zig/lib/std / json/Stringify.zig

Writes JSON ([RFC8259](https://tools.ietf.org/html/rfc8259)) formatted data to a stream. The sequence of method calls to write JSON content must follow this grammar: = = | | | write | print | = beginObject ( )* endObject = objectField | objectFieldRaw | = beginArray ( )* endArray = beginWriteRaw ( stream.writeAll )* endWriteRaw = beginObjectFieldRaw ( stream.writeAll )* endObjectFieldRaw

//! Writes JSON ([RFC8259](https://tools.ietf.org/html/rfc8259)) formatted data
//! to a stream.
//!
//! The sequence of method calls to write JSON content must follow this grammar:
//! ```
//!  <once> = <value>
//!  <value> =
//!    | <object>
//!    | <array>
//!    | write
//!    | print
//!    | <writeRawStream>
//!  <object> = beginObject ( <field> <value> )* endObject
//!  <field> = objectField | objectFieldRaw | <objectFieldRawStream>
//!  <array> = beginArray ( <value> )* endArray
//!  <writeRawStream> = beginWriteRaw ( stream.writeAll )* endWriteRaw
//!  <objectFieldRawStream> = beginObjectFieldRaw ( stream.writeAll )* endObjectFieldRaw
//! ```

Error

The safety_checks_hint parameter determines how much memory is used to enable assertions that the above grammar is being followed, e.g. tripping an assertion rather than allowing endObject to emit the final } in [[[]]}. "Depth" in this context means the depth of nested [] or {} expressions (or equivalently the amount of recursion on the grammar expression above). For example, emitting the JSON [[[]]] requires a depth of 3. If .checked_to_fixed_depth is used, there is additionally an assertion that the nesting depth never exceeds the given limit. .checked_to_fixed_depth embeds the storage required in the Stringify struct. .assumed_correct requires no space and performs none of these assertions. In ReleaseFast and ReleaseSmall mode, the given safety_checks_hint is ignored and is always treated as .assumed_correct.


const std = @import("../std.zig");
const assert = std.debug.assert;
const Allocator = std.mem.Allocator;
const ArrayList = std.ArrayList;
const BitStack = std.BitStack;
const Stringify = @This();
const Writer = std.io.Writer;

beginArray()

Rounded up to the nearest multiple of 8.


const IndentationMode = enum(u1) {
    object = 0,
    array = 1,
};

beginObject()

An alternative to calling write that formats a value with std.fmt. This function does the usual punctuation and indentation formatting assuming the resulting formatted string represents a single complete value; e.g. "1", "[]", "[1,2]", not "1,2". This function may be useful for doing your own number formatting.


writer: *Writer,
options: Options = .{},
indent_level: usize = 0,
next_punctuation: enum {
    the_beginning,
    none,
    comma,
    colon,
} = .the_beginning,

endArray()

An alternative to calling write that allows you to write directly to the .writer field, e.g. with .writer.writeAll(). Call beginWriteRaw(), then write a complete value (including any quotes if necessary) directly to the .writer field, then call endWriteRaw(). This can be useful for streaming very long strings into the output without needing it all buffered in memory.


nesting_stack: switch (safety_checks) {
    .checked_to_fixed_depth => |fixed_buffer_size| [(fixed_buffer_size + 7) >> 3]u8,
    .assumed_correct => void,
} = switch (safety_checks) {
    .checked_to_fixed_depth => @splat(0),
    .assumed_correct => {},
},

endObject()

See beginWriteRaw.


raw_streaming_mode: if (build_mode_has_safety)
    enum { none, value, objectField }
else
    void = if (build_mode_has_safety) .none else {},

print()

See Stringify for when to call this method. key is the string content of the property name. Surrounding quotes will be added and any special characters will be escaped. See also objectFieldRaw.


const build_mode_has_safety = switch (@import("builtin").mode) {
    .Debug, .ReleaseSafe => true,
    .ReleaseFast, .ReleaseSmall => false,
};

Test: print

See Stringify for when to call this method. quoted_key is the complete bytes of the key including quotes and any necessary escape sequences. A few assertions are performed on the given value to ensure that the caller of this function understands the API contract. See also objectField.


/// The `safety_checks_hint` parameter determines how much memory is used to enable assertions that the above grammar is being followed,
/// e.g. tripping an assertion rather than allowing `endObject` to emit the final `}` in `[[[]]}`.
/// "Depth" in this context means the depth of nested `[]` or `{}` expressions
/// (or equivalently the amount of recursion on the `<value>` grammar expression above).
/// For example, emitting the JSON `[[[]]]` requires a depth of 3.
/// If `.checked_to_fixed_depth` is used, there is additionally an assertion that the nesting depth never exceeds the given limit.
/// `.checked_to_fixed_depth` embeds the storage required in the `Stringify` struct.
/// `.assumed_correct` requires no space and performs none of these assertions.
/// In `ReleaseFast` and `ReleaseSmall` mode, the given `safety_checks_hint` is ignored and is always treated as `.assumed_correct`.
const safety_checks_hint: union(enum) {
    /// Rounded up to the nearest multiple of 8.
    checked_to_fixed_depth: usize,
    assumed_correct,
} = .{ .checked_to_fixed_depth = 256 };

beginWriteRaw()

In the rare case that you need to write very long object field names, this is an alternative to objectField and objectFieldRaw that allows you to write directly to the .writer field similar to beginWriteRaw. Call endObjectFieldRaw() when you're done.


