To extract the last sub string from strings
75 views (last 30 days)
Show older comments
Hello everybody,
I would like to extract the last string just before 'EP'.
I tried first split the string with strsplit function. and then get the last right string using for-loop.
I assume that it can be done without using for-loop.
Please give me some help how to make the code without using for-loop.
load input.mat
temp = cellfun(@(x) strsplit(x, {' ','EP'}), input, 'UniformOutput', false); % split the string
for i=1:length(temp)
num(i,:) = str2double(temp{i}(end-1)); % fill the last right string in cells
end
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1114430/image.png)
0 Comments
Accepted Answer
Dyuman Joshi
on 2 Sep 2022
Edited: Dyuman Joshi
on 2 Sep 2022
You can use extractBetween with index of last space character and the common denominator at the end of the string ('EP')
load input.mat
num=cellfun(@(x) str2double(extractBetween(x,find(x==' ',1,'last')+1,'EP')), input)
More Answers (5)
Stephen23
on 2 Sep 2022
Avoid slow CELLFUN, STR2NUM, REGEXP, etc.
One SSCANF call is probably the most efficient approach by far, as well as being very simple:
S = load('input.mat');
C = S.input
V = sscanf([C{:}],'%*s %*s%fEP')
2 Comments
Dyuman Joshi
on 2 Sep 2022
Stephen, how do you know that cellfun, str2num, regexp etc are slower functions? And are there any resources where I can read more about this?
Stephen23
on 2 Sep 2022
Edited: Stephen23
on 2 Sep 2022
"how do you know that cellfun, str2num, regexp etc are slower functions?"
- many years of reading this and other forums, learning from the combined knowledge of many users.
- many years of writing unit tests for my own code (i.e. making modifications and comparing).
- reading the documentation, to know what features functions have and how to use them.
- knowledge about functions, e.g. STR2NUM calls EVAL inside (so is not optimised by the JIT engine), and CELLFUN by design must call a function handle repeatedly (slower than a loop).
Lets compare the answers given so far on this thread:
S = load('input.mat');
C = repmat(S.input,1e3,1); % bigger array -> easier to compare
timeit(@()funAtsushiUeno(C)) % REGEXP, CELLFUN, STR2NUM
timeit(@()funDyumanJoshi(C)) % CELLFUN, EXTRACTBETWEEN, STR2DOUBLE
timeit(@()funChunru(C)) % REGEXP, CELLFUN, STR2NUM
timeit(@()funImageAnalyst(C)) % loop and indexing
timeit(@()funKSSV(C)) % REGEXP, STRREP, STR2DOUBLE
timeit(@()funS23(C)) % SSCANF
So, my function is more than eight times faster than the next fastest function (from KSSV), as well as being the simplest. And I had a fair idea that would be the case, even before writing this test code.
"And are there any resources where I can read more about this?"
If you want to learn how to use MATLAB efficiently, my advice is to read this forum a lot. And when I write "a lot", I don't mean "just a little bit". And not just new threads: there are some really important topics that have been discussed in some old yet canonical threads on this forum.
V1 = funAtsushiUeno(C);
V2 = funDyumanJoshi(C);
V3 = funChunru(C);
V4 = funImageAnalyst(C);
V5 = funKSSV(C);
V6 = funS23(C);
isequal(V1(:), V2(:), V3(:), V4(:), V5(:), V6(:)) % checking the function outputs:
function num = funAtsushiUeno(C)
num = regexp(C,'([\d+-e.]+)EP\s*$','tokens');
num = [num{:}];
num = [num{:}];
num = cellfun(@str2num, num);
end
function num = funDyumanJoshi(C)
num=cellfun(@(x) str2double(extractBetween(x,find(x==' ',1,'last')+1,'EP')), C);
end
function y = funChunru(C)
s = regexp(C, '\s([\d\.]*)EP$', 'tokens');
y = cellfun(@(s) str2num(s{1}{1}), s);
end
function numbers = funImageAnalyst(data)
rows = numel(data);
numbers = zeros(rows, 1);
for row = 1 : rows
words = strsplit(data{row});
lastWord = words{end};
numbers(row) = str2double(lastWord(1:end-2));
end
end
function s = funKSSV(C)
expression = '\d+\.?\d*EP';
s = regexp(C,expression,'match') ;
s = strrep([s{:}]','EP','') ;
s = str2double(s);
end
function V = funS23(C);
V = sscanf([C{:}],'%*s %*s%fEP');
end
Chunru
on 2 Sep 2022
load(websave("input.mat", "https://www.mathworks.com/matlabcentral/answers/uploaded_files/1114425/input.mat"))
input
s = regexp(input, '\s([\d\.]*)EP$', 'tokens');
y = cellfun(@(s) str2num(s{1}{1}), s)
Atsushi Ueno
on 2 Sep 2022
Moved: Image Analyst
on 2 Sep 2022
This is a bit of a challenge because we need to deal with cell arrays with irregular number sizes.
I would use the regexp function.
load input.mat
num = regexp(input,'([\d+-e.]+)EP\s*$','tokens');
num = [num{:}];
num = [num{:}];
num = cellfun(@str2num, num)
0 Comments
Image Analyst
on 2 Sep 2022
Here's one way. Simple, straightforward, and intuitive. There may be more compact but cryptic methods though.
s = load('input.mat')
data = s.input; % DON'T call your variables input, which is a build in function!
rows = numel(data)
numbers = zeros(rows, 1);
for row = 1 : rows
words = strsplit(data{row});
lastWord = words{end};
numbers(row) = str2double(lastWord(1:end-2));
end
KSSV
on 2 Sep 2022
expression = '\d+\.?\d*EP';
s = regexp(input,expression,'match') ;
s = strrep([s{:}]','EP','') ;
s = str2double(s)
See Also
Categories
Find more on Loops and Conditional Statements 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!