477: feat: Allow types to provide default values

avro/src/serde/derive.rs

@@ -81,6 +81,10 @@ use std::collections::{HashMap, HashSet};
 ///
 ///    Set the `doc` attribute of the schema. Defaults to the documentation of the type.
 ///
+///  - `#[avro(default = r#"{"field": 42, "other": "Spam"}"#)]`
+///
+///    Provide the default value for this type when it is used in a field.
+///
 ///  - `#[avro(alias = "name")]`
 ///
 ///    Set the `alias` attribute of the schema. Can be specified multiple times.
@@ -113,11 +117,22 @@ use std::collections::{HashMap, HashSet};
 ///
 ///    Set the `doc` attribute of the field. Defaults to the documentation of the field.
 ///
-///  - `#[avro(default = "null")]`
+///  - `#[avro(default = ..)]`
+///
+///    Control the `default` attribute of the field. When not used, it will use [`AvroSchemaComponent::field_default`]
+///    to get the default value for a type. To remove the `default` attribute for a field, set `default` to `false`: `#[avro(default = false)]`.
 ///
-///    Set the `default` attribute of the field.
+///    To override or set a default value, provide a JSON string:
 ///
-///    _Note:_ This is a JSON value not a Rust value, as this is put in the schema itself.
+///      - Null: `#[avro(default = "null")]`
+///      - Boolean: `#[avro(default = "true")]`.
+///      - Number: `#[avro(default = "42")]` or `#[avro(default = "42.5")]`
+///      - String: `#[avro(default = r#""String needs extra quotes""#)]`.
+///      - Array: `#[avro(default = r#"["One", "Two", "Three"]"#)]`.
+///      - Object: `#[avro(default = r#"{"One": 1}"#)]`.
+///
+///    See [the specification](https://avro.apache.org/docs/++version++/specification/#schema-record)
+///    for details on how to map a type to a JSON value.
 ///
 ///  - `#[serde(alias = "name")]`
 ///
@@ -220,6 +235,11 @@ pub trait AvroSchema {
 ///     fn get_record_fields_in_ctxt(_: usize, _: &mut HashSet<Name>, _: &Namespace) -> Option<Vec<RecordField>> {
 ///         None // A Schema::Int is not a Schema::Record so there are no fields to return
 ///     }
+///
+///     fn field_default() -> Option<serde_json::Value> {
+///         // Zero as default value. Can also be None if you don't want to provide a default value
+///         Some(0u8.into())
+///     }
 ///}
 /// ```
 ///
@@ -242,6 +262,10 @@ pub trait AvroSchema {
 ///     fn get_record_fields_in_ctxt(first_field_position: usize, named_schemas: &mut HashSet<Name>, enclosing_namespace: &Namespace) -> Option<Vec<RecordField>> {
 ///         T::get_record_fields_in_ctxt(first_field_position, named_schemas, enclosing_namespace)
 ///     }
+///
+///     fn field_default() -> Option<serde_json::Value> {
+///         T::field_default()
+///     }
 ///}
 /// ```
 ///
@@ -256,6 +280,7 @@ pub trait AvroSchema {
 ///  - Implement `get_record_fields_in_ctxt` as the default implementation has to be implemented
 ///    with backtracking and a lot of cloning.
 ///      - Even if your schema is not a record, still implement the function and just return `None`
+///  - Implement `field_default()` if you want to use `#[serde(skip_serializing{,_if})]`.
 ///
 /// ```
 /// # use apache_avro::{Schema, serde::{AvroSchemaComponent}, schema::{Name, Namespace, RecordField, RecordSchema}};
@@ -305,6 +330,11 @@ pub trait AvroSchema {
 ///                 .build(),
 ///         ])
 ///     }
+///
+///     fn field_default() -> Option<serde_json::Value> {
+///         // This type does not provide a default value
+///         None
+///     }
 ///}
 /// ```
 pub trait AvroSchemaComponent {
@@ -332,6 +362,16 @@ pub trait AvroSchemaComponent {
             Self::get_schema_in_ctxt,
         )
     }
+
+    /// The default value of this type when used for a record field.
+    ///
+    /// `None` means no default value, which is also the default implementation.
+    ///
+    /// Implementations of this trait provided by this crate return `None` except for `Option<T>`
+    /// which returns `Some(serde_json::Value::Null)`.
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 /// Get the record fields from `schema_fn` without polluting `named_schemas` or causing duplicate names
@@ -487,6 +527,10 @@ macro_rules! impl_schema (
             fn get_record_fields_in_ctxt(_: usize, _: &mut HashSet<Name>, _: &Namespace) -> Option<Vec<RecordField>> {
                 None
             }
+
+            fn field_default() -> Option<serde_json::Value> {
+                None
+            }
         }
     );
 );
@@ -515,6 +559,10 @@ macro_rules! impl_passthrough_schema (
             fn get_record_fields_in_ctxt(first_field_position: usize, named_schemas: &mut HashSet<Name>, enclosing_namespace: &Namespace) -> Option<Vec<RecordField>> {
                 T::get_record_fields_in_ctxt(first_field_position, named_schemas, enclosing_namespace)
             }
+
+            fn field_default() -> Option<serde_json::Value> {
+                T::field_default()
+            }
         }
     );
 );
@@ -535,6 +583,10 @@ macro_rules! impl_array_schema (
             fn get_record_fields_in_ctxt(_: usize, _: &mut HashSet<Name>, _: &Namespace) -> Option<Vec<RecordField>> {
                 None
             }
+
+            fn field_default() -> Option<serde_json::Value> {
+                None
+            }
         }
     );
 );
@@ -562,6 +614,11 @@ where
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    /// If `T` has a field default, this will return an array of elements with that default. Otherwise there is no default.
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 impl<T> AvroSchemaComponent for HashMap<String, T>
@@ -582,6 +639,10 @@ where
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 impl<T> AvroSchemaComponent for Option<T>
@@ -609,6 +670,10 @@ where
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    fn field_default() -> Option<serde_json::Value> {
+        Some(serde_json::Value::Null)
+    }
 }
 
