Ignore format errors using textscan

11 views (last 30 days)
I been using textscan to read large text files containing data. Here's some lines to get an idea of what it looks like:
2015,1,1,23,1,23,100,9034
2015,1,1,23,1,23,203,8940
2015,1,1,23,1,23,313,8807
There's several million lines in the .txt file and every now and then there's a small error. Some strange symbols in the line or missing data (early end) of a line and very inconsistent. Whenever textscan comes across one of these lines it gives me the following error;
Error using horzcat
Dimensions of matrices being concatenated are not consistent.
I would like it to ignore these lines and continue reading all data. Anybody can give me advice on what I can do best?
Thanks!
  1 Comment
Stephen23
Stephen23 on 12 Jan 2015
The error message that you posted does not refer to textscan, but rather to horzcat...

Sign in to comment.

Accepted Answer

Star Strider
Star Strider on 12 Jan 2015
You don’t supply a lot of specific information, so I can’t provide a specific answer.
You can deal with the missing data fields by specifying a value for the 'EmptyValue' parameter. See the section of textscan ‘Name-Value Pair Arguments’ under 'EmptyValue'.
Stopping at a string character may be more difficult, because you do not specify what the non-numeric values are. An example of one way to deal with that is (with ‘fidi’ being the input file ID):
D1 = textscan(fidi, '%f %f', 'HeaderLines',2, 'Delimiter','\n', 'CollectOutput',1);
fseek(fidi,0,0); % Position Start Of Second Part Of File
D2 = textscan(fidi, '%f %f', 'HeaderLines',2, 'Delimiter','\n', 'CollectOutput',1);
This instructs textscan to read to the first interruption, the start itself again and read through the rest of the file. (This example is from my archived code. In this instance, the file had only one header in the middle of the file, with the same number of header lines as the beginning of the file. You can combine the ‘D1’ and ‘D2’ variables here, or keep them separate, depending on the nature of your data.)

More Answers (1)

Marcel
Marcel on 12 Jan 2015
Edited: Marcel on 12 Jan 2015
Thanks. Apparently the error didn't happen using textscan, but a function afterwards. Thanks to your info I was able to figure out a way to ignore invalid lines and continue reading the .txt file.
For anybody curious;
a1 = [];
while feof(fidi) == 0
a2 = textscan(fidi,format,'delimiter',',','HeaderLines',1,'CollectOutput',1);
a1 = [a1;a2{1}];
end
  1 Comment
Star Strider
Star Strider on 12 Jan 2015
My pleasure!
Yours is a new approach — at least not one I’ve seen before — so +1 for your Answer and +1 your Question.

Sign in to comment.

Categories

Find more on Characters and Strings in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by