function [port_bonds, status] = abg2cbg(system_name, system_type, full_name,
					port_bonds,
					port_bond_direction,
					port_status,
					typefile, infofile, errorfile)

% abg2cbg - acausal to causal bg conversion
%
%     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 
%     %%%%% Model Transformation Tools %%%%%
%     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% 
% Matlab function  abg2cbg.m
% Acausal bond graph to causal bond graph: mfile format
% [bonds,status] = abg2cbg(system_name, system_type, full_name, port_bonds, infofile)

% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% %% Version control history
% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% %% $Id$
% %% $Log$
% %% Revision 1.32  1998/07/28 19:06:11  peterg
% %% *** empty log message ***
% %%
% %% Revision 1.31  1998/07/28 13:15:10  peterg
% %% Vector SS ports included.
% %%
% %% Revision 1.30  1998/07/27 20:29:49  peterg
% %% Had another go at causality ....
% %%   1. Impose external causality onto all port bonds
% %%   2. Set C_cause.m so that it DOESN'T set causality
% %%       -- I_cause is already ok !
% %%
% %% Revision 1.29  1998/07/10 09:01:42  peterg
% %% Added error + info file in new form
% %%
% %% Revision 1.28  1998/07/08 09:23:42  peterg
% %% Undid the previous change -- needs more thought.
% %%
% %% Revision 1.27  1998/07/03 19:03:31  peterg
% %% Always override the causality of port bonds!
% %%
% %% Revision 1.26  1998/06/27 13:24:04  peterg
% %% Causality now set correctly for:
% %% 	multi-port C and I
% %% 	C and I with arrows pointing in
% %%
% %% Revision 1.25  1998/06/25 18:53:30  peterg
% %% Actually, the previous comment was optimistic.
% %% The port causalities on a compound are now forced to be the same as
% %% that specified by a a _cuase.m file (if it exists)
% %%
% %% Revision 1.24  1998/06/25 17:45:03  peterg
% %% No change -- but checked that explicit causality works!
% %%
% %% Revision 1.23  1998/04/04 10:46:37  peterg
% %% Coerces port bonds to have smae direction as the imposing bonds
% %% _cbg now generates the (coerced) components as welll as the (coerced)
% %% causality.
% %%
% %% Revision 1.22  1997/08/19 10:21:09  peterg
% %% Only copy port cuaslity info if not already set within the
% %% subsystem. I thought I'd done this already ....
% %%
% %% Revision 1.21  1997/08/18 16:25:25  peterg
% %% Minor bug fixes
% %%
% %% Revision 1.20  1997/08/18 12:45:24  peterg
% %% Replaced: comp_bonds = bonds(bond_list,:)
% %% by: 	for kk = 1:n_comp
% %% 	  comp_bonds(kk,:) = bonds(comp(kk),:);
% %% 	end;
% %%
% %% to avoid an octave bug in 1.92.
% %%
% %% Revision 1.19  1997/08/18 11:23:59  peterg
% %% Main component loop now misses out the ports (SS:[]) -- the causality
% %% is merely passed through these components.
% %%
% %% Revision 1.18  1997/08/08 08:11:04  peterg
% %% Suppress compoment trace.
% %%
% %% Revision 1.17  1997/08/07 16:10:13  peterg
% %% Move the if status ..  to the beginning of the main loop.
% %%
% %% Revision 1.16  1997/08/04 13:11:19  peterg
% %% Only change to component-orientated causality for simple components
% %% NOT for compound components.
% %%
% %% Revision 1.15  1997/01/05 12:25:59  peterg
% %% More informative message about port bonds incompatible with  ports
% %%
% %% Revision 1.14  1996/12/31 16:20:42  peterg
% %% Just write causality information at top level -- it gets a bit
% %% voluminous if written at deeper levels.
% %%
% %% Revision 1.13  1996/12/31 11:49:09  peterg
% %% Don't copy port bond causality if already set -- allows subsystem
% %% causality to be preset directely on named SS.
% %%
% %% Revision 1.12  1996/12/31 11:42:36  peterg
% %% *** empty log message ***
% %%
% %% Revision 1.11  1996/12/07  17:10:48  peterg
% %% Allows port SS at top level - ie takes it to be an ardianry SS at top
% %% level.
% %%
% %% Revision 1.10  1996/12/04 21:48:55  peterg
% %% Compares full-name with empty string (instead of testing for zero
% %% length.
% %%
% %% Revision 1.9  1996/08/30  12:55:40  peter
% %% More heirachical stuff added.
% %%
% %% Revision 1.8  1996/08/26  10:04:25  peterg
% %% Fixed error due to a line wrap.
% %%
% %%Revision 1.7  1996/08/16  12:58:58  peter
% %% Now does preferred causality of I and C.
% %% Revision 1.6  1996/08/09 08:27:29  peter
% %% Added a few deguging lines
% %%
% %% Revision 1.5  1996/08/08 18:06:18  peter
% %% Unified file naming scheme
% %%
% %% Revision 1.4  1996/08/08 08:30:06  peter
% %% The cbg filename contains the system name - this makes things easier
% %% when setting up the m to fig translation and m to ese translation
% %%
% %% Revision 1.3  1996/08/05 18:53:21  peter
% %% Fixed bug passing causality from subsystems.
% %%
% %% Revision 1.2  1996/08/05 15:41:41  peter
% %% Now recursively does causality on subsystems.
% %%
% %% Revision 1.1  1996/08/04 17:55:55  peter
% %% Initial revision
% %%
% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

pc = '%';
if nargin<1
  system_name = 'no_name';
end;

if nargin<4
  port_bonds = [];
end;

if nargin<5
  infofile = 'stdout';
end;

