Setfield vs dot indexing for deeply nested struct

Supose i have the following struct and an associated string:

s.f1.f2.f3.f4.f5 = data;
tmp_string = 's.f1.f2.f3.f4.f5'; 

So, just to be clear, eval(tmp_string) yields data, But eval is bad and i dont use it.

Now suppose I wish to programmatically replace data with newdata. Currently I do this by

sparts = strsplit(tmp_string,'.'); % yields sparts = [{s},{f1},{f2},{f3},{f4},{f5}];
s = setfield(s,sparts{2:end}) = newdata;

However, Matalb suggests I use dot notation and dynamic fieldnames instead of setfield and getfield whenever possible. I think this is a situation where setfield and getfield actually are prefereable, but I'm not sure. Can I do this more cleanly and efficiently with dot notation?

Answer 1

[Stephen23]

"However, Matalb suggests I use dot notation and dynamic fieldnames instead of setfield and getfield whenever possible."

Ignore it, it is just a suggestion. The static code analyzer cannot understand the entire context of why particular code is being used, it just follows relatively simple rules. Disable the warning on that line (right click, "Supress... on this line").

"I think this is a situation where setfield and getfield actually are prefereable.."

Yes, that seems reasonable.

"Can I do this more cleanly and efficiently with dot notation?"

No.

[Walter Roberson]

Though it is not immediately obvious to me that you should permit such a situation to arise.