Overall design ============== * The document (KWDocument) has a list of framesets (KWFrameSet) * Framesets include: text frameset, picture/clipart frameset, table frameset etc. * A frameset has a list of frames, KWFrame. A KWFrame is basically a rectangle (KoRect) with a bunch of attributes. The frameset contains the contents, like the text or a picture. The frames visualize the contents of the frameset. Imagine a 10 pages document, with 1 text-frameset and 10 frames, because our example author wants to have one frame per page. This means that there is one text, and the frames define where on the pages this text is displayed. * About text framesets: they hold a text document (KWTextDocument, which is a QTextDocument from the KoText classes - more on KoText later). * About Tables; one table is a KWTableFrameset. This tableframeset tqcontains all the cells as its frames. But because the cells have a different text per cell, we contain the cells in their own KWTextFrameset. KWTableFrameSet | | | TextFrames Cells +--------+ | | + ---> KWTextFrameSet | | +-> KWFrame <-+ | | + ---> KWTextFrameSet | | +-> KWFrame <-+ | | + ---> KWTextFrameSet | +-> KWFrame <-+ * The z-order of the frames (i.e. which one is on top of which) is defined by m_zOrder in KWFrame. This number is relative to the other frames *on the same page*. Frame tqlayout and pages ====================== Generally, the user is free to position frames where he/she wants to. KWTextFrameSet::slotAfterFormatting() takes care of triggering frame resizing (AutoExtend frames), creating a new page (AutoCreateNewFrame), etc. However, in Word Processing mode, some frames are special: main text frameset (which might include columns), headers, footers, and footnotes. All of those are laid out by the KWFrameLayout class, which is triggered by KWDocument::recalcFrames. When creating a new page, KWDocument::insertPage takes care of creating the followup frames for all frames with newFrameBehavior=Reconnect. The newFrameBehavior called "Copy" means "exact copy of the contents of the frame". This is used for headers/footers (but the user can also use it for logos and stuff). In such a case we don't copy anything, we simply paint the same contents in many pages. In WP mode, removing pages is automatically done by KWDocument::tryRemovingPages(). In DTP mode pages are user-controlled (insert and delete page actions), but overflowing text can still create a new page as well. The difficulty with frame tqlayout is that in most cases the frame size and number depends on the text inside, and the text tqlayout depends on the frames (frame width, frames on top, etc.). This is why text tqlayout (formatMore()) and frame tqlayout (e.g. KWFrameLayout) often call each other, resulting in infinite loops when some code goes wrong. Editing ======= To edit the contents of a text frameset (including table cells), the user clicks on it. This creates an edit object for the frameset, which will be destroyed as soon as we go edit another frameset. This means, in a given view (canvas), there is only one edit object at a given moment (for text objects, it's the one with the cursor). There is one type of edit object per type of frameset. For instance: KWTextFrameSetEdit. The kotext library provides a base class for the 'text editing' object, it's KoTextView. Custom items ============ A custom item is anything (like an inline frame, a variable etc.) that is treated as a special character. QTextCustomItem is the base class for it, we have KoTextCustomItem in kotext. Inline frames are one kind of custom item. The frame (and its englobing frameset) exist as usual, but a KWAnchor instance is created, inserted into the text, and linked to the frame. The anchor always has the same size as the frame. When the text is flowed, the anchor is positioned, it tells the frame to move to the new position. Painting ======== Here's how the painting (drawing the stuff on screen, and on printer) works: Each frameset is responsible for drawing itself. The base class, KWFrameSet, handles the generic code for iterating through the frames, applying the "frame is a copy" stuff, drawing the borders, etc. So each specific frameset class only has to implement drawFrame (exception for KWTableFrameset though). Painting of text: KWTextFrameset paints the text for each frame, calling KoTextDocument::drawWYSIWYG (with flags for "only the changed paragraphs" etc). When printing this method delegates to drawWithoutDoubleBuffer. Both methods iterate through the paragraphs, and call drawParagWYSIWYG for each one to be redrawn. This method implements the double-buffering, paints the background color and the space on the right of the paragraph, and calls QTextParag::paint(). Its role is to draw series of characters with the same formatting at once (in one go). For each set of chars to be painted, KoTextParag::drawParagString is called. In short: KWFrame::drawContents -> KWTextFrameSet::drawFrame -> KoTextDocument::drawWYSIWYG -> KoTextDocument::drawParagWYSIWYG -> KoTextParag::paint -> KoTextParag::drawParagString -> KoTextParag::drawParagStringInternal The cursor is also drawn by KoTextParag::paint, which calls KoTextParag::drawCursor. The blinking cursor is implemented in KoTextView, which calls a virtual drawCursor() method, implemented in KWTextFrameSetEdit. More low-level things ===================== .. such as coordinate systems, font sizes, justified spaces etc. See lib/kotext/DESIGN. Wow, you managed to read this until the end ? Good. "GOTO 10" now ;) David Faure