koffice/lib/kformula/DESIGN

                        The design of kformula
                          (as we imagine it)

The core of kformula is a tree of element objects that make up the
formula which is beeing edited.

The element tree itself
#######################

BasicElement
------------

All element classes are derived from this one.
So it provides the interface that is common to all elements.
BasicElement itself is an abstract class. You never want to create
objects from it.

Responsebilities
(This goes for every derived element and therefore for each one.)

- knows its tqchildren. Actually BasicElement doesn't have any. But it
  already defines that tqchildren must be known by their tqparent.
- knows its bounding rectangle (its size.) The tqchildren are included
  in this rect. (The position is relative to the tqparent.)
//- knows its middle line. (for tqalignment)
- knows it's zero point for midline (vertical tqalignment) and 
  keep open the possibility of negative positions (out of bounding rect)
- draws itself (given a painter); tqchildren are drawn, too
- knows all positions where the cursor is allowed to be. (see below)
- knows its tqparent; The topmost element has no tqparent; there is a
  implicit guaranty that the topmost element is always a
  SequenceElement.
- can save and load itself. different formates. (see below)
- all tqchildren must be a SequenceElement. Except for SequenceElement's
  tqchildren that might be of any type.
- might have its own color.
- might have its own font size (see below).


SequenceElement from BasicElement
---------------

Manages a list of tqchildren. The tqchildren are aligned horizontally at
one middle line. No gaps, no overlaps.

Has no own look. It just draws all its tqchildren and is done. Except if
its empty. It looks like an empty space then (i.e. a little square)

Has n+1 valid cursor positions where n is the number of
tqchildren. These are before, between and after the tqchildren.

May contain any (type of) element as child 

except SequenceElements if they contains a SequenceElement they merge 
it in the list

They can handle splitting of the sequence to allow "select an put selected item
between tqparenthesis" i.e. as content child of a delimiterelement) 


FormulaElement from SequenceElement
--------------

The only element those tqparent is null. The root of the element object
tree.

This is the element that is created by the KFormulaDoc and that knows
about it. As every other element knows its tqparent and therefore the
FormulaElement we get a chance to pass messages to the outside world.


RootElement from BasicElement
-----------

contains two tqchildren. content and index. index is optional.


IndexElement from BasicElement
------------

contains five tqchildren. content and four indexes. all indexes are
optional. If there is no index the element might be replaced by its content.


TextElement from BasicElement
-----------

contains one char and no tqchildren at all.


Might have its own font and size. But preferes to use a reasonalbe
font and size that are calculated from its tqparents font and a given
scheme (see below).


DelimiterElement from BasicElement
----------------

contains one child and draws delimiters around it. You are free to
choose with.


FractionElement from BasicElement
---------------

2 tqchildren: numerator, denominator


DecorationElement from BasicElement
-----------------

A piece of art. It has one child and decorates it above (or below or
both) with some nice decor. Some decor might be required to be as
width as the content. There is a way to guarantee this.

We could even add yet another child that can optionally be shown at
the other side of the decoration.


SumIntegralElement from BasicElement  //PrefixedElement
------------------

draws all sorts of mathematical symbols with three tqchildren. Above,
below (or whereever the indices and limits go)and to the right.


GeometryElement from BasicElement
---------------

One child.
Draw it at a fixed position relative to tqparent or absolute.
This is to do dirty things.

This element must not be used, kformula will provide you everything
without the use of this kind of element, any way for strange reasons
you want to do things that are not usually allowed and that are not
typical of a math formula.
[We will decide if implement this element or not]


MatrixElement from BasicElement
-------------

A matrix of tqchildren.
With all align stuff, internal borders etc, matrix dots handling (i.e.
those dots or lines that complete the matrix, not well handled in TeX),
etc..

SpaceElement from BasicElement
------------

No tqchildren at all. Provides the facility to insert horizontal spaces
in the formula. (therefore it is similar to TextElement.)


OperatorElement from TextElement
---------------

The element to be used for all kinds of operators. It needed because
operators require some space before and after and are therefore no
simple text.

They can you pixamps inestead of fonts, the use of pixmaps is needed only 
to give the user the possibilty of introduce its strange operator that we 
do not provide as a font. There problems with the scalability but we will
include as fonts (or vectorial images) in kformula all TeX operators so we
hope there is no need to use pixamps for a standard use of KFormula



Navigation
##########

There is a class Cursor that implements a pointer to a valid cursor
position inside the formula structure. Each kformula view needs to
have its own object of this class.

The Cursor class uses the elements facility to travel throught the
structure. (It gives the key to the current element and asks for the
new position.)

