43
$\begingroup$

Say I have a group of functions:

f1[a_] := a * -1;
f2[a_] := a * 100;
f3[a_] := a / 10.0;

and some data in a list:

data := Range[1, 20];

I would like to apply this group of functions to the data: the first function applied to the first item of data, the second to the second, and so on. Because there are more data elements than there are functions, the first function is also applied to the fourth data element, and so on.

A simple work-round is this:

Flatten[{f1[#[[1]]], f2[#[[2]]], f3[#[[3]]]}  & /@ Partition[data, 3]]

giving

{-1, 200, 0.3, -4, 500, 0.6, -7, 800, 0.9, -10, 1100, 1.2, -13, 1400, 
  1.5, -16, 1700, 1.8}

but this isn't an ideal solution: the slots have been 'hard-wired', and it wouldn't be possible to modify the list of functions easily.

Is there a Map-related function that could do this elegantly? I've not been able to discover it yet.

(This is a toy example, of course!)

$\endgroup$
3
  • 3
    $\begingroup$ your current solution only goes up to 18, is that intentional? $\endgroup$
    – acl
    Apr 3, 2012 at 10:22
  • 6
    $\begingroup$ Congrats on the 1024th question of this site! $\endgroup$
    – F'x
    Apr 3, 2012 at 12:02
  • 8
    $\begingroup$ Oh good, a nice round number: xkcd.com/1000 $\endgroup$
    – tkott
    Apr 3, 2012 at 14:04

7 Answers 7

27
$\begingroup$

Here's how you can do it in a simple way:

functionMap[funcs_List, data_] := Module[{fn = RotateRight[funcs]}, 
    First[(fn = RotateLeft[fn])][#] & /@ data]

Use it as:

functionMap[{f1, f2, f3}, Range[20]]
(* {f1[1], f2[2], f3[3], f1[4], f2[5], f3[6], f1[7], f2[8], f3[9], f1[10],
    f2[11], f3[12], f1[13], f2[14], f3[15], f1[16], f2[17], f3[18], f1[19], f2[20]} *)
$\endgroup$
3
  • 2
    $\begingroup$ Oooo, clever. +1 (you know I like that approach.) $\endgroup$
    – Mr.Wizard
    Apr 9, 2012 at 15:22
  • 2
    $\begingroup$ You could write it as a closure: makeRotator[funcs_] := Module[{fn = funcs}, First[(fn = RotateLeft[fn])][#] &] then makeRotator[{f1,f2,f3}] /@ Range[10]. $\endgroup$
    – Szabolcs
    May 25, 2012 at 6:50
  • 1
    $\begingroup$ Shorter/simpler: {fn = funcs} and Last[(fn = RotateLeft[fn])] $\endgroup$
    – Mr.Wizard
    Jan 4, 2013 at 18:01
32
$\begingroup$

Use PadRight with the cyclical padding setup:

funcs = {f1, f2, f3};
data = Range[1, 20];
MapThread[#1[#2] &, {PadRight[funcs, Length@data, funcs], data}]

or

MapThread[Compose, {PadRight[funcs, Length@data, funcs], data}]

{f1[1], f2[2], f3[3], f1[4], f2[5], f3[6], f1[7], f2[8], f3[9], f1[10], f2[11], f3[12], f1[13], f2[14], f3[15], f1[16], f2[17], f3[18], f1[19], f2[20]}

$\endgroup$
3
  • 1
    $\begingroup$ You'll get a populist badge pretty soon :) Congrats in advance! $\endgroup$
    – rm -rf
    May 24, 2012 at 21:56
  • $\begingroup$ I wrote a solution myself before looking at the answers. It was exactly this one, nearly character by character. :) $\endgroup$
    – Szabolcs
    May 25, 2012 at 6:44
  • $\begingroup$ @Szabolcs: Great minds when meet... :) Thanks R.M, you should know that usually I use your method. $\endgroup$ May 25, 2012 at 8:02
21
$\begingroup$
MapIndexed[{f1, f2, f3}[[Mod[First@#2, 3, 1]]][#1] &, data]

does what you want (thanks to Sjoerd for pointing out a silly inefficiency).

