wolffd@0: function data = csv2cell(varargin)
wolffd@0: % CSV2CELL - parses a Windows CSV file into an NxM cell array, where N is
wolffd@0: % the number of lines in the CSV text and M is the number of fields in the
wolffd@0: % longest line of the CSV file. Lines are delimited by carriage returns
wolffd@0: % and/or newlines.
wolffd@0: %
wolffd@0: % A Windows CSV file format allows for commas (,) and double quotes (") to
wolffd@0: % be contained within fields of the CSV file. Regular fields are just text
wolffd@0: % separated by commas (e.g. foo,bar,hello world). Fields containing commas
wolffd@0: % or double quotes are surrounded by double quotes (e.g.
wolffd@0: % foo,bar,"item1,item2,item3",hello world). In the previous example,
wolffd@0: % "item1,item2,item3" is one field in the CSV text. For double quotes to be
wolffd@0: % represented, they are written in pairs in the file, and contained within
wolffd@0: % a quoted field, (e.g. foo,"this field contains ""quotes""",bar). Spaces
wolffd@0: % within fields (even leading and trailing) are preserved.
wolffd@0: %
wolffd@0: % All fields from the CSV file are returned as strings. If the CSV text
wolffd@0: % contains lines with different numbers of fields, then the "missing"
wolffd@0: % fields with appear as empty arrays, [], in the returned data. You can
wolffd@0: % easily convert the data you expect to be numeric using str2num() and
wolffd@0: % num2cell(). 
wolffd@0: %
wolffd@0: % Examples:
wolffd@0: %  >> csv2cell('foo.csv','fromfile') % loads and parses entire file
wolffd@0: %  >> csv2cell(',,,') % returns cell array {'','','',''}
wolffd@0: %  >> csv2cell(',,,','text') % same as above, declaring text input
wolffd@0: %  >> csv2cell(sprintf('%s\r\n%s',...
wolffd@0: %     '"Ten Thousand",10000,,"10,000","""It''s ""10 Grand"", baby",10k',...
wolffd@0: %     ',foo,bar,soo'))
wolffd@0: %  ans = 
wolffd@0: %    'Ten Thousand'    '10000'       ''    '10,000'    [1x22 char]    '10k'
wolffd@0: %                ''    'foo'      'bar'    'soo'                []       []
wolffd@0: %  >> % note the two empty [] cells, because the second line has two fewer
wolffd@0: %  >> % fields than the first. The empty field '' at the beginning of the
wolffd@0: %  >> % second line is due to the leading comma on that line, which is
wolffd@0: %  >> % correct behavior. A trailing comma will do the same to the end of a
wolffd@0: %  >> % line.
wolffd@0: % 
wolffd@0: % Limitations/Exceptions:
wolffd@0: %   * This code is untested on large files. It may take a long time due to
wolffd@0: %   variables growing inside loops (yes, poor practice, but easy coding).
wolffd@0: %   * This code has been minimally tested to work with a variety of weird
wolffd@0: %   Excel files that I have.
wolffd@0: %   * Behavior with improperly formatted CSV files is untested.
wolffd@0: %   * Technically, CSV files from Excel always separate lines with the pair
wolffd@0: %   of characters \r\n. This parser will also separate lines that have only
wolffd@0: %   \r or \n as line terminators. 
wolffd@0: %   * Line separation is the first operation. I don't think the Excel CSV
wolffd@0: %   format has any allowance for newlines or carriage returns within
wolffd@0: %   fields. If it does, then this parser does not support it and would not
wolffd@0: %   return bad output.
wolffd@0: %
wolffd@0: % Copyright 2008 Arthur Hebert
wolffd@0: 
wolffd@0: % Process arguments
wolffd@0: if nargin == 1
wolffd@0:     text = varargin{1};
wolffd@0: elseif nargin == 2
wolffd@0:     switch varargin{2}
wolffd@0:         case 'fromfile'
wolffd@0:             filename = varargin{1};
wolffd@0:             fid = fopen(filename);
wolffd@0:             text = char(fread(fid))';
wolffd@0:             fclose(fid);
wolffd@0:         case 'text'
wolffd@0:             text = varargin{1};
wolffd@0:         otherwise
wolffd@0:             error('Invalid 2nd argument %s. Valid options are ''fromfile'' and ''text''',varargin{2})
wolffd@0:     end
wolffd@0: else
wolffd@0:     error('CSV2CELL requires 1 or 2 arguments.')
