Simplify display format of AggregateFunctionExpr, add Expr::sql_name (#15253)

* Simplify display format of `AggregateFunctionExpr`

* add more info for SqlDisplay

* simplify aggr expr

* add doc

* add table name

* fix issues

* Update datafusion/physical-expr/src/aggregate.rs

Co-authored-by: Andrew Lamb <[email protected]>

* Update datafusion/physical-expr/src/aggregate.rs

Co-authored-by: Andrew Lamb <[email protected]>

* Update datafusion/physical-plan/src/aggregates/mod.rs

Co-authored-by: Andrew Lamb <[email protected]>

* Update datafusion/physical-plan/src/aggregates/mod.rs

Co-authored-by: Andrew Lamb <[email protected]>

* fmt

* fix build

* improve docs

* Add documentation about sql_format

* Rename to `human_display`

---------

Co-authored-by: Andrew Lamb <[email protected]>

Author: irenjj

datafusion/core/src/physical_planner.rs

@@ -1588,6 +1588,7 @@ type AggregateExprWithOptionalArgs = (
 pub fn create_aggregate_expr_with_name_and_maybe_filter(
     e: &Expr,
     name: Option<String>,
+    human_displan: String,
     logical_input_schema: &DFSchema,
     physical_input_schema: &Schema,
     execution_props: &ExecutionProps,
@@ -1642,6 +1643,7 @@ pub fn create_aggregate_expr_with_name_and_maybe_filter(
                         .order_by(ordering_reqs)
                         .schema(Arc::new(physical_input_schema.to_owned()))
                         .alias(name)
+                        .human_display(human_displan)
                         .with_ignore_nulls(ignore_nulls)
                         .with_distinct(*distinct)
                         .build()
@@ -1664,15 +1666,22 @@ pub fn create_aggregate_expr_and_maybe_filter(
     execution_props: &ExecutionProps,
 ) -> Result<AggregateExprWithOptionalArgs> {
     // unpack (nested) aliased logical expressions, e.g. "sum(col) as total"
-    let (name, e) = match e {
-        Expr::Alias(Alias { expr, name, .. }) => (Some(name.clone()), expr.as_ref()),
-        Expr::AggregateFunction(_) => (Some(e.schema_name().to_string()), e),
-        _ => (None, e),
+    let (name, human_display, e) = match e {
+        Expr::Alias(Alias { expr, name, .. }) => {
+            (Some(name.clone()), String::default(), expr.as_ref())
+        }
+        Expr::AggregateFunction(_) => (
+            Some(e.schema_name().to_string()),
+            e.human_display().to_string(),
+            e,
+        ),
+        _ => (None, String::default(), e),
     };
 
     create_aggregate_expr_with_name_and_maybe_filter(
         e,
         name,
+        human_display,
         logical_input_schema,
         physical_input_schema,
         execution_props,

