How to convert string to variable name ?

I try to do like this

if true
  Varname=matlab.lang.makeValidName(strcat('Indiv_Reg_',Month))
end

where Month is a input variable represent char '01,02,...,12'

My problem is I cannot refer variable such as 'Indiv_Reg_01' to get a data from this variable. For example, I cannot get data from Indiv_Reg_01{1,1} by using this code

if true
  Varname{1,1}
end
How Can I fix this problem ?

thanks in advance.

[Stephen23]

"My problem is I cannot refer variable such as 'Indiv_Reg_01' to get a data from this variable"

The problem is that you are hiding pesudo-indices in variable names. Turn them into real indices and your "problem" instantly goes away. Solving problems through better code design is always preferred to writing slow, buggy, hack code accessing variable names.

https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

Answer 1

[Jan]

This question is an evergreen. The solution is easy: Don't do this. It is a shot in your knee.

See:

• https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval
• https://www.mathworks.com/matlabcentral/answers/57445-faq-how-can-i-create-variables-a1-a2-a10-in-a-loop
• and hundreds of discussions concerning eval in this forum.

Do not hide indices in the names of variables, but use indices instead: Let Indiv_Reg be a cell array and use the numerical value as index. This is much better than the indirection over the conversion to a string and hiding it in the names of dynamically created variables.

[Vladislav Kargin]

typical idiotic answer by matlab experts. The question "How can I do it?" The answer "We are clever and we know better what you should want to do."

[Jan]

@Vladislav Kargin: Exactly. Matlab is a tool and you can use it the wrong way. Many beginners in programming struggle with the problem of hiding indices in the names of variables, because they do not have enough experiences with the efficient using of arrays. This is equivalent to the question, how to use a hammer to push a screw into a piece of wood. Of course an expert will recommend a rotating screwdriver instead. This is the well known XY-problem. As usual, there is always another beginner, who does not understand the power of efficient programming methods.

If all Matlab experts and the original documentation give the same advice, it is time to rethink your impression, if this advice is "idiotic".

This is your first contribution in this forum and you start with a trial to be offensive. Good luck with your hammer.

[monkeyquant]

When we visit the site like this and post our questions, we are looking for a direct 'answer' to our question. We in general spend enough time (clicking, reading, and testing) and we still wants to do it for various reasons. I think those (even new, well probably new) who program, have reasons to pursue an approach if any even though there is a significant drawback by doing so. In other words, any consequences by applying a method against warnings and advices are their own. I wish those experts trust that we are also capable and understand your advices but we have reasons. I am sure we would learn a lesson as you have learned before. In my humble opinion, a single line warning is good enough because there is a very thourough article by Stephen23 (impossible to miss this.)

In my case, I use assignin against all warnings by many including Jan and Stephen23 for now to manage versions of model parameters. I rather unify parameters for all possible variations, still small size, and key inputs to a funciton are these versions predefined like m##### and p#####, respectively. I then load model and parameter mat file like M##### and P##### and pick elements from variables in the mat file loaded. I separately manage parameters, called parameter generator and save it to mat file by utilizing P#####.

I read the advices from experts but still determined to use assignin. I am taking possible consequences until a better approach I find in the future. I believe I certainly understand certain payoffs from using assignin (it is like a polution issue - penalty vs the polution level, lol.

Lastly, I thank you all contributors and enjoy various ideas.

[Jan]

"When we visit the site like this and post our questions, we are looking for a direct 'answer' to our question."

Of course. And the "direct" answer is often not the optimal solution, if the problem was formulated in a suboptimal way. This is the core of the already mentioned "XY-problem".

"We in general spend enough time (clicking, reading, and testing) and we still wants to do it for various reasons."

I cannot confirm this. The majority of users asking questions in this forum do not search before. Many newcomers tend to use eval and assignin folloing the wrong opinion, that this is useful. In the last 25 years I had many discussion about this topic and in all of them other methods have been fater, more elegant and reduce the effort for debugging. I'd be happy if all beginners read and consider Stephen's great summary about this topic,before they ask a concerning question, but this is not the case in general. If you a an exception, this is really nice for you. But as far as I can see, this is not your thread, but the question of Pradya Panyainkaew. He has accepted this answer and therefore I assume, it was useful.

Feel free to use assign in however you want. I'm not your teacher and you will not get any grades from me. But if you ask me for a good solution, I will suggest another method due to my experiences with Matlab and other programming languages. I have no problem, if you have another opinion. Feel free to post answers with your preferred method. Remember that I am a Matlab user as you are. There is no certificate or other sign, that makes me an "expert" - except that a certain number of users found my answers useful in the past.

"I rather unify parameters for all possible variations, still small size, and key inputs to a funciton are these versions predefined like m##### and p#####, respectively. I then load model and parameter mat file like M##### and P##### and pick elements from variables in the mat file loaded."