const safety_checks: @TypeOf(safety_checks_hint) = if (build_mode_has_safety)
    safety_checks_hint
else
    .assumed_correct;

endWriteRaw()

See beginObjectFieldRaw.


pub const Error = Writer.Error;

objectField()

Renders the given Zig value as JSON. Supported types: * Zig bool -> JSON true or false. * Zig ?T -> null or the rendering of T. * Zig i32, u64, etc. -> JSON number or string. * When option emit_nonportable_numbers_as_strings is true, if the value is outside the range +-1<<53 (the precise integer range of f64), it is rendered as a JSON string in base 10. Otherwise, it is rendered as JSON number. * Zig floats -> JSON number or string. * If the value cannot be precisely represented by an f64, it is rendered as a JSON string. Otherwise, it is rendered as JSON number. * TODO: Float rendering will likely change in the future, e.g. to remove the unnecessary "e+00". * Zig []const u8, []u8, *[N]u8, @Vector(N, u8), and similar -> JSON string. * See Options.emit_strings_as_arrays. * If the content is not valid UTF-8, rendered as an array of numbers instead. * Zig []T, [N]T, *[N]T, @Vector(N, T), and similar -> JSON array of the rendering of each item. * Zig tuple -> JSON array of the rendering of each item. * Zig struct -> JSON object with each field in declaration order. * If the struct declares a method pub fn jsonStringify(self: *@This(), jw: anytype) !void, it is called to do the serialization instead of the default behavior. The given jw is a pointer to this Stringify. See std.json.Value for an example. * See Options.emit_null_optional_fields. * Zig union(enum) -> JSON object with one field named for the active tag and a value representing the payload. * If the payload is void, then the emitted value is {}. * If the union declares a method pub fn jsonStringify(self: *@This(), jw: anytype) !void, it is called to do the serialization instead of the default behavior. The given jw is a pointer to this Stringify. * Zig enum -> JSON string naming the active tag. * If the enum declares a method pub fn jsonStringify(self: *@This(), jw: anytype) !void, it is called to do the serialization instead of the default behavior. The given jw is a pointer to this Stringify. * If the enum is non-exhaustive, unnamed values are rendered as integers. * Zig untyped enum literal -> JSON string naming the active tag. * Zig error -> JSON string naming the error. * Zig *T -> the rendering of T. Note there is no guard against circular-reference infinite recursion. See also alternative functions print and beginWriteRaw. For writing object field names, use objectField instead.


pub fn beginArray(self: *Stringify) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    try self.valueStart();
    try self.writer.writeByte('[');
    try self.pushIndentation(.array);
    self.next_punctuation = .none;

valueAlloc()

Controls the whitespace emitted. The default .minified is a compact encoding with no whitespace between tokens. Any setting other than .minified will use newlines, indentation, and a space after each ':'. .indent_1 means 1 space for each indentation level, .indent_2 means 2 spaces, etc. .indent_tab uses a tab for each indentation level.

}

beginObjectFieldRaw()

Should optional fields with null value be written?


pub fn beginObject(self: *Stringify) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    try self.valueStart();
    try self.writer.writeByte('{');
    try self.pushIndentation(.object);
    self.next_punctuation = .none;

valueAlloc()

Arrays/slices of u8 are typically encoded as JSON strings. This option emits them as arrays of numbers instead. Does not affect calls to objectField*().

}

write()

Should unicode characters be escaped in strings?


pub fn endArray(self: *Stringify) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    self.popIndentation(.array);
    switch (self.next_punctuation) {
        .none => {},
        .comma => {
            try self.indent();
        },
        .the_beginning, .colon => unreachable,
    }
    try self.writer.writeByte(']');
    self.valueDone();

valueAlloc()

When true, renders numbers outside the range +-1<<53 (the precise integer range of f64) as JSON strings in base 10.

}

value()

Writes the given value to the Writer writer. See Stringify for how the given value is serialized into JSON. The maximum nesting depth of the output JSON document is 256.


pub fn endObject(self: *Stringify) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    self.popIndentation(.object);
    switch (self.next_punctuation) {
        .none => {},
        .comma => {
            try self.indent();
        },
        .the_beginning, .colon => unreachable,
    }
    try self.writer.writeByte('}');
    self.valueDone();

valueAlloc()

Calls value and stores the result in dynamically allocated memory instead of taking a writer. Caller owns returned memory.

}

valueAlloc()

Write string to writer as a JSON encoded string.


fn pushIndentation(self: *Stringify, mode: IndentationMode) !void {
    switch (safety_checks) {
        .checked_to_fixed_depth => {
            BitStack.pushWithStateAssumeCapacity(&self.nesting_stack, &self.indent_level, @intFromEnum(mode));
        },
        .assumed_correct => {
            self.indent_level += 1;
        },
    }
}
fn popIndentation(self: *Stringify, expected_mode: IndentationMode) void {
    switch (safety_checks) {
        .checked_to_fixed_depth => {
            assert(BitStack.popWithState(&self.nesting_stack, &self.indent_level) == @intFromEnum(expected_mode));
        },
        .assumed_correct => {
            self.indent_level -= 1;
        },
    }
}

Test: valueAlloc

Write chars to writer as JSON encoded string characters.


