deserialize_large_vec.rs

/// Helper to be used with `#[serde(deserialize_with = "deserialize_large_vec")]` to efficiently
/// deserialize large `Vec<_>` that includes a trusted `size_hint`.
///
/// Ordinarily, serde uses a `cautious` size hint that's at most 4096 entries:
/// https://github.com/serde-rs/serde/blob/10e4839f8325dab5472b0ebf5551f4d607f14a33/serde/src/de/impls.rs#L1035.
/// For large Vec's, this causes lots of reallocations as the Vec is grown during deserialization.
/// This function avoids the cautious size hint and immediately allocates a Vec of the hinted size
/// to avoid unnecessary reallocations.
pub fn deserialize_large_vec<'de, D, T>(deserializer: D) -> Result<Vec<T>, D::Error>
where
  D: Deserializer<'de>,
  T: Deserialize<'de>,
{
  struct VecVisitor<T> {
    marker: PhantomData<T>,
  }

  impl<'de, T> Visitor<'de> for VecVisitor<T>
  where
    T: Deserialize<'de>,
  {
    type Value = Vec<T>;

    fn expecting(&self, formatter: &mut Formatter) -> std::fmt::Result {
      formatter.write_str("a sequence")
    }

    fn visit_seq<A>(self, mut seq: A) -> Result<Self::Value, A::Error>
    where
      A: SeqAccess<'de>,
    {
      // NB: This is the main line that differs from the serde default impl for Vec<_>.
      let mut values = Vec::with_capacity(seq.size_hint().unwrap_or(0));

      while let Some(value) = seq.next_element()? {
        values.push(value);
      }

      Ok(values)
    }
  }

  let visitor = VecVisitor {
    marker: PhantomData,
  };
  deserializer.deserialize_seq(visitor)
}