tobixdev commented on code in PR #18552: URL: https://github.com/apache/datafusion/pull/18552#discussion_r2514530669
########## datafusion/common/src/types/canonical.rs: ########## @@ -0,0 +1,198 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +use crate::error::_internal_err; +use crate::types::{ + LogicalType, NativeType, TypeParameter, TypeSignature, ValuePrettyPrinter, +}; +use crate::ScalarValue; +use crate::{Result, _internal_datafusion_err}; +use arrow_schema::extension::{ExtensionType, Opaque, Uuid}; +use std::sync::{Arc, LazyLock}; +use uuid::Bytes; + +impl LogicalType for Uuid { + fn native(&self) -> &NativeType { + &NativeType::FixedSizeBinary(16) + } + + fn signature(&self) -> TypeSignature<'_> { + TypeSignature::Extension { + name: Uuid::NAME, + parameters: vec![], + } + } + + fn pretty_printer(&self) -> &Arc<dyn ValuePrettyPrinter> { + static PRETTY_PRINTER: LazyLock<Arc<dyn ValuePrettyPrinter>> = + LazyLock::new(|| Arc::new(UuidValuePrettyPrinter {})); + &PRETTY_PRINTER + } +} + +#[derive(Debug, Default, Clone, Copy, PartialEq, Eq, Hash)] +struct UuidValuePrettyPrinter; + +impl ValuePrettyPrinter for UuidValuePrettyPrinter { + fn pretty_print_scalar(&self, value: &ScalarValue) -> Result<String> { + match value { + ScalarValue::FixedSizeBinary(16, value) => match value { + Some(value) => { + let bytes = Bytes::try_from(value.as_slice()).map_err(|_| { + _internal_datafusion_err!( + "Invalid UUID bytes even though type is correct." + ) + })?; + let uuid = uuid::Uuid::from_bytes(bytes); + Ok(format!("arrow.uuid({uuid})")) + } + None => Ok("arrow.uuid(NULL)".to_owned()), + }, + _ => _internal_err!("Wrong scalar given to "), + } + } +} + +/// Represents the canonical [Opaque extension type](https://arrow.apache.org/docs/format/CanonicalExtensions.html#opaque). +/// +/// In the context of DataFusion, a common use case of the opaque type is when an extension type +/// is unknown to DataFusion. Contrary to [UnresolvedExtensionType], the extension type has +/// already been checked against the extension type registry and was not found. +impl LogicalType for Opaque { + fn native(&self) -> &NativeType { + &NativeType::FixedSizeBinary(16) + } + + fn signature(&self) -> TypeSignature<'_> { + let parameter = TypeParameter::Type(TypeSignature::Extension { + name: self.metadata().type_name(), + parameters: vec![], + }); + TypeSignature::Extension { + name: Opaque::NAME, + parameters: vec![parameter], + } + } + + fn pretty_printer(&self) -> &Arc<dyn ValuePrettyPrinter> { + static PRETTY_PRINTER: LazyLock<Arc<dyn ValuePrettyPrinter>> = + LazyLock::new(|| Arc::new(OpaqueValuePrettyPrinter {})); + &PRETTY_PRINTER + } +} + +// TODO Other canonical extension types. + +#[derive(Debug, Default, Clone, Copy, PartialEq, Eq, Hash)] +struct OpaqueValuePrettyPrinter; + +impl ValuePrettyPrinter for OpaqueValuePrettyPrinter { + fn pretty_print_scalar(&self, value: &ScalarValue) -> Result<String> { + Ok(format!("arrow.opaque({value})")) + } +} + +/// Represents an unresolved extension type with a given native type and name. +/// +/// This does not necessarily indicate that DataFusion does not understand the extension type. For +/// this purpose, see [OpaqueType]. However, it does indicate that the extension type was not yet +/// checked against the extension type registry. +/// +/// This extension type exists because it is often challenging to gain access to an extension type +/// registry. Especially because extension type support is relatively new, and therefore this +/// consideration was not taken into account by users. This provides a workaround such that +/// unresolved extension types can be resolved at a later point in time where access to the registry +/// is available. +pub struct UnresolvedExtensionType { + /// The name of the underlying extension type. + name: String, + /// The metadata of the underlying extension type. + metadata: Option<String>, + /// The underlying native type. + native_type: NativeType, +} + +impl UnresolvedExtensionType { + /// Creates a new [UnresolvedExtensionType]. + pub fn new(name: String, metadata: Option<String>, native_type: NativeType) -> Self { + Self { + name, + metadata, + native_type, + } + } + + /// The name of the unresolved extension type. + pub fn name(&self) -> &str { + &self.name + } + + /// The metadata of the unresolved extension type. + pub fn metadata(&self) -> Option<&str> { + self.metadata.as_deref() + } +} + +impl LogicalType for UnresolvedExtensionType { + fn native(&self) -> &NativeType { + &self.native_type + } + + fn signature(&self) -> TypeSignature<'_> { + let inner_type = TypeParameter::Type(TypeSignature::Extension { + name: &self.name, + parameters: vec![], + }); + TypeSignature::Extension { + name: "datafusion.unresolved", + parameters: vec![inner_type], + } + } + + fn pretty_printer(&self) -> &Arc<dyn ValuePrettyPrinter> { + static PRETTY_PRINTER: LazyLock<Arc<dyn ValuePrettyPrinter>> = + LazyLock::new(|| Arc::new(UnresolvedValuePrettyPrinter {})); + &PRETTY_PRINTER + } +} + +#[derive(Debug, Default, Clone, Copy, PartialEq, Eq, Hash)] +struct UnresolvedValuePrettyPrinter {} + +impl ValuePrettyPrinter for UnresolvedValuePrettyPrinter { + fn pretty_print_scalar(&self, value: &ScalarValue) -> Result<String> { + Ok(format!("datafusion.unresolved({value})")) + } Review Comment: I am not sure. I think this will depend on how we want to handle the whole unresolved situation. I was more thinking into the direction that we have a logical plan invariant (maybe only on the executable level) that ensures that no unresolved type is part of the plan / expressions when creating the execution plan. I think pretty printing is somewhat of a special case as it doesn't directly introduce wrong behavior and printing a value in, for example, error messages should maybe include the information that it is unresolved. Once we introduce extensions necessary for planning (e.g., the sorting stuff I was experimenting with), an error should probably be returned as you mentioned. Unifying with `Opaque` may be another possibility. Maybe distinguishing between unresolved and unknown isn't necessary. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
