}

you get something like:
\begin{verbatim}
std
  Add 
  ... 
user
  workspace
> 
\end{verbatim}

Now \bbi knows the package \texttt{std} and the black boxes it provides,
such as the \texttt{'Add'} box. Remark that the 
content of \texttt{std} may vary from one version to another 
as new black boxes might be added to it. 
If you type:  
\begin{verbatim}
> help Add
\end{verbatim}

You get:
\begin{verbatim}
Black Box <std::Add>
 Adds its inputs
 By: laurent.guigues at creatis.insa-lyon.fr
 Categories : atomic box;math;
 * Inputs: 
    'BoxExecute'     <bbtk::Void> : Any signal received by this input executes the box
    'BoxProcessMode' <String>     : Sets the processing mode of the box (Pipeline | Always | Reactive)
    'In1'            <Double>     : First number to add
    'In2'            <Double>     : Second number to add
  * Outputs : 
    'BoxChange'      <bbtk::Void> : Signal modifications of the box
    'Out'            <Double>     : Result'In1'         <double>: First number to add
\end{verbatim}

Like previously, 
the \texttt{std::} prepended to the name \texttt{Add} 
means that the box \texttt{Add} 
belongs to the \texttt{std} package. 
Then comes a description 
(the one which was provided by the author of the box), 
the author(s) of the box (usually e-mail adress(es)) and 
the categories to which the box belong. 
Finally comes the lists of inputs and outputs of the box.
For each input or output, \bbi provides 
its \emph{name} (between quotes, e.g. \texttt{'ProcessMode'}), 
its \emph{type} (between \texttt{<>}, e.g. \texttt{<Int>})
and a description.    
Remark that the box \texttt{Add} is not a 'complex' black box 
but an 'atomic' box, hence its help does not 
mention any information concerning possible internal boxes.

You can create an \emph{instance} of an \texttt{Add} box by 
the command \texttt{new}:
\begin{verbatim}
> new Add a
\end{verbatim}

The \texttt{'a'} at the end is the \emph{name} of the instance, 
which will be used to reference it later. 
It is important to make the difference between box \emph{types} 
and \emph{instances} of box types. 
The \texttt{Add} box of the package \texttt{std} is in fact 
a \emph{type} of box, like \texttt{int} is a type of data 
in \texttt{C} langage. The \texttt{new} command allows to create 
an instance of a box type, exactly like \texttt{int i;} in 
a \texttt{C} code declares a variable of type \texttt{int} whose 
name is \texttt{i}. 
Of course, like in \texttt{C} Language, you can declare multiple boxes of the 
same type in \bbi. 

After the creation of the box \texttt{a}, type:
\begin{verbatim}
> help workspace
\end{verbatim}

you get:
\begin{verbatim}
Complex Black Box <user::workspace>
 User's workspace
 By: bbi (internal)
 Category(s) : complex box;
 * No inputs
 * No outputs
 * Boxes: 
    'a' <std::Add>
\end{verbatim}

which means that \bbi's workspace now contains a black box 
of type \texttt{std::Add} and name \texttt{a}.

Now look back at the help on \texttt{Add} boxes: 
you can see that this type of box has two inputs, 
with name \texttt{In1} and \texttt{In2},
and an output, with name \texttt{Out}.

You can set the input \texttt{In1} 
of the \texttt{Add} box \texttt{a} to the value $1$ 
by the command:
\begin{verbatim}
> set a.In1 1 
\end{verbatim}

Similarly, setting the input \texttt{In2} of \texttt{a} to the value $2$
is done with:
\begin{verbatim}
> set a.In2 2
\end{verbatim}
 
And you print the output \texttt{Out} of the box \texttt{a} with:
\begin{verbatim}
> print "result=$a.Out$"
result=3
\end{verbatim}

In the string passed to the \texttt{print} command, 
each substring enclosed between a couple of \$ is considered 
as the name of an output of a box. 
To process this special substrings, \bbi does:
\begin{enumerate}
\item Processes the box if needed (see below)
\item Converts the output of the box to a string if possible 
(see below)
\item Substitutes the result in the string to print
\end{enumerate}

Box processing is needed if:
\begin{itemize}
\item at least input has changed since last processing or
\item the input \texttt{'BoxProcessMode'} of the box is set to 
\texttt{'Always'}, which forces box reprocessing. 
\end{itemize}

Note that all boxes have the input \texttt{'BoxProcessMode'}. 

