std.sort: add pdqsort and heapsort

pdqsort (Pattern-defeating quicksort):
    A novel sorting algorithm that combines the fast average case of randomized quicksort
    with the fast worst case of heapsort, while achieving linear time on inputs with certain patterns.
    This implementation is based on https://github.com/zhangyunhao116/pdqsort
    which later replaced Go's original sort algorithm

also:
- compareFn now returns `std.math.Order`
- moved *blocksort* (default stable algorithm) and *pdqsort* (default unstable algorithm) to lib/std/sort/
- made tests run for each algorithm

Co-authored-by: VÖRÖSKŐI András <[email protected]>

Author: alichraghi

lib/std/array_hash_map.zig

@@ -2291,8 +2291,8 @@ test "sort" {
     const C = struct {
         keys: []i32,
 
-        pub fn lessThan(ctx: @This(), a_index: usize, b_index: usize) bool {
-            return ctx.keys[a_index] < ctx.keys[b_index];
+        pub fn compare(ctx: @This(), a_index: usize, b_index: usize) std.math.Order {
+            return std.math.order(ctx.keys[a_index], ctx.keys[b_index]);
         }
     };
 

lib/std/compress/deflate/huffman_code.zig

@@ -346,17 +346,16 @@ pub fn generateFixedOffsetEncoding(allocator: Allocator) !HuffmanEncoder {
     return h;
 }
 
-fn byLiteral(context: void, a: LiteralNode, b: LiteralNode) bool {
+fn byLiteral(context: void, a: LiteralNode, b: LiteralNode) math.Order {
     _ = context;
-    return a.literal < b.literal;
+    return math.order(a.literal, b.literal);
 }
 
-fn byFreq(context: void, a: LiteralNode, b: LiteralNode) bool {
-    _ = context;
+fn byFreq(context: void, a: LiteralNode, b: LiteralNode) math.Order {
     if (a.freq == b.freq) {
-        return a.literal < b.literal;
+        return byLiteral(context, a, b);
     }
-    return a.freq < b.freq;
+    return math.order(a.freq, b.freq);
 }
 
 test "generate a Huffman code from an array of frequencies" {

lib/std/compress/zstandard/decode/huffman.zig

@@ -128,7 +128,7 @@ fn assignSymbols(weight_sorted_prefixed_symbols: []LiteralsSection.HuffmanTree.P
         LiteralsSection.HuffmanTree.PrefixedSymbol,
         weight_sorted_prefixed_symbols,
         weights,
-        lessThanByWeight,
+        compareByWeight,
     );
 
     var prefix: u16 = 0;
@@ -222,13 +222,13 @@ pub fn decodeHuffmanTreeSlice(
     return buildHuffmanTree(&weights, symbol_count);
 }
 
-fn lessThanByWeight(
+fn compareByWeight(
     weights: [256]u4,
     lhs: LiteralsSection.HuffmanTree.PrefixedSymbol,
     rhs: LiteralsSection.HuffmanTree.PrefixedSymbol,
-) bool {
+) std.math.Order {
     // NOTE: this function relies on the use of a stable sorting algorithm,
     //       otherwise a special case of if (weights[lhs] == weights[rhs]) return lhs < rhs;
     //       should be added
-    return weights[lhs.symbol] < weights[rhs.symbol];
+    return std.math.order(weights[lhs.symbol], weights[rhs.symbol]);
 }

lib/std/comptime_string_map.zig

@@ -9,16 +9,16 @@ const mem = std.mem;
 /// You can pass `struct { []const u8 }` (only keys) tuples if `V` is `void`.
 pub fn ComptimeStringMap(comptime V: type, comptime kvs_list: anytype) type {
     const precomputed = comptime blk: {
-        @setEvalBranchQuota(2000);
+        @setEvalBranchQuota(4000);
         const KV = struct {
             key: []const u8,
             value: V,
         };
         var sorted_kvs: [kvs_list.len]KV = undefined;
         const lenAsc = (struct {
-            fn lenAsc(context: void, a: KV, b: KV) bool {
+            fn lenAsc(context: void, a: KV, b: KV) std.math.Order {
                 _ = context;
-                return a.key.len < b.key.len;
+                return std.math.order(a.key.len, b.key.len);
             }
         }).lenAsc;
         for (kvs_list, 0..) |kv, i| {

