Merge pull request #148 from chmp/feature/remove-bytecode-deserializer

Start to imlement basic bytecode-less deserializer

Changes.md

@@ -1,18 +1,47 @@
 # Change log
 
-## 0.10.1
+## 0.11.0
 
+`0.11.0` does contain any anticipated breaking changes. However it's a major
+refactoring and may change some untested behavior.
+
+- Remove the bytecode deserializer and use the serde API directly
+  - Easier to understand and extend
+  - The `Deserialization` implementation can ask for its expected type, e.g.,
+    `chrono::DateTime<Utc>` can now be used with  `serde_arrow` without
+    explicitly specifying the strategy
+- Add `Date32` and `Time64` support
 - Allow to use `arrow` schemas in `SchemaLike::from_value()`, e.g., `let fields
   = Vec::<Field>::from_value(&batch.schema())`.
 - Fix bug in `SchemaLike::from_type()` for nested unions
 
+### Thanks
+
+The following people contributed to this release:
+
+- [@gz](https://github.com/gz) added `Date32` and `Time64` support
+  ([PR](https://github.com/chmp/serde_arrow/pull/147))
+- [@progval](https://github.com/progval) added additional error messages
+  ([PR](https://github.com/chmp/serde_arrow/pull/142))
+
 ## 0.10.0
 
 - Remove deprecated APIs
 - Use the serde serialization APIs directly, instead of using the bytecode
   serializer. Serialization will be about `2x` faster
 - Fix bug in `SchemaLike::from_value` with incorrect strategy deserialization
 
+### Thanks
+
+The following people contributed to this release:
+
+- [@Ten0](https://github.com/Ten0) motivated the rewrite to use the serde API
+  directly and contributed additional benchmarks for JSON transcoding
+  ([PR](https://github.com/chmp/serde_arrow/pull/130))
+- [@alamb](https://github.com/alamb) added improved documentation on how to use
+  `serde_arrow` with the `arrow` crate
+  ([PR](https://github.com/chmp/serde_arrow/pull/131))
+
 ## 0.9.1
 
 - `Decimal128` support: serialize / deserialize
@@ -132,6 +161,13 @@ Bug fixes:
   - nested options (`Option<Option<T>>`)
   - creating `float16` arrays
 
+### Thanks
+
+The following people contributed to this release:
+
+- [@elbaro](https://github.com/elbaro) updated the readme example
+  ([PR](https://github.com/chmp/serde_arrow/pull/33))
+
 ## 0.6.1
 
 - Add support for `arrow=37` with the `arrow-37` feature

serde_arrow/src/_impl/docs/quickstart.rs

@@ -66,8 +66,9 @@
 //! # #[cfg(not(has_arrow))] fn main() { }
 //! ```
 //!
-//! Integer fields containing timestamps in milliseconds since the epoch can be
-//! directly stored as `Date64` without any configuration:
+//! Integer fields containing timestamps in milliseconds since the epoch or
+//! `DateTime<Utc>` objects can be directly stored as `Date64` without any
+//! configuration:
 //!
 //! ```rust
 //! # #[cfg(has_arrow)]

serde_arrow/src/arrow2_impl/api.rs

@@ -7,13 +7,15 @@ use serde::{Deserialize, Serialize};
 use crate::{
     _impl::arrow2::{array::Array, datatypes::Field},
     internal::{
+        common::Mut,
         error::Result,
         schema::{GenericField, SerdeArrowSchema},
-        serialization_ng::OuterSequenceBuilder,
-        source::deserialize_from_source,
+        serialization::OuterSequenceBuilder,
     },
 };
 
+use super::deserialization::build_deserializer;
+
 /// Build arrow2 arrays record by record (*requires one of the `arrow2-*`
 /// features*)
 ///
@@ -176,28 +178,16 @@ where
     T: Deserialize<'de>,
     A: AsRef<dyn Array>,
 {
-    use crate::internal::{
-        common::{BufferExtract, Buffers},
-        deserialization,
-    };
-
     let fields = fields
         .iter()
         .map(GenericField::try_from)
         .collect::<Result<Vec<_>>>()?;
-
-    let num_items = arrays
+    let arrays = arrays
         .iter()
-        .map(|a| a.as_ref().len())
-        .min()
-        .unwrap_or_default();
-
-    let mut buffers = Buffers::new();
-    let mut mappings = Vec::new();
-    for (field, array) in fields.iter().zip(arrays.iter()) {
-        mappings.push(array.as_ref().extract_buffers(field, &mut buffers)?);
-    }
+        .map(|array| array.as_ref())
+        .collect::<Vec<_>>();
 
-    let interpreter = deserialization::compile_deserialization(num_items, &mappings, buffers)?;
-    deserialize_from_source(interpreter)
+    let mut deserializer = build_deserializer(&fields, &arrays)?;
+    let res = T::deserialize(Mut(&mut deserializer))?;
+    Ok(res)
 }