Dynamic import with evalin not possible

Hi,

My previous question (link) led me to a conclusion that the only way to allow calling of static methods from within the same classdef was to introduce a self-import statement at the beginning of each static method, which would in turn import the class along with the surrounding pacakge structure into the scope of the static method (seems very redundant and broken by design).

However, I ran across yet another weird behavior, namely that evalin("caller","import ...") imports into the current scope instead of the caller's scope. This is true for methods, functions, as well as scripts. For example:

function testImport

! mkdir +TestPackage

x = 'main';

fprintf("Outside subfunction import list count: %d\n",length(import));

fprintf("Outside subfunction normal variable: %s\n\n",x);

testsubf()

fprintf("\nOutside subfunction import list count: %d (should increase by one)\n",length(import));

fprintf("Outside subfunction normal variable: %s\n",x);

function testsubf()

fprintf("Before subfunction evalin import list count: %d\n",length(import));

fprintf("Before subfunction evalin normal variable: %s\n",x);

evalin("caller","import +TestPackage.*")

evalin("caller","x = 'sub';")

fprintf("After subfunction evalin import list count: %d (should remain the same)\n",length(import));

fprintf("After subfunction evalin normal variable: %s\n",x);

end

end

produces an output:

>> testImport

Outside subfunction import list count: 0

Outside subfunction normal variable: main

Before subfunction evalin import list count: 0

Before subfunction evalin normal variable: main

After subfunction evalin import list count: 1 (should remain the same)

After subfunction evalin normal variable: sub

Outside subfunction import list count: 0 (should increase by one)

Outside subfunction normal variable: sub

I have found a very similar unuanswered issue but regarding evalin("base",...) here.

Is there any way to solve this? The above behavior makes it harded to dynamically import packages, which would be especially helpful to make functions/classess independent of the structure of packages in which they're located.

On top of that, I see that "in a future release [I use R2022b], IMPORT will not accept variable names, function calls, or operators [...]", making it necessary to do eval("import ...") even in case of simple dynamic import to local scope (which currently can still be done with import(pkg) call). I sincerely hope that eval("import ...") will still be possible or else each packaged function or static method will have to have the package prefix hard-coded in front of each call.

[Robert Borkowski]

The above is also misbehaving if I nest eval("import +TestPackage.*") inside of evalin:

evalin("caller","eval(""import TestPackage.*"")")

Answer 1

[Neha]

Hi Robert,

I understand that you want to dynamically import packages into the caller’s scope. "evalin" is only used for evaluating in a specific workspace. Local/nested functions and imports, which are specific to another frame's (like the caller's) context, are inaccessible. This is the documented behaviour. You can refer to the following documentation link for more information on the "evalin" function:

https://www.mathworks.com/help/matlab/ref/evalin.html

Hope this helps!