fn indent(self: *Stringify) !void {
    var char: u8 = ' ';
    const n_chars = switch (self.options.whitespace) {
        .minified => return,
        .indent_1 => 1 * self.indent_level,
        .indent_2 => 2 * self.indent_level,
        .indent_3 => 3 * self.indent_level,
        .indent_4 => 4 * self.indent_level,
        .indent_8 => 8 * self.indent_level,
        .indent_tab => blk: {
            char = '\t';
            break :blk self.indent_level;
        },
    };
    try self.writer.writeByte('\n');
    try self.writer.splatByteAll(char, n_chars);
}

encodeJsonString()


fn valueStart(self: *Stringify) !void {
    if (self.isObjectKeyExpected()) |is_it| assert(!is_it); // Call objectField*(), not write(), for object keys.
    return self.valueStartAssumeTypeOk();
}
fn objectFieldStart(self: *Stringify) !void {
    if (self.isObjectKeyExpected()) |is_it| assert(is_it); // Expected write(), not objectField*().
    return self.valueStartAssumeTypeOk();
}
fn valueStartAssumeTypeOk(self: *Stringify) !void {
    assert(!self.isComplete()); // JSON document already complete.
    switch (self.next_punctuation) {
        .the_beginning => {
            // No indentation for the very beginning.
        },
        .none => {
            // First item in a container.
            try self.indent();
        },
        .comma => {
            // Subsequent item in a container.
            try self.writer.writeByte(',');
            try self.indent();
        },
        .colon => {
            try self.writer.writeByte(':');
            if (self.options.whitespace != .minified) {
                try self.writer.writeByte(' ');
            }
        },
    }
}
fn valueDone(self: *Stringify) void {
    self.next_punctuation = .comma;
}

encodeJsonStringChars()


// Only when safety is enabled:
fn isObjectKeyExpected(self: *const Stringify) ?bool {
    switch (safety_checks) {
        .checked_to_fixed_depth => return self.indent_level > 0 and
            BitStack.peekWithState(&self.nesting_stack, self.indent_level) == @intFromEnum(IndentationMode.object) and
            self.next_punctuation != .colon,
        .assumed_correct => return null,
    }
}
fn isComplete(self: *const Stringify) bool {
    return self.indent_level == 0 and self.next_punctuation == .comma;
}

Test:

json write stream


/// An alternative to calling `write` that formats a value with `std.fmt`.
/// This function does the usual punctuation and indentation formatting
/// assuming the resulting formatted string represents a single complete value;
/// e.g. `"1"`, `"[]"`, `"[1,2]"`, not `"1,2"`.
/// This function may be useful for doing your own number formatting.
pub fn print(self: *Stringify, comptime fmt: []const u8, args: anytype) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    try self.valueStart();
    try self.writer.print(fmt, args);
    self.valueDone();
}

Test:

stringify null optional fields


test print {
    var out_buf: [1024]u8 = undefined;
    var out: Writer = .fixed(&out_buf);

Test:

stringify basic types


    var w: Stringify = .{ .writer = &out, .options = .{ .whitespace = .indent_2 } };

Test:

stringify string


    try w.beginObject();
    try w.objectField("a");
    try w.print("[  ]", .{});
    try w.objectField("b");
    try w.beginArray();
    try w.print("[{s}] ", .{"[]"});
    try w.print("  {}", .{12345});
    try w.endArray();
    try w.endObject();

Test:

stringify many-item sentinel-terminated string


    const expected =
        \\{
        \\  "a": [  ],
        \\  "b": [
        \\    [[]] ,
        \\      12345
        \\  ]
        \\}
    ;
    try std.testing.expectEqualStrings(expected, out.buffered());
}

Test:

stringify enums


/// An alternative to calling `write` that allows you to write directly to the `.writer` field, e.g. with `.writer.writeAll()`.
/// Call `beginWriteRaw()`, then write a complete value (including any quotes if necessary) directly to the `.writer` field,
/// then call `endWriteRaw()`.
/// This can be useful for streaming very long strings into the output without needing it all buffered in memory.
pub fn beginWriteRaw(self: *Stringify) !void {
    if (build_mode_has_safety) {
        assert(self.raw_streaming_mode == .none);
        self.raw_streaming_mode = .value;
    }
    try self.valueStart();
}

Test:

stringify non-exhaustive enum


/// See `beginWriteRaw`.
pub fn endWriteRaw(self: *Stringify) void {
    if (build_mode_has_safety) {
        assert(self.raw_streaming_mode == .value);
        self.raw_streaming_mode = .none;
    }
    self.valueDone();
}

Test:

stringify enum literals


/// See `Stringify` for when to call this method.
/// `key` is the string content of the property name.
/// Surrounding quotes will be added and any special characters will be escaped.
/// See also `objectFieldRaw`.
pub fn objectField(self: *Stringify, key: []const u8) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    try self.objectFieldStart();
    try encodeJsonString(key, self.options, self.writer);
    self.next_punctuation = .colon;
}
/// See `Stringify` for when to call this method.
/// `quoted_key` is the complete bytes of the key including quotes and any necessary escape sequences.
/// A few assertions are performed on the given value to ensure that the caller of this function understands the API contract.
/// See also `objectField`.
pub fn objectFieldRaw(self: *Stringify, quoted_key: []const u8) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    assert(quoted_key.len >= 2 and quoted_key[0] == '"' and quoted_key[quoted_key.len - 1] == '"'); // quoted_key should be "quoted".
    try self.objectFieldStart();
    try self.writer.writeAll(quoted_key);
    self.next_punctuation = .colon;
}

