copy-on-write, performance, and code readability

I use a global, multi-level nested struct() to hold a lot of data. For example, I have:

global DATA;
DATA.US.AAPL.date = <array of dates>
DATA.US.AAPL.px = <array of prices>

Exactly why I am using a structure like this or what I am storing in it is not relevant for this question. Above is just an example.

Inside of functions that operate on this structure, I often create an "alias":

function m = doStuff(region,ticker)
   global DATA;
   ldata = DATA.(region).(ticker);
   <work on and possibly update ldata>
   DATA.(region).(ticker) = ldata;
end

It's obvious that using an alias like ldata improves readability - however this appears to come at a cost. When DATA is large, replace ldata with direct DATA.(region).(ticker) leads to material performance improvement.

Presumably, the Matlab copy-on-write semantics is somehow to blame here.

Any thoughts as to how I can get readability and performance?

Answer 1

[Muneef]

so are you using struct. inside a function?

[matal]

yes - is the code sample unclear somehow? let me know.

Answer 2

[The Matlab Spot]

Check this out Using Pointers in MATLAB