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
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
//! Procedural macros to derive minicbor's `Encode` and `Decode` traits.
//!
//! Deriving is supported for `struct`s and `enum`s. The encoding is optimised
//! for forward and backward compatibility and the overall approach is
//! influenced by [Google's Protocol Buffers][1].
//!
//! The goal is that ideally a change to a type still allows older software,
//! which is unaware of the changes, to decode values of the changed type
//! (forward compatibility) and newer software, to decode values of types
//! encoded by older software, which do not include the changes made to the
//! type (backward compatibility).
//!
//! In order to reach this goal, the encoding has the following characteristics:
//!
//! 1. The encoding does not contain any names, i.e. no field names, type names
//! or variant names. Instead, every field and every constructor needs to be
//! annotated with an (unsigned) index number, e.g. `#[n(1)]`.
//!
//! 2. Unknown fields are ignored during decoding.
//!
//! 3. Optional types default to `None` if their value is not present during
//! decoding.
//!
//! 4. Optional enums default to `None` if an unknown variant is encountered
//! during decoding.
//!
//! Item **1** ensures that names can be changed freely without compatibility
//! concerns. Item **2** ensures that new fields do not affect older software.
//! Item **3** ensures that newer software can stop producing optional values.
//! Item **4** ensures that enums can get new variants that older software is
//! not aware of. By "fields" we mean the elements of structs and tuple structs
//! as well as enum structs and enum tuples. In addition, it is a compatible
//! change to turn a unit variant into a struct or tuple variant if all fields
//! are optional.
//!
//! From the above it should be obvious that *non-optional fields need to be
//! present forever*, so they should only be part of a type after careful
//! consideration.
//!
//! It should be emphasised that an `enum` itself can not be changed in a
//! compatible way. An unknown variant causes an error. It is only when they
//! are declared as an optional field type that unknown variants of an enum
//! are mapped to `None`. In other words, *only structs can be used as
//! top-level types in a forward and backward compatible way, enums can not.*
//!
//! # Example
//!
//! ```
//! use minicbor::{Encode, Decode};
//!
//! #[derive(Encode, Decode)]
//! struct Point {
//!     #[n(0)] x: f64,
//!     #[n(1)] y: f64
//! }
//!
//! #[derive(Encode, Decode)]
//! struct ConvexHull {
//!     #[n(0)] left: Point,
//!     #[n(1)] right: Point,
//!     #[n(2)] points: Vec<Point>,
//!     #[n(3)] state: Option<State>
//! }
//!
//! #[derive(Encode, Decode)]
//! enum State {
//!     #[n(0)] Start,
//!     #[n(1)] Search { #[n(0)] info: u64 }
//! }
//! ```
//!
//! In this example the following changes would be compatible in both
//! directions:
//!
//! - Renaming every identifier.
//!
//! - Adding optional fields to `Point`, `ConvexHull`, `State::Start` or
//! `State::Search`.
//!
//! - Adding more variants to `State` *iff* `State` is only decoded as part of
//! `ConvexHull`. Direct decoding of `State` would produce an `UnknownVariant`
//! error for those new variants.
//!
//! [1]: https://developers.google.com/protocol-buffers/
//!
//! # Attributes and borrowing
//!
//! Each field and variant needs to be annotated with an index number, which is
//! used instead of the name, using either **`n`** or **`b`** as attribute names.
//! For the encoding it makes no difference which one to choose. For decoding,
//! `b` indicates that the value borrows from the decoding input, whereas `n`
//! produces non-borrowed values (except for implicit borrows).
//!
//! ## Encoding format
//!
//! The actual CBOR encoding to use can be selected by attaching either the
//! **`#[cbor(array)]`** or **`#[cbor(map)]`** attribute to structs, enums or
//! enum variants. By default `#[cbor(array)]` is implied. The attribute
//! attached to an enum applies to all its variants but can be overriden per
//! variant with another such attribute.
//!
//! ## Implicit borrowing
//!