Test:

stringify tagged unions


/// In the rare case that you need to write very long object field names,
/// this is an alternative to `objectField` and `objectFieldRaw` that allows you to write directly to the `.writer` field
/// similar to `beginWriteRaw`.
/// Call `endObjectFieldRaw()` when you're done.
pub fn beginObjectFieldRaw(self: *Stringify) !void {
    if (build_mode_has_safety) {
        assert(self.raw_streaming_mode == .none);
        self.raw_streaming_mode = .objectField;
    }
    try self.objectFieldStart();
}

Test:

stringify struct


/// See `beginObjectFieldRaw`.
pub fn endObjectFieldRaw(self: *Stringify) void {
    if (build_mode_has_safety) {
        assert(self.raw_streaming_mode == .objectField);
        self.raw_streaming_mode = .none;
    }
    self.next_punctuation = .colon;
}

Test:

emit_strings_as_arrays


/// Renders the given Zig value as JSON.
///
/// Supported types:
///  * Zig `bool` -> JSON `true` or `false`.
///  * Zig `?T` -> `null` or the rendering of `T`.
///  * Zig `i32`, `u64`, etc. -> JSON number or string.
///      * When option `emit_nonportable_numbers_as_strings` is true, if the value is outside the range `+-1<<53` (the precise integer range of f64), it is rendered as a JSON string in base 10. Otherwise, it is rendered as JSON number.
///  * Zig floats -> JSON number or string.
///      * If the value cannot be precisely represented by an f64, it is rendered as a JSON string. Otherwise, it is rendered as JSON number.
///      * TODO: Float rendering will likely change in the future, e.g. to remove the unnecessary "e+00".
///  * Zig `[]const u8`, `[]u8`, `*[N]u8`, `@Vector(N, u8)`, and similar -> JSON string.
///      * See `Options.emit_strings_as_arrays`.
///      * If the content is not valid UTF-8, rendered as an array of numbers instead.
///  * Zig `[]T`, `[N]T`, `*[N]T`, `@Vector(N, T)`, and similar -> JSON array of the rendering of each item.
///  * Zig tuple -> JSON array of the rendering of each item.
///  * Zig `struct` -> JSON object with each field in declaration order.
///      * If the struct declares a method `pub fn jsonStringify(self: *@This(), jw: anytype) !void`, it is called to do the serialization instead of the default behavior. The given `jw` is a pointer to this `Stringify`. See `std.json.Value` for an example.
///      * See `Options.emit_null_optional_fields`.
///  * Zig `union(enum)` -> JSON object with one field named for the active tag and a value representing the payload.
///      * If the payload is `void`, then the emitted value is `{}`.
///      * If the union declares a method `pub fn jsonStringify(self: *@This(), jw: anytype) !void`, it is called to do the serialization instead of the default behavior. The given `jw` is a pointer to this `Stringify`.
///  * Zig `enum` -> JSON string naming the active tag.
///      * If the enum declares a method `pub fn jsonStringify(self: *@This(), jw: anytype) !void`, it is called to do the serialization instead of the default behavior. The given `jw` is a pointer to this `Stringify`.
///      * If the enum is non-exhaustive, unnamed values are rendered as integers.
///  * Zig untyped enum literal -> JSON string naming the active tag.
///  * Zig error -> JSON string naming the error.
///  * Zig `*T` -> the rendering of `T`. Note there is no guard against circular-reference infinite recursion.
///
/// See also alternative functions `print` and `beginWriteRaw`.
/// For writing object field names, use `objectField` instead.
pub fn write(self: *Stringify, v: anytype) Error!void {
    if (build_mode_has_safety) assert(self.raw_streaming_mode == .none);
    const T = @TypeOf(v);
    switch (@typeInfo(T)) {
        .int => {
            try self.valueStart();
            if (self.options.emit_nonportable_numbers_as_strings and
                (v <= -(1 << 53) or v >= (1 << 53)))
            {
                try self.writer.print("\"{}\"", .{v});
            } else {
                try self.writer.print("{}", .{v});
            }
            self.valueDone();
            return;
        },
        .comptime_int => {
            return self.write(@as(std.math.IntFittingRange(v, v), v));
        },
        .float, .comptime_float => {
            if (@as(f64, @floatCast(v)) == v) {
                try self.valueStart();
                try self.writer.print("{}", .{@as(f64, @floatCast(v))});
                self.valueDone();
                return;
            }
            try self.valueStart();
            try self.writer.print("\"{}\"", .{v});
            self.valueDone();
            return;
        },

Test:

stringify struct with indentation


        .bool => {
            try self.valueStart();
            try self.writer.writeAll(if (v) "true" else "false");
            self.valueDone();
            return;
        },
        .null => {
            try self.valueStart();
            try self.writer.writeAll("null");
            self.valueDone();
            return;
        },
        .optional => {
            if (v) |payload| {
                return try self.write(payload);
            } else {
                return try self.write(null);
            }
        },
        .@"enum" => |enum_info| {
            if (std.meta.hasFn(T, "jsonStringify")) {
                return v.jsonStringify(self);
            }

Test:

stringify array of structs


            if (!enum_info.is_exhaustive) {
                inline for (enum_info.fields) |field| {
                    if (v == @field(T, field.name)) {
                        break;
                    }
                } else {
                    return self.write(@intFromEnum(v));
                }
            }

