This is all I see left that needs to move out of PSLM.

I now agree though that column balancing should stay
in PSLM, because it's a dynamic layout task and
because its implementation--unlike the rote setting up
of regions for a PageViewport--would most likely
differ between PSLM implementations (our pluggable
LM's allowing for a different PSLMs to be used). 
Also, CB will most probably have
interactive/corresponding effects with other Spans
and/or pages--another signal that this is PSLM's
responsibility.

Glen

--- Jeremias Maerki <[EMAIL PROTECTED]> wrote:
> Not an objection but a concern: I get the impression
> that more an more
> logic wanders off into the area package which should
> IMO only be a data
> structure plus serialization/deserialization and
> reference resolution.
> And I believe that column balancing is clearly the
> LMs responsibility.
> Especially with the Knuth approach this can't even
> be done otherwise. I
> don't know what the others think.
> 
> On 30.03.2005 02:49:08 Glen Mazza wrote:
> > Team,
> > 
> > I'd like to recommend next that we move the
> > initialization (*not* population) of the
> > page-reference-areas, region-viewport-areas, and
> > region-reference-areas from PSLM to
> area.PageViewport
> > (and possibly some in area.RegionViewport).  IOW,
> > everything from line 685 to the end of the file
> > gone[1], with the remaining few lines of
> > createPageAreas() moved to makeNewPage().  
> > 
> > This is all boilerplate routine setup--predefined
> > margin widths, etc-- independent of layout
> mechanisms,
> > and I suspect this work can be more cleanly and
> > orderly implemented in area.PageViewport anyway.
> > 
> > Replacing this logic in PSLM eventually will be
> more
> > flow-mapping logic, both of the future fo:flow-map
> and
> > even our current 1.0 requirements to be able to
> place
> > fo:flow data into any of the region areas. 
> (Perhaps
> > the column balancing as well, if the Span object
> > cannot handle it by itself.)  This is more of
> PSLM's
> > responsibilities, and it will be quite large
> enough in
> > accomplishing all of them.  But PSLM should more
> or
> > less be able to just call curPage = new
> > PageViewport(), and get a PageViewport with
> already
> > initialized child region areas.  The coding to
> create
> > this IMHO will be a distraction if kept in PSLM.
> > 
> > Any objections?
> > 
> > Thanks,
> > Glen
> > 
> > [1] http://tinyurl.com/4qca3
> > 
> > --- [EMAIL PROTECTED] wrote:
> > >
> > > gmazza      2005/03/29 16:06:30
> > > 
> > >   Modified:    src/java/org/apache/fop/area Tag:
> > >                        
> Temp_KnuthStylePageBreaking
> > > PageViewport.java
> > >                src/java/org/apache/fop/layoutmgr
> > > Tag:
> > >                        
> Temp_KnuthStylePageBreaking
> > >                        
> > > PageSequenceLayoutManager.java
> > >   Log:
> > >   Removed the curSpan instance variable -- now
> > > obtainable via curPage.
> > >   
> 
> 
> 
> Jeremias Maerki
> 
> 

Reply via email to