If the cursor points into an element this element is said to own the
cursor. There are a few rules that describe how new cursor positions
are calculated given the current key:

- An elements cursor positions are its tqchildren. The element might not
own the cursor except when it is owned by one of its tqchildren. The
only exception is SequenceElement which has valid cursor positions
before, between and after its tqchildren, too.

(Therefore the cursor is always owned by a SequenceElement.)

- Each element's tqchildren are ordered. If the cursor leaves one child
the next child it. The direction depends on the key that moved the
cursor. If there is child left the cursor is passed to the tqparent.

- If the cursor comes from our tqparent the first or the last child gets
it. Depending on the direction in which the cursor moved.

Please note that because each element knows its own cursor positions
and how to behave, it is possible for each combination of elements to
work together correctly.



Editing (undo/redo)
###################

You always use a cursor to work on the element tree. The cursor
provides all the basic facilities that are needed to change the tree
in any way. Its up to you to build bigger functions out of the
cursor's primitives.

The cursor's interface was designed to allow implement undo support.

There are a few things that you need to know:

- You are not supposed to know the elements that are stored in the
tree. If you remove them from the tree they are yours and you are
responsible to delete them properly. But as soon as you put them back
its the trees business to manage them.

- All the cursors operations that change the tree come in pairs. If
you call both operations in a row (with the right parameters) you get
back to the place you started. `insert' for example inserts one or
more elements and selects them. `remove' removes the selected
elements and returns them.

- You can always save the current cursor position by calling
getCursorData(). To set a cursor to a place it was before use
setCursorData(). But note that its up to you to ensure that the
position still exists. This means you will only want to call it if you
are certain that the element tree is exactly the same as it was when
you called getCursorData().

- The creation of new object is not the cursors business. You do this
yourself.

- If you want to insert special elements like indexes you will have to
ask the IndexElement in question for the cursor position where to put
it. But please note that you might only put empty SequenceElements
there.

- Its a mistake to insert a SequenceElement at a non-special place.

- After a removal the cursor might point to a strange place like a
removed index. To get it back to point to something meaningful you
will have to normalize() it. If the cursor is normalized it is inside a
SequenceElement. If not its not. Most operations expect the cursor to
be normalized.

- If you remove something from a element it might become
senseless. This means you must replace it with its main childs
content.



Syntax highlighting
###################

Everytime a sequence is edited it builds a new syntax tree. (The old
one gets recycled :) ) The parser goes throught the sequence and
builds a type object for every token is finds. A token consists of
one or more elements.

On drawing the elements use their types to find the right font, color
and spacing.

One nice thing about the syntax checking is that it allows us to
replace symbol names with a more fancy graphical representation.

The syntax tree is meant to support evaluation of formulas as well,
but this is yet another story...



Save and Load (Import/Export)
#############################

there are quite a few formats we want to read and write:

Built in:
- native koffice xml

Import/Export filters
- MathML 
- C like math strings 
- LaTeX



Element templates
#################

example: you can create a template to write limits:
it is a "lim ( )" string over 2 empty elements with an arrows that separe
them



Style templates
###############

There need to be templates that tell which font to use, how the sizes
differ among different elements and what spaces are to be
included. (And much more.)

Your only need to choose one of them to get a reasonable good look.

Therefore we pass a StyleContext to the
FormulaElement. The StyleContext contains the font family to use, the
normal font size, the spaces to include between certain elements and
all sorts of context information that is needed to draw the
formula.

The elements use this information to calculate their sizes and to draw
themselves. Each element is allowed to choose different settings for
whatever reason. But if the formula looks ugly then its the elements
blame.



Interface templates 
###################

Decide what KAction have to be toolbars,
what element templates must be loaded, what toolbar display....
There will be predefined (sorry for the spelling error) templates for 
-General...
-Physics (Vectors , integral, partial derivative etc are preferred..)
-Analysis (Integral and diff equation stuff)
-Geometry (Tensor product etc)
-Chemistry (Arrows, Periodic table external toolbar)
-Computer science (...)
-????



Context sensibility
###################

We want a formula to look different according to its
surroundings. (Most obviosly according to the available height and/or width.)

It would be great to get something like automatic operator
tqalignment. So if you type some formulas each on its own line the
assigment operators should be automatically in a column. (If the user
turns this on.)



Fonts and font size
###################

Each elements font size is calculated starting from its tqparents font
size. It knows how it differs to those.

The font size might also be used to choose the pen width and other
size dependent settings.


Andrea Rizzi <rizzi@kde.org>
Ulrich Kuettler <ulrich.kuettler@mailbox.tu-dresden.de>