Another way to process the box \texttt{a} is to issue the command:
\begin{verbatim}
> exec a
\end{verbatim}

however this command does not display anything (except if the 
box itself displays something in its processing).
It just processes the box if needed. 
This command is used to execute boxes that do not have any output,  
such as boxes that write something to a file or, display a 
graphical interface, and so on. 

To exit \bbi, type:
\begin{verbatim}
> quit
Good bye !
\end{verbatim}

% ==========================================
\hrule
\paragraph{Summary}
%\hrule
\begin{itemize}
\item The \texttt{include} command allows to load a package.
\item \texttt{help} gives help on:
\begin{itemize} 
\item Available commands if you just type \texttt{help}.
\item A particular command if you type \texttt{help <command-name>}.
\item All available packages and their boxes (without description) if you type \texttt{help packages}.
\item A particular package and its boxes (with brief description) if you type \texttt{help <package-name>}.
\item A particular black box type (with full description) if you type \texttt{help <box-type-name>}. In particular, \texttt{help workspace} displays information on the content of the \texttt{'workspace'} black box, which stores the boxes created by the user (by \texttt{new}).
\end{itemize}
%\item \texttt{list} displays the list of black box instances created so far (by \texttt{new}).
\item \texttt{new} creates an instance of a black box. 
\item \texttt{set} sets the value of an input of a black box. 
\item In all \bbi, to reference the input called \texttt{i} 
of a black box called \texttt{b} you must type \texttt{'b.i'}. 
The same syntax holds for outputs.
\item \texttt{print} prints a string, substituting each substring of the form \$b.o\$ by the value of the output \texttt{o} of the black box \texttt{b}. 
\item \texttt{exec} runs the process of a box if needed. 
\item \texttt{quit} quits \bbi.
\end{itemize}
\hrule
% ==========================================

% ==========================================
\subsubsection{Connecting black boxes}
\label{bbi-connecting-black-boxes}
% ==========================================

LG : THE GUIDE IS UP TO DATE UNTIL HERE

\BBTK allows to create 
and execute processing chains, 
also called \emph{pipelines}, 
by connecting black boxes.
This section explains how to do it with examples. 
Read section \ref{bbi-more-on-pipeline-processing} to get 
more information on pipeline processing.

First start \bbi and load the package \texttt{std}:
\begin{verbatim}
> include std
\end{verbatim}

Assume you want to compute $1+2+3$. You can do it by 
chaining two \texttt{Add} boxes, as shown in figure 
\ref{bbi-fig-connecting-black-boxes-1}. 

\begin{figure}[!ht]
\caption{\label{bbi-fig-connecting-black-boxes-1}
A simple pipeline which adds 3 numbers}
\begin{center}
\includegraphics[width=0.5\textwidth]{1plus2plus3.png}
\end{center}
\end{figure}

The \bbi instructions to create and execute this pipeline are:
\begin{verbatim}
> new Add a
> new Add b
> connect a.Out b.In1
> set a.In1 1
> set a.In2 2
> set b.In2 3
> print $b.Out$
6
\end{verbatim}

The first three commands build the pipeline, 
the next three set its inputs and the last one 
executes it and prints its output.
 
The command \texttt{'connect a.Out b.In1'} ``plugs'' the output 
\texttt{Out} of the box \texttt{a} into the input \texttt{In1} of the 
box \texttt{b}. 
Once the boxes connected, the processing of the two boxes are chained:
getting the output of \texttt{b} requires getting its inputs, 
hence getting the output of \texttt{a} which is connected to it. 
This pipeline mechanism can recurse into arbitrary long 
chains of boxes (see \ref{bbi-more-on-pipeline-processing} 
for details).

Of course, to be able to connect two boxes, 
the output and the input must be compatibles. 
You can always connect an output to an input of the \emph{same} type, 
but you can do more, thanks to particular black boxes called {\bf adaptors}.

An adaptor is a black box which has at least one input, called \texttt{In}, 
and at least one ouput called \texttt{Out} and whose role is to convert 
a data of the type of \texttt{In} 
into a data of the type of \texttt{Out} (other inputs or outputs may serve 
to parameter the adaptor or retreive other usefull information).

