function X = seqread (infile)

% SEQREAD - reads a sequence object from a file
%
% caveats : doesnt deal with empty sequences within tags
% doesnt check if length in tags is correct

X.data = {};
X.labels = {};
X.origlabels = [];
X.comments = {};

HAS_DIM_TAG = 0;
HAS_LABEL_TAG = 0;
DENSE = 0;
SPARSE = 0;
TAGSEP = 0; % sequences separated by tags
SPACESEP = 0; % sequences separated by blank lines

DIM = 0;
LABS = [];

z = textread(infile,'%s','headerlines',0,'delimiter','\n','bufsize',4096*100);
LZ = length(z);

if ~LZ
  return;
  warning('no lines in input file');
end


%%%%%%%%%%%%%%%%%%%%% first check for any dimension tag %%%%%%%%%%%

f=findstr(z{1},'dimension');
if length(f)
  L=length(z{1});
  g=findstr(z{1},'>');
  if (~length(g))
    error('no ending triangular bracket for dimension tag');
  end

  f2=findstr(z{1},'=');
  if (length(f2))
    if (f2>length(z{1})) 
      error('what is the dimension in the dimension tag?');
    else
      DIM = str2num(z{1}(f2+1:g-1));
    end
  else
    DIM = str2num(z{1}(f+1:g-1));
  end

  if ~DIM
    error ('what is the dimension?');
  else
    HAS_DIM_TAG = 1;
  end
end

%%%%%%% Now check for any label tag %%%%%%%%%%%%%%%

if (LZ < HAS_DIM_TAG+1)
  return;
  warning('very short file!');
end

f=findstr(z{HAS_DIM_TAG+1},'label');
if length(f)
  L=length(z{HAS_DIM_TAG+1});
  g=findstr(z{HAS_DIM_TAG+1},'>');
  if (~length(g))
    error('no ending triangular bracket for label tag');
  end

  f2=findstr(z{HAS_DIM_TAG+1},'=');
  if (length(f2))
    if (f2>length(z{HAS_DIM_TAG+1})) 
      error('what is the label in the label tag?');
    else
      LABS = strread(z{HAS_DIM_TAG+1}(f2+1:g-1),  '%d', 'delimiter', ' ');
    end
  else
    LABS = strread(z{HAS_DIM_TAG+1}(f+1:g-1),  '%d', 'delimiter', ' ');
  end

  if ~length(LABS)
    error ('Are there no labels?');
  else
    HAS_LABEL_TAG = 1;
    X.origlabels = LABS;
  end
end

%%%%%%%%%%%%%%%% Now keep going till you get to the first sequence data line %%%%%%

i = HAS_LABEL_TAG + HAS_DIM_TAG + 1;

found_data_line = 0;
while ((i <= length(z)) & (~ found_data_line))
  if (length(z{i})>0)              % if this line isnt empty
    if (z{i}(1) == '%')            % if this line starts with a % it is a comment
      X.comments{ 1+length(X.comments) } = z{i}(2:length(z{i}));
    else
      if length(strfind(z{i},'<sequence'))
	TAGSEP = 1;
	found_data_line = 1;
      else
	if hasnumber(z{i})
	  if (~ TAGSEP)
	    SPACESEP = 1;
	  end
  	  found_data_line = 1;
        end
      end
    end
  end
  if (~found_data_line) 
    i=i+1;
  end
end


if (found_data_line)
    if TAGSEP
      i=i+1;      % this line is <sequence...> so go to next one
    end
    if length(findstr(z{i},':')) % if this line is sparse
      SPARSE = 1;
    else
      DENSE = 1;
      % now determine DIM since all data lines must have the same number of elements
      a = strread(z{i},'%s','delimiter',' ');
      if (mod(str2num(a{1}),1))
	z{i}
	error('first element of each data line should be an integer label');
      end
      DIM = length(a)-1;
    end
    if TAGSEP
      i=i-1;      % now go back to first line
    end
else
  error ('cannot find data line');
end

if (~DENSE) & (~SPARSE)
  error ('is this file in dense or sparse format?');
end

if SPARSE
  Xi = {};
  Xj = {};
  Xv = {};
end

anynumbershere = -1;

inspace = 0;  % inspace is not used
vecnum = 0;   % number of next current/vector in current sequence
seqnum = -2;

if TAGSEP 
  seqnum = 0;   % number of current sequence 
elseif SPACESEP
  seqnum = 1;   
end

while i<=LZ
  if (TAGSEP) & (findstr(z{i},'<sequence'))
    seqnum = seqnum+1;
    vecnum = 0;
    anynumbershere = 0;
  else
    anynumbershere = hasnumber(z{i});
    if (anynumbershere)
      vecnum = vecnum+1;
      inspace = 0;
      
