completeData.m

%%Performs the following function
%(1)Completes the missing data from the edf file using rplparallel marker
%   data
%(2)In case rplparallel is also missing markers, it calls callEyelink
%   function to complete the missing data from rplparallel using eyelink, or
%   delete common missing trials
%(3)Returns elTrials that contains the completed and corrected eyelink
%   times, missingData that contains all the timestamps that were missing
%   from the edf file (which is then saved in .csv form in eyelink.m)
%(4)Returns a flag that is set 0 if the session is valid, and 1 if the
%   the session is a dummy session with too few trials (and so it is
%   skipped)
function [elTrials, missingData, flag] = completeData(edfdata, m, messageEvent, sessionName, moreSessionsFlag)

	% set default flag to 0
	flag = 0;

    %Extract the start time of the session in case the start time of the 
    %1st trial is missing (the 1st index of messageEvent is the session
    %index
    sessionStart = edfdata.FEVENT(messageEvent(1,1)).sttime; 

    %the start of the experiment is taken to normalise the data 
    expTime = edfdata.FEVENT(1).sttime;
    expTime = uint32(expTime);
   
    %%Correct the data that has been extracted from edfdata for ONE sessiom
    %create a new matrix that contains all trial messages only

    disp(size(messageEvent));
    messageEvent(contains(m, 'Trigger')) = [];
    m(contains(m, 'Trigger')) = [];
    
    messageEvent(contains(m, 'end')) = [];
    m(contains(m, 'end')) = [];
    
    messageEvent(contains(m, 'ERROR')) = [];
    m(contains(m, 'ERROR')) = [];
    messages = m;
    disp(size(messageEvent));
    
    %stores the starting time of all the events 
    eltimes = {edfdata.FEVENT(messageEvent(1:end)).sttime}'; 
    eltimes = cell2mat(eltimes); 
    eltimes = uint32(eltimes);
    
    %Make the ripple object to access its fields
    cd (sessionName)
    rpl = rplparallel('auto');
   
        
    if (rpl.data.numSets ~= 0 && isempty(rpl.data.timeStamps)~=1) %no missing rplparallel.mat
        
        %markers will store all the event numbers in the trial, as taken 
        %from the ripple object. This will be used to mark which events are missing
        %in th eyelink object. (1-start, 2-cue, 3/4 - end/timeout)
        markers = rpl.data.markers;
        rpltimeStamps = rpl.data.timeStamps; %stores the event time for ripple
        n = size(markers, 1);
        
        %CHECK IF THE RPLPARALLEL OBJECT IS FORMATTED CORRECTLY OR IS
        %MISSING INFORMATION
        if(n==1) %if the formatting is 1xSIZE
            df = rpl; 
            save('rplparallel0.mat', 'df');
            markers(1) = [];
            rpltimeStamps(1) = [];
            rpltimeStamps(find(~markers)) = [];
            markers(markers==0) = [];
            n = size(markers,2)/3;
            if (rem(size(markers),3) ~= 0)
               markers = rpl.data.markers;
               rpltimeStamps = rpl.data.timeStamps;
               [markers,rpltimeStamps] = callEyelink(markers, messages, eltimes-expTime, rpltimeStamps);
               
            else %if the rplparallel object dimension is divisibly by 3 - better to use arrangeMarkers directly.
                markers = reshape (markers, [3 n]);
                rpltimeStamps = reshape(rpltimeStamps, [3 n]);
                markers = markers';
                rpltimeStamps = rpltimeStamps';
            end 
            %Save the newly created rplparallel object into the folder 
             n = size(markers, 1);
             df.data.markers = markers;
             df.data.timeStamps = rpltimeStamps;
             save('rplparallel.mat', 'df');
             
        elseif (n*3 < size(m,1)) %If rplparallel obj is missing data, use callEyelink 
            if (exist('rplparallel0.mat', 'file')~=2)
                df = rpl;  
                [markers,rpltimeStamps] = callEyelink(markers, messages, eltimes-expTime, rpltimeStamps);
                %save object and then return
                n = size(markers, 1);
                df.data.markers = markers;
                df.data.timeStamps = rpltimeStamps;
                save('rplparallel.mat', 'df');
            end
        end
        
        cd ..;
        
        noOfmessages = size(messages, 1); %stores the number of messages recorded by eyelink in the session 
         
        missing = zeros (n, size(markers,2)); %stores the event that is missing
        rpldurations = zeros (n, size(markers,2)); %stores the rpl durations for filling missing eyelink timestamps
        elTrials = zeros (n, size(markers,2)); %stores the eyelink timestamps for the events in a trial
        
        %To check if there are more sessions than must be
        if(moreSessionsFlag ~= 0)
           if ((n*3) - size(messageEvent,1)-2 >= 100)
               flag = 1;
               missingData = cell(n,3);
               elTrials = cell (n,3);
               missingData = cell2table(missingData, 'VariableNames', {'Type', 'Timestamps', 'Messages'});
               return; 
           end 
        end 
        
