Support alternate formats for unparsing datetime to timestamp and interval (apache#11466)

* Unparser rule for datatime cast (#10)

* use timestamp as the identifier for date64

* rename

* implement CustomDialectBuilder

* fix

* dialect with interval style (#11)

---------

Co-authored-by: Phillip LeBlanc <phillip@leblanc.tech>

* fmt

* clippy

* doc

* Update datafusion/sql/src/unparser/expr.rs

Co-authored-by: Andrew Lamb <andrew@nerdnetworks.org>

* update the doc for CustomDialectBuilder

* fix doc test

---------

Co-authored-by: Phillip LeBlanc <phillip@leblanc.tech>
Co-authored-by: Andrew Lamb <andrew@nerdnetworks.org>

Author: 3 people

datafusion-examples/examples/plan_to_sql.rs

@@ -19,7 +19,7 @@ use datafusion::error::Result;
 
 use datafusion::prelude::*;
 use datafusion::sql::unparser::expr_to_sql;
-use datafusion_sql::unparser::dialect::CustomDialect;
+use datafusion_sql::unparser::dialect::CustomDialectBuilder;
 use datafusion_sql::unparser::{plan_to_sql, Unparser};
 
 /// This example demonstrates the programmatic construction of SQL strings using
@@ -80,7 +80,9 @@ fn simple_expr_to_pretty_sql_demo() -> Result<()> {
 /// using a custom dialect and an explicit unparser
 fn simple_expr_to_sql_demo_escape_mysql_style() -> Result<()> {
     let expr = col("a").lt(lit(5)).or(col("a").eq(lit(8)));
-    let dialect = CustomDialect::new(Some('`'));
+    let dialect = CustomDialectBuilder::new()
+        .with_identifier_quote_style('`')
+        .build();
     let unparser = Unparser::new(&dialect);
     let sql = unparser.expr_to_sql(&expr)?.to_string();
     assert_eq!(sql, r#"((`a` < 5) OR (`a` = 8))"#);

datafusion/sql/src/unparser/dialect.rs

@@ -35,7 +35,33 @@ pub trait Dialect {
     fn supports_nulls_first_in_sort(&self) -> bool {
         true
     }
+
+    // Does the dialect use TIMESTAMP to represent Date64 rather than DATETIME?
+    // E.g. Trino, Athena and Dremio does not have DATETIME data type
+    fn use_timestamp_for_date64(&self) -> bool {
+        false
+    }
+
+    fn interval_style(&self) -> IntervalStyle {
+        IntervalStyle::PostgresVerbose
+    }
 }
+
+/// `IntervalStyle` to use for unparsing
+///
+/// <https://www.postgresql.org/docs/current/datatype-datetime.html#DATATYPE-INTERVAL-INPUT>
+/// different DBMS follows different standards, popular ones are:
+/// postgres_verbose: '2 years 15 months 100 weeks 99 hours 123456789 milliseconds' which is
+/// compatible with arrow display format, as well as duckdb
+/// sql standard format is '1-2' for year-month, or '1 10:10:10.123456' for day-time
+/// <https://www.contrib.andrew.cmu.edu/~shadow/sql/sql1992.txt>
+#[derive(Clone, Copy)]
+pub enum IntervalStyle {
+    PostgresVerbose,
+    SQLStandard,
+    MySQL,
+}
+
 pub struct DefaultDialect {}
 
 impl Dialect for DefaultDialect {
@@ -57,6 +83,10 @@ impl Dialect for PostgreSqlDialect {
     fn identifier_quote_style(&self, _: &str) -> Option<char> {
         Some('"')
     }
+
+    fn interval_style(&self) -> IntervalStyle {
+        IntervalStyle::PostgresVerbose
+    }
 }
 
 pub struct MySqlDialect {}
@@ -69,6 +99,10 @@ impl Dialect for MySqlDialect {
     fn supports_nulls_first_in_sort(&self) -> bool {
         false
     }
+
+    fn interval_style(&self) -> IntervalStyle {
+        IntervalStyle::MySQL
+    }
 }
 
