This is an automated email from the ASF dual-hosted git repository.
HappenLee pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris-website.git
The following commit(s) were added to refs/heads/master by this push:
new 1635355180e [doc](function) Document map support for inner_product
(#4113)
1635355180e is described below
commit 1635355180e23d5a8a95e9c0abb166548b31fd07
Author: Pxl <[email protected]>
AuthorDate: Sun Sep 6 02:16:29 2026 +0800
[doc](function) Document map support for inner_product (#4113)
---
.../distance-functions/inner-product.md | 51 +++++++++++++++++----
.../distance-functions/inner-product.md | 53 ++++++++++++++++++----
2 files changed, 87 insertions(+), 17 deletions(-)
diff --git
a/docs/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
b/docs/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
index c64ca9ea074..2d5828cfbf5 100644
---
a/docs/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
+++
b/docs/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
@@ -2,34 +2,39 @@
{
"title": "INNER_PRODUCT",
"language": "en",
- "description": "Computes the scalar product of two vectors of the same
size"
+ "description": "Computes the scalar product of two dense or sparse vectors"
}
---
## Description
-Computes the scalar product of two vectors of the same size
+Computes the scalar product of two dense vectors represented by arrays, or two
sparse vectors represented by maps. For sparse vectors, each map key identifies
a dimension and its value is the coordinate. Only dimensions present in both
maps contribute to the result.
## Syntax
```sql
INNER_PRODUCT(<array1>, <array2>)
+INNER_PRODUCT(<map1>, <map2>)
```
## Parameters
| Parameter | Description |
-| -- |--|
-| `<array1>` | The first vector, the subtype of the input array supports:
TINYINT, SMALLINT, INT, BIGINT, LARGEINT, FLOAT, DOUBLE, the number of elements
must be consistent with array2 |
-| `<array2>` | The second vector, the subtype of the input array supports:
TINYINT, SMALLINT, INT, BIGINT, LARGEINT, FLOAT, DOUBLE, the number of elements
must be consistent with array1 |
+|---|---|
+| `<array1>` | The first dense vector. Supported element types are TINYINT,
SMALLINT, INT, BIGINT, LARGEINT, FLOAT, and DOUBLE. The number of elements must
equal that of `<array2>`. |
+| `<array2>` | The second dense vector. Supported element types are TINYINT,
SMALLINT, INT, BIGINT, LARGEINT, FLOAT, and DOUBLE. The number of elements must
equal that of `<array1>`. |
+| `<map1>` | The first sparse vector. Its type must be `MAP<K, FLOAT>`, where
`K` is an integer type (TINYINT, SMALLINT, INT, BIGINT, or LARGEINT) or a
string type (CHAR, VARCHAR, or STRING). Its key type must be from the same type
family as that of `<map2>`. |
+| `<map2>` | The second sparse vector. It has the same type constraints as
`<map1>`. The two maps can contain different numbers of entries. |
## Return Value
-Returns the scalar product of two vectors of the same size. The return type is
`FLOAT`.
+Returns the scalar product as a `FLOAT`.
-If either input array is `NULL`, or contains a `NULL` element, the function
returns an error.
+For array inputs, corresponding elements are multiplied and summed. If either
array is `NULL`, contains a `NULL` element, or the arrays have different
lengths, the function returns an error.
-## Examples
+For map inputs, values with the same key are multiplied and summed. A key that
occurs in only one map has no effect, which is equivalent to treating the
missing coordinate as zero. `NULL` keys match other `NULL` keys. If a key
occurs more than once, its last value is used. If either map is `NULL`, or the
value retained for any key is `NULL`, the function returns an error.
