textfile

Provides a one-call interface to read/write flat text files or line-based text files.

Description

This convenience function provides one-call read/write access to a flat text or line-based text file using calls of the following form.

content = textfile(filename);
textfile(filename, content);

The content can be returned as a char array, or a cell array of lines in the file. Some additional options are provided (see help).

• MATLAB release: MATLAB 7.10 (2010a)

[Jan Simon]

This program is useful for a very frequently used job.
The documentation is exact and descriptive, but does not follow the standard design of Matlab:
  function XY
  % H1-line
  % Help text...
This impedes e.g. LOOKFOR. The documentatoion mentions author and date. The code is well commented. Inputs are checked.

"~exist('arg1')" is not efficient to test, if an input was defined: besides variables, it accepts "arg1.m", "arg1.p", "arg1.mex...", "arg1.mdl", a folder or java class called "arg1" anywhere in the Matlab path also! Therefore it is horribly slow, if "arg1" is not a variable. Either use "exist('arg1', 'var')", or NARGIN.
The conversion from a cell string to a string without pre-allocation gets very slow for large data. SPRINTF('%s\n', content{:}) is much faster.
Pre-allocation would be a good idea for splitting a read text to a cell string also: content = cell(N, 1);
Finding CHAR(10) using "lf = find(str == 10);" has drawbacks: 1) It compares different types. 2) It creates a temporary LOGICAL array. Therefore this is ~30% faster for a 10kB string: "lf = find(str == char(10));", and "lf = findstr(str, char(10))" consumes just the half processing time. Same for "content(content ~= 13);".
Nevertheless, TEXTSCAN(fid, '%s', 'delimiter', '\n') is much more efficient to read a file into a cell string, so there is no need to reinvent the wheel.
Perhaps these ideas help to improve the efficiency of this submission.

[Ben Mitch]

Thanks Jan, have copied your text into my TODO file :)