% Are we at the top level of the heirarchy?
at_top_level = strcmp(full_name, '');

% Create a file to contain shell commands which map subsystem types 
% onto components
if at_top_level % Then at top level
  fprintf(typefile, '# Commands to map types onto subsystems\n');
  fprintf(typefile, '# File %s_type.sh\n', system_name);
  fprintf(typefile, '# Generated by MTT on %s\n\n', date);
end;

% Create the (full) system name
if at_top_level
  full_name = system_name;
  system_type = system_name;
else
  full_name = [full_name, '_', system_name];
end;


% Add to file to contain shell commands which map subsystem types 
% onto components
fprintf(typefile, '$1%s$2%s$3\n', system_type, full_name);

fun_name = [system_type, '_abg'];

disp('====================================');
disp(['BEGIN: ', full_name, ' (', fun_name, ')']);
disp('====================================');

% If no such function - then there is nothing to be done.
if exist(fun_name)~=2
  mtt_error(['m-file ', fun_name, ' does not exist'], errorfile);
  bonds = [];
  status = [];
  return
end;

% Evaluate the system function to get the bonds and number of ports
eval(['[bonds,components,n_ports,N_ports]=', fun_name, ';']);

% Find number of bonds
[n_bonds,columns] = size(bonds);
if (columns ~= 2)&(n_bonds>0)
  error('Incorrect bonds matrix: must have 2 columns');
end;

 
% Find number of components
[n_components,m_components] = size(components);
if n_components==0 % there is nothing to be done
  return
end;

port_bond_direction

% Coerce the port (SS:[]) component bonds to have the same direction as
% of the bonds in the encapsulating system -- but not at top level
if (n_ports>0)&&(~at_top_level)
  port_bond_index = abs(components(1:n_ports,1:m_components))# relevant
							     # bond
							     # numbers
  port_bond_index = nozeros(reshape(\
                          port_bond_index',n_ports*m_components,1))'# vectorise

  for i=1:n_ports
    % Is the direction different?
    if (sign(components(i,1))~=port_bond_direction(i))
      mtt_info(sprintf("Flip port %i",i),infofile);
      % Flip direction at port
      components(i,1) = - components(i,1);
      % and at the other end
      for j=n_ports+1:n_components
	for k=1:m_components
	  if (abs(components(j,k))==port_bond_index(i))
	    components(j,k) = - components(j,k);
	  end
	end
      end;
      % Flip the bond causalities (these are arrow-orientated)
      bonds(port_bond_index(i),:) = -bonds(port_bond_index(i),:);
    end
  end
end



% Set initial status
status = -ones(n_components,1);

% If not at top level, then copy the port bonds.
if ~at_top_level 
  % Find number of port bonds
  [n_port_bonds,columns] = size(port_bonds);

  % Check compatibility - if ok copy port bonds to the internal bonds list.
  if n_port_bonds~=N_ports
  mtt_error(sprintf('%s: %1.0f port bonds incompatible with %1.0f ports', ...
               full_name, n_port_bonds, n_ports), errorfile);

  else % Copy the port bonds & status
    for j = 1:n_port_bonds
      jj = port_bond_index(j);
      for k = 1:2
#	if bonds(jj,k)==0 % only copy if not already set
	bonds(jj,k) = port_bonds(j,k);
#	end;
      end;
#      status(1:N_ports) = port_status;
    end
  end
else
  n_port_bonds=0;
end;
% bonds,port_bonds

% Causality indicator
total = 2*n_bonds;
done = sum(sum(abs(bonds)))/total*100;

% Outer while loop sets preferred causality
ci_index=1;

while( ci_index>0)
  ci_index
  old_done = inf;
  % Inner loop propagates causality
  while done~=old_done
   % disp(sprintf('Causality is %3.0f%s complete.', done, pc));
    old_done = done;
done,ci_index  
    for i = n_ports+1:n_components % Miss out the ports 
      if status(i) ~= 0 % only do this if causality not yet complete

	% Get the bonds on this component
	comp = nozeros(components(i,:));
	bond_list = abs(comp);
	direction = sign(comp)'*[1 1];
        n_bonds = length(bond_list);

	% Get the component details
	eval([ '[comp_type,name,cr,arg] = ', system_type, '_cmp(i);' ]);

	% Change name of 0 and 1 components -- matlab doesn't like numbers here
	if strcmp(comp_type,'0')
	  comp_type = 'zero';
	end;
	if strcmp(comp_type,'1')
	  comp_type = 'one';
	end;
      