In \bbi, if you type:
\begin{verbatim}
> load std
> help std all
\end{verbatim}
you get:
\begin{verbatim}
 Package std v1.0.0 - laurent.guigues@creatis.insa-lyon.fr
 Basic useful black boxes
 Black boxes: 
   Add                                    : Adds its inputs
   Cast<double,float>                [DA] : Casts a double into a float
   Cast<double,int>                  [DA] : Casts a double into a int
    ... 
   Convert<unsigned int,string>      [DA] : Converts a unsigned int into a s...
   Convert<unsigned short,string>    [DA] : Converts a unsigned short into a...
   Print                                  : Prints its input to standard out...
>
\end{verbatim}

The \texttt{Cast<?,?>} and \texttt{Convert<?,?>} boxes are \emph{default adaptors}, which is signaled by the tag \texttt{[DA]} before their descriptions.

Once you have loaded the package \texttt{std}, you can 
plug an output of type \texttt{char} into an input of type \texttt{double}. 
When \bbi encounters the \texttt{connect} command, 
it looks for an adequate \emph{adaptor} in the  loaded packages. 
In our case, as the package \texttt{std} provides the 
\texttt{Cast<char,double>} adaptor, \bbi automatically creates an 
instance of this adaptor and place it \emph{between} 
the output and the input you want to connect 
(however this adaptor is hidden to you, 
it is embedded into the created connection and does not appear 
as an existing black box). 
When the pipeline is processed the 
adaptor converts the output data into the required input type, 
in a totally transparent way.
In our example, the \texttt{Cast<char,double>} adaptor 
would simply cast the value of the \texttt{char} into a \texttt{double}, 
however arbitrarily complex type conversion can be done.

Question: 
if two adaptors with the same input and output types exist 
in the packages loaded, 
which one is chosen by \bbi at connection ? 
-> Role of default adaptors

Note that the \texttt{set} and \texttt{print} commands of \bbi 
work with adaptors from \texttt{string} to the type of the input to set 
or from the type of the output to print to \texttt{string}. 
Hence in order to \texttt{set} or \texttt{print} values the adequate 
adaptors must be available in the packages currently loaded.
 
% ==========================================
\hrule
\paragraph{Summary}
%\hrule
\begin{itemize}
\item The \texttt{connect} command allows to connect two black boxes
\item You can connect two black boxes if (and only if): 
\begin{itemize}
\item The output and the input are of the same type, or
\item There is an adaptor black box in the packages loaded which 
converts data of the output type into data of the input type
\end{itemize}
\item \texttt{help <package name>} does not display the adaptors of the package. To see them use: \texttt{help <package name> all}.
including adaptors
\end{itemize}
\hrule
% ==========================================

% ==========================================
\subsubsection{Creating complex black boxes}
\label{bbi-complex-black-boxes}
% ==========================================

Remember the pipeline of figure 
\ref{bbi-fig-connecting-black-boxes-1}, which 
computed the sum of three doubles ?
You can view it as a whole and define 
a new black box type, which will be a \emph{complex black box}, 
having three inputs and one output, 
as shown in figure \ref{bbi-fig-complex-black-box-1}.

\begin{figure}[!ht]
\caption{\label{bbi-fig-complex-black-box-1}
Creating the complex black box \texttt{Add3}}
\begin{center}
\includegraphics[width=0.5\textwidth]{Add3.png}
\end{center}
\end{figure}

The \bbi commands to define this complex black box are 
the following:

\begin{verbatim}
> load std
>
> define Add3
>
> new Add a
> new Add b
> connect a.Out b.In1
>
> author "myself"
> description "adds 3 doubles"
> input x a.In1 "first double to add"
> input y a.In2 "second double to add"
> input z b.In2 "third double to add"
> output result b.Out "output"
>
> endefine
\end{verbatim}

Explainations:

As we will use \texttt{Add} boxes, we need to load the package \texttt{std}, which is done in first line.

The command \texttt{define} then starts the definition 
of the complex box type, which will be called \texttt{Add3}. 

The next three lines define the pipeline, 
exactly in the same way than outside a complex box definition. 

The commands \texttt{author}, \texttt{description}, \texttt{input} 
and \texttt{output} are commands specific to complex boxes definition:

\texttt{author} and \texttt{description} are used for the documentation 
of the new box. You can provide multiple \texttt{author} or 
\texttt{description} commands, the arguments of the commands will 
be concatenated to produce the final author and description strings.