datafusion/expr/src/expr.rs

@@ -64,6 +64,15 @@ use sqlparser::ast::{
 ///
 /// [`ExprFunctionExt`]: crate::expr_fn::ExprFunctionExt
 ///
+/// # Printing Expressions
+///
+/// You can print `Expr`s using the the `Debug` trait, `Display` trait, or
+/// [`Self::human_display`]. See the [examples](#examples-displaying-exprs) below.
+///
+/// If you need  SQL to pass to other systems, consider using [`Unparser`].
+///
+/// [`Unparser`]: https://docs.rs/datafusion/latest/datafusion/sql/unparser/struct.Unparser.html
+///
 /// # Schema Access
 ///
 /// See [`ExprSchemable::get_type`] to access the [`DataType`] and nullability
@@ -76,9 +85,9 @@ use sqlparser::ast::{
 /// `Expr` and [`TreeNode::transform`] can be used to rewrite an expression. See
 /// the examples below and [`TreeNode`] for more information.
 ///
-/// # Examples
+/// # Examples: Creating and Using `Expr`s
 ///
-/// ## Column references and literals
+/// ## Column References and Literals
 ///
 /// [`Expr::Column`] refer to the values of columns and are often created with
 /// the [`col`] function. For example to create an expression `c1` referring to
@@ -104,6 +113,7 @@ use sqlparser::ast::{
 /// // All literals are strongly typed in DataFusion. To make an `i64` 42:
 /// let expr = lit(42i64);
 /// assert_eq!(expr, Expr::Literal(ScalarValue::Int64(Some(42))));
+/// assert_eq!(expr, Expr::Literal(ScalarValue::Int64(Some(42))));
 /// // To make a (typed) NULL:
 /// let expr = Expr::Literal(ScalarValue::Int64(None));
 /// // to make an (untyped) NULL (the optimizer will coerce this to the correct type):
@@ -171,7 +181,51 @@ use sqlparser::ast::{
 /// ]);
 /// ```
 ///
-/// # Visiting and Rewriting `Expr`s
+/// # Examples: Displaying `Exprs`
+///
+/// There are three ways to print an `Expr` depending on the usecase.
+///
+/// ## Use `Debug` trait
+///
+/// Following Rust conventions, the `Debug` implementation prints out the
+/// internal structure of the expression, which is useful for debugging.
+///
+/// ```
+/// # use datafusion_expr::{lit, col};
+/// let expr = col("c1") + lit(42);
+/// assert_eq!(format!("{expr:?}"), "BinaryExpr(BinaryExpr { left: Column(Column { relation: None, name: \"c1\" }), op: Plus, right: Literal(Int32(42)) })");
+/// ```
+///
+/// ## Use the `Display` trait  (detailed expression)
+///
+/// The `Display` implementation prints out the expression in a SQL-like form,
+/// but has additional details such as the data type of literals. This is useful
+/// for understanding the expression in more detail and is used for the low level
+/// [`ExplainFormat::Indent`] explain plan format.
+///
+/// [`ExplainFormat::Indent`]: crate::logical_plan::ExplainFormat::Indent
+///
+/// ```
+/// # use datafusion_expr::{lit, col};
+/// let expr = col("c1") + lit(42);
+/// assert_eq!(format!("{expr}"), "c1 + Int32(42)");
+/// ```
+///
+/// ## Use [`Self::human_display`] (human readable)
+///
+/// [`Self::human_display`]  prints out the expression in a SQL-like form, optimized
+/// for human consumption by end users. It is used for the
+/// [`ExplainFormat::Tree`] explain plan format.
+///
+/// [`ExplainFormat::Tree`]: crate::logical_plan::ExplainFormat::Tree
+///
+///```
+/// # use datafusion_expr::{lit, col};
+/// let expr = col("c1") + lit(42);
+/// assert_eq!(format!("{}", expr.human_display()), "c1 + 42");
+/// ```
+///
+/// # Examples: Visiting and Rewriting `Expr`s
 ///
 /// Here is an example that finds all literals in an `Expr` tree:
 /// ```
@@ -1147,6 +1201,31 @@ impl Expr {
         SchemaDisplay(self)
     }
 
+    /// Human readable display formatting for this expression.
+    ///
+    /// This function is primarily used in printing the explain tree output,
+    /// (e.g. `EXPLAIN FORMAT TREE <query>`), providing a readable format to
+    /// show how expressions are used in physical and logical plans. See the
+    /// [`Expr`] for other ways to format expressions
+    ///
+    /// Note this format is intended for human consumption rather than SQL for
+    /// other systems. If you need  SQL to pass to other systems, consider using
+    /// [`Unparser`].
+    ///
+    /// [`Unparser`]: https://docs.rs/datafusion/latest/datafusion/sql/unparser/struct.Unparser.html
+    ///
+    /// # Example
+    /// ```
+    /// # use datafusion_expr::{col, lit};
+    /// let expr = col("foo") + lit(42);
+    /// // For EXPLAIN output:
+    /// // "foo + 42"
+    /// println!("{}", expr.human_display());
+    /// ```
+    pub fn human_display(&self) -> impl Display + '_ {
+        SqlDisplay(self)
+    }
+
     /// Returns the qualifier and the schema name of this expression.
     ///
     /// Used when the expression forms the output field of a certain plan.
@@ -2596,6 +2675,187 @@ impl Display for SchemaDisplay<'_> {
     }
 }
 
