Bug 6621 – Superimposition amount for std.range.chunks

std.range.chunks is quite useful. For it a rather useful third optional argument (that defaults to 0) is how many items are repeated from the precedent chunk: auto data = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]; chunks(data, 3, 0) ==> (default, as now) [[1, 2, 3], [4, 5, 6], [7, 8, 9], [10]] chunks(data, 3, 1) ==> [[1, 2, 3], [3, 4, 5], [5, 6, 7], [7, 8, 9], [9, 10]] chunks(data, 3, 2) ==> [[1, 2, 3], [2, 3, 4], [3, 4, 5], [4, 5, 6], [5, 6, 7], [6, 7, 8], [7, 8, 9], [8, 9, 10]] In Python I use now and then a lazy generator with such extra argument. The third argument can't be higher than the second (if they are equal it's like cycle on a slice).

A simple use case, count the number of local maxima in an array (a number is a local maxima if it is greater than the number before and after it): size_t countMaxima(T)(in T[] xs) pure nothrow if (__traits(compiles, T.init > T.init)) { if (xs.length < 2) return xs.length; typeof(return) count = 0; foreach (immutable i; 1 .. xs.length - 1) count += xs[i] > xs[i - 1] && xs[i] > xs[i + 1]; return count + (xs[0] > xs[1]) + (xs[$ - 1] > xs[$ - 2]); } Using chunks() with the second argument the code avoids some bug-prone indexes (this is component programming): size_t countMaxima(T)(in T[] xs) pure nothrow if (__traits(compiles, T.init > T.init)) { if (xs.length < 2) return xs.length; auto count = xs .chunks(3, 2) .filter!(c => c[1] > c[0] && c[1] > c[2]) .walkLength; return count + (xs[0] > xs[1]) + (xs[$ - 1] > xs[$ - 2]); } Another use case: http://forum.dlang.org/thread/[email protected]

Another usage example. Here I have implemented a basic version of chunks with the third (optional) argument to solve the Euler Problem 8: https://projecteuler.net/problem=8 import std.stdio, std.range, std.algorithm, std.string; auto chunks(R)(R data, in size_t n, in size_t m=0) pure @safe if (hasLength!R && hasSlicing!R) in { assert(n > 0 && m < n); } body { return iota(n, data.length + 1, n - m) .map!(i => data[i - n .. i]); } void main() { enum K = 13; "euler008.txt" .File .byLine .map!(r => r.strip.representation) .joiner .map!(d => d - '0') .array .chunks(K, K - 1) .map!(c => reduce!q{a * b}(1UL, c)) .reduce!max .writeln; } That outputs the correct result: 235146240000 Where the file euler008.txt is: 73167176531330624919225119674426574742355349194934 96983520312774506326239578318016984801869478851843 85861560789112949495459501737958331952853208805511 12540698747158523863050715693290963295227443043557 66896648950445244523161731856403098711121722383113 62229893423380308135336276614282806444486645238749 30358907296290491560440772390713810515859307960866 70172427121883998797908792274921901699720888093776 65727333001053367881220235421809751254540594752243 52584907711670556013604839586446706324415722155397 53697817977846174064955149290862569321978468622482 83972241375657056057490261407972968652414535100474 82166370484403199890008895243450658541227588666881 16427171479924442928230863465674813919123162824586 17866458359124566529476545682848912883142607690042 24219022671055626321111109370544217506941658960408 07198403850962455444362981230987879927244284909188 84580156166097919133875499200524063689912560717606 05886116467109405077541002256983155200055935729725 71636269561882670428252483600823257530420752963450 Currently without the improved chunks you have to use something like this, you can't use a single UFCS chain: void main() { import std.stdio, std.range, std.algorithm, std.string; enum K = 13; auto digits = "euler008.txt" .File .byLine .map!(r => r.strip.representation) .joiner .map!(d => d - '0') .array; iota(digits.length - K + 1) .map!(i => reduce!q{a * b}(1UL, digits[i .. i + K])) .reduce!max .writeln; } Note: if chunks() gains yet another (optional) argument, allowing the programmer to give a buffer of length n, then chunks() becomes able to work with just input ranges (but it has to copy lot of data, so it's slower): void main() { enum K = 13; int[K] buf; "euler008.txt" .File .byLine .map!(r => r.strip.representation) .joiner .map!(d => d - '0') //.array // No need of this. .chunks(K, K - 1, buf) .map!(c => reduce!q{a * b}(1UL, c)) .reduce!max .writeln; }

See also Sliding window (repeats previous chunks): https://github.com/D-Programming-Language/phobos/pull/4027 std.range.chunks with mere input ranges (buffered) https://issues.dlang.org/show_bug.cgi?id=15759

Commit pushed to master at https://github.com/dlang/phobos https://github.com/dlang/phobos/commit/db2bf8fcfb494bdf62c4d7b6ff1199e1cfa5c06f Fix issue 6621 - add sliding window iterator

Commit pushed to stable at https://github.com/dlang/phobos https://github.com/dlang/phobos/commit/db2bf8fcfb494bdf62c4d7b6ff1199e1cfa5c06f Fix issue 6621 - add sliding window iterator