Implement MapSkipError, analogous to VecSkipError, but for map-like data structures. (#763)

Author: jonasbb

serde_with/src/de/impls.rs

@@ -908,80 +908,6 @@ where
     }
 }
 
-#[cfg(feature = "alloc")]
-impl<'de, T, U> DeserializeAs<'de, Vec<T>> for VecSkipError<U>
-where
-    U: DeserializeAs<'de, T>,
-{
-    fn deserialize_as<D>(deserializer: D) -> Result<Vec<T>, D::Error>
-    where
-        D: Deserializer<'de>,
-    {
-        enum GoodOrError<T, TAs> {
-            Good(T),
-            // Only here to consume the TAs generic
-            Error(PhantomData<TAs>),
-        }
-
-        impl<'de, T, TAs> Deserialize<'de> for GoodOrError<T, TAs>
-        where
-            TAs: DeserializeAs<'de, T>,
-        {
-            fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
-            where
-                D: Deserializer<'de>,
-            {
-                let is_hr = deserializer.is_human_readable();
-                let content: content::de::Content<'de> = Deserialize::deserialize(deserializer)?;
-
-                Ok(
-                    match <DeserializeAsWrap<T, TAs>>::deserialize(
-                        content::de::ContentDeserializer::<D::Error>::new(content, is_hr),
-                    ) {
-                        Ok(elem) => GoodOrError::Good(elem.into_inner()),
-                        Err(_) => GoodOrError::Error(PhantomData),
-                    },
-                )
-            }
-        }
-
-        struct SeqVisitor<T, U> {
-            marker: PhantomData<T>,
-            marker2: PhantomData<U>,
-        }
-
-        impl<'de, T, TAs> Visitor<'de> for SeqVisitor<T, TAs>
-        where
-            TAs: DeserializeAs<'de, T>,
-        {
-            type Value = Vec<T>;
-
-            fn expecting(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
-                formatter.write_str("a sequence")
-            }
-
-            fn visit_seq<A>(self, seq: A) -> Result<Self::Value, A::Error>
-            where
-                A: SeqAccess<'de>,
-            {
-                utils::SeqIter::new(seq)
-                    .filter_map(|res: Result<GoodOrError<T, TAs>, A::Error>| match res {
-                        Ok(GoodOrError::Good(value)) => Some(Ok(value)),
-                        Ok(GoodOrError::Error(_)) => None,
-                        Err(err) => Some(Err(err)),
-                    })
-                    .collect()
-            }
-        }
-
-        let visitor = SeqVisitor::<T, U> {
-            marker: PhantomData,
-            marker2: PhantomData,
-        };
-        deserializer.deserialize_seq(visitor)
-    }
-}
-
 impl<'de, Str> DeserializeAs<'de, Option<Str>> for NoneAsEmptyString
 where
     Str: FromStr,

serde_with/src/de/mod.rs

@@ -10,6 +10,8 @@
 #[cfg(feature = "alloc")]
 mod duplicates;
 mod impls;
+#[cfg(feature = "alloc")]
+mod skip_error;
 
 use crate::prelude::*;
 