+/// A helper struct for displaying an `Expr` as an SQL-like string.
+struct SqlDisplay<'a>(&'a Expr);
+
+impl Display for SqlDisplay<'_> {
+    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
+        match self.0 {
+            Expr::Literal(scalar) => scalar.fmt(f),
+            Expr::Alias(Alias { name, .. }) => write!(f, "{name}"),
+            Expr::Between(Between {
+                expr,
+                negated,
+                low,
+                high,
+            }) => {
+                if *negated {
+                    write!(
+                        f,
+                        "{} NOT BETWEEN {} AND {}",
+                        SqlDisplay(expr),
+                        SqlDisplay(low),
+                        SqlDisplay(high),
+                    )
+                } else {
+                    write!(
+                        f,
+                        "{} BETWEEN {} AND {}",
+                        SqlDisplay(expr),
+                        SqlDisplay(low),
+                        SqlDisplay(high),
+                    )
+                }
+            }
+            Expr::BinaryExpr(BinaryExpr { left, op, right }) => {
+                write!(f, "{} {op} {}", SqlDisplay(left), SqlDisplay(right),)
+            }
+            Expr::Case(Case {
+                expr,
+                when_then_expr,
+                else_expr,
+            }) => {
+                write!(f, "CASE ")?;
+
+                if let Some(e) = expr {
+                    write!(f, "{} ", SqlDisplay(e))?;
+                }
+
+                for (when, then) in when_then_expr {
+                    write!(f, "WHEN {} THEN {} ", SqlDisplay(when), SqlDisplay(then),)?;
+                }
+
+                if let Some(e) = else_expr {
+                    write!(f, "ELSE {} ", SqlDisplay(e))?;
+                }
+
+                write!(f, "END")
+            }
+            Expr::Cast(Cast { expr, .. }) | Expr::TryCast(TryCast { expr, .. }) => {
+                write!(f, "{}", SqlDisplay(expr))
+            }
+            Expr::InList(InList {
+                expr,
+                list,
+                negated,
+            }) => {
+                write!(
+                    f,
+                    "{}{} IN {}",
+                    SqlDisplay(expr),
+                    if *negated { " NOT" } else { "" },
+                    ExprListDisplay::comma_separated(list.as_slice())
+                )
+            }
+            Expr::GroupingSet(GroupingSet::Cube(exprs)) => {
+                write!(
+                    f,
+                    "ROLLUP ({})",
+                    ExprListDisplay::comma_separated(exprs.as_slice())
+                )
+            }
+            Expr::GroupingSet(GroupingSet::GroupingSets(lists_of_exprs)) => {
+                write!(f, "GROUPING SETS (")?;
+                for exprs in lists_of_exprs.iter() {
+                    write!(
+                        f,
+                        "({})",
+                        ExprListDisplay::comma_separated(exprs.as_slice())
+                    )?;
+                }
+                write!(f, ")")
+            }
+            Expr::GroupingSet(GroupingSet::Rollup(exprs)) => {
+                write!(
+                    f,
+                    "ROLLUP ({})",
+                    ExprListDisplay::comma_separated(exprs.as_slice())
+                )
+            }
+            Expr::IsNull(expr) => write!(f, "{} IS NULL", SqlDisplay(expr)),
+            Expr::IsNotNull(expr) => {
+                write!(f, "{} IS NOT NULL", SqlDisplay(expr))
+            }
+            Expr::IsUnknown(expr) => {
+                write!(f, "{} IS UNKNOWN", SqlDisplay(expr))
+            }
+            Expr::IsNotUnknown(expr) => {
+                write!(f, "{} IS NOT UNKNOWN", SqlDisplay(expr))
