1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63

#![allow(bad_style)]

extern crate proc_macro;
use core::str::FromStr;
use proc_macro::{TokenStream, TokenTree};

mod char_escape;
use char_escape::perform_the_escaping;

#[proc_macro]
pub fn utf16_lit(stream: TokenStream) -> TokenStream {
  const USAGE: &str = "Usage: utf16_lit!(ident, string_lit)";

  // This "parsing" system is janky as hell, but it doesn't depend on the
  // `quote` or `syn` crates, so we save a lot on compile time at the expense of
  // having slightly worse errors. However, since the user usually calls the
  // macro correctly and doesn't have an error, it's probably a good trade.
  let mut tt_iter = stream.into_iter();
  let ident = match tt_iter.next().expect(USAGE) {
    TokenTree::Ident(i) => i,
    _ => panic!(USAGE),
  };
  match tt_iter.next().expect(USAGE) {
    TokenTree::Punct(p) if p.as_char() == ',' => (),
    _ => panic!(USAGE),
  };
  let lit = match tt_iter.next().expect(USAGE) {
    TokenTree::Literal(lit) => lit,
    _ => panic!(USAGE),
  };
  assert!(tt_iter.next().is_none(), USAGE);

  // The literal, if it's a string literal, will have the double quotes on the
  // edges, which we want to strip off of the value we're going to encode as
  // utf16. The end we can cheaply strip off with `pop`, but for the start it
  // would be expensive to delete the 0th char and move all the rest. Instead,
  // we want to simply start the encoding just after the double quote position.
  let mut lit_string = lit.to_string();
  assert!(lit_string.pop() == Some('"'), USAGE);
  assert!(lit_string.chars().nth(0) == Some('"'), USAGE);

  // Also we have to convert any escape sequences within the string into the
  // correct characters ourselves because the `proc_macro` crate doesn't do that
  // for us.
  let all_the_chars = perform_the_escaping(&lit_string[1..]);
  let mut units: Vec<u16> = Vec::with_capacity(all_the_chars.len() * 2);
  let mut encode_buf = [0_u16; 2];
  for ch in all_the_chars {
    for u in ch.encode_utf16(&mut encode_buf) {
      units.push(*u);
    }
  }

  // Finally, instead of trying to fiddle a TokenStream with extend and all
  // that, we just write down the text of the code we wanted to have had the
  // whole time and then we use `from_str` and let the system handle it for us.
  let buf = format!(
    "pub const {ident}: &[u16] = &{units:?};",
    ident = ident,
    units = &units[..],
  );
  TokenStream::from_str(&buf).unwrap()
}