Part extraction vs functions like First, Last?

Question

There are many ways to extract different parts of lists in Mathematica. For example, the first part of a list v can be accessed either as v[[1]] or First[v] or Take[v,1]. Likewise the last element can be accessed as v[[-1]], Last[v] or Take[v,-1]. Similarly v[[2;;]] is equivalent to Rest[v] and to Drop[v,1], and v[[;;-2]] is equivalent to Most[v] and Drop[v,-1].

The more specialized functions First, Last, Rest and Most are marginally more efficient. On the other hand, it makes sense to use the more general Part ([[ ]]) or Take/Drop if it is part of a calculation that also requires parts to be accessed that do not have a specialist function.

Beyond these considerations, are there any reasons to prefer Part over Take/Drop or the more specialized functions, other than personal preferences over coding style?

Answer 1

You cannot make assignments to First, Last, Rest, or Most the way you can with Part. Therefore, there is greater consistency in using Part for all operations. See this answer for an example of and argument for this consistency.

Also, you must change functions if you need to update your code to index a different element or change an element to a Span. By using Part from the beginning you have the greatest flexibility without changing functions. This can make comparing different code revisions easier. It can also make a user defined function more flexible, e.g.: f[x_, part__ : 1] := foo @ x[[part]]

Less importantly x[[1]] is more terse than First@x, especially in the FrontEnd using 〚 〛, and I am a fan of terse coding.

Answer 2

Putting aside any minor performance differences, the specialized functions look cleaner and clear in intent in cases where you know for sure that you will not be needing any arbitrary part. Some examples that come to mind are

1: expression//Timing//First            (* Get only the timing *)
2: (... Sow[x]...)//Reap//Last          (* Get only sowed values *)
3: Most@#/Last@#&@CoefficientList[poly] (* normalize by coefficient of highest degree term *)
4: Rest@#/First@#&@Eigenvalues[matrix]  (* normalize by largest eigenvalue *)
5: Most@#/Rest@#&@list                  (* divide each element by the next *)

I'm not very into golfing, but I do like minimizing my keystrokes that are spent doing nothing (and avoid the mouse). If I'm at the end of a long expression and I need to take the first element, I'd simply use Part to extract it rather than move the cursor all the way back to insert a First@. In other words:

something + (insert very long expression here)[[1]]

instead of

something + First@(insert very long expression here)

If I change my mind and decide I need a different element, I have the flexibility to do so and also include all elements with [[;;]]. In such cases, you can't simply stick in a //First at the end, because of the very low precedence of Postfix — the expression returned might not be what you intended to get.