 impl AvroSchemaComponent for core::time::Duration {
@@ -644,6 +709,10 @@ impl AvroSchemaComponent for core::time::Duration {
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 impl AvroSchemaComponent for uuid::Uuid {
@@ -679,6 +748,10 @@ impl AvroSchemaComponent for uuid::Uuid {
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 impl AvroSchemaComponent for u64 {
@@ -712,6 +785,10 @@ impl AvroSchemaComponent for u64 {
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 impl AvroSchemaComponent for u128 {
@@ -745,6 +822,10 @@ impl AvroSchemaComponent for u128 {
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 impl AvroSchemaComponent for i128 {
@@ -778,12 +859,18 @@ impl AvroSchemaComponent for i128 {
     ) -> Option<Vec<RecordField>> {
         None
     }
+
+    fn field_default() -> Option<serde_json::Value> {
+        None
+    }
 }
 
 #[cfg(test)]
 mod tests {
-    use crate::schema::{FixedSchema, Name};
-    use crate::{AvroSchema, Schema};
+    use crate::{
+        AvroSchema, Schema,
+        schema::{FixedSchema, Name},
+    };
     use apache_avro_test_helper::TestResult;
 
     #[test]

avro/src/types.rs

@@ -766,6 +766,13 @@ impl Value {
                 }
                 Value::Uuid(Uuid::from_slice(bytes).map_err(Details::ConvertSliceToUuid)?)
             }
+            (Value::String(ref string), UuidSchema::Fixed(_)) => {
+                let bytes = string.as_bytes();
+                if bytes.len() != 16 {
+                    return Err(Details::ConvertFixedToUuid(bytes.len()).into());
+                }
+                Value::Uuid(Uuid::from_slice(bytes).map_err(Details::ConvertSliceToUuid)?)
+            }
             (other, _) => return Err(Details::GetUuid(other).into()),
         };
         Ok(value)

avro_derive/src/attributes/avro.rs

@@ -21,6 +21,7 @@
 //! Although a user will mostly use the Serde attributes, there are some Avro specific attributes
 //! a user can use. These add extra metadata to the generated schema.
 
