Bill Allombert on Thu, 17 Oct 2013 13:42:22 +0200 |
[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]
Re: gp/pari version of the "Map" datatype and "reduce" function |
On Wed, Oct 16, 2013 at 11:32:50PM +0100, Richard Heylen wrote: > When writing gp/pari programs I often have a function that produces > lots of different discrete values and I want to record some data about > these values. For example, I might want to accumulate sets of input > values that result in the different output values. The key feature is > that I don't really know what the output values are before I start > running the function and they can't conveniently be converted into > integers to be used as an index into an array. > In other languages this is often addressed by providing a data type > which can take an arbitrary value as a key and lets you set or return > an arbitrary value associated with that key. As a bonus, it might be > useful to be able to interate over the data in the order in which the > keys were first used. This is difficult to implement efficiently in the GP design which lacks a garbage collector. Karim managed to implement usable lists, but they do not work with GP2C. > Also, pari/gp has functions "apply" and "select" which implement > common functional programming ideas. A key one which seems to be > missing is "reduce" (or "fold") which takes a vector, a function and > an initial value and applies the function repeatedly to the value and > the next element of the vector, using the return value as the value > for the next iteration. For example reduce((a,b)->a+b,v,0) would > return the sum of the elements of v. it is a ggood suggestion, but you can implement reduce in GP easily: reduce(f,v)=my(s=v[1]);for(i=2,#v,s=f(s,v[i]));s I could add it to PARI but we need to settle on a name. Cheers, Bill.