Test:

stringify struct with custom stringifier


            return self.stringValue(@tagName(v));
        },
        .enum_literal => {
            return self.stringValue(@tagName(v));
        },
        .@"union" => {
            if (std.meta.hasFn(T, "jsonStringify")) {
                return v.jsonStringify(self);
            }

jsonStringify()


            const info = @typeInfo(T).@"union";
            if (info.tag_type) |UnionTagType| {
                try self.beginObject();
                inline for (info.fields) |u_field| {
                    if (v == @field(UnionTagType, u_field.name)) {
                        try self.objectField(u_field.name);
                        if (u_field.type == void) {
                            // void v is {}
                            try self.beginObject();
                            try self.endObject();
                        } else {
                            try self.write(@field(v, u_field.name));
                        }
                        break;
                    }
                } else {
                    unreachable; // No active tag?
                }
                try self.endObject();
                return;
            } else {
                @compileError("Unable to stringify untagged union '" ++ @typeName(T) ++ "'");
            }
        },
        .@"struct" => |S| {
            if (std.meta.hasFn(T, "jsonStringify")) {
                return v.jsonStringify(self);
            }

Test:

raw streaming


            if (S.is_tuple) {
                try self.beginArray();
            } else {
                try self.beginObject();
            }
            inline for (S.fields) |Field| {
                // don't include void fields
                if (Field.type == void) continue;

                var emit_field = true;

                // don't include optional fields that are null when emit_null_optional_fields is set to false
                if (@typeInfo(Field.type) == .optional) {
                    if (self.options.emit_null_optional_fields == false) {
                        if (@field(v, Field.name) == null) {
                            emit_field = false;
                        }
                    }
                }

                if (emit_field) {
                    if (!S.is_tuple) {
                        try self.objectField(Field.name);
                    }
                    try self.write(@field(v, Field.name));
                }
            }
            if (S.is_tuple) {
                try self.endArray();
            } else {
                try self.endObject();
            }
            return;
        },
        .error_set => return self.stringValue(@errorName(v)),
        .pointer => |ptr_info| switch (ptr_info.size) {
            .one => switch (@typeInfo(ptr_info.child)) {
                .array => {
                    // Coerce `*[N]T` to `[]const T`.
                    const Slice = []const std.meta.Elem(ptr_info.child);
                    return self.write(@as(Slice, v));
                },
                else => {
                    return self.write(v.*);
                },
            },
            .many, .slice => {
                if (ptr_info.size == .many and ptr_info.sentinel() == null)
                    @compileError("unable to stringify type '" ++ @typeName(T) ++ "' without sentinel");
                const slice = if (ptr_info.size == .many) std.mem.span(v) else v;

                if (ptr_info.child == u8) {
                    // This is a []const u8, or some similar Zig string.
                    if (!self.options.emit_strings_as_arrays and std.unicode.utf8ValidateSlice(slice)) {
                        return self.stringValue(slice);
                    }
                }

                try self.beginArray();
                for (slice) |x| {
                    try self.write(x);
                }
                try self.endArray();
                return;
            },
            else => @compileError("Unable to stringify type '" ++ @typeName(T) ++ "'"),
        },
        .array => {
            // Coerce `[N]T` to `*const [N]T` (and then to `[]const T`).
            return self.write(&v);
        },
        .vector => |info| {
            const array: [info.len]info.child = v;
            return self.write(&array);
        },
        else => @compileError("Unable to stringify type '" ++ @typeName(T) ++ "'"),
    }
    unreachable;
}

fn stringValue(self: *Stringify, s: []const u8) !void {
    try self.valueStart();
    try encodeJsonString(s, self.options, self.writer);
    self.valueDone();
}

pub const Options = struct {
    /// Controls the whitespace emitted.
    /// The default `.minified` is a compact encoding with no whitespace between tokens.
    /// Any setting other than `.minified` will use newlines, indentation, and a space after each ':'.
    /// `.indent_1` means 1 space for each indentation level, `.indent_2` means 2 spaces, etc.
    /// `.indent_tab` uses a tab for each indentation level.
    whitespace: enum {
        minified,
        indent_1,
        indent_2,
        indent_3,
        indent_4,
        indent_8,
        indent_tab,
    } = .minified,

    /// Should optional fields with null value be written?
    emit_null_optional_fields: bool = true,

    /// Arrays/slices of u8 are typically encoded as JSON strings.
    /// This option emits them as arrays of numbers instead.
    /// Does not affect calls to `objectField*()`.
    emit_strings_as_arrays: bool = false,

    /// Should unicode characters be escaped in strings?
    escape_unicode: bool = false,

    /// When true, renders numbers outside the range `+-1<<53` (the precise integer range of f64) as JSON strings in base 10.
    emit_nonportable_numbers_as_strings: bool = false,
};

