@voxgig/sdkgen 4.10.0 → 4.11.0

package/project/.sdk/tm/zig/feature/secrets/plugin/value.zig

@@ -0,0 +1,662 @@
+// VENDORED: @voxgig/plugin sdk-20260908-1556-0 (zig/src/value.zig)
+// Source: https://github.com/voxgig/plugin @ 48392f5e2b6d1434ee9b1a4a9a11f4480aaeb46a  [tag: sdk-20260908-1556-0]
+// License: MIT (c) voxgig - see repository LICENSE. Do not edit: resync from upstream.
+//! The dynamic value, and the JSON reader that fills it (§16).
+//!
+//! HAND-WRITTEN, THOUGH `std.json` EXISTS. That is a choice, not a
+//! constraint: zig's `std.json` is the standard library, so §16 would
+//! permit it, and its `ObjectMap` even preserves insertion order. The
+//! reason to write the reader here is symmetry — `c`, `cpp`, `ocaml`
+//! and `haskell` all carry the same 200-line reader, the corpus is the
+//! only thing it ever reads, and a port whose value type is its own is
+//! a port whose ordering rules are visible in one place.
+//!
+//! ARENA ALLOCATION, for the reason `c` gives: every Value lives in one
+//! arena and the whole arena is freed at once. Nothing frees an
+//! individual value, so nothing can double-free one. Zig hands the
+//! arena over in the standard library rather than making the port build
+//! it.
+
+const std = @import("std");
+const builtin = @import("builtin");
+
+/// TWO TOOLCHAINS, ONE SOURCE. This port's own CI builds it on zig 0.13,
+/// and its first consumer (voxgig/sekreto) builds it on zig 0.16, and
+/// the two standard libraries disagree in exactly three places: the
+/// managed list, file I/O, and the environment. `modern` is the one
+/// switch, and every difference is a comptime branch on it, so the
+/// untaken branch is never analysed against a standard library that
+/// lacks it.
+pub const modern = builtin.zig_version.major > 0 or builtin.zig_version.minor >= 15;
+
+/// The list whose allocator travels with it. Zig 0.15 made
+/// `std.ArrayList` unmanaged and moved the managed one to
+/// `std.array_list.Managed`; the API is the same on both sides of the
+/// move, so this alias is the whole of the difference.
+pub fn List(comptime T: type) type {
+    return if (modern) std.array_list.Managed(T) else std.ArrayList(T);
+}
+
+pub const Kind = enum { nul, boolean, num, str, list, map };
+
+pub const Entry = struct {
+    key: []const u8,
+    val: *Value,
+};
+
+pub const Value = struct {
+    kind: Kind = .nul,
+    boolean: bool = false,
+    num: f64 = 0,
+    str: []const u8 = "",
+    /// A list preserves order because it IS the order.
+    items: List(*Value) = undefined,
+    /// A MAP PRESERVES INSERTION ORDER AND SORTS ON DEMAND. §4 rule 4
+    /// makes order observable in several places (`keys` is sorted, `pos`
+    /// is the sorted-ref index), so both orders have to be available and
+    /// the code has to say which it means at each use. An ArrayList of
+    /// entries rather than a hash map: these maps hold a handful of keys.
+    entries: List(Entry) = undefined,
+};
+
+// --- the arena --------------------------------------------------------
+
+var arena_state: ?std.heap.ArenaAllocator = null;
+
+pub fn arena() std.mem.Allocator {
+    if (arena_state == null) {
+        arena_state = std.heap.ArenaAllocator.init(std.heap.page_allocator);
+    }
+    return arena_state.?.allocator();
+}
+
+/// Free EVERYTHING. Call between corpus entries, never inside one.
+pub fn arenaReset() void {
+    if (arena_state) |*a| {
+        _ = a.reset(.free_all);
+    }
+}
+
+/// Allocation failure is fatal: a library that limps on after it cannot
+/// be reasoned about and the corpus cannot express it.
