% CONSOLIDATE resamples all input AOs onto the same time grid.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% CONSOLIDATE resamples all input AOs onto the same time grid and truncates all
%             time-series to start at the maximum start time of the inputs and end
%             at the minimum stop time of the inputs.
% 
% ALGORITHM:
%             1) Drop duplicate samples (ao/dropduplicates)
%             2) Interpolate missing samples (ao/interpmissing)
%             3) Fix uneven sample rate using interpolate (ao/fixfs)
%             4) Resample to same fs, either max or specified (ao/resample
%                or ao/interp depending on ratio of old and new sample
%                rate)
%             5) Truncate all vectors to minimum overlap of time-series
%                (ao/split)
%             6) Resample on to the same timing grid (ao/interp)
%             7) Truncate all vectors to same number of samples to correct for 
%                any rounding errors in previous steps (ao/select)
%
% CALL:       >> bs = consolidate(as)
%
% INPUTS:     as  - array of at least two time-series analysis objects
%             pl  - parameter list (see below)
%
% OUTPUTS:    bs  - array of analysis objects, one for each input
%
% <a href="matlab:utils.helper.displayMethodInfo('ao', 'consolidate')">Parameters Description</a>
%
% VERSION:     $Id: consolidate.m,v 1.32 2011/04/08 08:56:13 hewitson Exp $
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%           't'     - specify a new time vector to resample on to. This
%                     will be truncated to fit within the maximum start
%                     time and minimum stop time of the inputs.
%      or

function varargout = consolidate(varargin)

  % Check if this is a call for parameters
  if utils.helper.isinfocall(varargin{:})
    varargout{1} = getInfo(varargin{3});
    return
  end

  import utils.const.*
  utils.helper.msg(msg.PROC3, 'running %s/%s', mfilename('class'), mfilename);
  
  % Collect input variable names
  in_names = cell(size(varargin));
  for ii = 1:nargin,in_names{ii} = inputname(ii);end

  % Collect all AOs and plists
  [as, ao_invars] = utils.helper.collect_objects(varargin(:), 'ao', in_names);
  [pl, pl_invars] = utils.helper.collect_objects(varargin(:), 'plist', in_names);

  if numel(as) < 2
    error('### Consolidate requires at least two time-series AOs to work.');
  end
  
  if nargout == 0
    error('### Consolidate cannot be used as a modifier. Please give an output variable.');
  end
  
  % Decide on a deep copy or a modify
  bs = copy(as, nargout);
  na = numel(bs);

  % Combine plists
  pl = parse(pl, getDefaultPlist);

  % Get only tsdata AOs
  inhists = [];
  for j=1:na
    if ~isa(bs(j).data, 'tsdata')
      bs(j) = [];
      warning('!!! Skipping AO %s - it''s not a time-series AO.', bs(j).name);
    else
      % gather the input history objects
      inhists = [inhists bs(j).hist];
    end
  end

  % If fs is specified, use it. Otherwise, use max of all
  % input AOs.
  fs = find(pl, 'fs');
  if isempty(fs)
    % compute max fs
    fs = 0;
    for j=1:na
      if bs(j).data.fs > fs
        fs = bs(j).data.fs;
      end
    end
  end
  utils.helper.msg(msg.PROC2, 'resampling all time-series to an fs of %f', fs);
  
  %----------------- Drop all repeated samples
  utils.helper.msg(msg.PROC1, 'drop duplicates');
  for j=1:na
    utils.helper.msg(msg.PROC2, 'processing %s', bs(j).name);
    dropduplicates(bs(j),pl);
  end

  %----------------- Interpolate all missing samples
  utils.helper.msg(msg.PROC1, 'interpolate missing samples');
  for j=1:na
    utils.helper.msg(msg.PROC2, 'processing %s', bs(j).name);    
    interpmissing(bs(j),pl.pset('method', find(pl, 'interp_method')));
  end