\texttt{input} and \texttt{output} are used to define the inputs and outputs 
of the new complex box. 
Their syntax is the same: for each new input/output you need to say 
to which internal input/output it corresponds and to provide 
a help string documenting the input/output.
In our example, we define that the box \texttt{Add3} has 
three inputs: \texttt{x}, \texttt{y} and \texttt{z}. 
The input \texttt{x} corresponds to the input \texttt{In1} of the 
internal box \texttt{a}. 
In the same way, the external input \texttt{y} 
corresponds to the internal input \texttt{a.In2}, and 
the external input \texttt{In3} to \texttt{b.In2}. 
The only output of the new box is called \texttt{result}
and corresponds to \texttt{b.Out}. 
The figure \ref{bbi-fig-complex-black-box-1} 
illustrates the external to internal 
input/output correspondence.

Finally, the \texttt{endefine} command ends the definition of the 
new box type.

After this definition, if you ask for help 
on packages, you get:
\begin{verbatim}
> help packages
std
  Add
  ...
user
  Add3
  workspace
\end{verbatim}

The \texttt{user} package now contains a new black box type, called 
\texttt{Add3}. If you ask for help on this type of box, you get:
\begin{verbatim}
> help Add3
Complex Black Box <user::Add3>
 adds 3 doubles
 By: myself
 * Inputs: 
    'x'      <double>: first double to add
    'y'      <double>: second double to add
    'z'      <double>: third double to add
 * Outputs: 
    'result' <double>: output
 * Boxes: 
    'a' <std::Add>
    'b' <std::Add>
\end{verbatim}

and you can use it like any other box, for example type:

\begin{verbatim}
> new Add3 a
> set a.x 1
> set a.y 2
> set a.z 3
> print $a.result$
6
\end{verbatim}


% ==========================================
\hrule
\paragraph{Summary}
%\hrule
\begin{itemize}
\item The \texttt{define/endefine} commands allows to define complex black box types, i.e. types of black boxes made up of other black boxes. 
Inside a \texttt{define/endefine} block:
\begin{itemize}
\item The \texttt{author} and \texttt{description} commands allow to document the new type of box
\item The \texttt{input} and \texttt{output} commands allow to define the inputs and outputs of the new type of box, that is to which inputs and outputs 
of internal boxes they correspond.  
\end{itemize}
\end{itemize}
\hrule
% ==========================================

% ==========================================
\subsubsection{Writing scripts}
\label{bbi-writing-scripts}
% ==========================================

Once you have defined a new type of complex box, you 
may like to reuse it. To do this, you can simply 
write the \bbi commands defining the new box 
into a text file and afterwards include that file in \bbi. 
Doing this, you start writing \bbi scripts.
The conventionnal extension for such scripts is \texttt{bbs} 
(black box script).

For example, the \texttt{Add3} complex box we previously worked on 
can be defined in the \texttt{Add3.bbs} file:

\begin{file}{Add3.bbs}
\begin{verbatim}
# Defines the Add3 black box which adds 3 doubles 
load std

define Add3
  # I am the author 
  author "myself"
  description "adds 3 doubles"
  # Pipeline creation
  new Add a
  new Add b
  connect a.Out b.In1
  # Inputs definition
  input x a.In1 "first double to add
  input y a.In2 "second double to add
  input z b.In2 "third double to add"
  # Output definition
  output result b.Out "output"
endefine
\end{verbatim}
\end{file}

Lines starting with a \texttt{\#} character are ignored, they 
are considered as comments by \bbi.
To use this file in \bbi, use the \texttt{include} command:

\begin{verbatim}
> include Add3.bbs
> help Add3
Complex Black Box <user::Add3>
 adds 3 doubles
 By: myself
 * Inputs: 
    'x'      <double>: first double to add
    'y'      <double>: second double to add
    'z'      <double>: third double to add
 * Outputs: 
    'result' <double>: output
 * Boxes: 
    'a' <std::Add>
    'b' <std::Add>
>
and so on ...
\end{verbatim}

If the file has the \texttt{bbs} extension, you can ommit it and just type:
\begin{verbatim}
> include Add3
\end{verbatim}