+fn alloc(comptime T: type, n: usize) []T {
+    return arena().alloc(T, n) catch @panic("plugin: out of memory");
+}
+
+pub fn dupe(s: []const u8) []const u8 {
+    const out = alloc(u8, s.len);
+    @memcpy(out, s);
+    return out;
+}
+
+pub fn print(comptime fmt: []const u8, args: anytype) []const u8 {
+    return std.fmt.allocPrint(arena(), fmt, args) catch @panic("plugin: out of memory");
+}
+
+// --- construction -----------------------------------------------------
+
+fn make() *Value {
+    const v = arena().create(Value) catch @panic("plugin: out of memory");
+    v.* = .{};
+    return v;
+}
+
+pub fn vnull() *Value {
+    return make();
+}
+
+pub fn vbool(b: bool) *Value {
+    const v = make();
+    v.kind = .boolean;
+    v.boolean = b;
+    return v;
+}
+
+pub fn vnum(n: f64) *Value {
+    const v = make();
+    v.kind = .num;
+    v.num = n;
+    return v;
+}
+
+pub fn vstr(s: []const u8) *Value {
+    const v = make();
+    v.kind = .str;
+    v.str = s;
+    return v;
+}
+
+pub fn vlist() *Value {
+    const v = make();
+    v.kind = .list;
+    v.items = List(*Value).init(arena());
+    return v;
+}
+
+pub fn vmap() *Value {
+    const v = make();
+    v.kind = .map;
+    v.entries = List(Entry).init(arena());
+    return v;
+}
+
+// --- kinds ------------------------------------------------------------
+
+/// A NULL POINTER MEANS "NOTHING"; a `.nul` Value means "JSON null".
+/// They are different answers and several places need both: a `bail`
+/// binding declining is not one answering null, and a missing export is
+/// not an export of null. Every accessor tolerates a null pointer.
+pub fn isNull(v: ?*const Value) bool {
+    return v == null or v.?.kind == .nul;
+}
+pub fn isBool(v: ?*const Value) bool {
+    return v != null and v.?.kind == .boolean;
+}
+pub fn isNum(v: ?*const Value) bool {
+    return v != null and v.?.kind == .num;
+}
+pub fn isStr(v: ?*const Value) bool {
+    return v != null and v.?.kind == .str;
+}
+pub fn isList(v: ?*const Value) bool {
+    return v != null and v.?.kind == .list;
+}
+pub fn isMap(v: ?*const Value) bool {
+    return v != null and v.?.kind == .map;
+}
+
+// --- access -----------------------------------------------------------
+
+/// `get` answers a null Value for a missing key AND for a key holding
+/// JSON null; `has` distinguishes them, which is what §9.1's "an
+/// authored null is not an absent key" needs.
+pub fn get(m: ?*const Value, key: []const u8) *Value {
+    if (!isMap(m)) return vnull();
+    for (m.?.entries.items) |e| {
+        if (std.mem.eql(u8, e.key, key)) return e.val;
+    }
+    return vnull();
+}
+
+pub fn has(m: ?*const Value, key: []const u8) bool {
+    if (!isMap(m)) return false;
+    for (m.?.entries.items) |e| {
+        if (std.mem.eql(u8, e.key, key)) return true;
+    }
+    return false;
+}
+
+pub fn set(m: ?*Value, key: []const u8, val: ?*Value) void {
+    if (!isMap(m)) return;
+    const v = val orelse vnull();
+    for (m.?.entries.items) |*e| {
+        if (std.mem.eql(u8, e.key, key)) {
+            e.val = v;
+            return;
+        }
+    }
+    m.?.entries.append(.{ .key = dupe(key), .val = v }) catch @panic("plugin: out of memory");
+}
+
+pub fn del(m: ?*Value, key: []const u8) void {
+    if (!isMap(m)) return;
+    var i: usize = 0;
+    while (i < m.?.entries.items.len) : (i += 1) {
+        if (std.mem.eql(u8, m.?.entries.items[i].key, key)) {
+            _ = m.?.entries.orderedRemove(i);
+            return;
+        }
+    }
+}
+
+pub fn at(l: ?*const Value, i: usize) *Value {
+    if (!isList(l) or i >= l.?.items.items.len) return vnull();
+    return l.?.items.items[i];
+}
+