serde_with/src/de/skip_error.rs

@@ -0,0 +1,144 @@
+use super::impls::macros::foreach_map;
+use crate::prelude::*;
+
+#[cfg(feature = "hashbrown_0_14")]
+use hashbrown_0_14::HashMap as HashbrownMap014;
+#[cfg(feature = "indexmap_1")]
+use indexmap_1::IndexMap;
+#[cfg(feature = "indexmap_2")]
+use indexmap_2::IndexMap as IndexMap2;
+
+enum GoodOrError<T, TAs> {
+    Good(T),
+    // Only here to consume the TAs generic
+    Error(PhantomData<TAs>),
+}
+
+impl<'de, T, TAs> Deserialize<'de> for GoodOrError<T, TAs>
+where
+    TAs: DeserializeAs<'de, T>,
+{
+    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        let is_hr = deserializer.is_human_readable();
+        let content: content::de::Content<'de> = Deserialize::deserialize(deserializer)?;
+
+        Ok(
+            match <DeserializeAsWrap<T, TAs>>::deserialize(content::de::ContentDeserializer::<
+                D::Error,
+            >::new(content, is_hr))
+            {
+                Ok(elem) => GoodOrError::Good(elem.into_inner()),
+                Err(_) => GoodOrError::Error(PhantomData),
+            },
+        )
+    }
+}
+
+impl<'de, T, U> DeserializeAs<'de, Vec<T>> for VecSkipError<U>
+where
+    U: DeserializeAs<'de, T>,
+{
+    fn deserialize_as<D>(deserializer: D) -> Result<Vec<T>, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        struct SeqVisitor<T, U> {
+            marker: PhantomData<T>,
+            marker2: PhantomData<U>,
+        }
+
+        impl<'de, T, TAs> Visitor<'de> for SeqVisitor<T, TAs>
+        where
+            TAs: DeserializeAs<'de, T>,
+        {
+            type Value = Vec<T>;
+
+            fn expecting(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+                formatter.write_str("a sequence")
+            }
+
+            fn visit_seq<A>(self, seq: A) -> Result<Self::Value, A::Error>
+            where
+                A: SeqAccess<'de>,
+            {
+                utils::SeqIter::new(seq)
+                    .filter_map(|res: Result<GoodOrError<T, TAs>, A::Error>| match res {
+                        Ok(GoodOrError::Good(value)) => Some(Ok(value)),
+                        Ok(GoodOrError::Error(_)) => None,
+                        Err(err) => Some(Err(err)),
+                    })
+                    .collect()
+            }
+        }
+
+        let visitor = SeqVisitor::<T, U> {
+            marker: PhantomData,
+            marker2: PhantomData,
+        };
+        deserializer.deserialize_seq(visitor)
+    }
+}
+
+struct MapSkipErrorVisitor<MAP, K, KAs, V, VAs>(PhantomData<(MAP, K, KAs, V, VAs)>);
+
+impl<'de, MAP, K, KAs, V, VAs> Visitor<'de> for MapSkipErrorVisitor<MAP, K, KAs, V, VAs>
+where
+    MAP: FromIterator<(K, V)>,
+    KAs: DeserializeAs<'de, K>,
+    VAs: DeserializeAs<'de, V>,
+{
+    type Value = MAP;
+
+    fn expecting(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+        formatter.write_str("a map")
+    }
+
+    #[inline]
+    fn visit_map<A>(self, access: A) -> Result<Self::Value, A::Error>
+    where
+        A: MapAccess<'de>,
+    {
+        type KVPair<K, KAs, V, VAs> = (GoodOrError<K, KAs>, GoodOrError<V, VAs>);
+        utils::MapIter::new(access)
+            .filter_map(|res: Result<KVPair<K, KAs, V, VAs>, A::Error>| match res {
+                Ok((GoodOrError::Good(key), GoodOrError::Good(value))) => Some(Ok((key, value))),
+                Ok(_) => None,
+                Err(err) => Some(Err(err)),
+            })
+            .collect()
+    }
+}
+
+#[cfg(feature = "alloc")]
+macro_rules! map_impl {
+    (
+        $ty:ident < K $(: $kbound1:ident $(+ $kbound2:ident)*)?, V $(, $typaram:ident : $bound1:ident $(+ $bound2:ident)*)* >,
+        $with_capacity:expr
+    ) => {
+        impl<'de, K, V, KAs, VAs $(, $typaram)*> DeserializeAs<'de, $ty<K, V $(, $typaram)*>>
+            for MapSkipError<KAs, VAs>
+        where
+            KAs: DeserializeAs<'de, K>,
+            VAs: DeserializeAs<'de, V>,
+            $(K: $kbound1 $(+ $kbound2)*,)?
+            $($typaram: $bound1 $(+ $bound2)*),*
+        {
+            fn deserialize_as<D>(deserializer: D) -> Result<$ty<K, V $(, $typaram)*>, D::Error>
+            where
+                D: Deserializer<'de>,
+            {
+                deserializer.deserialize_map(MapSkipErrorVisitor::<
+                    $ty<K, V $(, $typaram)*>,
+                    K,
+                    KAs,
+                    V,
+                    VAs,
+                >(PhantomData))
+            }
+        }
+    };
+}
+foreach_map!(map_impl);

