textscan, problem with the treatasempty when the is the signal minus (-)

I have a problem to get this .tsv file (below) in. The character chosen by the Economic Institute as the empty value was the signal minus (-). When I use the function textscan and define the Treatasempty with '-' I wrongly damage the items with negative values.

my function request is:

a=textscan(fid,'%s%s%s%n%n%n%n%n%n%n%n%n%n%n%n','delimiter',',','headerlines',4,'ReturnOnError',1,'treatAsEmpty',{'-'});%

Can anybody help me?

headline1
headline2
headline3
headline4
Índice base fixa com ajuste sazonal (2011=100) (Número índice),"Tecidos, vestuário e calçados",103.17,-99.69,99.64,101.25,102.79,-104.80,105.25,104.64,104.85,102.74,104.85,105.47
Índice base fixa com ajuste sazonal (2011=100) (Número índice),"Móveis e eletrodomésticos",108.69,109.91,110.48,111.66,108.12,113.13,112.44,113.57,111.46,113.69,113.32,116.62
Índice base fixa com ajuste sazonal (2011=100) (Número índice),"Móveis",-,-,-,-,-,-,-,-,-,-,-,-

[Cedric]

Please attach a typical file to the question.

EDIT: well, disregard this comment if my solution below works. If it doesn't please do attach a file or send it to me by email if it's too confidential to go fully public.

Answer 1

[Cedric]

Here is a proposal that I can refine when/if you attach a file to your question:

 content = fileread( 'myFile.tsv' ) ;
 content = regexprep( content, '-(,|$)', ',' ) ;
 data    = textscan( content, '%s%s%s%n%n%n%n%n%n%n%n%n%n%n%n', ...
                     'Delimiter', ',', 'HeaderLines', 4, 'ReturnOnError', 1 ) ;

The idea is to eliminate minus signs which are before a comma or at the end of the file, before using TEXTSCAN.

Answer 2

[Ricardo MF]

Dear Cedric, tks so much for your answer. Actually, it has been my first post (question) here at the www.mathworks.com. I tried your sugestion, but the function has not removed the minus from the end of each line. Maybe the signal for the end of line is not that one!? Tks a lot.

[Cedric]

Dear Ricardo, please perform the following update: change the call to REGEXPREP for..

 content = regexprep( buffer, '-(?=[,\r\n]|$)', ',' ) ;

As I didn't have your file I tried on a one line string, and, as you pointed out, I forgot to implement the pattern for the end of line. Let me know if it is too slow; there are other patterns which could apply and it's difficult to know in advance which are faster. If it doesn't work still, feel free to send me the file or a chunk of it so I can perform tests.