+pub fn push(l: ?*Value, item: ?*Value) void {
+    if (!isList(l)) return;
+    l.?.items.append(item orelse vnull()) catch @panic("plugin: out of memory");
+}
+
+pub fn len(v: ?*const Value) usize {
+    if (v == null) return 0;
+    return switch (v.?.kind) {
+        .list => v.?.items.items.len,
+        .map => v.?.entries.items.len,
+        else => 0,
+    };
+}
+
+pub fn items(v: ?*const Value) []*Value {
+    if (!isList(v)) return &[_]*Value{};
+    return v.?.items.items;
+}
+
+pub fn asStr(v: ?*const Value) []const u8 {
+    return if (isStr(v)) v.?.str else "";
+}
+pub fn asNum(v: ?*const Value) f64 {
+    return if (isNum(v)) v.?.num else 0;
+}
+pub fn asBool(v: ?*const Value) bool {
+    return isBool(v) and v.?.boolean;
+}
+
+/// Keys in INSERTION order.
+pub fn keys(m: ?*const Value) [][]const u8 {
+    if (!isMap(m)) return &[_][]const u8{};
+    const out = alloc([]const u8, m.?.entries.items.len);
+    for (m.?.entries.items, 0..) |e, i| out[i] = e.key;
+    return out;
+}
+
+fn lessStr(_: void, a: []const u8, b: []const u8) bool {
+    return std.mem.order(u8, a, b) == .lt;
+}
+
+/// Keys SORTED by byte order — §4 rule 4's deterministic walk.
+pub fn sortedKeys(m: ?*const Value) [][]const u8 {
+    const out = keys(m);
+    std.mem.sort([]const u8, out, {}, lessStr);
+    return out;
+}
+
+/// §4 rule 4: truthiness is JSON's, not zig's.
+pub fn truthy(v: ?*const Value) bool {
+    if (v == null) return false;
+    return switch (v.?.kind) {
+        .nul => false,
+        .boolean => v.?.boolean,
+        .num => v.?.num != 0,
+        .str => v.?.str.len != 0,
+        else => true,
+    };
+}
+
+/// Deep equality INCLUDING JSON type, which is the half that matters:
+/// half the ports are written in languages whose `==` says `true == 1`,
+/// and `capability/match` exists to catch exactly that.
+pub fn same(a: ?*const Value, b: ?*const Value) bool {
+    const anull = isNull(a);
+    const bnull = isNull(b);
+    if (anull or bnull) return anull and bnull;
+    if (a.?.kind != b.?.kind) return false;
+    return switch (a.?.kind) {
+        .boolean => a.?.boolean == b.?.boolean,
+        .num => a.?.num == b.?.num,
+        .str => std.mem.eql(u8, a.?.str, b.?.str),
+        .list => blk: {
+            if (len(a) != len(b)) break :blk false;
+            for (a.?.items.items, 0..) |x, i| {
+                if (!same(x, at(b, i))) break :blk false;
+            }
+            break :blk true;
+        },
+        .map => blk: {
+            if (len(a) != len(b)) break :blk false;
+            for (a.?.entries.items) |e| {
+                if (!has(b, e.key)) break :blk false;
+                if (!same(e.val, get(b, e.key))) break :blk false;
+            }
+            break :blk true;
+        },
+        else => true,
+    };
+}
+
+pub fn clone(v: ?*const Value) ?*Value {
+    if (v == null) return null;
+    return switch (v.?.kind) {
+        .list => blk: {
+            const out = vlist();
+            for (v.?.items.items) |x| push(out, clone(x));
+            break :blk out;
+        },
+        .map => blk: {
+            const out = vmap();
+            for (v.?.entries.items) |e| set(out, e.key, clone(e.val));
+            break :blk out;
+        },
+        .boolean => vbool(v.?.boolean),
+        .num => vnum(v.?.num),
+        .str => vstr(v.?.str),
+        else => vnull(),
+    };
+}
+
+// --- json -------------------------------------------------------------
+
+/// An integral float renders as an integer: the corpus's expected values
+/// are written `1`, not `1.0`, and a port that emits the latter fails
+/// every comparison for a reason that has nothing to do with the
+/// behaviour under test.