This is not clear enough to be discussed in detail. "m#####" looks like information is hidden in the name of the variables. As you know, I consider this as a bad code design, but this does not matter as long as you do not ask me for my opinion :-)

Have a nice day and I hope you find all needed asnwers in this forum.

[Walter Roberson]

When we visit the site like this and post our questions, we are looking for a direct 'answer' to our question.

This site gets quite a lot of questions from students. A fair portion of the students are "looking for" complete working code for their homework or honours project, that they can submit as their own work, without having to learn how to solve the problems themselves. Does it follow that we should provide that to them?

Answer 2

[Luna]

Hi all,

I agree with Jan & Stephen what they said about eval function. Don't use it. Instead use struct and arrays with dynamic naming.

For this example I have created a solution you can check:

Month = '01,02,03,04,05,06,07,08,09,10,11,12';
monthsArray = strsplit(Month,',');

for i = 1:numel(monthsArray)
Varnames{i} = matlab.lang.makeValidName(strcat('Indiv_Reg_',monthsArray{i}));
myStruct.(Varnames{i}) = randi(20,1,1);
end

myStruct.(Varnames{1,1}) % should give you a value of a random number
myStruct.Indiv_Reg_01 % same result above

ps: if you are forced to name your variables use this. Otherwise use directly indices of your Nx12 array.

[Steven Lord]

A struct is one possible solution. Depending on what data the poster wants to store associated with each name, another possible solution would be to use a table where you can give each row or each variable (column) its own name.

A = magic(3);
t = array2table(A, 'VariableNames', {'Huey', 'Dewey', 'Louie'})

To retrieve the third element in the second variable you can use several different approaches, two of which are:

nine1 = t{3, 2}
nine2 = t.Dewey(3)

[Luna]

Thanks for additional information, Steven! :)

[joe zhong]

In fact, this two solutions should be the "Accepted" answer.

We all know we should don't do something, but things just happened, and we need a real solution rather than "just redo all scripts you got".

[Luna]

Believe me you will put much more effort to the solution you are asking for in your way rather than "just redoing all scripts you got".

You will spend time just for once redoing your code with indexing and implementing arrays. After that you will be free. Think that when you want to change a single variable name, that will force you to "redo all your scripts" again.

I only use eval if I'm running Simulink models with from Workspace blocks.

Don't use it if you really don't know what eval is used for. Here is some informations about justifiable usages of eval. These advices are from Matlab experts, at least they know some programming better than most.

https://www.mathworks.com/matlabcentral/answers/325760-does-justifiable-use-of-eval-exist

Answer 3

[Gabor]

Your answer does not work if you only have hammer and a screw available at hand (curvfitting tool does not accept structs as variable for eg.). In that case its perfect to use the hammer to pop that screw in fast :D

I found this:

assignin('base',var_name, value)

[Walter Roberson]

Curvefitting toolbox accepts function handles for the fit model; you can pass whatever you like to those handles.

functions such as fit() accept data that might have been referenced such as FileStruct.(Xvar) where Xvar is a scalar string object or character vector that indicates which field to look for.

cftool() accepts passing x, y, z in on the command line.

[paul harder]

Fun thread! I agree with everything that has been said about the faults in indexing variable names, but I also believe it's rarely but sometimes necessary to create variables in a loop. In my case, I needed to create a list of variables in a very specific format so that they can be exported out to a different piece of software. Here's how to initialize vairalbes in a loop using the method shown above. But as stated by others, this should not be used as common practice, especially if further analysis is to be done on the data within matlab.

for i=1:3
    assignin('base',['data_' num2str(i)],0)
end

[Stephen23]

"In my case, I needed to create a list of variables in a very specific format so that they can be exported out to a different piece of software."

You write that you need "a very specific format" (by which I presume that you mean that the variable name needs to have a very specific format, as of course the variable format itself is an unrelated topic). The most commonly-used MATLAB function for exporting data which relies on the variable names is SAVE, but of course SAVE has the -STRUCT option which completely avoids the need for dynamically named variables, so you must be using some other export function. Which one?

In the interest of helping other users understand the need you have, would you be so kind as to provide some sample code, or a description of the file format, or something similar so that we can understand why your use-case has this particular "need" which cannot be fulfilled with simpler, more efficient code.

[Govind Narayan Sahu]

This works perfectly fine for me. Thanks a lot.

data = rand(1,8);
vars = {'a', 'b', 'd', 'e', 'f', 'g', 'h', 'i'};
index_vars = 1:8;

for k = 1:length(vars)
    assignin('base',vars{k}, data(1, index_vars(k)))
end

Answer 4

[Anshuman Agrawal]

Check out: https://www.mathworks.com/help/matlab/matlab_prog/string-evaluation.html

Hopefully, what you wanted to do was covered here.