how to read a file like a table

Hello,

I'm looking to read double precision data from a file, in however many columns it is, and to the end of file. The result should be a N x M double.

i.e. if I paste this in the Matlab window, I get what I need:

table=[
1016.6 0 0 0 0 0 0 0
1016.6 0 0 0 0 0 0 0
0 -479.528 0 0 0 0 0 0
0 468.718 0 0 0 0 0 0
0 0 -385.346 0 0 0 0 0
0 0 696.118 0 0 0 0 0
0 0 0 -7591 0 0 0 0
0 0 0 7591 0 0 0 0
0 0 0 0 -7573.16 0 0 0
0 0 0 0 7573.16 0 0 0
0 0 0 0 0 -2620.704 0 0
0 0 0 0 0 4134 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 1022.074 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 7701.638
];
That gives a 16x8 double.
But in the file would be just the values, as shown.

Some form of fscan maybe, but with generalized format?

Thanks for unconfusing me...

Dave

Answer 1

[dpb]

Use textscan without an explicit format string--that's a magic incantation that will return the data in the same shape as is in the file for regular files of all numeric data.

data=cell2mat(textscan(fid,'','collectoutput',1));

Add any additional parameters needed for delimiter, missing value, headerlines, etc., etc., ...

Or, dlmread will also return the proper shape for those files which it can handle (which generally includes those for which the above textscan "trick" will work.

Since R2013b there's readtable which will leave you with a table dataset rather than an array so whether that is/is not an advantage depends on what you're going to be doing with the data.

[David Pesetsky]

Thank you. I like tricks of the trade :)

[dpb]

It's undocumented for textscan; there's an example with empty string for the format string in textread still altho it's not mentioned there. I learned it functioned for textscan by accident when using the same idiom had learned from textread by habit, not by design initially.

It's a very important feature that deserves to be "shouted from the rooftop" as it ameliorates an otherwise painful problem if one doesn't know the number of elements/record plus the need for the abomination of a format string of some N repeating fields to force the proper count. In lieu of that, it's do another reshape after reading into a single column.