Of course, you can include script files in other script files, 
like in the following example:

\begin{file}{Add4.bbs}
\begin{verbatim}
# Defines the Add4 black box which adds 4 doubles 
include Add3

define Add4
  author "myself"
  description "adds 4 doubles"
  new Add3 a
  new Add b
  connect a.Out b.In1
  input In1 a.In1 "first double to add
  input In2 a.In2 "second double to add
  input In3 a.In3 "third double to add"
  input In4 b.In2 "fourth double to add"
  output Out b.Out "output"
endefine
\end{verbatim}
\end{file}

TO DO: 

- naming conventions:  one cbb per file with the same name
- search paths 

% ==========================================
\hrule
\paragraph{Summary}
%\hrule
\begin{itemize}
\item The \texttt{include} command allows to include a script file in \bbi.
\item Lines starting with a \texttt{\#} are treated as comments in \bbi scripts.
\end{itemize}
\hrule
% ==========================================

% ==========================================
\subsubsection{Creating command line applications}
\label{bbi-command-line-app}
% ==========================================

Now that you now how to create complex black boxes 
(with \texttt{define/endefine}), think 
back to the \texttt{workspace} object. 
Remember that it is also 
a \texttt{complex black box}. 
In fact what you are doing when you type \bbi commands 
outside a \texttt{define/endefine} block 
is to progressively define the \texttt{workspace} 
complex black box.
You can think of it like if at start 
\bbi was issuing a command \texttt{'define workspace'} 
and then letting you define the interior of the box 
\texttt{workspace}.

Remember that the command \texttt{inputs} 
allows to define an input of a complex box. 
Now, if you use the command \texttt{input} 
outside a \texttt{define/endefine} block then 
it defines an input of the \texttt{workspace} box, 
that is an input of the \emph{main program}. 
This input will then be connected to the 
parameters that the user passes to the command line.

For example, consider the script: 

\begin{file}{add.bbs}
\begin{verbatim}
load std
new Add a
input x a.In1 "first number to add"
input y a.In2 "second number to add"
print "x+y=$a.Out$"
\end{verbatim}
\end{file}

The third and fourth lines define two inputs \texttt{x} 
and \texttt{y}. When you execute this script, 
you can pass these two arguments on the command line, 
like this:

\begin{verbatim}
> bbi add x=1 y=1
x+y=2
\end{verbatim}

You can also invoke \bbi the option \texttt{-h}, 
which gives help on the \texttt{workspace} box:

\begin{verbatim}
> bbi add -h
 User's workspace
 By: bbi (internal)
 * Inputs: 
    'x' <double>: first number to add
    'y' <double>: second number to add
\end{verbatim}

To get a better help, use the \texttt{description} 
and \texttt{author} commands:

\begin{file}{add.bbs}
\begin{verbatim}
description "Adds two numbers"
author "foo@bar.com"
load std
new Add a
input x a.In1 "first number to add"
input y a.In2 "second number to add"
print "x+y=$a.Out$"
\end{verbatim}
\end{file}

Now if you ask for help on the \texttt{add} script, you get:

\begin{verbatim}
> bbi add -h
 Adds two numbers
 By: foo@bar.com
 * Inputs: 
    'x' <double>: first number to add
    'y' <double>: second number to add
\end{verbatim}

Rather than getting the inputs of a script 
from the command line, you can ask \bbi to 
prompt the user for the values, using the \texttt{-t}
commutator:

\begin{verbatim}
> bbi add -t
x=[the program waits for user answer]2
y=[the program waits for user answer]5
x+y=7
\end{verbatim}

If \bbi is compiled in graphical mode (with \wx), 
you can also use the \texttt{-g} commutator. 
\bbi then prompts the user in graphical mode, 
displaying a dialog box for each input,
like in fig. \ref{bb-input-dialog-box}.

\begin{figure}[!ht]
\caption{\label{bb-input-dialog-box}Input dialog box}
\begin{center}
\includegraphics[width=0.6\textwidth]{enter-the-value-of-x.png}
\end{center}
\end{figure}

% ==========================================
\hrule
\paragraph{Summary}
%\hrule
\begin{itemize}
\item The \texttt{input}, \texttt{description} and \texttt{author} commands,
when they are used outside a \texttt{define/endefine} block allow 
to define the inputs, description and author of the main program.
\item Inputs of the main program can be passed on the command line 
using the syntax \texttt{<input-name>=<value>}. 
No white space is allowed, if the value or the input name 
contains white spaces, enclose them 
between double quotes, e.g. \texttt{"parameter with white spaces = gnu's not unix"}.
\item The \texttt{-h} option of \bbi prints help on the main program.
\item The \texttt{-t} option of \bbi orders the program to prompt for its inputs in text mode.
\item The \texttt{-g} option of \bbi orders the program to prompt for its inputs in graphical mode.
\end{itemize}
\hrule
% ==========================================