$\endgroup$
2
  • 1
    $\begingroup$ This is more or less what I was thinking about too, but I wouldn't give the arguments to the functions first. In this way you're doing 3 times more calculations than necessary. Just put a [#1] after the Part and leave the function list alone. $\endgroup$ Apr 3, 2012 at 10:44
  • $\begingroup$ @Sjoerd yes good point, also, this way it generalizes better. thanks $\endgroup$
    – acl
    Apr 3, 2012 at 10:58
18
$\begingroup$

Another approach

#1@#2 & @@@ 
  Partition[Riffle[Range[20], {f1, f2, f3}, {1, -1, 2}], 2] 

Comparing with Acl's solution:

#1@#2 & @@@ 
  Partition[Riffle[Range[20], {f1, f2, f3}, {1, -1, 2}], 2] == 
 MapIndexed[{f1@#1, f2@#1, f3@#1}[[Mod[First@#2, 3, 1]]] &, data

==> True

$\endgroup$
1
  • 1
    $\begingroup$ nice one +1. Can save few key strokes with #2@#1 & @@@ Partition[Riffle[Range[20], {f1, f2, f3}], 2]:) $\endgroup$
    – kglr
    Apr 3, 2012 at 14:02
17
$\begingroup$

Another approach is to use Outer, as follows

Flatten@Outer[#1[#2]&, {f1, f2, f3}, Range[1,5]]
(*
{f1[1], f1[2], f1[3], f1[4], f1[5], 
 f2[1], f2[2], f2[3], f2[4], f2[5],
 f3[1], f3[2], f3[3], f3[4], f3[5]}
*)

Unfortunately, Outer causes problems if the data points are vectors, for instance

Flatten@Outer[#1[#2] &, {f1, f2, f3}, {#, #} & /@ Range[1, 3]]

produces

{f1[1], f1[1], f1[2], f1[2], f1[3], f1[3], 
 f2[1], f2[1], f2[2], f2[2], f2[3], f2[3], 
 f3[1], f3[1], f3[2], f3[2], f3[3], f3[3]}

To work around this, you need to use the 3$^\text{rd}$ and subsequent parameters of Outer which are level specifications:

Outer[#1[#2] &, {f1, f2, f3}, {#, #} & /@ Range[1, 3], Infinity, 1] //
  Flatten
(*
{f1[{1, 1}], f1[{2, 2}], f1[{3, 3}], 
 f2[{1, 1}], f2[{2, 2}], f2[{3, 3}], 
 f3[{1, 1}], f3[{2, 2}], f3[{3, 3}]}
*)

Or, if you prefer

Outer[#1 @@ #2 &, {f1, f2, f3}, {#, #} & /@ Range[1, 3], Infinity, 1] //
  Flatten
(*
{f1[1, 1], f1[2, 2], f1[3, 3], 
 f2[1, 1], f2[2, 2], f2[3, 3], 
 f3[1, 1], f3[2, 2], f3[3, 3]}
*)
$\endgroup$
16
$\begingroup$

Another approach using Fold in combination with Sow/Reap:

Reap[Fold[(Sow[#[[1]][#2]]; RotateLeft[#]) &, {f1, f2, f3}, data];][[2, 1]]
$\endgroup$
8
$\begingroup$

Would the following qualify as elegant or not?

mapFunctions[funcs_, list_] := Module[{r = list, u},
  applyFunc[f_, l_, i_, t_] := 
   MapAt[f, l, Table[{u}, {u, i, Length[l], t}]];
  Do[r = applyFunc[funcs[[u]], r, u, Length[funcs]],
   {u, Length[funcs]}];
  Return[r];
  ]

which gives:

In[9]:= MapFunctions[{f1, f2, f3}, data]

Out[9]= {f1[1], f2[2], f3[3], f1[4], f2[5], f3[6], f1[7], f2[8], 
 f3[9], f1[10], f2[11], f3[12], f1[13], f2[14], f3[15], f1[16], 
 f2[17], f3[18], f1[19], f2[20]}
$\endgroup$
2
  • 3
    $\begingroup$ Well, it has a Do. What would you say? ;-) $\endgroup$ Apr 3, 2012 at 10:46
  • 1
    $\begingroup$ @SjoerdC.deVries but it has no Goto, I swear! $\endgroup$
    – F'x
    Apr 3, 2012 at 11:14

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.