+use crate::attributes::FieldDefault;
 use crate::case::RenameRule;
 use darling::FromMeta;
 use proc_macro2::Span;
@@ -53,6 +54,9 @@ pub struct ContainerAttributes {
     /// [`serde::ContainerAttributes::rename_all`]: crate::attributes::serde::ContainerAttributes::rename_all
     #[darling(default)]
     pub rename_all: RenameRule,
+    /// Set the default value if this schema is used as a field
+    #[darling(default)]
+    pub default: Option<String>,
 }
 
 impl ContainerAttributes {
@@ -125,7 +129,7 @@ pub struct FieldAttributes {
     ///
     /// This is also used as the default when `skip_serializing{_if}` is used.
     #[darling(default)]
-    pub default: Option<String>,
+    pub default: FieldDefault,
     /// Deprecated. Use [`serde::FieldAttributes::alias`] instead.
     ///
     /// Adds the `aliases` field to the schema.

avro_derive/src/attributes/mod.rs

@@ -17,7 +17,8 @@
 
 use crate::case::RenameRule;
 use darling::{FromAttributes, FromMeta};
-use proc_macro2::Span;
+use proc_macro2::{Span, TokenStream};
+use quote::quote;
 use syn::{AttrStyle, Attribute, Expr, Ident, Path, spanned::Spanned};
 
 mod avro;
@@ -30,6 +31,7 @@ pub struct NamedTypeOptions {
     pub aliases: Vec<String>,
     pub rename_all: RenameRule,
     pub transparent: bool,
+    pub default: TokenStream,
 }
 
 impl NamedTypeOptions {
@@ -116,12 +118,29 @@ impl NamedTypeOptions {
 
         let doc = avro.doc.or_else(|| extract_rustdoc(attributes));
 
+        let default = match avro.default {
+            None => quote! { None },
+            Some(default_value) => {
+                let _: serde_json::Value =
+                    serde_json::from_str(&default_value[..]).map_err(|e| {
+                        vec![syn::Error::new(
+                            ident.span(),
+                            format!("Invalid Avro `default` JSON: \n{e}"),
+                        )]
+                    })?;
+                quote! {
+                    Some(serde_json::from_str(#default_value).expect(format!("Invalid JSON: {:?}", #default_value).as_str()))
+                }
+            }
+        };
+
         Ok(Self {
             name: full_schema_name,
             doc,
             aliases: avro.alias,
             rename_all: serde.rename_all.serialize,
             transparent: serde.transparent,
+            default,
         })
     }
 }
@@ -210,11 +229,38 @@ impl With {
         }
     }
 }
+/// How to get the default value for a value.
+#[derive(Debug, PartialEq, Default)]
+pub enum FieldDefault {
+    /// Use `<T as AvroSchemaComponent>::field_default`.
+    #[default]
+    Trait,
+    /// Don't set a default.
+    Disabled,
+    /// Use this JSON value.
+    Value(String),
+}
+
+impl FromMeta for FieldDefault {
+    fn from_string(value: &str) -> darling::Result<Self> {
+        Ok(Self::Value(value.to_string()))
+    }
+
+    fn from_bool(value: bool) -> darling::Result<Self> {
+        if value {
+            Err(darling::Error::custom(
+                "Expected `false` or a JSON string, got `true`",
+            ))
+        } else {
+            Ok(Self::Disabled)
+        }
+    }
+}
 
 #[derive(Default)]
 pub struct FieldOptions {
     pub doc: Option<String>,
-    pub default: Option<String>,
+    pub default: FieldDefault,
     pub alias: Vec<String>,
     pub rename: Option<String>,
     pub skip: bool,
@@ -274,11 +320,11 @@ impl FieldOptions {
         }
         if ((serde.skip_serializing && !serde.skip_deserializing)
             || serde.skip_serializing_if.is_some())
-            && avro.default.is_none()
+            && avro.default == FieldDefault::Disabled
         {
             errors.push(syn::Error::new(
                 span,
-                r#"`#[serde(skip_serializing)]` and `#[serde(skip_serializing_if)]` require `#[avro(default = "..")]`"#
+                "`#[serde(skip_serializing)]` and `#[serde(skip_serializing_if)]` are incompatible with `#[avro(default = false)]`"
             ));
         }
         let with = match With::from_avro_and_serde(&avro.with, &serde.with, span) {