Can a Table iterator “leak” into a Module?

Question

I want to repeatedly call a module by using a Table statement. When I use an i iterator in the Table command calling the module, I receive the error message "Table::itraw: Raw object 1 cannot be used as an iterator. >>" The module also contains a separate Table statement with the i iterator.

I noticed that when I change the iterator in the Table statement calling the module to something else, say v I no longer get the error message above. This makes me think the i iterator is leaking into the module, so ....

Question 1: Am I correct that the value of the i iterator from the Table statement calling the Module is leaking into the module?

Question 2: Is this expected behavior?

Question 3: Why does this happen?

Question 4: Is there another way to solve this problem, other than to avoid the use of i as an iterator in both the Table calling the module and the module itself?

Thank you for any advice! Todd

---

Here is a snippet of the offending code:

errorcorrection[signaltype_,chipid_,probesetpostocheck_]:=Module[{clusters,positions,loopindex,groupminuscurrent,similarprobesets},

loopindex=1; (* initialize while loop index to 1; this loop continues for the length of the number of probesets that need checked for unusual signal - which is the length of the probesetpostocheck *)

While[loopindex<=2(*Length[probesetpostocheck[chipid]]*),

clusters=Flatten[Drop[Sort[FindClusters[signaltype[chipid][[probesetpostocheck[chipid][[loopindex]]]],3,DistanceFunction->EuclideanDistance,Method->"Agglomerate"]],-1]]; (* locate probe signals within the current probeset being evaluated that are unreliable and need replaced *)

Print[clusters];

With[{i=i},positions=Flatten[Table[Position[signaltype[chipid][[probesetpostocheck[chipid][[loopindex]]]],clusters[[i]]],{i,1,Length[clusters]}]]]

Print[positions];

If[MemberQ[controlchips,chipid],

(groupminuscurrent=DeleteCases[controlchips,x_/;x==chipid];
similarprobesets=Table[signaltype[groupminuscurrent[[i]]][[probesetpostocheck[chipid][[loopindex]]]],{i,1,Length[groupminuscurrent]}]),

(groupminuscurrent=DeleteCases[experimentchips,x_/;x==chipid];
similarprobesets=Table[signaltype[groupminuscurrent[[i]]][[probesetpostocheck[chipid][[loopindex]]]],{i,1,Length[groupminuscurrent]}])]

Print[similarprobesets];

loopindex++]]

---

run2 = Table[
  errorcorrection[pmsignal, experimentchips[[i]], pmpostocheck], {i, 
   1, 1}]


{11719.,1567.,2149.,2001.,1781.,1693.}

Table::itraw: Raw object 1 cannot be used as an iterator. >>

Table[Position[pmsignal[GSM356796][[pmpostocheck[GSM356796][[loopindex$112290]]]],clusters$112290[[1]]],{1,1,Length[clusters$112290]}]

---

run3 = Table[
  errorcorrection[pmsignal, experimentchips[[v]], pmpostocheck], {v, 
   1, 1}]

{11719.,1567.,2149.,2001.,1781.,1693.}

{7,6,8,9,17,20}

---

After considering everyone's responses below, I have been able to get my code to work better by replacing the With[{i=i},positions=Flatten .... statement inside the errorcorrection module with Module[{i=i},positions=Flatten .... As noted Block also works fine.

I am still puzzled, however, why i takes on a value of 1 inside the errorcorrection module, when the module is called by this statement:

run2 = Table[errorcorrection[pmsignal, experimentchips[[i]], pmpostocheck], {i, 1, 1}]

An example of output showing the iterator i is here:

run2 = Table[errorcorrection14[pmsignal, experimentchips[[i]], pmpostocheck], {i, 1, 1}]


{11719.,1567.,2149.,2001.,1781.,1693.}

i equals 1

{7,6,8,9,17,20}

Question 5: (continuing from numbering above): Doesn't the Table command in run2=Table[errorcorrection14[pmsignal, experimentchips[[i]], pmpostocheck], {i, 1, 1}] localize the iterator?

Question 6: If so, why does that value show up in the errorcorrection module?

Question 7: If i did NOT take on the value of 1 (or any value) in the errorcorrection module, would my original With[{i=i},positions=Flatten statement have worked ok?

It seems to me (in my naivety) that the problem with using With in this case, is that i is taking on a value in the module from the Table statement calling the module .... but I thought Table should localize i and prevent its value from going into the module??

Thank you again, as I struggle to understand at a "deeper" level.

Answer 1

I know the reason for the second giving you the error: i has been set to 1. Under normal operations, this will not happen as Table has the Attribute HoldAll, so even if i has a value, Table should be shielded from it. For example,

i = 5
Table[ i, {i , 3}]
(* Out[1] = 5 
   Out[2] = {1, 2, 3}
*)

However, in the large block of code, you use this construct:

With[{i = i}, Table[ ..., {i, ...} ]

So, it is likely that the second Table is executed in a similar manner, and that is where the problem lies. For instance,

With[{i = 1}, Table[i, {i, 5}]]
(* Table::itraw: Raw object 1 cannot be used as an iterator. >>
   Out[3] = Table[1, {1, 5}]
*)

With operates by replacing i in Table with its value prior to Table doing anything which circumvents the HoldAll. Neither, Block nor Module have the same effect, so I would advise using them, instead, for this type of thing.

---

In your code for run2, you effectively have this code

Table[ With[{i2 = i1}, Table[ ..., {i2, ...}]], {i1, ...}]
(* I have renamed the variables for clarity, but the results are the
   same with them all set to i*)

Here, i1 is localized to the outer Table and the With sets i2 to the value of i1 which causes the error you're seeing. The reason this shows up inside errorcorrection is that i1 is not localized to errorcorrection, but takes on the global (to it) value.

Incidentally, this reveals that Table is using Block or Internal`InheritedBlock to perform its scoping. Consider,

f[x_] := {x, i}
Block[{i = 3}, f[q]]
Internal`InheritedBlock[{i = 3}, f[q]]
Module[{i = 3}, f[q]]
With[{i = 3}, f[q]]
(*
Out[131]= {q, 3}
Out[132]= {q, 3}
Out[133]= {q, i}
Out[134]= {q, i}
*)

which reveals one of the key differences between the different scoping constructs. Here i in f is interpreted as Global`i, but Module and With effectively use Unique[i], instead. So, they return Global`i unevaluated. (With actually performs a rewriting operation, but the effect is the same.) Block and Internal`InheritedBlock effectively define Unique[]`i and without any context qualifier on i in f, it is subsumed into the new context. The difference between them is that Block assumes that no definitions of i exists. But, Internal`InheritedBlock pulls in the global definitions into Unique[]` allowing for small changes to be made easily.

In the code, if i was not specified in the outer Table (as in the code for run3), then yes With would work as expected. But, the moral of the story is that Table need not be wrapped in any of scoping constructs as it does the scoping itself.