/// Writes the given value to the `Writer` writer.
/// See `Stringify` for how the given value is serialized into JSON.
/// The maximum nesting depth of the output JSON document is 256.
pub fn value(v: anytype, options: Options, writer: *Writer) Error!void {
    var s: Stringify = .{ .writer = writer, .options = options };
    try s.write(v);
}

test value {
    var out: std.io.Writer.Allocating = .init(std.testing.allocator);
    const writer = &out.writer;
    defer out.deinit();

    const T = struct { a: i32, b: []const u8 };
    try value(T{ .a = 123, .b = "xy" }, .{}, writer);
    try std.testing.expectEqualSlices(u8, "{\"a\":123,\"b\":\"xy\"}", out.getWritten());

    try testStringify("9999999999999999", 9999999999999999, .{});
    try testStringify("\"9999999999999999\"", 9999999999999999, .{ .emit_nonportable_numbers_as_strings = true });

    try testStringify("[1,1]", @as(@Vector(2, u32), @splat(1)), .{});
    try testStringify("\"AA\"", @as(@Vector(2, u8), @splat('A')), .{});
    try testStringify("[65,65]", @as(@Vector(2, u8), @splat('A')), .{ .emit_strings_as_arrays = true });

    // void field
    try testStringify("{\"foo\":42}", struct {
        foo: u32,
        bar: void = {},
    }{ .foo = 42 }, .{});

    const Tuple = struct { []const u8, usize };
    try testStringify("[\"foo\",42]", Tuple{ "foo", 42 }, .{});

    comptime {
        testStringify("false", false, .{}) catch unreachable;
        const MyStruct = struct { foo: u32 };
        testStringify("[{\"foo\":42},{\"foo\":100},{\"foo\":1000}]", [_]MyStruct{
            MyStruct{ .foo = 42 },
            MyStruct{ .foo = 100 },
            MyStruct{ .foo = 1000 },
        }, .{}) catch unreachable;
    }
}

/// Calls `value` and stores the result in dynamically allocated memory instead
/// of taking a writer.
///
/// Caller owns returned memory.
pub fn valueAlloc(gpa: Allocator, v: anytype, options: Options) error{OutOfMemory}![]u8 {
    var aw: std.io.Writer.Allocating = .init(gpa);
    defer aw.deinit();
    value(v, options, &aw.writer) catch return error.OutOfMemory;
    return aw.toOwnedSlice();
}

test valueAlloc {
    const allocator = std.testing.allocator;
    const expected =
        \\{"foo":"bar","answer":42,"my_friend":"sammy"}
    ;
    const actual = try valueAlloc(allocator, .{ .foo = "bar", .answer = 42, .my_friend = "sammy" }, .{});
    defer allocator.free(actual);

    try std.testing.expectEqualStrings(expected, actual);
}

fn outputUnicodeEscape(codepoint: u21, w: *Writer) Error!void {
    if (codepoint <= 0xFFFF) {
        // If the character is in the Basic Multilingual Plane (U+0000 through U+FFFF),
        // then it may be represented as a six-character sequence: a reverse solidus, followed
        // by the lowercase letter u, followed by four hexadecimal digits that encode the character's code point.
        try w.writeAll("\\u");
        try w.printInt(codepoint, 16, .lower, .{ .width = 4, .fill = '0' });
    } else {
        assert(codepoint <= 0x10FFFF);
        // To escape an extended character that is not in the Basic Multilingual Plane,
        // the character is represented as a 12-character sequence, encoding the UTF-16 surrogate pair.
        const high = @as(u16, @intCast((codepoint - 0x10000) >> 10)) + 0xD800;
        const low = @as(u16, @intCast(codepoint & 0x3FF)) + 0xDC00;
        try w.writeAll("\\u");
        try w.printInt(high, 16, .lower, .{ .width = 4, .fill = '0' });
        try w.writeAll("\\u");
        try w.printInt(low, 16, .lower, .{ .width = 4, .fill = '0' });
    }
}

fn outputSpecialEscape(c: u8, writer: *Writer) Error!void {
    switch (c) {
        '\\' => try writer.writeAll("\\\\"),
        '\"' => try writer.writeAll("\\\""),
        0x08 => try writer.writeAll("\\b"),
        0x0C => try writer.writeAll("\\f"),
        '\n' => try writer.writeAll("\\n"),
        '\r' => try writer.writeAll("\\r"),
        '\t' => try writer.writeAll("\\t"),
        else => try outputUnicodeEscape(c, writer),
    }
}

/// Write `string` to `writer` as a JSON encoded string.
pub fn encodeJsonString(string: []const u8, options: Options, writer: *Writer) Error!void {
    try writer.writeByte('\"');
    try encodeJsonStringChars(string, options, writer);
    try writer.writeByte('\"');
}

/// Write `chars` to `writer` as JSON encoded string characters.
pub fn encodeJsonStringChars(chars: []const u8, options: Options, writer: *Writer) Error!void {
    var write_cursor: usize = 0;
    var i: usize = 0;
    if (options.escape_unicode) {
        while (i < chars.len) : (i += 1) {
            switch (chars[i]) {
                // normal ascii character
                0x20...0x21, 0x23...0x5B, 0x5D...0x7E => {},
                0x00...0x1F, '\\', '\"' => {
                    // Always must escape these.
                    try writer.writeAll(chars[write_cursor..i]);
                    try outputSpecialEscape(chars[i], writer);
                    write_cursor = i + 1;
                },
                0x7F...0xFF => {
                    try writer.writeAll(chars[write_cursor..i]);
                    const ulen = std.unicode.utf8ByteSequenceLength(chars[i]) catch unreachable;
                    const codepoint = std.unicode.utf8Decode(chars[i..][0..ulen]) catch unreachable;
                    try outputUnicodeEscape(codepoint, writer);
                    i += ulen - 1;
                    write_cursor = i + 1;
                },
            }
        }
    } else {
        while (i < chars.len) : (i += 1) {
            switch (chars[i]) {
                // normal bytes
                0x20...0x21, 0x23...0x5B, 0x5D...0xFF => {},
                0x00...0x1F, '\\', '\"' => {
                    // Always must escape these.
                    try writer.writeAll(chars[write_cursor..i]);
                    try outputSpecialEscape(chars[i], writer);
                    write_cursor = i + 1;
                },
            }
        }
    }
    try writer.writeAll(chars[write_cursor..chars.len]);
}

