How to define data input in an elegant way?

Hi all,

I have a newbie question: in complex programs, we may have lots of input data like this (for instance in my program):

time.max = 5;
time.step = 1;
no.incl = 3; 
no.pre.hat = 4;
fce.time = 1;
fce.node = 4;
domain.length.I1 = 17;
domain.length.I2 = 17;
domain.length.S = 17;
domain.bond.L.I1 = 1;
domain.bond.R.I1 = 2;
domain.bond.L.I2 = 1;
domain.bond.R.I2 = 2;
pmVal.fix.I3 = 1000;

and some empty sets like this:

err.max.store = []; 
err.loc.store = [];
err.max.store_exactwRB = []; 
err.loc.store_exactwRB = [];
err.max.store_hat = []; % ehat
err.loc.store_hat = [];
err.max.store_hhat = []; % ehhat
err.loc.store_hhat = [];

I have a feeling that this is extremely inelegant and unnecessary, also hard to read. Is there any ways to define a big data input elegantly and efficiently?

Many thanks!

[Stephen23]

@Xh Du: that looks like a very reasonable way to specify input data. What is your concern?

Answer 1

[Adam]

Grouping data together like that is certainly far better than hundreds of individual variables.

I tend to use classes myself, but purely from the point of view of grouping data together structs work equally. Make sure your variable, struct, field names are as clear and intuitive as they can be though, even if they are quite long. Nobody wants to look at code with 27 variables all with names like ya, yb, ys, A, B, C1, C2, etc

Perhaps a bigger question is do you need all those things in the same place? Can you not break down your algorithm into smaller parts where you don't need all those things at once? Possibly you do at the very top level to feed them into the system in the first place, but later on parts of the program should be as self-contained as you can make them.

[Xh Du]

Hi Adam,

Thanks for your reply!

When you say classes do you mean use OOP commands like classdef? Also when you say break down to smaller parts you mean into smaller scripts? I did this but only for the main body of the algorithm, like you said I group data input in the very top, this is why I have this readability problem.

[Adam]

Yes, I use classdef to create classes. That then naturally leads to a breakdown of the program, but obviously it depends what the program is. Some algorithms are hard to breakdown into meaningful components that you can give names to (naming the components in my programs is often one of the hardest aspects of the design!).

You should really get used to using functions, not scripts though. You are very limited to what and how you can breakdown a program if you are only using scripts. Functions are a natural way to create self-contained parts of code with their own workspace that isn't contaminated by 'everything else' as a script is.

[Xh Du]

I do use loads of functions including anonymous functions. Sometimes I group data into scripts when the data are simple inputs or definitions, in these cases I found it meaningless to use functions because these scripts basically do things like

time.max = 5;

[Adam]

How does that script get called though? From a code readability and understandability point of view a function that returns 'time' (though preferably with a name that doesn't hide a builtin function) would make it clear that the variable exists within whichever function it was called from.

Calling a script from inside a function means that your function workspace will suddenly acquire whatever variables were in the script's workspace when it finishes and none of this is visually obvious in the function without opening the script and understanding it.

If

time.max = 5;

is literally all that is in a script though then surely it could just go in-place instead of the line to call the script?

[Stephen23]

@Xh Du: avoid scripts. Functions and classes make code much more robust:

https://www.mathworks.com/matlabcentral/answers/228557-experts-of-matlab-how-did-you-learn-any-advice-for-beginner-intermediate-users