serde_with/src/guide/serde_as_transformations.md

@@ -11,26 +11,27 @@ This page lists the transformations implemented in this crate and supported by `
 7. [Convert to an intermediate type using `TryInto`](#convert-to-an-intermediate-type-using-tryinto)
 8. [`Default` from `null`](#default-from-null)
 9. [De/Serialize into `Vec`, ignoring errors](#deserialize-into-vec-ignoring-errors)
-10. [De/Serialize with `FromStr` and `Display`](#deserialize-with-fromstr-and-display)
-11. [`Duration` as seconds](#duration-as-seconds)
-12. [Hex encode bytes](#hex-encode-bytes)
-13. [Ignore deserialization errors](#ignore-deserialization-errors)
-14. [`Maps` to `Vec` of enums](#maps-to-vec-of-enums)
-15. [`Maps` to `Vec` of tuples](#maps-to-vec-of-tuples)
-16. [`NaiveDateTime` like UTC timestamp](#naivedatetime-like-utc-timestamp)
-17. [`None` as empty `String`](#none-as-empty-string)
-18. [One or many elements into `Vec`](#one-or-many-elements-into-vec)
-19. [Overwrite existing set values](#overwrite-existing-set-values)
-20. [Pick first successful deserialization](#pick-first-successful-deserialization)
-21. [Prefer the first map key when duplicates exist](#prefer-the-first-map-key-when-duplicates-exist)
-22. [Prevent duplicate map keys](#prevent-duplicate-map-keys)
-23. [Prevent duplicate set values](#prevent-duplicate-set-values)
-24. [Struct fields as map keys](#struct-fields-as-map-keys)
-25. [Timestamps as seconds since UNIX epoch](#timestamps-as-seconds-since-unix-epoch)
-26. [Value into JSON String](#value-into-json-string)
-27. [`Vec` of tuples to `Maps`](#vec-of-tuples-to-maps)
-28. [Well-known time formats for `OffsetDateTime`](#well-known-time-formats-for-offsetdatetime)
-29. [De/Serialize depending on `De/Serializer::is_human_readable`](#deserialize-depending-on-deserializeris_human_readable)
+10. [De/Serialize into a map, ignoring errors](#deserialize-into-a-map-ignoring-errors)
+11. [De/Serialize with `FromStr` and `Display`](#deserialize-with-fromstr-and-display)
+12. [`Duration` as seconds](#duration-as-seconds)
+13. [Hex encode bytes](#hex-encode-bytes)
+14. [Ignore deserialization errors](#ignore-deserialization-errors)
+15. [`Maps` to `Vec` of enums](#maps-to-vec-of-enums)
+16. [`Maps` to `Vec` of tuples](#maps-to-vec-of-tuples)
+17. [`NaiveDateTime` like UTC timestamp](#naivedatetime-like-utc-timestamp)
+18. [`None` as empty `String`](#none-as-empty-string)
+19. [One or many elements into `Vec`](#one-or-many-elements-into-vec)
+20. [Overwrite existing set values](#overwrite-existing-set-values)
+21. [Pick first successful deserialization](#pick-first-successful-deserialization)
+22. [Prefer the first map key when duplicates exist](#prefer-the-first-map-key-when-duplicates-exist)
+23. [Prevent duplicate map keys](#prevent-duplicate-map-keys)
+24. [Prevent duplicate set values](#prevent-duplicate-set-values)
+25. [Struct fields as map keys](#struct-fields-as-map-keys)
+26. [Timestamps as seconds since UNIX epoch](#timestamps-as-seconds-since-unix-epoch)
+27. [Value into JSON String](#value-into-json-string)
+28. [`Vec` of tuples to `Maps`](#vec-of-tuples-to-maps)
+29. [Well-known time formats for `OffsetDateTime`](#well-known-time-formats-for-offsetdatetime)
+30. [De/Serialize depending on `De/Serializer::is_human_readable`](#deserialize-depending-on-deserializeris_human_readable)
 