% ==========================================
\subsubsection{Using graphical interface boxes (widget boxes)}
\label{bbi-widget}
% ==========================================

If \bbi is compiled in graphical mode 
(option \texttt{BUILD\_bbi\_GRAPHICAL} of \cmake, requires \wx),
then you can use special black boxes which are 
graphical interface components (widgets). 
Basic components are provided in the package \texttt{wx}, 
such as buttons, sliders, file open/save dialogs, etc.

As first example, type the following commands in \bbi:
\begin{verbatim}
> load wx 
> new TextCtrl t
> print $t.Out$\n
\end{verbatim}

When you type \texttt{enter} after the last line, 
a window pops up in which you can entrer a text.
When you close the window, the text you entered is printed by 
the \texttt{print} command.

Type \texttt{help wx}, you get something like:
\begin{verbatim}
Package wx v1.0.0- eduardo.davila/laurent.guigues@creatis.insa-lyon.fr
 Basic graphical interface elements (slider, button ...) based on wxWidgets
 Black boxes: 
   Button      : Button that gives a string 
   FileDialog  : FileDialog widget (wxFileDialog)
   RadioButton : RadioButton group widget (wxRadioButton)  0-9 entries
   Sizer       : Sizer widget (wxSizer)
   Slider      : Slider widget (wxSlider)
   Split       : Split widget (wxSplitterWindow)
   StaticText  : wxWidget Static text
   TextCtrl    : TextCtrl widget (wxTextCtrl)
\end{verbatim}

You can reproduce the same experiment as above using a 
\texttt{Slider} or a \texttt{FileDialog} rather than a \texttt{TextCtrl}.
See the files \texttt{test*.bbs} in the \texttt{scripts/test} directory.

There are two kind of widgets: ``terminal'' widgets and ``container'' widgets.
The \texttt{TextCtrl}, \texttt{FileDialog} or \texttt{Slider} widgets 
are ``terminal'' widgets. 
``container'' widgets are of another kind: they are made to 
contain other widgets in order to build larger dialog boxes. 
For example, the \texttt{Split} widget is a container which 
``splits'' horizontally a window into two parts, 
each part including another widget. 
The size of the two parts can be adjusted by the user thanks 
to a ``handle''.

The script \texttt{scripts/test/testSplit.bbs} demonstrate its use. 
Run it: it displays a window with two sliders. 
Move the sliders and close the window. 
The final positions of the sliders are printed out. 
Now edit the file to see how this is done:

\begin{file}{scripts/test/testSplit.bbs}
\begin{verbatim}
load std
load wx

new Slider s1
new Slider s2

new Split s
connect s.Child s1.Parent
connect s.Child s2.Parent

print s1=$s1.Out$\\n
print s2=$s2.Out$\\n
\end{verbatim}
\end{file}

First, the two sliders \texttt{s1} and \texttt{s2} are created.
A \texttt{Split} box \texttt{s} is also created. 
The \texttt{connect} commands then ``includes'' the sliders in the 
split ``container''. 
The input \texttt{Parent} is common to all widget boxes: 
every widget can be inserted into another widget. 
The output \texttt{Child} is specific of \emph{container} 
widgets 
(in \bbi type \texttt{help Slider}: 
you will see the input \texttt{Parent}; 
type \texttt{help Split}: 
you will see the input \texttt{Parent} 
and the output \texttt{Child}). 
When you connect the \texttt{Child} output of a container 
to the \texttt{Parent} input of a widget, 
you order to include the widget in the container.
Of course, the order of connection is important. 
In our case, the slider \texttt{s1} is included first, 
then the slider \texttt{s2}: \texttt{s1} will be placed 
on top of \texttt{s2} (the \texttt{Split} box is 
implemented that way, but this is arbitrary choice).

For the moment, there are only \emph{two} container widgets in the \texttt{wx} package: 
the \texttt{Split} widget we just described and the \texttt{Sizer} 
widget, which can have multiple children and 
divides its window into as much parts as children, 
each part of equal size. 
The orientation of the sizer can be changed by the input \texttt{Orientation}.
See the example \texttt{test/testSizer.bbs}.
With only those two containers you can already create 
complex dialog boxes (of course containers can be nested, which 
leads to tree-like structures of widgets). 
See the script \texttt{test/testSizerSplit.bbs} for an example.

