fediglam/src/http/multipart.zig

const std = @import("std");
const util = @import("util");

const max_boundary = 70;

const FormFieldResult = struct {
    value: []const u8,
    params: FormDataParams,
    more: bool,
};

const ParamIter = struct {
    str: []const u8,
    index: usize = 0,

    const Param = struct {
        name: []const u8,
        value: []const u8,
    };

    pub fn from(str: []const u8) ParamIter {
        return .{ .str = str, .index = std.mem.indexOfScalar(u8, str, ';') orelse str.len };
    }

    pub fn next(self: *ParamIter) ?Param {
        if (self.index >= self.str.len) return null;

        const start = self.index + 1;
        const new_start = std.mem.indexOfScalarPos(u8, self.str, start, ';') orelse self.str.len;
        self.index = new_start;

        const param = std.mem.trim(u8, self.str[start..new_start], " \t");
        var split = std.mem.split(u8, param, "=");
        const name = split.first();
        const value = std.mem.trimLeft(u8, split.rest(), " \t");
        // TODO: handle quoted values
        // TODO: handle parse errors

        return Param{
            .name = name,
            .value = value,
        };
    }
};

const FormDataParams = struct {
    name: ?[]const u8 = null,
    filename: ?[]const u8 = null,
    charset: ?[]const u8 = null,
};

fn parseParams(alloc: std.mem.Allocator, comptime T: type, str: []const u8) !T {
    var result = T{};
    errdefer util.deepFree(alloc, result);

    var iter = ParamIter.from(str);
    while (iter.next()) |param| {
        inline for (comptime std.meta.fieldNames(T)) |f| {
            if (std.mem.eql(u8, param.name, f)) {
                @field(result, f) = try util.deepClone(alloc, param.value);
            }
        }
    }

    return result;
}

fn isFinalPart(peek_stream: anytype) !bool {
    const reader = peek_stream.reader();
    var buf: [2]u8 = undefined;
    const end = try reader.readAll(&buf);
    const end_line = buf[0..end];
    const terminal = std.mem.eql(u8, end_line, "--");
    if (!terminal) try peek_stream.putBack(end_line);

    // Skip whitespace
    while (true) {
        const b = reader.readByte() catch |err| switch (err) {
            error.EndOfStream => {
                if (terminal) break else return error.InvalidMultipartBoundary;
            },
            else => return err,
        };

        if (std.mem.indexOfScalar(u8, " \t\r\n", b) == null) {
            try peek_stream.putBackByte(b);
            break;
        }
    }

    return terminal;
}

fn parseFormField(boundary: []const u8, peek_stream: anytype, alloc: std.mem.Allocator) !FormFieldResult {
    const reader = peek_stream.reader();

    // TODO: refactor
    var headers = try @import("./request/parser.zig").parseHeaders(alloc, reader);
    defer headers.deinit();

    var value = std.ArrayList(u8).init(alloc);
    errdefer value.deinit();

    line_loop: while (true) {
        // parse crlf--
        var buf: [4]u8 = undefined;
        try reader.readNoEof(&buf);
        if (!std.mem.eql(u8, &buf, "\r\n--")) {
            try value.append(buf[0]);
            try peek_stream.putBack(buf[1..]);
            var ch = try reader.readByte();
            while (ch != '\r') : (ch = try reader.readByte()) try value.append(ch);

            try peek_stream.putBackByte(ch);
            continue;
        }

        for (boundary) |ch, i| {
            const b = try reader.readByte();

            if (b != ch) {
                try value.appendSlice("\r\n--");
                try value.appendSlice(boundary[0 .. i + 1]);
                continue :line_loop;
            }
        }

        // Boundary parsed. See if its a terminal or not
        break;
    }

    const terminal = try isFinalPart(peek_stream);
    const disposition = headers.get("Content-Disposition") orelse return error.NoDisposition;

    return FormFieldResult{
        .value = value.toOwnedSlice(),
        .params = try parseParams(alloc, FormDataParams, disposition),
        .more = !terminal,
    };
}

pub fn parseFormData(boundary: []const u8, reader: anytype, alloc: std.mem.Allocator) !void {
    if (boundary.len > max_boundary) return error.BoundaryTooLarge;

    var stream = std.io.peekStream(72, reader);
    {
        var buf: [72]u8 = undefined;
        const count = try stream.reader().readAll(buf[0 .. boundary.len + 2]);
        var line = buf[0..count];
        if (line.len != boundary.len + 2) return error.InvalidMultipartBoundary;
        if (!std.mem.startsWith(u8, line, "--")) return error.InvalidMultipartBoundary;
        if (!std.mem.endsWith(u8, line, boundary)) return error.InvalidMultipartBoundary;

        if (try isFinalPart(&stream)) return;
    }

    while (true) {
        const field = try parseFormField(boundary, &stream, alloc);
        defer util.deepFree(alloc, field);

        std.debug.print("{any}\n", .{field});

        if (!field.more) return;
    }
}
fn toCrlf(comptime str: []const u8) []const u8 {
    comptime {
        var buf: [str.len * 2]u8 = undefined;

        @setEvalBranchQuota(@intCast(u32, str.len * 2)); // TODO: why does this need to be *2

        var buf_len: usize = 0;
        for (str) |ch| {
            if (ch == '\n') {
                buf[buf_len] = '\r';
                buf_len += 1;
            }

            buf[buf_len] = ch;
            buf_len += 1;
        }

        return buf[0..buf_len];
    }
}

test "parseFormData" {
    const body = toCrlf(
        \\--abcd
        \\Content-Disposition: form-data; name=first; charset=utf8
        \\
        \\content
        \\--abcd
        \\content-Disposition: form-data; name=second
        \\
        \\no content
        \\--abcd
        \\content-disposition: form-data; name=third
        \\
        \\
        \\--abcd--
        \\
    );

    var stream = std.io.StreamSource{ .const_buffer = std.io.fixedBufferStream(body) };
    try parseFormData("abcd", stream.reader(), std.testing.allocator);
}