+pub fn numStr(n: f64) []const u8 {
+    if (std.math.isFinite(n) and n == @trunc(n) and @abs(n) < 1e18) {
+        return print("{d}", .{@as(i64, @intFromFloat(n))});
+    }
+    return print("{d}", .{n});
+}
+
+fn escape(s: []const u8, buf: *List(u8)) void {
+    buf.append('"') catch @panic("oom");
+    for (s) |c| {
+        switch (c) {
+            '"' => buf.appendSlice("\\\"") catch @panic("oom"),
+            '\\' => buf.appendSlice("\\\\") catch @panic("oom"),
+            '\n' => buf.appendSlice("\\n") catch @panic("oom"),
+            '\r' => buf.appendSlice("\\r") catch @panic("oom"),
+            '\t' => buf.appendSlice("\\t") catch @panic("oom"),
+            0x08 => buf.appendSlice("\\b") catch @panic("oom"),
+            0x0C => buf.appendSlice("\\f") catch @panic("oom"),
+            else => {
+                if (c < 0x20) {
+                    buf.appendSlice(print("\\u{x:0>4}", .{c})) catch @panic("oom");
+                } else {
+                    buf.append(c) catch @panic("oom");
+                }
+            },
+        }
+    }
+    buf.append('"') catch @panic("oom");
+}
+
+fn render(v: ?*const Value, buf: *List(u8)) void {
+    if (isNull(v)) {
+        buf.appendSlice("null") catch @panic("oom");
+        return;
+    }
+    switch (v.?.kind) {
+        .boolean => buf.appendSlice(if (v.?.boolean) "true" else "false") catch @panic("oom"),
+        .num => buf.appendSlice(numStr(v.?.num)) catch @panic("oom"),
+        .str => escape(v.?.str, buf),
+        .list => {
+            buf.append('[') catch @panic("oom");
+            for (v.?.items.items, 0..) |x, i| {
+                if (i > 0) buf.append(',') catch @panic("oom");
+                render(x, buf);
+            }
+            buf.append(']') catch @panic("oom");
+        },
+        .map => {
+            // SORTED, so two values that are `same` render identically.
+            buf.append('{') catch @panic("oom");
+            for (sortedKeys(v), 0..) |k, i| {
+                if (i > 0) buf.append(',') catch @panic("oom");
+                escape(k, buf);
+                buf.append(':') catch @panic("oom");
+                render(get(v, k), buf);
+            }
+            buf.append('}') catch @panic("oom");
+        },
+        else => buf.appendSlice("null") catch @panic("oom"),
+    }
+}
+
+pub fn json(v: ?*const Value) []const u8 {
+    var buf = List(u8).init(arena());
+    render(v, &buf);
+    return buf.items;
+}
+
+pub const ParseError = error{BadJson};
+
+const Parser = struct {
+    text: []const u8,
+    i: usize = 0,
+    err: []const u8 = "",
+
+    fn digit(p: *Parser) bool {
+        return p.i < p.text.len and p.text[p.i] >= '0' and p.text[p.i] <= '9';
+    }
+
+    fn skip(p: *Parser) void {
+        while (p.i < p.text.len) {
+            const c = p.text[p.i];
+            if (c == ' ' or c == '\t' or c == '\n' or c == '\r') p.i += 1 else break;
+        }
+    }
+
+    fn lit(p: *Parser, word: []const u8) bool {
+        if (p.i + word.len > p.text.len) return false;
+        if (!std.mem.eql(u8, p.text[p.i .. p.i + word.len], word)) return false;
+        p.i += word.len;
+        return true;
+    }
+
+    fn hex4(p: *Parser) ?u21 {
+        if (p.i + 4 > p.text.len) return null;
+        var acc: u21 = 0;
+        var k: usize = 0;
+        while (k < 4) : (k += 1) {
+            const d = std.fmt.charToDigit(p.text[p.i + k], 16) catch return null;
+            acc = acc * 16 + d;
+        }
+        p.i += 4;
+        return acc;
+    }
+
+    fn string(p: *Parser, buf: *List(u8)) bool {
+        if (p.i >= p.text.len or p.text[p.i] != '"') {
+            p.err = "expected a string";
+            return false;
+        }
+        p.i += 1;
+        while (p.i < p.text.len and p.text[p.i] != '"') {