lib/std/debug.zig

@@ -1211,7 +1211,7 @@ fn readMachODebugInfo(allocator: mem.Allocator, macho_file: File) !ModuleDebugIn
     // Even though lld emits symbols in ascending order, this debug code
     // should work for programs linked in any valid way.
     // This sort is so that we can binary search later.
-    std.sort.sort(MachoSymbol, symbols, {}, MachoSymbol.addressLessThan);
+    std.sort.sort(MachoSymbol, symbols, {}, MachoSymbol.addressCompare);
 
     return ModuleDebugInfo{
         .base_address = undefined,
@@ -1269,9 +1269,9 @@ const MachoSymbol = struct {
         return self.addr;
     }
 
-    fn addressLessThan(context: void, lhs: MachoSymbol, rhs: MachoSymbol) bool {
+    fn addressCompare(context: void, lhs: MachoSymbol, rhs: MachoSymbol) std.math.Order {
         _ = context;
-        return lhs.addr < rhs.addr;
+        return std.math.order(lhs.addr, rhs.addr);
     }
 };
 

lib/std/enums.zig

@@ -1289,9 +1289,9 @@ test "std.enums.ensureIndexer" {
     });
 }
 
-fn ascByValue(ctx: void, comptime a: EnumField, comptime b: EnumField) bool {
+fn ascByValue(ctx: void, comptime a: EnumField, comptime b: EnumField) std.math.Order {
     _ = ctx;
-    return a.value < b.value;
+    return std.math.order(a.value, b.value);
 }
 pub fn EnumIndexer(comptime E: type) type {
     if (!@typeInfo(E).Enum.is_exhaustive) {

lib/std/http/Headers.zig

@@ -36,11 +36,10 @@ pub const Field = struct {
     name: []const u8,
     value: []const u8,
 
-    fn lessThan(ctx: void, a: Field, b: Field) bool {
+    fn compare(ctx: void, a: Field, b: Field) std.math.Order {
         _ = ctx;
-        if (a.name.ptr == b.name.ptr) return false;
-
-        return ascii.lessThanIgnoreCase(a.name, b.name);
+        if (a.name.ptr == b.name.ptr) return .eq;
+        return ascii.orderIgnoreCase(a.name, b.name);
     }
 };
 
@@ -191,7 +190,7 @@ pub const Headers = struct {
 
     /// Sorts the headers in lexicographical order.
     pub fn sort(headers: *Headers) void {
-        std.sort.sort(Field, headers.list.items, {}, Field.lessThan);
+        std.sort.sort(Field, headers.list.items, {}, Field.compare);
         headers.rebuildIndex();
     }
 

lib/std/multi_array_list.zig

@@ -155,12 +155,12 @@ pub fn MultiArrayList(comptime T: type) type {
                 };
             }
             const Sort = struct {
-                fn lessThan(context: void, lhs: Data, rhs: Data) bool {
+                fn compare(context: void, lhs: Data, rhs: Data) std.math.Order {
                     _ = context;
-                    return lhs.alignment > rhs.alignment;
+                    return std.math.order(rhs.alignment, lhs.alignment);
                 }
             };
-            std.sort.sort(Data, &data, {}, Sort.lessThan);
+            std.sort.sort(Data, &data, {}, Sort.compare);
             var sizes_bytes: [fields.len]usize = undefined;
             var field_indexes: [fields.len]usize = undefined;
             for (data, 0..) |elem, i| {
@@ -483,8 +483,8 @@ pub fn MultiArrayList(comptime T: type) type {
                     }
                 }
 
