Can regexp return key/value pairs in a structure?

Question

Let's say I have a character array with key/value pairs:

ch = sprintf('name: John\nsex: M\n')
ch =
    'name: John
     sex: M
     '

This is just a sample. The actual data is in a file and has many pairs. I can use regexp to get tokens, then use a for loop to assign them to a structure:

lns = regexp(ch, '(\w*):\s(.*?)\n', 'tokens');
for i = 1:numel(lns)
    myStruct.(lns{i}{1}) = lns{i}{2};
end

myStruct = 

  struct with fields:

    name: 'John'
     sex: 'M'

Is there a simpler way to achieve this, like using regexp(ch, expr, 'names')?

Do you know the keys in advance? That is, will they always be 'name' and 'sex'? — informaton, Sep 21 '17 at 03:14
If we avoid _for_ loop, it may potentially improve performance at least. — Xiangrui Li, Sep 21 '17 at 03:15
@informaton I don't know keys in advance. From regexp documentation, it seems that known keys are needed. — Xiangrui Li, Sep 21 '17 at 03:17
You are right. Sorry, I was about to give example using known keys. — informaton, Sep 21 '17 at 03:19

gnovice · Accepted Answer · 2017-09-21T04:09:23.450

5

You can avoid the for loop by collecting your key/value pairs into a single cell array and then passing the contents of that as a comma-separated list to struct:

args = [lns{:}];
myStruct = struct(args{:});

And the output:

myStruct = 

  struct with fields:

    name: 'John'
     sex: 'M'

Alternatively, you could avoid the use of regexp and use textscan to read from the file, which may improve performance:

strs = textscan(fileID, '%s', 'Delimiter', ':');
myStruct = struct(strs{1}{:});

edited Sep 21 '17 at 04:09

answered Sep 21 '17 at 03:21

gnovice

125,304
15
256
359

Thanks! I tested the performance, with 75 pairs, this seems a little faster than `for` loop, although the difference is very minor. – Xiangrui Li Sep 21 '17 at 03:35
@XiangruiLi: I added an option using `textscan` instead of `regexp`. Hopefully that might boost performance. – gnovice Sep 21 '17 at 03:39
That might be faster. But for my case, I can't use '%s', since the `value` can have space inside – Xiangrui Li Sep 21 '17 at 04:00
@XiangruiLi: I updated with a fix that allows spaces. :) – gnovice Sep 21 '17 at 04:10

Can regexp return key/value pairs in a structure?

1 Answers1