 ## Base64 encode bytes
 
@@ -181,6 +182,25 @@ colors: Vec<Color>,
 // => vec![Blue, Green]
 ```
 
+## De/Serialize into a map, ignoring errors
+
+[`MapSkipError`]
+
+For formats with heterogeneously typed maps, we can collect only the elements where both key and value are deserializable.
+This is also useful in conjunction to `#[serde(flatten)]` to ignore some entries when capturing additional fields.
+
+```ignore
+// JSON
+"value": {"0": "v0", "5": "v5", "str": "str", "10": 2},
+
+// Rust
+#[serde_as(as = "MapSkipError<DisplayFromStr, _>")]
+value: BTreeMap<u32, String>,
+
+// Only deserializes entries with a numerical key and a string value, i.e.,
+{0 => "v0", 5 => "v5"}
+```
+
 ## De/Serialize with `FromStr` and `Display`
 
 Useful if a type implements `FromStr` / `Display` but not `Deserialize` / `Serialize`.
@@ -614,3 +634,4 @@ value: u128,
 [`TimestampSecondsWithFrac`]: crate::TimestampSecondsWithFrac
 [`TryFromInto`]: crate::TryFromInto
 [`VecSkipError`]: crate::VecSkipError
+[`MapSkipError`]: crate::MapSkipError

serde_with/src/lib.rs

@@ -2205,6 +2205,64 @@ pub struct BorrowCow;
 #[cfg(feature = "alloc")]
 pub struct VecSkipError<T>(PhantomData<T>);
 
+/// Deserialize a map, skipping keys and values which fail to deserialize.
+///
+/// By default serde terminates if it fails to deserialize a key or a value when deserializing
+/// a map. Sometimes a map has heterogeneous keys or values but we only care about some specific
+/// types, and it is desirable to skip entries on errors.
+///
+/// It is especially useful in conjunction to `#[serde(flatten)]` to capture a map mixed in with
+/// other entries which we don't want to exhaust in the type definition.
+///
+/// The serialization behavior is identical to the underlying map.
+///
+/// The implementation supports both the [`HashMap`] and the [`BTreeMap`] from the standard library.
+///
+/// [`BTreeMap`]: std::collections::BTreeMap
+/// [`HashMap`]: std::collections::HashMap
+///
+/// # Examples
+///
+/// ```rust
+/// # #[cfg(feature = "macros")] {
+/// # use serde::{Deserialize, Serialize};
+/// # use std::collections::BTreeMap;
+/// # use serde_with::{serde_as, DisplayFromStr, MapSkipError};
+/// #
+/// #[serde_as]
+/// # #[derive(Debug, PartialEq)]
+/// #[derive(Deserialize, Serialize)]
+/// struct VersionNames {
+///     yanked: Vec<u16>,
+///     #[serde_as(as = "MapSkipError<DisplayFromStr, _>")]
+///     #[serde(flatten)]
+///     names: BTreeMap<u16, String>,
+/// }
+///
+/// let data = VersionNames {
+///     yanked: vec![2, 5],
+///     names: BTreeMap::from_iter([
+///         (0u16, "v0".to_string()),
+///         (1, "v1".to_string()),
+///         (4, "v4".to_string())
+///     ]),
+/// };
+/// let source_json = r#"{
+///   "0": "v0",
+///   "1": "v1",
+///   "4": "v4",
+///   "yanked": [2, 5],
+///   "last_updated": 1704085200
+/// }"#;
+/// let data_json = r#"{"yanked":[2,5],"0":"v0","1":"v1","4":"v4"}"#;
+/// // Ensure serialization and deserialization produce the expected results
+/// assert_eq!(data_json, serde_json::to_string(&data).unwrap());
+/// assert_eq!(data, serde_json::from_str(source_json).unwrap());
+/// # }
+/// ```
+#[cfg(feature = "alloc")]
+pub struct MapSkipError<K, V>(PhantomData<(K, V)>);
+
 /// Deserialize a boolean from a number
 ///
 /// Deserialize a number (of `u8`) and turn it into a boolean.

serde_with/src/ser/impls.rs

@@ -559,19 +559,6 @@ where
     }
 }
 
-#[cfg(feature = "alloc")]
-impl<T, U> SerializeAs<Vec<T>> for VecSkipError<U>
-where
-    U: SerializeAs<T>,
-{
-    fn serialize_as<S>(source: &Vec<T>, serializer: S) -> Result<S::Ok, S::Error>
-    where
-        S: Serializer,
-    {
-        Vec::<U>::serialize_as(source, serializer)
-    }
-}
-
 impl<T> SerializeAs<Option<T>> for NoneAsEmptyString
 where
     T: Display,