problem with xlsread:Bizarre results

 Dear all,

I have an excel file that contains 3 sheets

when I look at each of these sheets  inside the excel file they all have the same number of rows( 245)  but when I use the command

[num1 ,txt1 ,raw1 ] = xlsread(xlfilename,1);

[num2 ,txt2 ,raw2 ] = xlsread(xlfilename,2);

[num3 ,txt3 ,raw3 ] = xlsread(xlfilename,3);

I get that raw1 has 266 rows

raw2 has 267 rows

raw3 has 279 rows

I also noticed that inside the workspace matlab adds arbitrarily different number of rows of NaN at the end of the raw1 raw2 raw3

Bear also in mind that “raw1” “raw2” “raw3” contains string numbers and empty cells

Unfortunately it is not convenient for my purposes to use “num1““num1“ “num1“

Thank you

[Jan]

How did you check that the sheets have 245 rows? Perhaps there are some orphaned elements far far beyond the right margin of the window?

[Sabbas]

Is it possible to select what I "see" ( I mean the columns and the rows that are visible) from the excel file that contains the orinimal data? Because none of the three outputs [num1 ,txt1 ,raw1 ] help me do that.

Thanks

[Tom]

xlsread(filename,-1)

Lets you select the data by going into the spreadsheet.

[Sabbas]

I tried but does not help because then I have to select the area I am interested in. But my excel file contains39 sheets. So it does not make sense

Answer 1

[Mark Whirdy]

Hi Sabbas I think its maybe impossible to debug this from here - all I can suggest is that the rows below are not actually empty but contain some character somewhere, putting a breakpoint in xlsread and stepping through should clarify this.

In general though, I think that you shouldn't be using xlsread at all (especially when you are running several of them in quick succession) as it instantiates an excel COM application object with each run. Use the Activex object directly and capture the data by invoking the VBA-type properties and methods of the excel COM Object, its much faster and you have more visibility on what is happening.

Below is a code snippet that you can adapt:

All the best,

Mark Whirdy

xlApp = actxserver('Excel.Application');
xlApp.Visible = 1;
xlWorkbook = xlApp.workbooks.Open(fullfile(xlFilePath,xlFileName),0,true);
xlSheets = xlWorkbook.Sheets;
xlSheetNamesArray = cell(xlSheets.Count,1); 
for i = 1:xlSheets.Count
    xlSheetNamesArray{i} = xlSheets.Item(i).Name; % sheet-order is not guaranteed so must build array
end

[~,idx] = ismember('Price',xlSheetNamesArray);
xlSheets.Item(idx).Activate
xlActiveSheet = xlWorkbook.ActiveSheet;
xlCurrRange = xlActiveSheet.Range('C4:C33');
priceVector = xlCurrRange.Value2;
priceVector = cell2mat(priceVector);

[~,idx] = ismember('Portfolios',xlSheetNamesArray);
xlSheets.Item(idx).Activate
xlActiveSheet = xlWorkbook.ActiveSheet;
xlCurrRange = xlActiveSheet.Range('B4:B22');
isinVector = xlCurrRange.Value2;
xlCurrRange = xlActiveSheet.Range('C3:G3');
ptfNumVector = cell2mat(xlCurrRange.Value2)';
xlCurrRange = xlActiveSheet.Range('C4:G22');
dataMatrix = xlCurrRange.Value2;

isnanMatrixMask = strcmp(dataMatrix,'ActiveX VT_ERROR: '); % handle missing data - assume as no position
dataMatrix(isnanMatrixMask) = {0};
dataMatrix = cell2mat(dataMatrix);

[Sabbas]

thank you Mark. What worries me with the ActiveX is that I have to incert cells like B4:B22. But i want to insert the data in the form of matrices like Sheet1(1:10.2:end)

[Mark Whirdy]

Hi Sabbas

I’m not sure I understand. By “I have to incert cells like B4:B22.” are you saying that you are reading from excel into Matlab and don’t want to specify the exlce range reference bounds explicitly but want Matlab/excel to “know” what the relevant (excel cells containing filled values) range of the spreadsheet is in advance to avoid reading in the entire sheet as a cell-array?