+            }
+            Expr::IsTrue(expr) => write!(f, "{} IS TRUE", SqlDisplay(expr)),
+            Expr::IsFalse(expr) => write!(f, "{} IS FALSE", SqlDisplay(expr)),
+            Expr::IsNotTrue(expr) => {
+                write!(f, "{} IS NOT TRUE", SqlDisplay(expr))
+            }
+            Expr::IsNotFalse(expr) => {
+                write!(f, "{} IS NOT FALSE", SqlDisplay(expr))
+            }
+            Expr::Like(Like {
+                negated,
+                expr,
+                pattern,
+                escape_char,
+                case_insensitive,
+            }) => {
+                write!(
+                    f,
+                    "{} {}{} {}",
+                    SqlDisplay(expr),
+                    if *negated { "NOT " } else { "" },
+                    if *case_insensitive { "ILIKE" } else { "LIKE" },
+                    SqlDisplay(pattern),
+                )?;
+
+                if let Some(char) = escape_char {
+                    write!(f, " CHAR '{char}'")?;
+                }
+
+                Ok(())
+            }
+            Expr::Negative(expr) => write!(f, "(- {})", SqlDisplay(expr)),
+            Expr::Not(expr) => write!(f, "NOT {}", SqlDisplay(expr)),
+            Expr::Unnest(Unnest { expr }) => {
+                write!(f, "UNNEST({})", SqlDisplay(expr))
+            }
+            Expr::SimilarTo(Like {
+                negated,
+                expr,
+                pattern,
+                escape_char,
+                ..
+            }) => {
+                write!(
+                    f,
+                    "{} {} {}",
+                    SqlDisplay(expr),
+                    if *negated {
+                        "NOT SIMILAR TO"
+                    } else {
+                        "SIMILAR TO"
+                    },
+                    SqlDisplay(pattern),
+                )?;
+                if let Some(char) = escape_char {
+                    write!(f, " CHAR '{char}'")?;
+                }
+
+                Ok(())
+            }
+            Expr::AggregateFunction(AggregateFunction { func, params }) => {
+                match func.human_display(params) {
+                    Ok(name) => {
+                        write!(f, "{name}")
+                    }
+                    Err(e) => {
+                        write!(f, "got error from schema_name {}", e)
+                    }
+                }
+            }
+            _ => write!(f, "{}", self.0),
+        }
+    }
+}
+
 /// Get schema_name for Vector of expressions
 ///
 /// Internal usage. Please call `schema_name_from_exprs` instead
@@ -2607,6 +2867,38 @@ pub(crate) fn schema_name_from_exprs_comma_separated_without_space(
     schema_name_from_exprs_inner(exprs, ",")
 }
 
+/// Formats a list of `&Expr` with a custom separator using SQL display format
+pub struct ExprListDisplay<'a> {
+    exprs: &'a [Expr],
+    sep: &'a str,
+}
+
+impl<'a> ExprListDisplay<'a> {
+    /// Create a new display struct with the given expressions and separator
+    pub fn new(exprs: &'a [Expr], sep: &'a str) -> Self {
+        Self { exprs, sep }
+    }
+
+    /// Create a new display struct with comma-space separator
+    pub fn comma_separated(exprs: &'a [Expr]) -> Self {
+        Self::new(exprs, ", ")
+    }
+}
+
+impl Display for ExprListDisplay<'_> {
+    fn fmt(&self, f: &mut Formatter) -> fmt::Result {
+        let mut first = true;
+        for expr in self.exprs {
+            if !first {
+                write!(f, "{}", self.sep)?;
+            }
+            write!(f, "{}", SqlDisplay(expr))?;
+            first = false;
+        }
+        Ok(())
+    }
+}
+
 /// Get schema_name for Vector of expressions
 pub fn schema_name_from_exprs(exprs: &[Expr]) -> Result<String, fmt::Error> {
     schema_name_from_exprs_inner(exprs, ", ")