Skip to content

Apache DataFusion SQL Query Engine

docs: fix SQL examples that the default dialect cannot parse (#26373)

Roshan Ramani30f066d2275c
diff --git a/docs/source/user-guide/sql/ddl.md b/docs/source/user-guide/sql/ddl.md
index 8382795e74..e22898c6b1 100644
--- a/docs/source/user-guide/sql/ddl.md
+++ b/docs/source/user-guide/sql/ddl.md
@@ -194,9 +194,8 @@ It is also possible to use compressed files, such as `.csv.gz`:
```sql
CREATE EXTERNAL TABLE test
STORED AS CSV
-COMPRESSION TYPE GZIP
LOCATION '/path/to/aggregate_simple.csv.gz'
-OPTIONS ('has_header' 'true');
+OPTIONS ('has_header' 'true', 'format.compression' 'gzip');
```
### Example: Specifying Schema
diff --git a/docs/source/user-guide/sql/struct_coercion.md b/docs/source/user-guide/sql/struct_coercion.md
index 059f9fd009..7fe8a4b59d 100644
--- a/docs/source/user-guide/sql/struct_coercion.md
+++ b/docs/source/user-guide/sql/struct_coercion.md
@@ -61,8 +61,8 @@ SELECT [{x: 1, y: 2}, {y: 3, x: 4}];
When constructing arrays from table columns with different struct schemas:
```sql
-CREATE TABLE t_left (s struct(x int, y int)) AS VALUES ({x: 1, y: 2});
-CREATE TABLE t_right (s struct(y int, x int)) AS VALUES ({y: 3, x: 4});
+CREATE TABLE t_left (s struct<x int, y int>) AS VALUES ({x: 1, y: 2});
+CREATE TABLE t_right (s struct<y int, x int>) AS VALUES ({y: 3, x: 4});
-- Dynamically constructs unified array schema
SELECT [t_left.s, t_right.s] FROM t_left JOIN t_right;
@@ -123,8 +123,8 @@ CREATE TABLE t AS VALUES ({a: 1, b: 2}), ({b: 3, a: 4});
When joining tables where the JOIN condition involves structs with different field orders:
```sql
-CREATE TABLE orders (customer struct(name varchar, id int));
-CREATE TABLE customers (info struct(id int, name varchar));
+CREATE TABLE orders (customer struct<name varchar, id int>);
+CREATE TABLE customers (info struct<id int, name varchar>);
-- Join matches struct fields by name
SELECT * FROM orders
@@ -175,8 +175,8 @@ When structs have different field sets, missing fields are filled with **NULL**
-- Unified schema includes all fields: a, b, c
SELECT [
- CAST({a: 1, b: 2} AS STRUCT(a INT, b INT, c INT)),
- CAST({b: 3, c: 4} AS STRUCT(a INT, b INT, c INT))
+ CAST({a: 1, b: 2} AS STRUCT<a INT, b INT, c INT>),
+ CAST({b: 3, c: 4} AS STRUCT<a INT, b INT, c INT>)
];
-- Result:
@@ -203,8 +203,8 @@ To handle partial field overlap, explicitly cast structs to a unified schema:
```sql
SELECT [
- CAST(t_left.s AS STRUCT(x INT, y INT, z INT)),
- CAST(t_right.s AS STRUCT(x INT, y INT, z INT))
+ CAST(t_left.s AS STRUCT<x INT, y INT, z INT>),
+ CAST(t_right.s AS STRUCT<x INT, y INT, z INT>)
] FROM t_left JOIN t_right;
```
@@ -263,7 +263,7 @@ If you need precise control over struct field order and types, use explicit `CAS
```sql
-- Guarantee specific field order and types
-SELECT CAST({b: 3, a: 4} AS STRUCT(a INT, b INT));
+SELECT CAST({b: 3, a: 4} AS STRUCT<a INT, b INT>);
-- Result: {"a": 4, "b": 3}
```
@@ -275,7 +275,7 @@ When joining or combining structs, define target schemas explicitly:
```sql
-- Good: explicit schema definition
-SELECT CAST(data AS STRUCT(id INT, name VARCHAR, active BOOLEAN))
+SELECT CAST(data AS STRUCT<id INT, name VARCHAR, active BOOLEAN>)
FROM external_source;
```
@@ -308,8 +308,8 @@ When combining structs with partial field overlap, use explicit CAST:
```sql
-- Instead of relying on implicit coercion
SELECT [
- CAST(left_struct AS STRUCT(x INT, y INT, z INT)),
- CAST(right_struct AS STRUCT(x INT, y INT, z INT))
+ CAST(left_struct AS STRUCT<x INT, y INT, z INT>),
+ CAST(right_struct AS STRUCT<x INT, y INT, z INT>)
];
```
@@ -338,8 +338,8 @@ FROM customers c;
```sql
-- Use explicit CAST to handle missing fields
SELECT [
- CAST(struct1 AS STRUCT(a INT, b INT, c INT)),
- CAST(struct2 AS STRUCT(a INT, b INT, c INT))
+ CAST(struct1 AS STRUCT<a INT, b INT, c INT>),
+ CAST(struct2 AS STRUCT<a INT, b INT, c INT>)
];
```