  
  %----------------- Fix uneven sampling
  utils.helper.msg(msg.PROC1, 'fixing uneven sample rates');
  for j=1:na
    utils.helper.msg(msg.PROC2, 'processing %s', bs(j).name);
    fixfs(bs(j),pl.pset('method', find(pl, 'fixfs_method')));
  end
  %----------------- Resample all vectors to same fs
  utils.helper.msg(msg.PROC1, 'resample to same fs');

  for j=1:na
    % Check the resampling factor
    [P,Q] = utils.math.intfact(fs,bs(j).data.fs);
    if P > 100 || Q > 100
      utils.helper.msg(msg.PROC2, 'resampling factor too high [%g/%g]. Trying interpolation', P, Q);
      N  = length(bs(j).data.getX);
      t  = linspace(0, (P*N/Q-1)/fs, P*N/Q);
      interp(bs(j), plist('vertices', t));
    else
      resample(bs(j), plist('fsout', fs));
    end
  end

  %---------------- Time properties of AOs
  % Find max start time
  start = 0;
  for j=1:na
    dstart = bs(j).data.t0.utc_epoch_milli/1000 + bs(j).data.getX(1);
    if dstart > start
      start = dstart;
    end
  end

  % Find min stop time
  stop = 1e20;
  for j=1:na
    dstop = floor(bs(j).data.t0.utc_epoch_milli/1000 + bs(j).data.getX(end));
    if dstop < stop
      stop = dstop;
    end
  end

  %----------------- Truncate all vectors
  utils.helper.msg(msg.PROC1, 'truncate all vectors');
  utils.helper.msg(msg.PROC2, 'truncating vectors on interval [%.4f,%.4f]', start, stop);

  % split each ao
  bs = split(bs, plist('timespan', timespan(start, stop)));
  
  %----------------- Resample all vectors on to the same grid
  utils.helper.msg(msg.PROC1, 'resample to same grid');
  % compute new time grid
  
  % get the grid from the first AO
  for j=1:na
    toff = start - bs(j).t0.utc_epoch_milli/1000;
    N = length(bs(j).data.getX);
    t = linspace(toff, toff+(N-1)/fs, N);
    interp(bs(j), plist('vertices', t));
  end
  
  % Now ensure that we have the same data length
  ns = realmax;
  for jj=1:na
    if len(bs(jj)) < ns
      ns = len(bs(jj));
    end
  end
  
  bs = select(bs, 1:ns);
  
  nsecs = [];
  for j=1:na
    if isempty(nsecs)
      nsecs = bs(j).data.nsecs;
    end
    if nsecs ~= bs(j).data.nsecs
      error('### Something went wrong with the truncation. Vectors don''t span the same time period.');
    end
  end

  %----------------- Set history on output AOs

  for j=1:na
    bs(j).name = sprintf('%s(%s)', mfilename, ao_invars{j});
    bs(j).addHistory(getInfo('None'), pl, ao_invars(j), inhists(j));
  end

  % Set output
  if nargout == numel(bs)
    % List of outputs
    for ii = 1:numel(bs)
      varargout{ii} = bs(ii);
    end
  else
    % Single output
    varargout{1} = bs;
  end
end

%--------------------------------------------------------------------------
% Get Info Object
%--------------------------------------------------------------------------
function ii = getInfo(varargin)
  if nargin == 1 && strcmpi(varargin{1}, 'None')
    sets = {};
    pl   = [];
  else
    sets = {'Default'};
    pl   = getDefaultPlist;
  end
  % Build info object
  ii = minfo(mfilename, 'ao', 'ltpda', utils.const.categories.sigproc, '$Id: consolidate.m,v 1.32 2011/04/08 08:56:13 hewitson Exp $', sets, pl);
  ii.setModifier(false);
  ii.setArgsmin(2);
end

%--------------------------------------------------------------------------
% Get Default Plist
%--------------------------------------------------------------------------

function plout = getDefaultPlist()
  persistent pl;  
  if exist('pl', 'var')==0 || isempty(pl)
    pl = buildplist();
  end
  plout = pl;  
end

function pl_default = buildplist()
  pl_default = combine(...
    plist({'fs','The target sampling frequency for consolidate'}, paramValue.EMPTY_DOUBLE),...
    plist({'interp_method', 'The method for the interpolation step'}, {2, {'nearest', 'linear', 'spline', 'cubic'}, paramValue.SINGLE}), ...
    plist({'fixfs_method', 'The method for the fixfs step'}, {1, {'Time', 'Samples'}, paramValue.SINGLE}), ...
    ao.getInfo('dropduplicates').plists,...
    ao.getInfo('interpmissing').plists,...
    ao.getInfo('fixfs').plists);
end