+            const c = p.text[p.i];
+            if (c != '\\') {
+                buf.append(c) catch @panic("oom");
+                p.i += 1;
+                continue;
+            }
+            p.i += 1;
+            if (p.i >= p.text.len) {
+                p.err = "unterminated escape";
+                return false;
+            }
+            const e = p.text[p.i];
+            p.i += 1;
+            switch (e) {
+                '"' => buf.append('"') catch @panic("oom"),
+                '\\' => buf.append('\\') catch @panic("oom"),
+                '/' => buf.append('/') catch @panic("oom"),
+                'n' => buf.append('\n') catch @panic("oom"),
+                'r' => buf.append('\r') catch @panic("oom"),
+                't' => buf.append('\t') catch @panic("oom"),
+                'b' => buf.append(0x08) catch @panic("oom"),
+                'f' => buf.append(0x0C) catch @panic("oom"),
+                'u' => {
+                    const hi = p.hex4() orelse {
+                        p.err = "bad \\u escape";
+                        return false;
+                    };
+                    var cp: u21 = hi;
+                    if (hi >= 0xD800 and hi <= 0xDBFF and p.i + 1 < p.text.len and
+                        p.text[p.i] == '\\' and p.text[p.i + 1] == 'u')
+                    {
+                        p.i += 2;
+                        const lo = p.hex4() orelse {
+                            p.err = "bad \\u escape";
+                            return false;
+                        };
+                        cp = 0x10000 + ((hi - 0xD800) << 10) + (lo - 0xDC00);
+                    }
+                    var tmp: [4]u8 = undefined;
+                    const n = std.unicode.utf8Encode(cp, &tmp) catch {
+                        p.err = "bad code point";
+                        return false;
+                    };
+                    buf.appendSlice(tmp[0..n]) catch @panic("oom");
+                },
+                else => {
+                    p.err = "bad escape";
+                    return false;
+                },
+            }
+        }
+        if (p.i >= p.text.len) {
+            p.err = "unterminated string";
+            return false;
+        }
+        p.i += 1;
+        return true;
+    }
+
+    fn value(p: *Parser) ?*Value {
+        p.skip();
+        if (p.i >= p.text.len) {
+            p.err = "unexpected end of input";
+            return null;
+        }
+        const c = p.text[p.i];
+
+        if (c == 'n') {
+            if (!p.lit("null")) {
+                p.err = "bad literal";
+                return null;
+            }
+            return vnull();
+        }
+        if (c == 't') {
+            if (!p.lit("true")) {
+                p.err = "bad literal";
+                return null;
+            }
+            return vbool(true);
+        }
+        if (c == 'f') {
+            if (!p.lit("false")) {
+                p.err = "bad literal";
+                return null;
+            }
+            return vbool(false);
+        }
+        if (c == '"') {
+            var buf = List(u8).init(arena());
+            if (!p.string(&buf)) return null;
+            return vstr(buf.items);
+        }
+        if (c == '[') {
+            p.i += 1;
+            const out = vlist();
+            p.skip();
+            if (p.i < p.text.len and p.text[p.i] == ']') {
+                p.i += 1;
+                return out;
+            }
+            while (true) {
+                const item = p.value() orelse return null;
+                push(out, item);
+                p.skip();
+                if (p.i < p.text.len and p.text[p.i] == ',') {
+                    p.i += 1;
+                    continue;
+                }
+                if (p.i < p.text.len and p.text[p.i] == ']') {