 pub struct SqliteDialect {}
@@ -81,12 +115,29 @@ impl Dialect for SqliteDialect {
 
 pub struct CustomDialect {
     identifier_quote_style: Option<char>,
+    supports_nulls_first_in_sort: bool,
+    use_timestamp_for_date64: bool,
+    interval_style: IntervalStyle,
+}
+
+impl Default for CustomDialect {
+    fn default() -> Self {
+        Self {
+            identifier_quote_style: None,
+            supports_nulls_first_in_sort: true,
+            use_timestamp_for_date64: false,
+            interval_style: IntervalStyle::SQLStandard,
+        }
+    }
 }
 
 impl CustomDialect {
+    // create a CustomDialect
+    #[deprecated(note = "please use `CustomDialectBuilder` instead")]
     pub fn new(identifier_quote_style: Option<char>) -> Self {
         Self {
             identifier_quote_style,
+            ..Default::default()
         }
     }
 }
@@ -95,4 +146,93 @@ impl Dialect for CustomDialect {
     fn identifier_quote_style(&self, _: &str) -> Option<char> {
         self.identifier_quote_style
     }
+
+    fn supports_nulls_first_in_sort(&self) -> bool {
+        self.supports_nulls_first_in_sort
+    }
+
+    fn use_timestamp_for_date64(&self) -> bool {
+        self.use_timestamp_for_date64
+    }
+
+    fn interval_style(&self) -> IntervalStyle {
+        self.interval_style
+    }
+}
+
+/// `CustomDialectBuilder` to build `CustomDialect` using builder pattern
+///
+///
+/// # Examples
+///
+/// Building a custom dialect with all default options set in CustomDialectBuilder::new()
+/// but with `use_timestamp_for_date64` overridden to `true`
+///
+/// ```
+/// use datafusion_sql::unparser::dialect::CustomDialectBuilder;
+/// let dialect = CustomDialectBuilder::new()
+///     .with_use_timestamp_for_date64(true)
+///     .build();
+/// ```
+pub struct CustomDialectBuilder {
+    identifier_quote_style: Option<char>,
+    supports_nulls_first_in_sort: bool,
+    use_timestamp_for_date64: bool,
+    interval_style: IntervalStyle,
+}
+
+impl Default for CustomDialectBuilder {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl CustomDialectBuilder {
+    pub fn new() -> Self {
+        Self {
+            identifier_quote_style: None,
+            supports_nulls_first_in_sort: true,
+            use_timestamp_for_date64: false,
+            interval_style: IntervalStyle::PostgresVerbose,
+        }
+    }
+
+    pub fn build(self) -> CustomDialect {
+        CustomDialect {
+            identifier_quote_style: self.identifier_quote_style,
+            supports_nulls_first_in_sort: self.supports_nulls_first_in_sort,
+            use_timestamp_for_date64: self.use_timestamp_for_date64,
+            interval_style: self.interval_style,
+        }
+    }
+
+    /// Customize the dialect with a specific identifier quote style, e.g. '`', '"'
+    pub fn with_identifier_quote_style(mut self, identifier_quote_style: char) -> Self {
+        self.identifier_quote_style = Some(identifier_quote_style);
+        self
+    }
+
+    /// Customize the dialect to supports `NULLS FIRST` in `ORDER BY` clauses
+    pub fn with_supports_nulls_first_in_sort(
+        mut self,
+        supports_nulls_first_in_sort: bool,
+    ) -> Self {
+        self.supports_nulls_first_in_sort = supports_nulls_first_in_sort;
+        self
+    }
+
+    /// Customize the dialect to uses TIMESTAMP when casting Date64 rather than DATETIME
+    pub fn with_use_timestamp_for_date64(
+        mut self,
+        use_timestamp_for_date64: bool,
+    ) -> Self {
+        self.use_timestamp_for_date64 = use_timestamp_for_date64;
+        self
+    }
+
+    /// Customize the dialect with a specific interval style listed in `IntervalStyle`
+    pub fn with_interval_style(mut self, interval_style: IntervalStyle) -> Self {
+        self.interval_style = interval_style;
+        self
+    }
 }