support top level fields

[andrewrk]

Now that #1047 is done, and every file is implicitly an empty struct, we could allow fields at the top level. This would solve the redundant import problem:

zig/std/index.zig

Lines 3 to 10 in 96164ce

	pub const BufMap = @import("buf_map.zig").BufMap;
	pub const BufSet = @import("buf_set.zig").BufSet;
	pub const Buffer = @import("buffer.zig").Buffer;
	pub const BufferOutStream = @import("buffer.zig").BufferOutStream;
	pub const HashMap = @import("hash_map.zig").HashMap;
	pub const LinkedList = @import("linked_list.zig").LinkedList;
	pub const IntrusiveLinkedList = @import("linked_list.zig").IntrusiveLinkedList;
	pub const SegmentedList = @import("segmented_list.zig").SegmentedList;

Syntactically it looks a bit odd, but semantically it makes sense.

[Hejsil]

This seems to already work in stage 1:

a: u8,

test "" {
    var s = @This(){ .a = 22 };
    @import("std").testing.expectEqual(u8(22), s.a);
}

When I did the grammar rework I allowed fields at top level in the syntax, and now that files are empty structs, it seems to work it just haven't been tested. To close this, we probably need:

• Tests
• std refactor
• Stage 2 parser update

[andrewrk]

ah! I got tricked by vim integration with zig fmt 😆

[andrewrk]

Also just to repeat this here:

The "index.zig" convention is deprecated. The new pattern is the one set by

• std/os/darwin.zig (and corresponding std/os/darwin/ folder)
• std/event.zig (and corresponding std/event/ folder)

[Hejsil]

Also, files cannot be generic so the redundant import problem is only solved for none generic structs.

[ifreund]

I was in the mood for a little bit of bikeshedding, so I refactored my project to use toplevel fields wherever possible. No issues so far on zig 0.6.0. Repository here for the curious: https://github.com/ifreund/river.