a philosophical question regarding function (how many tasks should be wrapped inside a function?)

My data have two types of tables. They share common columns, and the second type has extra columns. I couldn't decide whether if I should write two separate functions to process the tables? Or should I have the extra columns processed outside one common function? I personally prefer the one-function approach. Below are toy examples to show what my question is about. What would be the best practice? What are the criteria to choose it? Thanks for any thoughts you like to share.

Two types of tables

T = cell2table({'matlab 101'; 'C++ 202'}, "VariableNames", "u");
T2 = cell2table({'algebra 101', 'jack 001'; 'calculus 202', 'jill 002'}, "VariableNames", ["u", "v"]);

Coding Phylosophy I. One function processes common tasks while dealing with extra columns outside the function. What I don't like about this approach is that it doesn't look as clean.

sp = func(T.u);
desired_output1 = cell2table(sp, "VariableNames", ["course", "level"])

desired_output1 = 2×2 table
      course       level 
    __________    _______

    {'matlab'}    {'101'}
    {'C++'   }    {'202'}

spu = func(T2.u);
spv = func(T2.v);
desired_output2 = cell2table([spu, spv], "VariableNames", ["course", "level", "teacher", "id"])

desired_output2 = 2×4 table
       course        level     teacher       id   
    ____________    _______    ________    _______

    {'algebra' }    {'101'}    {'jack'}    {'001'}
    {'calculus'}    {'202'}    {'jill'}    {'002'}

Coding Phylosophy II. Two separate functions handle two slightly diffrent tables separately. What I don't like about this approach is that the functions must know the tables' variable names. If the variable names are to be changed, the functions must be rewritten.

desired_output1 = gunc(T)

desired_output1 = 2×2 table
      course       level 
    __________    _______

    {'matlab'}    {'101'}
    {'C++'   }    {'202'}

desired_output2 = hunc(T2)

desired_output2 = 2×4 table
       course        level     teacher       id   
    ____________    _______    ________    _______

    {'algebra' }    {'101'}    {'jack'}    {'001'}
    {'calculus'}    {'202'}    {'jill'}    {'002'}

functions:

function sp = func(C)
sp = split(C);

% There are other processings here. For the sake of demo simplicity, they are
% ommited.
end

function out = gunc(T)
sp = split(T.u);
out = cell2table(sp, "VariableNames", ["course", "level"]);
end

function out = hunc(T)
spu = split(T.u);
spv = split(T.v);
out = cell2table([spu, spv], "VariableNames", ["course", "level", "teacher", "id"]);
end

Answer 1

[Ive J]

Assuming you're gonna stick with those variable names, you can do somethig like this:

T1 = array2table(["matlab 101"; "C++ 202"], "VariableNames", "u");
T2 = array2table(["algebra 101", "jack 001"; "calculus 202", "jill 002"], "VariableNames", ["u", "v"]);

function out = tableSplitter(tab)
    cols = string(tab.Properties.VariableNames);
    if width(tab) > 2 || ~any(ismember(cols, ["u", "v"]))
        error("input table must have max two columns with var names of u/[v]!")
    end
    
    out = splitvars(varfun(@split, tab));
    if width(tab) == 1
        out.Properties.VariableNames = ["course", "level"];
    else
        out.Properties.VariableNames = ["course", "level", "teacher", "id"];
    end

end

[Simon]

Thanks for your feedback. I had an idea in a similar line. But it was considered a bad practice idea to modify function/program state with conditional statement. I read about that advice somewhere (source forgotten). As shown by your algorithm, it becomes harder to read and debug.

[Ive J]

I've failed to grasp what you meant by bad practice idea to modify function/program state with conditional statement. This cannot be true (unless you have a clear example why it is unadvised) simply because without conditional statements, no one can write anything. You can of course use column names themselves to compare (instead of width(tab) ), but regardless, it's still a conditional statement!

[Simon]

I didn't exactly know what I mean with that statement. I guess I mean in the if-else condition, the 'if' part and the 'else' part have different input formats and different output formats. So it seems to put two functions together with if-else statement. I don't completely avoid using conditional statements; just that I think it would be better for the condition to handle homogeneous data, such as if (x is positive), if ( year is before 2020), if (diagnosis is negative), ... etc. I do appreciate your codes and thoughts!

[Simon]

@Ive J I see now some benefitis your idea has. Using conditional statement to wrap related processes in a function does have its advantage. Your answer is accepted.