      % now can ADD NUMBERS HERE to vecnum-th vector of seqnum-th sequence
      
    else
      if ((length(z{i})>0) & (z{i}(1) == '%'))
	% comment, add it
	X.comments{ 1+length(X.comments) } = z{i}(2:length(z{i}));
      else
	if (SPACESEP) 
	  if (~inspace)
	    inspace = 1;
	    seqnum = seqnum+1;
	    vecnum = 0;
	  end
	end      % if TAGSEP, ignore blank lines and any other lines
      end
    end
  end
  
  if (anynumbershere>0)
    if hasalpha(z{i})      %% its ok if there is a 'e' as that could be part of exponential notation
      z{i}
      error('found letters in a line that should have just numbers');
    end
    if ((length(findstr(z{i},':'))) & (DENSE))
      z{i}
      error('found a colon : in a line when previous lines are in dense format');
    end
      
    a = strread(z{i},'%s','delimiter',' ');

    if (length(a)==0)
      error('no entry in data line');
    elseif ((DENSE) & (length(a) ~= DIM+1))
      z{i}
      error (sprintf('This line has a dense %d-dimensional vector while the first line had a %d-dimensional vector\n',length(a)-1,DIM));
    elseif length(findstr(a{1},':'))
      error('first element of line should be a label');
    else
      lab = str2num(a{1});
      if (mod(lab,1))
	z{i}
	error(sprintf('label %f should be an integer',lab));
      end
      f = find(X.origlabels == lab);
      if (~length(f))   % couldnt find lab in original labels
	if HAS_LABEL_TAG
	  error(sprintf('found label %d not supplied in label tag',lab));
	else
	  X.origlabels = [X.origlabels lab];
	end
      end
      if (length(X.labels)<seqnum)
	X.labels{seqnum} = [];
      end
      X.labels{seqnum}(1,vecnum) = lab;
    end

    if DENSE
      x = zeros(1,DIM);
    elseif SPARSE
      x = sparse(1,DIM);
    else 
      error('not sparse or dense?');
    end
    
    for j=2:length(a)
      if SPARSE
	b = strread(a{j},'%f','delimiter',':');
	indx = b(1);
	valu = b(2);
	if (((mod(indx,1))) | (indx < 1))
	  error(sprintf('index %f should be a positive integer',indx));
	end
      elseif DENSE
	indx = j-1;
	valu = str2num(a{j});
      end

      if (indx  > DIM)
	if HAS_DIM_TAG
	  z{i}
	  error (sprintf('found index %d that is larger than dimension %d in dimension tag',indx,DIM));
	else
	  DIM = indx;
	end
      end
      
      if DENSE
	if length(X.data)<seqnum, X.data{seqnum} = []; end
	X.data{seqnum}(vecnum,indx) = valu;
      elseif SPARSE
	if (length(Xi)<seqnum), Xi{seqnum}=[]; end
	if (length(Xj)<seqnum), Xj{seqnum}=[]; end
	if (length(Xv)<seqnum), Xv{seqnum}=[]; end
	r = length(Xv{seqnum});
	Xi{seqnum}(r+1) = vecnum;
	Xj{seqnum}(r+1) = indx;
  	Xv{seqnum}(r+1) = valu;
      end
    end

  end
  i=i+1;
end

if SPARSE
  for i=1:length(Xi)
    X.data{i} = sparse(Xi{i},Xj{i},Xv{i},max(Xi{i}),DIM);
  end
end    



X.origlabels = sort(X.origlabels);
L = length(X.origlabels);

% now replace all labels by value between 1 and L

for i=1:length(X)
  N = length(X.labels{i});
  labs = zeros(1,N);
  for j=1:L
    x = find(X.labels{i} == X.origlabels(j));
    if length(x)
      labs(x) = j;
    end
  end
  X.labels{i} = labs;
end

%%%%%%%%%%%%%

function answer = hasnumber (s)

digits = '0123456789';
answer = 0;
for i=1:length(digits)
  if length(findstr(s,digits(i)))
    answer=1;
    break;
  end
end

function answer = hasalpha (s)

alpha = 'qwrtyuioplkjhgfdsazxcvbnm'; % no e
answer = 0;
for i=1:length(alpha)
  if length(findstr(s,alpha(i)))
    answer=1;
    break;
  end
end