3
$\begingroup$

The Problem

Consider the following code (1):

In[1]:= x = 2;
        Module[{x, y},
         y[x_] = x;
         y[1]
        ]
Out[2]= 2

I expected the answer to be 1 since I assumed that x would be replaced with a local variable. The code returns the expected result if I replace Module with Block. However, I do not see why x would be out of scope of a lexical scoping construct. The effect can be demonstrated in a more drastic way with the following code (2):

In[3]:= x = 2;
        Module[{x, y},
         y[x_] = (x = 1);
        ];
        x
Out[5]= 1

The x inside the Module is not replaced with a local version and the value of the variable on the outside of the module is changed instead.


My own attempt (Can be skipped)

to trace the problem:

In[6]:= x = 2;
        Module[{x, y},
         y[x_] = x;
         y[1]
        ] // Trace
Out[7]= {Module[{x,y},y[x_]=x;y[1]],{y$2016[x_]=x;y$2016[1],{{x,2},y$2016[x_]=2,2},{y$2016[1],2},2},2}

This shows that the only y is replaced with a local version.

Next I had a look at the Code (3) that the Kernel sees:

In[8]:= x = 2;
        Module[{x, y},
         y[x_] = x;
         y[1]
        ] // Hold // FullForm
Out[9]//FullForm= Hold[Module[List[x,y],CompoundExpression[Set[y[Pattern[x,Blank[]]],x],y[1]]]]

and contrasted it with the code (4):

In[10]:= x = 2;
         Module[{x, y},
          y = x;
          y[1]
         ] // Hold // FullForm
Out[11]//FullForm= Hold[Module[List[x,y],CompoundExpression[Set[y,x],y[1]]]]

From a lexical point of view the difference between code (3) and (4) is only y[Pattern[x,Blank[]]] vs y. However in (3) x is not a local variable and in (4) it is.

Since Module has the Attribute HoldAll, I expect it to treat its Body in a purely lexical way before any of the attributes of the expression inside the body or whatnot apply.

I am aware that Set has the attributes HoldFirst and SequenceHold, but I do not see why this would make a difference in this case since Module localizes the variables even for held expressions.

Does the Kernel see x as a Symbol?:

In[12]:= Clear[x];
         Hold[
         Module[{x, y},
          y[x_] = x;
          y[1]
         ]] // Cases[#, _Symbol, Infinity] &
Out[12]= {x,y,x,x}

This suggests to me that the x on the left and right hand side of the Set is a Symbol and that it should be replaced with a localized version.


Resources considered (Can be skipped)

Since the code is so simple I highly expect this question to be either a lack of my understanding or a duplicate. However unfortunately I could neither find the answer

nor in the various other questions I checked. A local function inside a Module is not uncommon, however, the combination of Set with a pattern on the left as a localized variable inside a Module seems quite specific. I did not encounter it elsewhere.


The Question

Please explain the result of code (1).

Thanks in advance for answers or helpful pointers.

$\endgroup$
4
$\begingroup$

ref / Module / Details

  • Before evaluating expr, Module substitutes new symbols for each of the local variables that appear anywhere in expr except as local variables in scoping constructs

ref / Set / Details

  • When it appears in an unevaluated symbolic form, Set is treated as a scoping construct so that variables in nested occurrences are renamed if necessary. (* Kuba: or not scoped by outer scoping constructs *)

The common workaround is to use different names, use Block or to trick Module so it will not 'see' Set[_ , _]:

x = 2;
Module[{x, y}, Set @@ {y[x_], x}; y[1]]
  (*before evaluation, Module can only see `Apply[Set, {y[x_], x}]`*)

The issue is very similar to the automatic renaming feature, except it is the other way around :)

Enforcing correct variable bindings and avoiding renamings for conflicting variables in nested scoping constructs

Here is a comparison of those two features:

Module[{x, y}, Hold[y[x_] = x]] (*nested scoping constructs*)
Hold[y$32040[x_] = x]
Module[{x, y, z}, Hold[y[x_] = x + z]] (*automatic renaming*)
Hold[y$32041[x$_] = x$ + z$32041] (*note x$*)
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.