test "json write stream" {
    var out_buf: [1024]u8 = undefined;
    var out: Writer = .fixed(&out_buf);
    var w: Stringify = .{ .writer = &out, .options = .{ .whitespace = .indent_2 } };
    try testBasicWriteStream(&w);
}

fn testBasicWriteStream(w: *Stringify) !void {
    w.writer.end = 0;

    try w.beginObject();

    try w.objectField("object");
    var arena_allocator = std.heap.ArenaAllocator.init(std.testing.allocator);
    defer arena_allocator.deinit();
    try w.write(try getJsonObject(arena_allocator.allocator()));

    try w.objectFieldRaw("\"string\"");
    try w.write("This is a string");

    try w.objectField("array");
    try w.beginArray();
    try w.write("Another string");
    try w.write(@as(i32, 1));
    try w.write(@as(f32, 3.5));
    try w.endArray();

    try w.objectField("int");
    try w.write(@as(i32, 10));

    try w.objectField("float");
    try w.write(@as(f32, 3.5));

    try w.endObject();

    const expected =
        \\{
        \\  "object": {
        \\    "one": 1,
        \\    "two": 2
        \\  },
        \\  "string": "This is a string",
        \\  "array": [
        \\    "Another string",
        \\    1,
        \\    3.5
        \\  ],
        \\  "int": 10,
        \\  "float": 3.5
        \\}
    ;
    try std.testing.expectEqualStrings(expected, w.writer.buffered());
}

fn getJsonObject(allocator: std.mem.Allocator) !std.json.Value {
    var v: std.json.Value = .{ .object = std.json.ObjectMap.init(allocator) };
    try v.object.put("one", std.json.Value{ .integer = @as(i64, @intCast(1)) });
    try v.object.put("two", std.json.Value{ .float = 2.0 });
    return v;
}

test "stringify null optional fields" {
    const MyStruct = struct {
        optional: ?[]const u8 = null,
        required: []const u8 = "something",
        another_optional: ?[]const u8 = null,
        another_required: []const u8 = "something else",
    };
    try testStringify(
        \\{"optional":null,"required":"something","another_optional":null,"another_required":"something else"}
    ,
        MyStruct{},
        .{},
    );
    try testStringify(
        \\{"required":"something","another_required":"something else"}
    ,
        MyStruct{},
        .{ .emit_null_optional_fields = false },
    );
}

test "stringify basic types" {
    try testStringify("false", false, .{});
    try testStringify("true", true, .{});
    try testStringify("null", @as(?u8, null), .{});
    try testStringify("null", @as(?*u32, null), .{});
    try testStringify("42", 42, .{});
    try testStringify("42", 42.0, .{});
    try testStringify("42", @as(u8, 42), .{});
    try testStringify("42", @as(u128, 42), .{});
    try testStringify("9999999999999999", 9999999999999999, .{});
    try testStringify("42", @as(f32, 42), .{});
    try testStringify("42", @as(f64, 42), .{});
    try testStringify("\"ItBroke\"", @as(anyerror, error.ItBroke), .{});
    try testStringify("\"ItBroke\"", error.ItBroke, .{});
}

test "stringify string" {
    try testStringify("\"hello\"", "hello", .{});
    try testStringify("\"with\\nescapes\\r\"", "with\nescapes\r", .{});
    try testStringify("\"with\\nescapes\\r\"", "with\nescapes\r", .{ .escape_unicode = true });
    try testStringify("\"with unicode\\u0001\"", "with unicode\u{1}", .{});
    try testStringify("\"with unicode\\u0001\"", "with unicode\u{1}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{80}\"", "with unicode\u{80}", .{});
    try testStringify("\"with unicode\\u0080\"", "with unicode\u{80}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{FF}\"", "with unicode\u{FF}", .{});
    try testStringify("\"with unicode\\u00ff\"", "with unicode\u{FF}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{100}\"", "with unicode\u{100}", .{});
    try testStringify("\"with unicode\\u0100\"", "with unicode\u{100}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{800}\"", "with unicode\u{800}", .{});
    try testStringify("\"with unicode\\u0800\"", "with unicode\u{800}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{8000}\"", "with unicode\u{8000}", .{});
    try testStringify("\"with unicode\\u8000\"", "with unicode\u{8000}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{D799}\"", "with unicode\u{D799}", .{});
    try testStringify("\"with unicode\\ud799\"", "with unicode\u{D799}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{10000}\"", "with unicode\u{10000}", .{});
    try testStringify("\"with unicode\\ud800\\udc00\"", "with unicode\u{10000}", .{ .escape_unicode = true });
    try testStringify("\"with unicode\u{10FFFF}\"", "with unicode\u{10FFFF}", .{});
    try testStringify("\"with unicode\\udbff\\udfff\"", "with unicode\u{10FFFF}", .{ .escape_unicode = true });
}