//! The following types implicitly borrow from the decoding input, which means
//! their lifetimes are constrained by the input lifetime:
//!
//! - `&'_ str`
//! - `&'_ [u8]`
//! - `Option<&'_ str>`
//! - `Option<&'_ [u8]>`
//!
//! ## Explicit borrowing
//!
//! If a type is annotated with **`#[b(...)]`**, all its lifetimes will be
//! constrained to the input lifetime.
//!
//! If the type is a `std::borrow::Cow<'_, str>` or `std::borrow::Cow<'_, [u8]>`
//! type, the generated code will decode the inner type and construct a
//! `Cow::Borrowed` variant, contrary to the `Cow` impl of `Decode` which
//! produces owned values.
//!
//! ## Other attributes
//!
//! ### `encode_with`, `decode_with` and `with`
//!
//! Fields in structs and enum variants may be annotated with
//! **`#[cbor(encode_with = "path")]`**, **`#[cbor(decode_with = "path")]`** or
//! **`#[cbor(with = "module-path")]`** where `path` is the full path to a
//! function which is used instead of `Encode::encode` to encode the field or
//! `Decode::decode` to decode the field respectively. The types of these
//! functions must be equivalent to `Encode::encode` or `Decode::decode`.
//! The `with` attribute combines the other two with `module-path` denoting the
//! full path to a module with two functions `encode` and `decode` as members,
//! which are used for encoding and decoding of the field. These three
//! attributes can either override an existing `Encode` or `Decode` impl or be
//! used for types which do not implement those traits at all.
//!
//! ### `transparent`
//!
//! A **`#[cbor(transparent)]`** attribute can be attached to structs with
//! exactly one field (aka newtypes). If present, the generated `Encode` and
//! `Decode` impls will just forward the `encode`/`decode` calls to the inner
//! type, i.e. the resulting CBOR representation will be identical to the one
//! of the inner type.
//!
//! # CBOR encoding
//!
//! The CBOR values produced by a derived `Encode` implementation are of the
//! following formats.
//!
//! ## Structs
//!
//! ### Array encoding
//!
//! By default or if a struct has the **`#[cbor(array)]`** attribute, it will
//! be represented as a CBOR array. Its index numbers are represened by the
//! position of the field value in this array. Any gaps between index numbers
//! are filled with CBOR NULL values and `Option`s which are `None` likewise
//! end up as NULLs in this array.
//!
//! ```text
//! <<struct-as-array encoding>> =
//!     `array(n)`
//!         item_0
//!         item_1
//!         ...
//!         item_n
//! ```
//!
//! ### Map encoding
//!
//! If a struct has the **`#[cbor(map)]`** attribute, then it will be
//! represented as a CBOR map with keys corresponding to the numeric index
//! value:
//!
//! ```text
//! <<struct-as-map encoding>> =
//!     `map(n)`
//!         `0` item_0
//!         `1` item_1
//!         ...
//!          n  item_n
//! ```
//!
//! Optional fields whose value is `None` are not encoded.
//!
//! ## Enums
//!
//! Each enum variant is encoded as a two-element array. The first element
//! is the variant index and the second the actual variant value:
//!
//! ```text
//! <<enum encoding>> =
//!     | `array(2)` n <<struct-as-array encoding>> ; if #[cbor(array)]
//!     | `array(2)` n <<struct-as-map encoding>>   ; if #[cbor(map)]
//! ```
//!
//! ## Which encoding to use?
//!
//! The map encoding needs to represent the indexes explicitly in the encoding
//! which costs at least one extra byte per field value, whereas the array
//! encoding does not need to encode the indexes. On the other hand, absent
//! values, i.e. `None`s and gaps between indexes are not encoded with maps but
//! need to be encoded explicitly with arrays as NULLs which need one byte each.
//! Which encoding to choose depends therefore on the nature of the type that
//! should be encoded:
//!
//! - *Dense types* are types which contain only few `Option`s or their `Option`s
//! are assumed to be `Some`s usually. They are best encoded as arrays.
//!
//! - *Sparse types* are types with many `Option`s and their `Option`s are usually
//! `None`s. They are best encoded as maps.
//!
//! When selecting the encoding, future changes to the type should be considered
//! as they may turn a dense type into a sparse one over time.

extern crate proc_macro;

mod decode;
mod encode;

use quote::{ToTokens, TokenStreamExt};
use proc_macro2::Span;
use syn::spanned::Spanned;
use std::collections::HashSet;

/// Derive the `minicbor::Decode` trait for a struct or enum.
///
/// See the [crate] documentation for details.
#[proc_macro_derive(Decode, attributes(n, b, cbor))]
pub fn derive_decode(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
    decode::derive_from(input)
}

/// Derive the `minicbor::Encode` trait for a struct or enum.
///
/// See the [crate] documentation for details.
#[proc_macro_derive(Encode, attributes(n, b, cbor))]
pub fn derive_encode(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
    encode::derive_from(input)
}

// Helpers ////////////////////////////////////////////////////////////////////

/// Check if the given type is an `Option` whose inner type matches the predicate.
fn is_option(ty: &syn::Type, pred: impl FnOnce(&syn::Type) -> bool) -> bool {
    if let syn::Type::Path(t) = ty {
        if let Some(s) = t.path.segments.last() {
            if s.ident == "Option" {
                if let syn::PathArguments::AngleBracketed(b) = &s.arguments {
                    if b.args.len() == 1 {
                        if let syn::GenericArgument::Type(ty) = &b.args[0] {
                            return pred(ty)
                        }
                    }
                }
            }
        }
    }
    false
}

/// Check if the given type is a `Cow` whose inner type matches the predicate.
fn is_cow(ty: &syn::Type, pred: impl FnOnce(&syn::Type) -> bool) -> bool {
    if let syn::Type::Path(t) = ty {
        if let Some(s) = t.path.segments.last() {
            if s.ident == "Cow" {
                if let syn::PathArguments::AngleBracketed(b) = &s.arguments {
                    if b.args.len() == 2 {
                        if let syn::GenericArgument::Lifetime(_) = &b.args[0] {
                            if let syn::GenericArgument::Type(ty) = &b.args[1] {
                                return pred(ty)
                            }
                        }
                    }
                }
            }
        }
    }
    false
}

/// Check if the given type is a `&str`.
fn is_str(ty: &syn::Type) -> bool {
    if let syn::Type::Path(t) = ty {
        t.qself.is_none() && t.path.segments.len() == 1 && t.path.segments[0].ident == "str"
    } else {
        false
    }
}

/// Check if the given type is a `&[u8]`.
fn is_byte_slice(ty: &syn::Type) -> bool {
    if let syn::Type::Slice(t) = ty {
        if let syn::Type::Path(t) = &*t.elem {
            t.qself.is_none() && t.path.segments.len() == 1 && t.path.segments[0].ident == "u8"
        } else {
            false
        }
    } else {
        false
    }
}

/// Get the lifetime of the given type if it is an `Option` whose inner type matches the predicate.
fn option_lifetime(ty: &syn::Type, pred: impl FnOnce(&syn::Type) -> bool) -> Option<syn::Lifetime> {
    if let syn::Type::Path(t) = ty {
        if let Some(s) = t.path.segments.last() {
            if s.ident == "Option" {
                if let syn::PathArguments::AngleBracketed(b) = &s.arguments {
                    if b.args.len() == 1 {
                        if let syn::GenericArgument::Type(syn::Type::Reference(ty)) = &b.args[0] {
                            if pred(&*ty.elem) {
                                return ty.lifetime.clone()
                            }
                        }
                    }
                }
            }
        }
    }
    None
}

/// Get all lifetimes of a type.
fn get_lifetimes(ty: &syn::Type, set: &mut HashSet<syn::Lifetime>) {
    match ty {
        syn::Type::Array(t) => get_lifetimes(&t.elem, set),
        syn::Type::Slice(t) => get_lifetimes(&t.elem, set),
        syn::Type::Paren(t) => get_lifetimes(&t.elem, set),
        syn::Type::Group(t) => get_lifetimes(&t.elem, set),
        syn::Type::Ptr(t)   => get_lifetimes(&t.elem, set),
        syn::Type::Reference(t) => {
            if let Some(l) = &t.lifetime {
                set.insert(l.clone());
            }
            get_lifetimes(&t.elem, set)
        }
        syn::Type::Tuple(t) => {
            for t in &t.elems {
                get_lifetimes(t, set)
            }
        }
        syn::Type::Path(t) => {
            for s in &t.path.segments {
                if let syn::PathArguments::AngleBracketed(b) = &s.arguments {
                    for a in &b.args {
                        match a {
                            syn::GenericArgument::Type(t)     => get_lifetimes(t, set),
                            syn::GenericArgument::Binding(b)  => get_lifetimes(&b.ty, set),
                            syn::GenericArgument::Lifetime(l) => { set.insert(l.clone()); }
                            _                                 => {}
                        }
                    }
                }
            }
        }
        _ => {}
    }
}

/// Get the lifetime of a reference if its type matches the predicate.
fn tyref_lifetime(ty: &syn::Type, pred: impl FnOnce(&syn::Type) -> bool) -> Option<syn::Lifetime> {
    if let syn::Type::Reference(p) = ty {
        if pred(&*p.elem) {
            return p.lifetime.clone()
        }
    }
    None
}

/// Get the set of lifetimes which need to be constrained to the decoding input lifetime.
fn lifetimes_to_constrain<'a, I>(types: I) -> HashSet<syn::Lifetime>
where
    I: Iterator<Item = (&'a Idx, &'a syn::Type)>
{
    let mut set = HashSet::new();
    for (i, t) in types {
        if let Some(l) = tyref_lifetime(t, is_str) {
            set.insert(l);
            continue
        }
        if let Some(l) = tyref_lifetime(t, is_byte_slice) {
            set.insert(l);
            continue
        }
        if let Some(l) = option_lifetime(t, is_str) {
            set.insert(l);
            continue
        }
        if let Some(l) = option_lifetime(t, is_byte_slice) {
            set.insert(l);
            continue
        }
        if i.is_b() {
            get_lifetimes(t, &mut set)
        }
    }
    set
}

/// The index attribute.
#[derive(Debug, Clone, Copy)]
enum Idx {
    /// A regular, non-borrowing index.
    N(u32),
    /// An index which indicates that the value borrows from the decoding input.
    B(u32)
}

impl ToTokens for Idx {
    fn to_tokens(&self, tokens: &mut proc_macro2::TokenStream) {
        tokens.append(proc_macro2::Literal::u32_unsuffixed(self.val()))
    }
}

impl Idx {
    /// Test if `Idx` is the `B` variant.
    fn is_b(self) -> bool {
        if let Idx::B(_) = self {
            true
        } else {
            false
        }
    }

    /// Get the numeric index value.
    fn val(self) -> u32 {
        match self {
            Idx::N(i) => i,
            Idx::B(i) => i
        }
    }
}

/// Get the index number from the list of attributes.
///
/// The first attribute `n` will be used and its argument must be an
/// unsigned integer literal that fits into a `u32`.
fn index_number(s: Span, attrs: &[syn::Attribute]) -> syn::Result<Idx> {
    for a in attrs {
        if a.path.is_ident("n") {
            let lit: syn::LitInt = a.parse_args()?;
            return lit.base10_digits().parse()
                .map_err(|_| syn::Error::new(a.tokens.span(), "expected `u32` value"))
                .map(Idx::N)
        }
        if a.path.is_ident("b") {
            let lit: syn::LitInt = a.parse_args()?;
            return lit.base10_digits().parse()
                .map_err(|_| syn::Error::new(a.tokens.span(), "expected `u32` value"))
                .map(Idx::B)
        }
    }
    Err(syn::Error::new(s, "missing `#[n(...)]` or `#[b(...)]` attribute"))
}

/// Check that there are no duplicate elements in `iter`.
fn check_uniq<I>(s: Span, iter: I) -> syn::Result<()>
where
    I: IntoIterator<Item = Idx>
{
    let mut set = HashSet::new();
    let mut ctr = 0;
    for u in iter {
        set.insert(u.val());
        ctr += 1;
    }
    if ctr != set.len() {
        return Err(syn::Error::new(s, "duplicate index numbers"))
    }
    Ok(())
}

/// Get the index number of every field.
fn field_indices<'a, I>(iter: I) -> syn::Result<Vec<Idx>>
where
    I: Iterator<Item = &'a syn::Field>
{
    iter.map(|f| index_number(f.span(), &f.attrs)).collect()
}

/// Get the index number of every variant.
fn variant_indices<'a, I>(iter: I) -> syn::Result<Vec<Idx>>
where
    I: Iterator<Item = &'a syn::Variant>
{
    iter.map(|v| index_number(v.span(), &v.attrs)).collect()
}

/// The encoding to use for structs and enum variants.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
enum Encoding {
    Array,
    Map
}

impl Default for Encoding {
    fn default() -> Self {
        Encoding::Array
    }
}

/// Determine attribute value of the `#[cbor(map|array)]` attribute.
fn encoding(a: &syn::Attribute) -> Option<Encoding> {
    match a.parse_meta().ok()? {
        syn::Meta::List(ml) if ml.path.is_ident("cbor") => {
            if let Some(syn::NestedMeta::Meta(syn::Meta::Path(arg))) = ml.nested.first() {
                if arg.is_ident("map") {
                    return Some(Encoding::Map)
                }
                if arg.is_ident("array") {
                    return Some(Encoding::Array)
                }
            }
        }
        _ => {}
    }
    None
}

/// Custom encode/decode functions.
enum CustomCodec {
    /// Custom encode function.
    ///
    /// Assumed to be of a type equivalent to:
    ///
    ///   `fn<T, W: Write>(&T, &mut Encoder<W>) -> Result<(), Error<W::Error>>`
    ///
    /// Declared with `#[cbor(encode_with = "...")]`.
    Encode(syn::ExprPath),
    /// Custom decode function.
    ///
    /// Assumed to be of a type equivalent to:
    ///
    ///   `fn<T>(&mut Decoder<'_>) -> Result<T, Error>`
    ///
    /// Declared with `#[cbor(decode_with = "...")]`.
    Decode(syn::ExprPath),
    /// A module which contains custom encode/decode functions.
    ///
    /// The module is assumed to contain two functions named `encode` and
    /// `decode` whose types match those declared with
    /// `#[cbor(encode_with = "...")]` or `#[cbor(decode_with = "...")]`
    /// respectively. Declared with `#[cbor(with = "...")]`.
    Both(syn::ExprPath)
}

impl CustomCodec {
    /// Is this a custom codec from `encode_with` or `with`?
    fn is_encode(&self) -> bool {
        matches!(self, CustomCodec::Encode(_) | CustomCodec::Both(_))
    }

    /// Is this a custom codec from `decode_with` or `with`?
    fn is_decode(&self) -> bool {
        matches!(self, CustomCodec::Decode(_) | CustomCodec::Both(_))
    }

    /// Extract the encode function unless this `CustomCodec` does not declare one.
    fn to_encode_path(&self) -> Option<syn::ExprPath> {
        match self {
            CustomCodec::Encode(p) => Some(p.clone()),
            CustomCodec::Decode(_) => None,
            CustomCodec::Both(p) => {
                let mut p = p.clone();
                let ident = syn::Ident::new("encode", proc_macro2::Span::call_site());
                p.path.segments.push(ident.into());
                Some(p)
            }
        }
    }

    /// Extract the decode function unless this `CustomCodec` does not declare one.
    fn to_decode_path(&self) -> Option<syn::ExprPath> {
        match self {
            CustomCodec::Encode(_) => None,
            CustomCodec::Decode(p) => Some(p.clone()),
            CustomCodec::Both(p) => {
                let mut p = p.clone();
                let ident = syn::Ident::new("decode", proc_macro2::Span::call_site());
                p.path.segments.push(ident.into());
                Some(p)
            }
        }
    }
}

/// Determine the attribute value of the `#[cbor(encode_with|decode_with|with)]` attribute.
fn custom_codec(a: &syn::Attribute) -> syn::Result<Option<CustomCodec>> {
    if let syn::Meta::List(ml) = a.parse_meta()? {
        if !ml.path.is_ident("cbor") {
            return Ok(None)
        }
        if let Some(syn::NestedMeta::Meta(syn::Meta::NameValue(arg))) = ml.nested.first() {
            if arg.path.is_ident("encode_with") {
                if let syn::Lit::Str(path) = &arg.lit {
                    return Ok(Some(CustomCodec::Encode(syn::parse_str(&path.value())?)))
                }
            }
            if arg.path.is_ident("decode_with") {
                if let syn::Lit::Str(path) = &arg.lit {
                    return Ok(Some(CustomCodec::Decode(syn::parse_str(&path.value())?)))
                }
            }
            if arg.path.is_ident("with") {
                if let syn::Lit::Str(path) = &arg.lit {
                    return Ok(Some(CustomCodec::Both(syn::parse_str(&path.value())?)))
                }
            }
        }
    }
    Ok(None)
}

/// Traverse all field types and collect all type parameters along the way.
fn collect_type_params<'a, I>(all: &syn::Generics, fields: I) -> HashSet<syn::TypeParam>
where
    I: Iterator<Item = &'a syn::Field>
{
    use syn::visit::Visit;

    struct Collector {
        all: Vec<syn::Ident>,
        found: HashSet<syn::TypeParam>
    }

    impl<'a> Visit<'a> for Collector {
        fn visit_field(&mut self, f: &'a syn::Field) {
            if let syn::Type::Path(ty) = &f.ty {
                if let Some(t) = ty.path.segments.first() {
                    if self.all.contains(&t.ident) {
                        self.found.insert(syn::TypeParam::from(t.ident.clone()));
                    }
                }
            }
            self.visit_type(&f.ty)
        }

        fn visit_path(&mut self, p: &'a syn::Path) {
            if p.leading_colon.is_none() && p.segments.len() == 1 {
                let id = &p.segments[0].ident;
                if self.all.contains(id) {
                    self.found.insert(syn::TypeParam::from(id.clone()));
                }
            }
            syn::visit::visit_path(self, p)
        }
    }

    let mut c = Collector {
        all: all.type_params().map(|tp| tp.ident.clone()).collect(),
        found: HashSet::new()
    };

    for f in fields {
        c.visit_field(f)
    }

    c.found
}

/// Check if the attribute matches the given identifier.
fn is_cbor_attr(a: &syn::Attribute, ident: &str) -> syn::Result<bool> {
    match a.parse_meta()? {
        syn::Meta::List(ml) if ml.path.is_ident("cbor") => {
            if let Some(syn::NestedMeta::Meta(syn::Meta::Path(arg))) = ml.nested.first() {
                if arg.is_ident(ident) {
                    return Ok(true)
                }
            }
            if let Some(syn::NestedMeta::Meta(syn::Meta::NameValue(arg))) = ml.nested.first() {
                if arg.path.is_ident(ident) {
                    return Ok(true)
                }
            }
        }
        _ => {}
    }
    Ok(false)
}

/// Find any of the attributes that matches the given identifier.
fn find_cbor_attr<'a, I>(attrs: I, ident: &str) -> syn::Result<Option<&'a syn::Attribute>>
where
    I: Iterator<Item = &'a syn::Attribute>
{
    for a in attrs {
        if is_cbor_attr(a, ident)? {
            return Ok(Some(a))
        }
    }
    Ok(None)
}