6
$\begingroup$

Consider a set of (here) 3 series


r`data = Thread[{Range@4, # Range@4}] & /@ (10 Range@3);
(*{
{{1, 10}, {2, 20}, {3, 30}, {4, 40}}, 
{{1, 20}, {2, 40}, {3, 60}, {4, 80}},
{{1, 30}, {2, 60}, {3, 90}, {4, 120}}
}*)

The intent is to divide the second coordinate of each point in each series by some divisor divs[[i]]

r`divs = 10 Range@3;
(*{10, 20, 30}*)

I am currently achieving this as follows

r`f[x_] := #/x &
MapThread[MapAt[r`f@#2, #1, {All, 2}] &, {r`data, r`divs}]
(*{
{{1, 1}, {2, 2}, {3, 3}, {4, 4}},
{{1, 1}, {2, 2}, {3, 3}, {4, 4}},
{{1, 1}, {2, 2}, {3, 3}, {4, 4}}
}*)

How to eliminate r`f in favour of directly nesting the division lambda #/x &, present in the MapAt, inside the MapThread?

In other words how to correctly write something of the kind

MapThread[MapAt[#/#2 &, #1, {All, 2}] &, {r`data, r`divs}]
                ^           
                |
                |
      should refer to second coord of point
             and not r`data[[i]]

Please note that I did go through similar questions (e.g. MapThread on a nested Map) but couldn't apply their solutions to the present case.

$\endgroup$
1
  • 5
    $\begingroup$ How about MapThread[#1.DiagonalMatrix[{1, 1/#2}] &, {r`data, r`divs}]? $\endgroup$
    – Carl Woll
    Commented Oct 9, 2020 at 18:03

3 Answers 3

7
$\begingroup$
MapThread[MapAt[z \[Function] z/#2, #1, {All, 2}] &, {r`data, r`divs}]

or

MapThread[MapAt[Function[z, z/#2], #1, {All, 2}] &, {r`data, r`divs}]

or

MapThread[Function[{x, y}, MapAt[#/y &, x, {All, 2}]], {r`data, r`divs}]
{{{1, 1}, {2, 2}, {3, 3}, {4, 4}}, {{1, 1}, {2, 2}, {3, 3}, {4, 4}}, 
{{1, 1}, {2, 2}, {3, 3}, {4, 4}}}

Alternatively,

Transpose[Transpose[#]/{1, #2}] & @@@ Transpose[{r`data, r`divs}]

MapThread[Transpose[Transpose[#]/{1, #2}] &, {r`data, r`divs}]

MapThread[ReplacePart[#, {i_, 2} :> #[[i, 2]]/#2] &, {r`data, r`divs}]

SubsetMap[Flatten[Partition[#, First[Length /@ r`data]]/r`divs] &, r`data, 
  {All, All, 2}] 

Module[{z = r`data}, z[[All, All, 2]] = z[[All, All, 2]]/r`divs;z]

all give

{{{1, 1}, {2, 2}, {3, 3}, {4, 4}}, {{1, 1}, {2, 2}, {3, 3}, {4, 4}},
 {{1, 1}, {2, 2}, {3, 3}, {4, 4}}} 

And... a Halloween special:

enter image description here

☺[{r`data, r`divs}]
{{{1, 1}, {2, 2}, {3, 3}, {4, 4}}, {{1, 1}, {2, 2}, {3, 3}, {4, 4}}, 
 {{1, 1}, {2, 2}, {3, 3}, {4, 4}}}
$\endgroup$
3
  • $\begingroup$ so one can't escape having to define the MapAt's func explicitly? $\endgroup$
    – lineage
    Commented Oct 9, 2020 at 18:08
  • $\begingroup$ thnx for introducing me to spliced assignment! $\endgroup$
    – lineage
    Commented Oct 9, 2020 at 18:42
  • $\begingroup$ @lineage, if you have to use MapThread + MapAt combination I don't know how we can avoid defining the function for the first argument of MapAt. If MapAt is not required, Carl's suggestion in the comments is a clean way to get the desired result. $\endgroup$
    – kglr
    Commented Oct 9, 2020 at 18:50
4
$\begingroup$

By using ReplacePart in place of any mapping function, there is no need to define any function, neither explicitly like your r`f nor as a pure function.

data = Thread[{Range @ 4, # Range @ 4}]& /@ (10 Range @ 3);
divisors = 10 Range @ 3;
ReplacePart[
  data, 
  {i_, j_} :> Module[{m = data[[i, j]]}, m[[2]] = m[[2]]/divisors[[i]]; m]]
{{{1, 1}, {2, 2}, {3, 3}, {4, 4}}, 
 {{1, 1}, {2, 2}, {3, 3}, {4, 4}}, 
 {{1, 1}, {2, 2}, {3, 3}, {4, 4}}}
$\endgroup$
3
$\begingroup$

In Mathematica version 12.1+ we can use OperatorApplied to inject the outer argument into the inner function:

MapThread[MapAt[OperatorApplied[#/#2&][#2], #1, {All, 2}] &, {r`data, r`divs}]

(*
{{1, 1}, {2, 2}, {3, 3}, {4, 4}},
{{1, 1}, {2, 2}, {3, 3}, {4, 4}},
{{1, 1}, {2, 2}, {3, 3}, {4, 4}}
*)

Curry from version 11.3+ does the same thing but is now considered obsolete. For more details on these and related constructions, see (197168).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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