test "stringify many-item sentinel-terminated string" {
    try testStringify("\"hello\"", @as([*:0]const u8, "hello"), .{});
    try testStringify("\"with\\nescapes\\r\"", @as([*:0]const u8, "with\nescapes\r"), .{ .escape_unicode = true });
    try testStringify("\"with unicode\\u0001\"", @as([*:0]const u8, "with unicode\u{1}"), .{ .escape_unicode = true });
}

test "stringify enums" {
    const E = enum {
        foo,
        bar,
    };
    try testStringify("\"foo\"", E.foo, .{});
    try testStringify("\"bar\"", E.bar, .{});
}

test "stringify non-exhaustive enum" {
    const E = enum(u8) {
        foo = 0,
        _,
    };
    try testStringify("\"foo\"", E.foo, .{});
    try testStringify("1", @as(E, @enumFromInt(1)), .{});
}

test "stringify enum literals" {
    try testStringify("\"foo\"", .foo, .{});
    try testStringify("\"bar\"", .bar, .{});
}

test "stringify tagged unions" {
    const T = union(enum) {
        nothing,
        foo: u32,
        bar: bool,
    };
    try testStringify("{\"nothing\":{}}", T{ .nothing = {} }, .{});
    try testStringify("{\"foo\":42}", T{ .foo = 42 }, .{});
    try testStringify("{\"bar\":true}", T{ .bar = true }, .{});
}

test "stringify struct" {
    try testStringify("{\"foo\":42}", struct {
        foo: u32,
    }{ .foo = 42 }, .{});
}

test "emit_strings_as_arrays" {
    // Should only affect string values, not object keys.
    try testStringify("{\"foo\":\"bar\"}", .{ .foo = "bar" }, .{});
    try testStringify("{\"foo\":[98,97,114]}", .{ .foo = "bar" }, .{ .emit_strings_as_arrays = true });
    // Should *not* affect these types:
    try testStringify("\"foo\"", @as(enum { foo, bar }, .foo), .{ .emit_strings_as_arrays = true });
    try testStringify("\"ItBroke\"", error.ItBroke, .{ .emit_strings_as_arrays = true });
    // Should work on these:
    try testStringify("\"bar\"", @Vector(3, u8){ 'b', 'a', 'r' }, .{});
    try testStringify("[98,97,114]", @Vector(3, u8){ 'b', 'a', 'r' }, .{ .emit_strings_as_arrays = true });
    try testStringify("\"bar\"", [3]u8{ 'b', 'a', 'r' }, .{});
    try testStringify("[98,97,114]", [3]u8{ 'b', 'a', 'r' }, .{ .emit_strings_as_arrays = true });
}

test "stringify struct with indentation" {
    try testStringify(
        \\{
        \\    "foo": 42,
        \\    "bar": [
        \\        1,
        \\        2,
        \\        3
        \\    ]
        \\}
    ,
        struct {
            foo: u32,
            bar: [3]u32,
        }{
            .foo = 42,
            .bar = .{ 1, 2, 3 },
        },
        .{ .whitespace = .indent_4 },
    );
    try testStringify(
        "{\n\t\"foo\": 42,\n\t\"bar\": [\n\t\t1,\n\t\t2,\n\t\t3\n\t]\n}",
        struct {
            foo: u32,
            bar: [3]u32,
        }{
            .foo = 42,
            .bar = .{ 1, 2, 3 },
        },
        .{ .whitespace = .indent_tab },
    );
    try testStringify(
        \\{"foo":42,"bar":[1,2,3]}
    ,
        struct {
            foo: u32,
            bar: [3]u32,
        }{
            .foo = 42,
            .bar = .{ 1, 2, 3 },
        },
        .{ .whitespace = .minified },
    );
}

test "stringify array of structs" {
    const MyStruct = struct {
        foo: u32,
    };
    try testStringify("[{\"foo\":42},{\"foo\":100},{\"foo\":1000}]", [_]MyStruct{
        MyStruct{ .foo = 42 },
        MyStruct{ .foo = 100 },
        MyStruct{ .foo = 1000 },
    }, .{});
}

test "stringify struct with custom stringifier" {
    try testStringify("[\"something special\",42]", struct {
        foo: u32,
        const Self = @This();
        pub fn jsonStringify(v: @This(), jws: anytype) !void {
            _ = v;
            try jws.beginArray();
            try jws.write("something special");
            try jws.write(42);
            try jws.endArray();
        }
    }{ .foo = 42 }, .{});
}

fn testStringify(expected: []const u8, v: anytype, options: Options) !void {
    var buffer: [4096]u8 = undefined;
    var w: Writer = .fixed(&buffer);
    try value(v, options, &w);
    try std.testing.expectEqualStrings(expected, w.buffered());
}

test "raw streaming" {
    var out_buf: [1024]u8 = undefined;
    var out: Writer = .fixed(&out_buf);

    var w: Stringify = .{ .writer = &out, .options = .{ .whitespace = .indent_2 } };
    try w.beginObject();
    try w.beginObjectFieldRaw();
    try w.writer.writeAll("\"long");
    try w.writer.writeAll(" key\"");
    w.endObjectFieldRaw();
    try w.beginWriteRaw();
    try w.writer.writeAll("\"long");
    try w.writer.writeAll(" value\"");
    w.endWriteRaw();
    try w.endObject();

    const expected =
        \\{
        \\  "long key": "long value"
        \\}
    ;
    try std.testing.expectEqualStrings(expected, w.writer.buffered());
}