Textscan reading in floats and str2num dropping ending zeros

MATLAB VERSION: R2007a - I know...my company is up on the times.

I'm trying to read in data from a tab delimited text file that looks like:

ABCDEF 2012.190.18465.356 0 TRUE
MKJIJN 2012.190.19305.376 b7 183

My current scan string is:

SCAN_STR = '%s\t%f.%f\t%s\t%s\t%*s';
TEMP = textscan(FID, SCAN_STR, 'delimiter', '\t');

Matlab reads in the data as (sample of first to save time):

TEMP{1} = 'ABCDEF'
TEMP{2} = 2012.19
TEMP{3} = 18465.365
TEMP{4} = '0'
TEMP{5} = 'TRUE'

Why does it drop the ending zero for TEMP{2}? 19 vs. 190 is a big difference. Is there any way for me to make sure the ending zero remains?

When I try to read it in as a string:

SCAN_STR = '%s\t%s\%s\t%s\t%*s';

All are strings. Now:

TEMP{2} = '2012.190.18465.356'

I split it using:

YRDY = cellfun(@(x)x(1:8),TEMP{2},'Un',0);
YRDY = '2012.190'

Using str2num:

YRDYnum = cellfun(@(x)str2num(x), YRDY, 'Un',0);
YRDYnum = 2012.19

Why does this happen? There has to be some way for me to keep the ending zero.

Thanks, Mike

Answer 1

[Kelly Kearney]

If you read the year-day combo string in as a number, or convert from string to number, it's going to treat the value like a number. And in that case, 2012.19 is the same as 2012.190. You probably want to split the year and day values apart before converting to a number:

 % Recreate your test

 str = sprintf('%s\t%s\t%s\t%s\n', 'ABCDE','2012.190.18465.356', '0', 'TRUE', 'MKJIJN', '2012.190.19305.376', 'b7', '183')

 str =

 ABCDE  2012.190.18465.356  0  TRUE
 MKJIJN  2012.190.19305.376  b7  183

 % Read everything as strings

 data = textscan(str, '%s%s%s%s', 'delimiter', '\t');

 % Parse dates

 yrday = data{2};
 yrday = cellfun(@(x) regexp(x, '\.', 'split'), yrday, 'uni', 0);
 yrday = cat(1, yrday{:});
 yrday = cellfun(@str2num, yrday);

[Mike K]

Ok. Thats what I was going forward with. Then that presents a problem with my TEMP{3}. For me TEMP{3} is time. Now if that ends in a zero it could mess up my results. Is the only way around this to separate my time into sec.ms? I guess I'm confused on how MATLAB sees 2012.19 the same as 2012.190. Why does MATLAB drop the ending zero? MATLAB doesn't see zero as a place holder/number so it assumes its not needed?

[Kelly Kearney]

It's a little roundabout, but you can put the seconds and milliseconds back together prior to converting to a number

 yrday = data{2};
 yrday = cellfun(@(x) regexp(x, '\.', 'split'), yrday, 'uni', 0);
 yr = cellfun(@str2num, yrday(:,1));
 day = cellfun(@str2num, yrday(:,2));
 sec = cellfun(@(x,y) str2num([x '.' y]), yrday(:,3), yrday(:,4));

There's probably a way to split only along the first two instances of \. in the regular expression, but I don't know it off the top of my head.

As for why it behaves this way, I think the confusion is arising over the use of the period as a delimiter vs its use as a decimal point. When converting from string to number, Matlab assumes . is the decimal, so there's no need to hold on to trailing zeros after a decimal point (Matlab doesn't assume significant digits or anything like that). In your case, it seems as though the . is being used for both purposes: to delimit years, days, and seconds; and as a decimal point in the seconds. So you need to break apart the strings first, then convert to numeric form once you have removed all non-decimal .'s.