	% Component causality procedure name
	cause_name = [comp_type, '_cause'];

	% Component equation procedure name
	eqn_name = [comp_type, '_eqn'];
	
        % Bonds on this component (arrow-orientated) -- these become the
        % port bonds on the ith component of this subsystem.
	
	comp_bonds=[];
	for kk = 1:n_bonds
	  comp_bonds(kk,:) = bonds(bond_list(kk),:);
	end;
	

      % Invoke  the appropriate causality procedure
      if exist(eqn_name)~=2 % Try a compound component
        % Port status depends on whether the corresponding bonds are
        %  fully causal at this stage.
        one = ones(n_bonds,1);
        port_status = (sum(abs(comp_bonds'))'==2*one) - one;

	% Direction of bonds on the ports (0 if next to port)
	port_bond_direction = -sign(components(i,1:n_bonds))';

	% If there is a predefined causality function; use it
        if exist(cause_name)==2
    	  % Convert from arrow orientated to component orientated causality
	  comp_bonds = comp_bonds.*(port_bond_direction*[1 1]);
	
          % Evaluate the built-in causality procedure
	  eval([ '[comp_bonds] = ', cause_name, '(comp_bonds);' ]);

         % and convert from component orientated to arrow orientated causality
         comp_bonds = comp_bonds.*(port_bond_direction*[1 1]); 
        end;

	[comp_bonds,s] = abg2cbg(name, comp_type, full_name, 
            comp_bonds, port_bond_direction, port_status, ...
	    typefile, infofile, errorfile);
	
	% Create a single status from the status vector s
	if max(abs(s)) == 0 % Causal
	  status(i) = 0;
	else
	  if max(s) == 1 % At least one component is overcausal
	    status(i) = 1;
	  else           % no component is overcausal but some are undercausal
	    status(i) = -1;
	  end;
	end;

      else % its a simple component -- or explicit causality defined
	disp(['---', name, ' (', cause_name, ') ---']);
	comp_bonds_in = comp_bonds

	% Convert from arrow orientated to component orientated causality
	comp_bonds = comp_bonds.*direction;
	
        % Evaluate the built-in causality procedure
	eval([ '[comp_bonds,status(i)] = ', cause_name, '(comp_bonds);' ]);

        % and convert from component orientated to arrow orientated causality
        comp_bonds = comp_bonds.*direction; 
       
       comp_bonds_out = comp_bonds
      end;

      bonds,bond_list,comp_bonds
      % Update the full bonds list
      bonds(bond_list,:) = comp_bonds;
    end;
    end;
    
    done = sum(sum(abs(bonds)))/total*100;
    %disp(sprintf('Causality is %3.0f%s complete.', done, pc), infofile));
    
  end;
  
  % Set causality of a C or I which is not already set
  [ci_index,prefered] = getdynamic(status,system_type)
  if ci_index>0
    disp('Set causality of a C or I which is not already set')
    ci_bond_index = nozeros(components(ci_index,:)); # Get all bonds
    ci_direction = sign(ci_bond_index);
    ci_bond_index = abs(ci_bond_index);
    bonds(ci_bond_index,1:2) = prefered*ci_direction'*[1 1];
  end;
  
end;

if n_ports>0
  status(1:n_ports) = zeros(n_ports,1); # Port status not relevant
endif;
	  
% Print final causality
final_done =  (sum(status==zeros(n_components,1))/n_components)*100;

if at_top_level
  mtt_info(sprintf('Final causality of %s is %3.0f%s complete.', ...
      full_name, final_done, pc), infofile);

  if final_done<100
	  mtt_error(sprintf("Unable to complete causality"),errorfile);
  end;
 
  % List overcausal bonds
  [over_causal_bonds,n] = getindex(status,1);
  if n>0
    for i=over_causal_bonds'
      eval([ '[comp_type,name] = ', system_type, '_cmp(i);' ]);
      mtt_info(sprintf('Component %s (%s) is overcausal', name, comp_type), ...
	  infofile);
    end;
  end;
  
  % List undercausal bonds
  [under_causal_bonds,n] = getindex(status,-1);
  if n>0
    for i=under_causal_bonds'
      eval([ '[comp_type,name] = ', system_type, '_cmp(i);' ]);
      mtt_info(sprintf('Component %s (%s) is undercausal', name, comp_type), ...
	  infofile);
    end;
  end;
end;

% $$$ file_name = [full_name, '_', system_type]
file_name = [full_name, '_cbg.m'];
disp(['Writing ', file_name]);
cbgfilenum = fopen(file_name,'w');
write_cbg(cbgfilenum,full_name,system_type,bonds,status,components);
fclose(cbgfilenum);

% Return the port bonds - arrow orientated causality - and the direction 
if ~at_top_level % Not at top level
  port_bonds = bonds(port_bond_index,:) # Return port bonds
end;

disp('====================================');
disp(['END: ', full_name, ' (', fun_name, ')']);
disp('====================================');