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: % First split it into lines
wolffd@0: lines = regexp(text,'(\r\n|[\r\n])','split'); % lines should now be a cell array of text split by newlines
wolffd@0: 
wolffd@0: % a character is either a delimiter or a field
wolffd@0: inField = true;
wolffd@0: inQuoteField = false;
wolffd@0: % if inField && ~inQuoteField --> then we're in a raw field
wolffd@0: 
wolffd@0: skipNext = false;
wolffd@0: data = {};
wolffd@0: field = '';
wolffd@0: for lineNumber = 1:length(lines)
wolffd@0:     nChars = length(lines{lineNumber}); % number of characters in this line
wolffd@0:     fieldNumber = 1;
wolffd@0:     for charNumber = 1:nChars
wolffd@0:         if skipNext
wolffd@0:             skipNext = false;
wolffd@0:             continue
wolffd@0:         end
wolffd@0:         thisChar = lines{lineNumber}(charNumber);
wolffd@0:         if thisChar == ','
wolffd@0:             if inField 
wolffd@0:                 if inQuoteField % this comma is part of the field
wolffd@0:                     field(end+1) = thisChar;
wolffd@0:                 else % this comma is the delimiter marking the end of the field
wolffd@0:                     data{lineNumber,fieldNumber} = field;
wolffd@0:                     field = '';
wolffd@0:                     fieldNumber = fieldNumber + 1;
wolffd@0:                 end
wolffd@0:             else % we are not currently in a field -- this is the start of a new delimiter
wolffd@0:                 inField = true;
wolffd@0:             end
wolffd@0:             if charNumber == nChars % this is a hanging comma, indicating the last field is blank
wolffd@0:                 data{lineNumber,fieldNumber} = '';
wolffd@0:                 field = '';
wolffd@0:                 fieldNumber = fieldNumber + 1;
wolffd@0:             end
wolffd@0:         elseif thisChar == '"' 
wolffd@0:             if inField
wolffd@0:                 if inQuoteField
wolffd@0:                     if charNumber == nChars % it's the last character, so this must be the closing delimiter?
wolffd@0:                         inField = false;
wolffd@0:                         inQuoteField = false;
wolffd@0:                         data{lineNumber,fieldNumber} = field;
wolffd@0:                         field = '';
wolffd@0:                         fieldNumber = fieldNumber + 1;
wolffd@0:                     else 
wolffd@0:                         if lines{lineNumber}(charNumber+1) == '"' % this is translated to be a double quote in the field
wolffd@0:                             field(end+1) = '"';
wolffd@0:                             skipNext = true;
wolffd@0:                         else % this " is the delimiter ending this field
wolffd@0:                             data{lineNumber,fieldNumber} = field;
wolffd@0:                             field = '';
wolffd@0:                             inField = false;
wolffd@0:                             inQuoteField = false;
wolffd@0:                             fieldNumber = fieldNumber + 1;
wolffd@0:                         end
wolffd@0:                     end
wolffd@0:                 else % this is a delimiter and we are in a new quote field
wolffd@0:                     inQuoteField = true;
wolffd@0:                 end
wolffd@0:             else % we are not in a field. This must be an opening quote for the first field?
wolffd@0:                 inField = true;
wolffd@0:                 inQuoteField = true;
wolffd@0:             end
wolffd@0:         else % any other character ought to be added to field
wolffd@0:             field(end+1) = thisChar;
wolffd@0:             if charNumber == nChars
wolffd@0:                 data{lineNumber,fieldNumber} = field;
wolffd@0:                 field = '';
wolffd@0:                 fieldNumber = fieldNumber + 1;
wolffd@0:             elseif charNumber == 1 % we are starting a new raw field
wolffd@0:                 inField = true;
wolffd@0:             end
wolffd@0:         end
wolffd@0:     end
wolffd@0: end
wolffd@0: