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
use crate::types::*;
#[cfg(not(feature = "std"))]
use alloc::vec::Vec;
use core::{default::Default, mem::MaybeUninit, slice};

/// This is essentially equivalent to the
/// [SeekFrom](https://doc.rust-lang.org/std/io/enum.SeekFrom.html) type from
/// the rust standard library, but reproduced here to avoid dependency on
/// `std::io`, as well as to use 32-bit integers to specify positions. This
/// saves some computation and space, and is adequate for the kind of data sizes
/// that are possible in smart contracts.
pub enum SeekFrom {
    Start(u32),
    End(i32),
    Current(i32),
}

/// The `Seek` trait provides a cursor which can be moved within a stream of
/// bytes. This is essentially a copy of
/// [std::io::Seek](https://doc.rust-lang.org/std/io/trait.Seek.html), but
/// avoiding its dependency on `std::io::Error`, and the associated code size
/// increase. Additionally, the positions are expressed in terms of 32-bit
/// integers since this is adequate for the sizes of data in smart contracts.
pub trait Seek {
    type Err;
    /// Seek to the new position. If successful, return the new position from
    /// the beginning of the stream.
    fn seek(&mut self, pos: SeekFrom) -> Result<u32, Self::Err>;

    /// Get the cursor position counted from the beginning of the stream.
    fn cursor_position(&self) -> u32;
}

/// The `HasSize` trait provides a function for getting the current byte size.
pub trait HasSize {
    /// Get the current byte size.
    fn size(&self) -> u32;
}

/// Reads `n` bytes from a given `source` without initializing the byte array
/// beforehand using MaybeUninit.
macro_rules! read_n_bytes {
    ($n:expr, $source:tt) => {{
        let mut bytes: MaybeUninit<[u8; $n]> = MaybeUninit::uninit();
        let write_bytes = unsafe { slice::from_raw_parts_mut(bytes.as_mut_ptr() as *mut u8, $n) };
        $source.read_exact(write_bytes)?;
        unsafe { bytes.assume_init() }
    }};
}

/// The `Read` trait provides a means of reading from byte streams.
pub trait Read {
    /// Read a number of bytes into the provided buffer. The returned value is
    /// `Ok(n)` if a read was successful, and `n` bytes were read (`n` could be
    /// 0).
    fn read(&mut self, buf: &mut [u8]) -> ParseResult<usize>;

    /// Read exactly the required number of bytes. If not enough bytes could be
    /// read the function returns `Err(_)`, and the contents of the given buffer
    /// is unspecified.
    fn read_exact(&mut self, buf: &mut [u8]) -> ParseResult<()> {
        let mut start = 0;
        while start < buf.len() {
            match self.read(&mut buf[start..]) {
                Ok(0) => break,
                Ok(n) => {
                    start += n;
                }
                Err(_e) => return Err(Default::default()),
            }
        }
        if start == buf.len() {
            Ok(())
        } else {
            Err(Default::default())
        }
    }

    /// Read a `u64` in little-endian format.
    fn read_u64(&mut self) -> ParseResult<u64> {
        let bytes = read_n_bytes!(8, self);
        Ok(u64::from_le_bytes(bytes))
    }

    /// Read a `u32` in little-endian format.
    fn read_u32(&mut self) -> ParseResult<u32> {
        let bytes = read_n_bytes!(4, self);
        Ok(u32::from_le_bytes(bytes))
    }

    /// Read a `u16` in little-endian format.
    fn read_u16(&mut self) -> ParseResult<u16> {
        let bytes = read_n_bytes!(2, self);
        Ok(u16::from_le_bytes(bytes))
    }

    /// Read a `u8`.
    fn read_u8(&mut self) -> ParseResult<u8> {
        let bytes = read_n_bytes!(1, self);
        Ok(u8::from_le_bytes(bytes))
    }

    /// Read a `i64` in little-endian format.
    fn read_i64(&mut self) -> ParseResult<i64> {
        let bytes = read_n_bytes!(8, self);
        Ok(i64::from_le_bytes(bytes))
    }

    /// Read a `i32` in little-endian format.
    fn read_i32(&mut self) -> ParseResult<i32> {
        let bytes = read_n_bytes!(4, self);
        Ok(i32::from_le_bytes(bytes))
    }

    /// Read a `i16` in little-endian format.
    fn read_i16(&mut self) -> ParseResult<i16> {
        let bytes = read_n_bytes!(2, self);
        Ok(i16::from_le_bytes(bytes))
    }

    /// Read a `i32` in little-endian format.
    fn read_i8(&mut self) -> ParseResult<i8> {
        let bytes = read_n_bytes!(1, self);
        Ok(i8::from_le_bytes(bytes))
    }

    /// Load an array of the given size.
    fn read_array<const N: usize>(&mut self) -> ParseResult<[u8; N]> { Ok(read_n_bytes!(N, self)) }
}

/// The `Write` trait provides functionality for writing to byte streams.
pub trait Write {
    type Err: Default;
    /// Try to write the given buffer into the output stream. If writes are
    /// successful returns the number of bytes written.
    fn write(&mut self, buf: &[u8]) -> Result<usize, Self::Err>;

    /// Attempt to write the entirety of the buffer to the output by repeatedly
    /// calling `write` until either no more output can written, or writing
    /// fails.
    fn write_all(&mut self, buf: &[u8]) -> Result<(), Self::Err> {
        let mut start = 0;
        while start < buf.len() {
            match self.write(&buf[start..]) {
                Ok(n) if n > 0 => start += n,
                _ => return Err(Default::default()),
            }
        }
        Ok(())
    }

    /// Write a single byte to the output.
    fn write_u8(&mut self, x: u8) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }

    /// Write a `u16` in little endian.
    fn write_u16(&mut self, x: u16) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }

    /// Write a `u32` in little endian.
    fn write_u32(&mut self, x: u32) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }

    /// Write a `u64` in little endian.
    fn write_u64(&mut self, x: u64) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }

    /// Write a `i8` to the output.
    fn write_i8(&mut self, x: i8) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }

    /// Write a `i16` in little endian.
    fn write_i16(&mut self, x: i16) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }

    /// Write a `i32` in little endian.
    fn write_i32(&mut self, x: i32) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }

    /// Write a `i64` in little endian.
    fn write_i64(&mut self, x: i64) -> Result<(), Self::Err> { self.write_all(&x.to_le_bytes()) }
}

/// The `write` method always appends data to the end of the vector.
impl Write for Vec<u8> {
    type Err = ();

    #[inline]
    fn write(&mut self, buf: &[u8]) -> Result<usize, Self::Err> {
        let _ = self.extend_from_slice(buf);
        Ok(buf.len())
    }
}

/// This implementation overwrite the contents of the slice and updates the
/// reference to point to the unwritten part. The slice is (by necessity) never
/// extended.
/// This is in contrast to the `Vec<u8>` implementation which always extends the
/// vector with the data that is written.
impl Write for &mut [u8] {
    type Err = ();

    #[inline]
    fn write(&mut self, buf: &[u8]) -> Result<usize, Self::Err> {
        let to_write = core::cmp::min(buf.len(), self.len());
        let (overwrite, rest) = core::mem::replace(self, &mut []).split_at_mut(to_write);
        overwrite.copy_from_slice(&buf[..to_write]);
        *self = rest;
        Ok(to_write)
    }
}

/// The `Serial` trait provides a means of writing structures into byte-sinks
/// (`Write`).
///
/// Can be derived using `#[derive(Serial)]` for most cases.
pub trait Serial {
    /// Attempt to write the structure into the provided writer, failing if
    /// only part of the structure could be written.
    ///
    /// NB: We use Result instead of Option for better composability with other
    /// constructs.
    fn serial<W: Write>(&self, _out: &mut W) -> Result<(), W::Err>;
}

/// The `Deserial` trait provides a means of reading structures from
/// byte-sources (`Read`).
///
/// Can be derived using `#[derive(Deserial)]` for most cases.
pub trait Deserial: Sized {
    /// Attempt to read a structure from a given source, failing if an error
    /// occurs during deserialization or reading.
    fn deserial<R: Read>(_source: &mut R) -> ParseResult<Self>;
}

/// The `Serialize` trait provides a means of writing structures into byte-sinks
/// (`Write`) or reading structures from byte sources (`Read`).
///
/// Can be derived using `#[derive(Serialized)]` for most cases.
pub trait Serialize: Serial + Deserial {}

/// Generic instance deriving Serialize for any type that implements both Serial
/// and Deserial.
impl<A: Deserial + Serial> Serialize for A {}

/// The `SerialCtx` trait provides a means of writing structures into byte-sinks
/// (`Write`) using contextual information.
/// The contextual information is:
///
///   - `size_length`: The number of bytes used to record the length of the
///     data.
pub trait SerialCtx {
    /// Attempt to write the structure into the provided writer, failing if
    /// if the length cannot be represented in the provided `size_length` or
    /// only part of the structure could be written.
    ///
    /// NB: We use Result instead of Option for better composability with other
    /// constructs.
    fn serial_ctx<W: Write>(
        &self,
        size_length: crate::schema::SizeLength,
        out: &mut W,
    ) -> Result<(), W::Err>;
}

/// The `DeserialCtx` trait provides a means of reading structures from
/// byte-sources (`Read`) using contextual information.
/// The contextual information is:
///
///   - `size_length`: The expected number of bytes used for the length of the
///     data.
///   - `ensure_ordered`: Whether the ordering should be ensured, for example
///     that keys in `BTreeMap` and `BTreeSet` are in strictly increasing order.
pub trait DeserialCtx: Sized {
    /// Attempt to read a structure from a given source and context, failing if
    /// an error occurs during deserialization or reading.
    fn deserial_ctx<R: Read>(
        size_length: crate::schema::SizeLength,
        ensure_ordered: bool,
        source: &mut R,
    ) -> ParseResult<Self>;
}

/// A more convenient wrapper around `Deserial` that makes it easier to write
/// deserialization code. It has a blanked implementation for any read and
/// serialize pair. The key idea is that the type to deserialize is inferred
/// from the context, enabling one to write, for example,
///
/// ```rust
/// # fn deserial<R: concordium_contracts_common::Read>(source: &mut R) -> concordium_contracts_common::ParseResult<(u8, u8)> {
/// #  use crate::concordium_contracts_common::Get;
///    let x = source.get()?;
///    let y = source.get()?;
/// #   Ok((x,y))
/// # }
/// ```
/// where `source` is any type that implements `Read`.
pub trait Get<T> {
    fn get(&mut self) -> ParseResult<T>;
}

impl<R: Read, T: Deserial> Get<T> for R {
    #[inline(always)]
    fn get(&mut self) -> ParseResult<T> { T::deserial(self) }
}

#[cfg(test)]
mod tests {
    use super::*;
    #[test]
    fn write_u8_slice() {
        let mut xs = [0u8; 10];
        let mut slice: &mut [u8] = &mut xs;
        assert!(0xAAAAAAAAu32.serial(&mut slice).is_ok(), "Writing u32 should succeed.");
        assert_eq!(slice.len(), 6, "The new slice should be of length 6 (= 10 - 4)");
        assert!(0xBBBBBBBBu32.serial(&mut slice).is_ok(), "Writing the second u32 should succeed.");
        assert_eq!(slice.len(), 2, "The new slice should be of length 2 (= 10 - 4 - 4)");
        assert!(0xCCCCu16.serial(&mut slice).is_ok(), "Writing the final u16 should succeed.");
        assert_eq!(slice.len(), 0, "The new slice should be of length 0 (= 10 - 4 - 4 - 2)");
        assert!(0u8.serial(&mut slice).is_err(), "Writing past the end should fail.");
        assert_eq!(
            xs,
            [0xAA, 0xAA, 0xAA, 0xAA, 0xBB, 0xBB, 0xBB, 0xBB, 0xCC, 0xCC],
            "The original array has incorrect content."
        );
    }
}