+                    p.i += 1;
+                    return out;
+                }
+                p.err = "expected , or ] in array";
+                return null;
+            }
+        }
+        if (c == '{') {
+            p.i += 1;
+            const out = vmap();
+            p.skip();
+            if (p.i < p.text.len and p.text[p.i] == '}') {
+                p.i += 1;
+                return out;
+            }
+            while (true) {
+                p.skip();
+                var kbuf = List(u8).init(arena());
+                if (!p.string(&kbuf)) return null;
+                p.skip();
+                if (p.i >= p.text.len or p.text[p.i] != ':') {
+                    p.err = "expected : in object";
+                    return null;
+                }
+                p.i += 1;
+                const val = p.value() orelse return null;
+                set(out, kbuf.items, val);
+                p.skip();
+                if (p.i < p.text.len and p.text[p.i] == ',') {
+                    p.i += 1;
+                    continue;
+                }
+                if (p.i < p.text.len and p.text[p.i] == '}') {
+                    p.i += 1;
+                    return out;
+                }
+                p.err = "expected , or } in object";
+                return null;
+            }
+        }
+
+        // JSON's number grammar is STRICT, and §9.5 makes that
+        // observable: env values "parse as JSON, falling back to
+        // string", so anything this reader accepts loosely silently
+        // becomes a number where the canonical keeps the authored
+        // string. `1e`, `1.`, `01`, `.5` and a bare `-` are all errors
+        // to JSON.parse; only `-0` and `1e5` are not.
+        //
+        //   number = [ '-' ] int [ frac ] [ exp ]
+        //   int    = '0' | digit1-9 *digit
+        //   frac   = '.' 1*digit
+        //   exp    = ('e'|'E') [ '+' | '-' ] 1*digit
+        const start = p.i;
+        if (p.i < p.text.len and p.text[p.i] == '-') p.i += 1;
+        if (!p.digit()) {
+            p.err = "bad number";
+            return null;
+        }
+        if (p.text[p.i] == '0') p.i += 1 else while (p.digit()) p.i += 1;
+        if (p.i < p.text.len and p.text[p.i] == '.') {
+            p.i += 1;
+            if (!p.digit()) {
+                p.err = "bad number";
+                return null;
+            }
+            while (p.digit()) p.i += 1;
+        }
+        if (p.i < p.text.len and (p.text[p.i] == 'e' or p.text[p.i] == 'E')) {
+            p.i += 1;
+            if (p.i < p.text.len and (p.text[p.i] == '+' or p.text[p.i] == '-')) p.i += 1;
+            if (!p.digit()) {
+                p.err = "bad number";
+                return null;
+            }
+            while (p.digit()) p.i += 1;
+        }
+        if (start == p.i) {
+            p.err = "unexpected character";
+            return null;
+        }
+        // A REPORTED failure, not a crash: a bare `-` or a truncated
+        // `1e` reaches here from `env`'s parse-or-string fallback (§9.5),
+        // and a reader that panics on a bad number rather than reporting
+        // one turns "this env value is a string" into a crash. The ocaml
+        // port hit this first.
+        const n = std.fmt.parseFloat(f64, p.text[start..p.i]) catch {
+            p.err = "bad number";
+            return null;
+        };
+        return vnum(n);
+    }
+};
+
+/// Parse, or null with `err` set to a message.
+pub fn parse(text: []const u8, err: *[]const u8) ?*Value {
+    var p = Parser{ .text = text };
+    const out = p.value() orelse {
+        err.* = p.err;
+        return null;
+    };
+    p.skip();
+    if (p.i != p.text.len) {
+        err.* = "trailing content";
+        return null;
+    }
+    return out;
+}