        %calculate the durations between rplparallel timestamps 
        %rpldurations(1,1) is assumed to be the time difference between the
        %start of the session and the trial start time
        rpldurations (:,1) = [rpltimeStamps(1,1); rpltimeStamps(2:end, 1) - rpltimeStamps(1:end-1, 3)]; %end of last trial - start of this trial
        rpldurations (:,2) = rpltimeStamps(:,2) - rpltimeStamps(:,1); %cue time - start time
        rpldurations (:,3) = rpltimeStamps (:,3) - rpltimeStamps(:,2); %end - cue time
        
        idx = 1;
        n = n*3; %size of the markers 
        newMessages = cell(n,1); %stores all the missing messages 
        
        %For loop that goes through the entire rplparallel markers matrix
        %(1) Checks if edf message markers are missing, and accordingly
        %corrects the missing time using the rpldurations
        %(2) ensures that missing trials common to eyelink and rplparallel
        %are deleted
        %(3) creates the missing array that is later addded to the
        %missingData table 

        [elTrials, missing, newMessages] = filleye(messages, eltimes, rpl);
%         save('should_be_correct.mat', 'elTrials', 'missing', 'newMessages');
        
        %Get rid of extra zeros 
        [row ~] = find(~elTrials);
        elTrials(row, :) = []; 
        n = size(find(missing),1);
        elTrials = uint32(elTrials); %convert to uint to get rid of floating points

        if (n ~= 0) %if there are missing messages in this session, make the missingData matrix to add to the .csv file 
            fprintf('Missing messages\n');
            type = cell(n,1);
            type(:,:) = {24};
            correctedTimes = num2cell(sortrows(elTrials(find(missing))));
            newMessages = newMessages(~cellfun('isempty',newMessages));
            missingData = horzcat(type, correctedTimes, newMessages); %this needs to be turned into a csv file 
            missingData = cell2table(missingData, 'VariableNames', {'Type', 'Timestamps', 'Messages'}); 
            clear correctedTimes;
            
        else %if there are not any missing messages, make the missingData matrix empty 
            fprintf('No missing messages\n');
            missingData = cell(n,3);
            missingData (:,:) = {0};
            missingData = cell2table(missingData, 'VariableNames', {'Type', 'Timestamps', 'Messages'});
        end 

        %%To ensure that the correction of the eyelink object went correctly, 
        %%we now plot a histogram of the discrepancies in the start-cue,
        %%cue-end and end-start durations for ripple and eyelink objects for
        %%the same session to ensure that the data recorded is consistent 
        elTrials = elTrials - expTime;
        eldurations (:,1) = [0; elTrials(2:end, 1) - elTrials(1:end-1, 3)]; %end of last trial - start of this trial
        eldurations (:,2) = elTrials(:,2) - elTrials(:,1); %cue time - start time
        eldurations (:,3) = elTrials (:,3) - elTrials(:,2); %end - cue time
        eldurations = double(eldurations);
        eldurations = eldurations/1000; %conversion to seconds
        
        discrepancies = abs(rpldurations - eldurations); %stores the discrepancy between the two times in seconds
        discrepancies(1,1) = 0; 

        %%Plot the distributions of the durations in ms
        figure (1)
        histogram (discrepancies, 50);
        hold on;
        title('Histogram for rplparallel and eyelink durations');
        xlabel ('s');
        ylabel ('occurence');
        hold off 
        
    else %missing rplparallel  
        %you are going to assume that there are no missing messages
        cd ..; 
        fprintf ('Empty Object. Just fill up time array\n');
        n = size(messages,1);
        elTrials = zeros (n, 3); %stores the eyelink timestamps for the events in a trial
        missing = zeros(n,3); 
        
        for i = 1:n
        
            r = floor(i/3) + 1; %row 
            c = rem(i,3); %column

            if (c == 0)
                r = r-1;
                c = 3;
            end 
           
            missing (r,c) = 0;  
            if (contains(messages(i,1), {'Start Trial'})==1)
                elTrials (r,1) = eltimes(i,1) - expTime;
            elseif (contains(messages(i,1), {'End Trial'}) == 1)
                elTrials (r,3) = eltimes(i,1) - expTime;
            elseif (contains(messages(i,1), {'Timeout'}) == 1)
                elTrials (r,3) = eltimes(i,1) - expTime;
            elseif (contains(messages(i,1), {'Cue Offset'}) == 1)
                elTrials (r,2) = eltimes(i,1) - expTime;
            end
            
        end
        
        disp('No missing messages');
        missingData = cell(n,3);
        missingData (:,:) = {0};
        missingData = cell2table(missingData, 'VariableNames', {'Type', 'Timestamps', 'Messages'});
    

    end