One word about a special widget in the package \texttt{wx}: 
the \texttt{Button}... to be continued.


TO DO: 
\begin{enumerate}
\item Make a tour of ``complex'' widgets of wxvtk 
\item Explain the role of ProcessMode to update widgets
\item Explain the creation of complex widgets (containers, contained...)
\item Explain the ``control'' mechanism in bbi (switch exec commands, e.g. Button)
\end{enumerate}


% ==========================================
\subsection{More on ...}
\label{bbi-more-on}
% ==========================================

% ==========================================
\subsubsection{Black box packages}
\label{bbi-more-on-packages}
% ==========================================

% ==========================================
\subsubsection{Pipeline processing}
\label{bbi-more-on-pipeline-processing}
% ==========================================

% ==========================================
\subsubsection{Complex black boxes}
\label{bbi-more-on-complex-black-boxes}

%\subsubsection{Advanced issues}
%\paragraph{Reducing the number of inputs of a box}

% ==========================================
\subsubsection{Errors}
\label{bbi-more-on-errors}

% ==========================================
\subsubsection{\bbtk configuration file and search pathes}
\label{bbi-more-on-configuration}

At start, \bbi tries to open an \texttt{xml} 
configuration file named \texttt{bbtk\_config.xml}. 
The search order is 
\begin{enumerate}
\item The current directory
\item The subdir \texttt{.bbtk} of the user's home directory. 
\begin{itemize} 
\item On \texttt{Unix}, the home directory is the
one stored by the environnement variable \texttt{HOME}, 
typically \texttt{/home/username}.
\item On \texttt{Windows}, the home directory is 
the user's profile directory stored by the environnement 
variable \texttt{USERPROFILE}, 
typically \texttt{C:\\...}.
\end{itemize}
\item If none of these two pathes contains the file then it creates 
a new one in the \texttt{.bbtk} directory. 
\end{enumerate}

Once created, you can edit the \texttt{bbtk\_config.xml} file located 
in your \texttt{.bbtk} directory. It contains:

\begin{file}{bbtk\_config.xml}
\begin{verbatim}
<?xml version="1.0" encoding="iso-8859-1"?>
<config>
  <bbtk_url> http://www.creatis.insa-lyon.fr/software/bbtk </bbtk_url>
  <bbs_path> </bbs_path>     
  <package_path> </package_path> 
  <data_path> </data_path>
</config>
\end{verbatim}
\end{file}

You can add pathes to 
\begin{itemize}
\item A custom folder in which to search for \texttt{.bbs} scripts (\texttt{include} command of \bbi) by adding an \texttt{xml} tag: 
\texttt{<bbs\_path>complete\_path\_to\_folder<\/bbs\_path>}.
\item A custom folder in which to search for packages (\texttt{load} command of \bbi) by adding an \texttt{xml} tag: 
\texttt{<package\_path>complete\_path\_to\_folder<\/package\_path>}.
\end{itemize}

% ==========================================
\subsection{Language reference}
\label{bbi-reference}
% ==========================================





% ==========================================
\begin{table}[!ht]
\caption{\label{bbi-reference-box}
\bbi pipeline creation and execution related commands.}
\small
\begin{tabular}{|lcm{6cm}|}
\hline
Command & Parameters & Effect \\ \hline

\texttt{new} & \texttt{<boxtype>} \texttt{<box-name>}& 
Creates a box of type \texttt{boxtype} and name  
\texttt{box-name}.
\\ \hline

\texttt{delete} & \texttt{<box-name>} & 
Destroys the box named \texttt{box-name}.
\\ \hline 

\texttt{connect} & \texttt{<box1.output>} \texttt{<box2.input>} & 
Connects the output 
\texttt{output} of the box named \texttt{box1} 
to the input \texttt{input} of the box named \texttt{box2} \\ \hline 

\texttt{set} & \texttt{<box.input>} \texttt{<value>} &
Sets the input \texttt{input} of 
the box named \texttt{box} to the value \texttt{value}.
There must exist an \texttt{adaptor} 
in the packages loaded which converts a \texttt{std::string} 
to the type of the input \texttt{input}.
 \\ \hline 

