# pointer like operations in mathematica and evaluation control

I would like to create a data structure of the type "header[pointer]" where pointer would point to a list. I will shown on an example how I'd like this to work, and where the problems are.

In[1]:= pointer = Unique[storage];
Out[1]= storage$113  So far so good. Now I want to assign some values to the storage In[3]:= Evaluate[pointer] = {v1, v2, v3} Out[1]={v1,v2,v3}  Evaluate is needed in order to enforce evaluation of the left hand side of the assignment. The pointer has not been touched/changed which is good: In: Trace[pointer] Out: {pointer, storage$113, {v1,v2,v3}}


it still points to the storage, which in turn points to the list with the actual data. Yet, now it starts evaluating to something beyond storage$113. This will turn up later as the main problem. Now I'd like to have something like header[storage$113] as a wrapper to pass this around functions, and here the problems start. If one tries

In: header[pointer]


the pointer evaluates down to the values. Thus we get bust since pointer evaluates way to much. If one tries

In: SetAttributes[header, HoldAll]


to prevent its evaluation one gets, alas, that it evaluates too little:

In: header[pointer]


I found a way

In: header[tmp] /. tmp -> Trace[pointer][[2]]
Out: header[storage$113]  that finally does the trick. I really do not understand why this works, actually. Trace[pointer][[2]] should also evaluate down to values v1, v2, ... This implementation might be problematic for several reasons: Trace might come up with a rather long list, and I just need the second element of it! Another problem, I suppose that the output of the Trace command might change dynamically depending on whether one puts in some extra definitions. I suppose that this is theoretically possible. Another way would be to issue header[pointer] command while pointer is still unspecified as pointer = Unique[storage]; object = header @@ {pointer} ... fix storage here...  but this forces me to do something I potentially might not want to do. For example, what if the command object = header... has to come at the end for whatever reasons? Anyway, I feel these both constructs looks very fragile I think and I'd like to understand how to controll these things better. Are there better ways? - ## migrated from stackoverflow.comFeb 8 '13 at 15:50 This question came from our site for professional and enthusiast programmers. I think I've done something very similar to what you want in this answer – Leonid Shifrin Feb 8 '13 at 15:53 For pointer-like behavior, could use "expr bags" as described here. – Daniel Lichtblau Feb 8 '13 at 16:20 ## 1 Answer In situations where you want to only evaluate a variable "one level", you can instead apply it's OwnValues so for instance when you have: pointer=Unique[storage]; SetAttributes[header,HoldAll]; header[pointer]/.OwnValues[pointer] (* header[storage$17156] *)


I would also recomend using extract when you need to operate on a held symbol. For instance to set the value of the above structure you could use:

Extract[header[storage$17156], 1, Function[v, v= 34, HoldAll]]  Using Extract and an anonymous function with HoldAll means the OwnValues for storage$17156 won't be applied prior to storage$17156=34 evaluates. A potential way of making this method look nicer is to use UpValues of a custom symbol to change how SetDelayed (:=) is evaluated: pointerValue /: SetDelayed[var_, pointerValue[value_]] /; Head@var === header := Extract[var, 1, Function[v, v = value, HoldAll]] pointerValue::whdr = "Attempted to use non-pointer object: 1 as a pointer during SetDelayed"; pointerValue /: SetDelayed[var_, pointerValue[value_]] := (Message[pointerValue::whdr, Defer[var]];$Failed)


This way you can set new values for your header object using:

myObj = header[pointer] /. OwnValues[pointer];
myObj := pointerValue[1];
myObj[[1]]
myObj := pointerValue[2];
myObj[[1]]

(*1*)
(*2*)


With the only change in syntax being that you have to specify that you are attempting to assign a value to a pointer using the pointerValue wrapper on the right hand side of SetDelayed.

-
Many thanks! If not too problematic, would it be hard to explain what you mean by the setdelayed code? I think you showed something that is beyond my understanding which seems very valuable. –  zorank Feb 5 '13 at 12:53
actually, I do not understad the syntax of the pointerValue /: ... command. I mean, I should, I am not a novice to Mathematica, but I do not understand it. /; is conditioning, but the := sign at the far right is very confusing. –  zorank Feb 11 '13 at 20:02