-                pub fn lessThan(sc: @This(), a_index: usize, b_index: usize) bool {
-                    return sc.sub_ctx.lessThan(a_index, b_index);
+                pub fn compare(sc: @This(), a_index: usize, b_index: usize) std.math.Order {
+                    return sc.sub_ctx.compare(a_index, b_index);
                 }
             };
 

lib/std/net.zig

@@ -1082,7 +1082,7 @@ fn linuxLookupName(
         key |= (MAXADDRS - @intCast(i32, i)) << DAS_ORDER_SHIFT;
         addr.sortkey = key;
     }
-    std.sort.sort(LookupAddr, addrs.items, {}, addrCmpLessThan);
+    std.sort.sort(LookupAddr, addrs.items, {}, addrCompare);
 }
 
 const Policy = struct {
@@ -1199,9 +1199,9 @@ fn IN6_IS_ADDR_SITELOCAL(a: [16]u8) bool {
 }
 
 // Parameters `b` and `a` swapped to make this descending.
-fn addrCmpLessThan(context: void, b: LookupAddr, a: LookupAddr) bool {
+fn addrCompare(context: void, b: LookupAddr, a: LookupAddr) std.math.Order {
     _ = context;
-    return a.sortkey < b.sortkey;
+    return std.math.order(a.sortkey, b.sortkey);
 }
 
 fn linuxLookupNameFromNull(

lib/std/priority_dequeue.zig

@@ -461,12 +461,12 @@ pub fn PriorityDequeue(comptime T: type, comptime Context: type, comptime compar
     };
 }
 
-fn lessThanComparison(context: void, a: u32, b: u32) Order {
+fn compare(context: void, a: u32, b: u32) Order {
     _ = context;
     return std.math.order(a, b);
 }
 
-const PDQ = PriorityDequeue(u32, void, lessThanComparison);
+const PDQ = PriorityDequeue(u32, void, compare);
 
 test "std.PriorityDequeue: add and remove min" {
     var queue = PDQ.init(testing.allocator, {});
@@ -977,11 +977,11 @@ fn generateRandomSlice(allocator: std.mem.Allocator, rng: std.rand.Random, size:
     return array.toOwnedSlice();
 }
 
-fn contextLessThanComparison(context: []const u32, a: usize, b: usize) Order {
+fn contextComparison(context: []const u32, a: usize, b: usize) Order {
     return std.math.order(context[a], context[b]);
 }
 
-const CPDQ = PriorityDequeue(usize, []const u32, contextLessThanComparison);
+const CPDQ = PriorityDequeue(usize, []const u32, contextComparison);
 
 test "std.PriorityDequeue: add and remove" {
     const context = [_]u32{ 5, 3, 4, 2, 2, 8, 0 };

lib/std/priority_queue.zig

@@ -284,7 +284,8 @@ fn lessThan(context: void, a: u32, b: u32) Order {
 }
 
 fn greaterThan(context: void, a: u32, b: u32) Order {
-    return lessThan(context, a, b).invert();
+    _ = context;
+    return std.math.order(b, a);
 }
 
 const PQlt = PriorityQueue(u32, void, lessThan);
@@ -616,11 +617,11 @@ test "std.PriorityQueue: siftUp in remove" {
     }
 }
 
-fn contextLessThan(context: []const u32, a: usize, b: usize) Order {
+fn contextCompare(context: []const u32, a: usize, b: usize) Order {
     return std.math.order(context[a], context[b]);
 }
 
-const CPQlt = PriorityQueue(usize, []const u32, contextLessThan);
+const CPQlt = PriorityQueue(usize, []const u32, contextCompare);
 
 test "std.PriorityQueue: add and remove min heap with contextful comparator" {
     const context = [_]u32{ 5, 3, 4, 2, 2, 8, 0 };