If so, then you may as one solution use the “UsedRange” property of the Sheet object as per http://msdn.microsoft.com/en-us/library/aa215923(v=office.11).aspx This is what the xlsread function itself does.

DataRange = xlApp.ActiveSheet.UsedRange;
DataRange.Value;

However, some funky behaviour around this UsedRange excel property (or in cells adjacent to the intended used-range) may be what was originally causing the 267, 279 arbitrary length problem in the first place. You should be able to see if those extra cells are really empty by validations such as cellfun(@isempty,raw(:,1)) and cellfun(@isnumeric,raw(:,1)) and maybe spot the problem from there.

Another useful VBA property you can invoke in Matlab is the “xlToRight” and “xldown” as per http://www.mathworks.co.uk/support/solutions/en/data/1-3QI7UR/index.html?solution=1-3QI7UR You can identify a range of filled cell values in excel this way also.

Honestly, I don’t see the problem with specifying the range-bounds explicitly though as you should have some knowledge of this at the start, and then be able to remove empty/useless rows from the bottom and side of the cell-array (e.g. A1:BA1000) as F. is saying below.

If I have misunderstood what you’re saying, please expand in as much detail & clarity as possible on your previous point.

P.S. forget Matlab for a minute, go into excel to cell A1 (or whatever the upper-rightmost filled cell is) in each sheet and hit Ctrl+Shift+DownCursor – does the active range extend beyond what appears to be visibly filled? If so, the cells are not empty.

P.P.S. If you still have a problem after this please email the xls to [e-mail]

Best, Mark

[Mark Whirdy]

typo: " upper-rightmost " --> " upper-leftmost "

[Sabbas]

thanks Mark!

Answer 2

[F.]

Hello,

I think the problem is with Excel.

I think you should have at a moment more than 245 lines, and you should have deleted them. You don't see these old lines in excel, but the application remains it. When you are using xlsread, you import these empty lines. You can see this with raw datas.

When you want import data from excel, 2 solutions:

• First, in excel you select all lines ( and after all columns ) after your data and you remove them (not only delete)
• !the other one, with raw data you can define lines and columns with no data or NaN data, and you delete them :

[ N, T, R ] = xlsread( ‘toto.xls’ );
Tmp = cellfun( @isnan, R, ‘UniformOuput’, true);
R( all( Tmp,1), : ) = [] ;
R( : , all( Tmp,2) ) = [] ;

Or something like this (I’m sorry but I don’t have Matlab on my Pc today) (it’s the same thing with “empty” )

[Sabbas]

thanks F. the only problem is that among the usefull columns some are empty By erasing them I create problem in the data file

[F.]

I don't undersand your comment. If you want just erase rows from your raw data :

   Tmp = cellfun( @isnan, R, 'UniformOutput', true)
   R(all(Tmp,2),: )=[]

but if you have some rows with NaN in your data and you can use :

   Tmp = cellfun( @isnan, R, 'UniformOutput', true)
   Tmp = find( ~all(Tmp,2),1, 'last')
   R((Tmp+1):end,: )=[]

otherwise, I'm sorry, I didn't undersand your problem because, for me, it's just a problem with NaN rows at the end of your data ...

[Sabbas]

I think that your last code

Tmp = cellfun( @isnan, R, 'UniformOutput', true)

   Tmp = find( ~all(Tmp,2),1, 'last')

   R((Tmp+1):end,: )=[]

is what I was looking for but it does not work

Error using cellfun Non-scalar in Uniform output, at index 4, output 1. Set 'UniformOutput' to false.

could you help me fix that? thanks

[F.]

The reason is some cells contain string. Try this :

Tmp = cellfun( @(C) all(isnan(C)), R, 'UniformOutput', true)

[Mark Whirdy]

or

Tmp = cellfun( @isnan, R, 'UniformOutput', false)

as per error?

Need to start looking at the data and removing unncessary rows with cellfun(@isnumeric, cellfun(@isnan etc etc. Just experiment

[F.]

If you use this command, you will have a cell array for "Tmp", and we need an array.

[Sabbas]

Thanks F!

[Mark Whirdy]

cell2mat