\texttt{print} & \texttt{<string>} & 
Prints the string after substituting each token of the form \texttt{\$box.output\$} by the adaptation to string of the value of the 
output \texttt{output} of the box named \texttt{box}. 
There must exist an \texttt{adaptor} 
in the packages loaded which converts 
the type of the output \texttt{output}
to a \texttt{std::string}.
\\ \hline 

\texttt{exec} & \texttt{<box-name>} & 
Executes the box named \texttt{box-name}.
If needed the boxes 
connected to its inputs 
are also processed recursively (pipeline processing).\\ \hline 
\end{tabular}
\end{table}
% ==========================================



% ==========================================
\begin{table}[!ht]
\caption{\label{bbi-reference-interpreter}\bbi intepreter related commands.}
\small
\begin{tabular}{|lcm{6cm}|}
\hline
Command & Parameters & Effect \\ \hline


\texttt{help} & - & 
Prints help on available commands \\ \hline 

& \texttt{<command-name>} & 
Prints help on the command \texttt{command-name} \\ \hline 

& \texttt{packages} & 
Prints help on available packages and their box types 
(without description)\\ \hline 

& \texttt{<package-name>} & 
Prints help on the package \texttt{package-name} and its boxes 
(with brief description). 
The package must have been previously loaded
\\ \hline 

& \texttt{<box-type>} & 
Prints help (with full description) on the type of box 
\texttt{box-type}. 
The box type must belong to a package which has been previously loaded
\\ \hline 

\texttt{include} & \texttt{<file-name>} & 
Includes and executes the content of the file named \texttt{file-name} 
exactly like if you were typing its content at the place were the 
\texttt{include} command is.
\\ \hline 

\texttt{load} & \texttt{<package-name>} & 
Loads the package \texttt{package-name}\\ \hline 

\texttt{unload} & \texttt{<package-name>}& 
Unloads the package \texttt{package-name}. 
The package must have been previously loaded. 
No box of a type defined in this package must still exist.
\\ \hline 

\texttt{message} & \texttt{<category>} \texttt{<level>} & 
Sets the level of verbosity of \bbi for the category of messages 
\texttt{category} to \texttt{level}.
%See \ref{verbosity}.
\\ \hline 

\texttt{config} & - & Displays the Configuration parameters\\ \hline 

\texttt{reset} & - & Deletes all boxes and unloads all packages so 
that \bbi gets back to its initial state \\ \hline 

\texttt{quit} & - & Exits the interpreter\\ \hline 

\end{tabular}
\end{table}
% ==========================================





% ==========================================
\begin{table}[!ht]
\caption{\label{bbi-reference-complex-box}
\bbi complex black box definition related commands.}
\small
\begin{tabular}{|lcm{6cm}|}
\hline
Command & Parameters & Effect \\ \hline


\texttt{define} & \texttt{<box-type>} & 
Starts the definition of a complex black box of type  
\texttt{box-type}\\ \hline 

\texttt{endefine} & - & 
Ends the definition of a complex black box type\\ \hline 


\texttt{author} & \texttt{<string>} & 
Sets the author(s) of the complex black box currently being defined \\ \hline 

\texttt{description} & \texttt{<string>} & 
Sets the description of the complex black box currently being defined 
\\ \hline 


\texttt{input} & \texttt{<name>} \texttt{<box.input>} \texttt{<help>} & 
Defines a new input for the current complex black box, 
named \texttt{name}. 
It is defined as corresponding to 
the input \texttt{input} of the box \texttt{box}. 
\texttt{<help>} is the help string for the new input.
The box \texttt{box} must already have been created in the complex box 
and of course have an input named \texttt{input}.
\\ \hline 


\texttt{output} & \texttt{<name>} \texttt{<box.output>} \texttt{<help>} & 
Defines a new output for the current complex black box, 
named \texttt{name}. 
It is defined as corresponding to 
the output \texttt{output} of the box \texttt{box}. 
\texttt{<help>} is the help string for the new output.
The box \texttt{box} must already have been created in the complex box and of course have an output named \texttt{output}. 
\\ \hline 


\end{tabular}
\end{table}
% ==========================================




% ==========================================
\vspace{0.5cm}\hrule
\section{Using black boxes in \CPP programs}
\label{cpp}
% ==========================================




%\bibliography{all}



%\section{Conclusion}
\end{document}