+
+## Example
```sql
SELECT INNER_PRODUCT([1, 2], [2, 3]),INNER_PRODUCT([3, 6], [4, 7]);
@@ -43,6 +48,23 @@ SELECT INNER_PRODUCT([1, 2], [2, 3]),INNER_PRODUCT([3, 6],
[4, 7]);
+-------------------------------+-------------------------------+
```
+The following example computes the inner product of two sparse vectors. Only
keys `a` and `b` occur in both maps, so the result is `2 * 5 + 3 * 4 = 22`:
+
+```sql
+SELECT INNER_PRODUCT(
+ MAP('a', CAST(2 AS FLOAT), 'b', CAST(3 AS FLOAT)),
+ MAP('b', CAST(4 AS FLOAT), 'c', CAST(100 AS FLOAT), 'a', CAST(5 AS FLOAT))
+) AS result;
+```
+
+```text
++--------+
+| result |
++--------+
+| 22 |
++--------+
+```
+
If an input array is `NULL`, the function returns an error:
```sql
@@ -62,3 +84,16 @@ SELECT INNER_PRODUCT([1, NULL], [1, 2]);
```text
ERROR 1105 (HY000): errCode = 2, detailMessage =
(127.0.0.1)[INVALID_ARGUMENT]First argument for function inner_product cannot
have null
```
+
+If a map value is `NULL`, the function returns an error:
+
+```sql
+SELECT INNER_PRODUCT(
+ MAP(1, CAST(NULL AS FLOAT)),
+ MAP(1, CAST(2 AS FLOAT))
+);
+```
+
+```text
+ERROR 1105 (HY000): errCode = 2, detailMessage =
(127.0.0.1)[INVALID_ARGUMENT]First argument for function inner_product cannot
have null
+```
diff --git
a/i18n/zh-CN/docusaurus-plugin-content-docs/current/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
b/i18n/zh-CN/docusaurus-plugin-content-docs/current/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
index c0defe3deeb..b70efea70b3 100644
---
a/i18n/zh-CN/docusaurus-plugin-content-docs/current/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
+++
b/i18n/zh-CN/docusaurus-plugin-content-docs/current/sql-manual/sql-functions/ai-functions/distance-functions/inner-product.md
@@ -2,34 +2,39 @@
{
"title": "INNER_PRODUCT",
"language": "zh-CN",
- "description": "计算两个大小相同的向量的标量积"
+ "description": "计算两个稠密向量或稀疏向量的标量积"
}
---
## 描述
-计算两个大小相同的向量的标量积
+计算两个由数组表示的稠密向量，或两个由 Map 表示的稀疏向量的标量积。对于稀疏向量，每个 Map 的 key 表示一个维度，value
表示该维度的坐标值。只有同时出现在两个 Map 中的维度才参与计算。
## 语法
```sql
INNER_PRODUCT(<array1>, <array2>)
+INNER_PRODUCT(<map1>, <map2>)
```
## 参数
-| 参数 | 说明 |
-| -- |--|
-| `<array1>` |
第一个向量，输入数组的子类型支持：TINYINT、SMALLINT、INT、BIGINT、LARGEINT、FLOAT、DOUBLE，元素数量需与
array2 保持一致 |
-| `<array2>` |
第二个向量，输入数组的子类型支持：TINYINT、SMALLINT、INT、BIGINT、LARGEINT、FLOAT、DOUBLE，元素数量需与
array1 保持一致 |
+| 参数 | 描述 |
+|---|---|
+| `<array1>` | 第一个稠密向量。数组元素支持 TINYINT、SMALLINT、INT、BIGINT、LARGEINT、FLOAT 和
DOUBLE 类型，元素数量必须与 `<array2>` 相同。 |
+| `<array2>` | 第二个稠密向量。数组元素支持 TINYINT、SMALLINT、INT、BIGINT、LARGEINT、FLOAT 和
DOUBLE 类型，元素数量必须与 `<array1>` 相同。 |
+| `<map1>` | 第一个稀疏向量。类型必须为 `MAP<K, FLOAT>`，其中 `K`
可以是整数类型（TINYINT、SMALLINT、INT、BIGINT 或 LARGEINT）或字符串类型（CHAR、VARCHAR 或 STRING）。其
key 类型必须与 `<map2>` 的 key 类型属于同一类型族。 |
+| `<map2>` | 第二个稀疏向量，类型约束与 `<map1>` 相同。两个 Map 的元素数量可以不同。 |
## 返回值
-返回两个大小相同的向量的标量积，返回类型为 `FLOAT`。
+返回标量积，返回类型为 `FLOAT`。
-如果任一输入数组为 `NULL`，或包含 `NULL` 元素，函数会报错。
+对于数组输入，将相同位置的元素相乘后求和。如果任一数组为 `NULL`、包含 `NULL` 元素，或两个数组的长度不同，函数会报错。
-## 举例
+对于 Map 输入，将相同 key 对应的 value 相乘后求和。只出现在一个 Map 中的 key 不影响结果，相当于缺失维度的坐标值为零。`NULL`
key 可以与另一个 `NULL` key 匹配。如果一个 key 重复出现，则使用它最后一次出现的 value。如果任一 Map 为 `NULL`，或任一
key 最终保留的 value 为 `NULL`，函数会报错。
+
+## 示例
```sql
SELECT INNER_PRODUCT([1, 2], [2, 3]),INNER_PRODUCT([3, 6], [4, 7]);
@@ -43,6 +48,23 @@ SELECT INNER_PRODUCT([1, 2], [2, 3]),INNER_PRODUCT([3, 6],
[4, 7]);
+-------------------------------+-------------------------------+
```
+以下示例计算两个稀疏向量的内积。只有 key `a` 和 `b` 同时出现在两个 Map 中，因此结果为 `2 * 5 + 3 * 4 = 22`：
+
+```sql
+SELECT INNER_PRODUCT(
+ MAP('a', CAST(2 AS FLOAT), 'b', CAST(3 AS FLOAT)),
+ MAP('b', CAST(4 AS FLOAT), 'c', CAST(100 AS FLOAT), 'a', CAST(5 AS FLOAT))
+) AS result;
+```
+
+```text
++--------+
+| result |
++--------+
+| 22 |
++--------+
+```
+
如果输入数组为 `NULL`，函数会报错：
```sql
@@ -62,3 +84,16 @@ SELECT INNER_PRODUCT([1, NULL], [1, 2]);
```text
ERROR 1105 (HY000): errCode = 2, detailMessage =
(127.0.0.1)[INVALID_ARGUMENT]First argument for function inner_product cannot
have null
```
+
+如果 Map 的 value 为 `NULL`，函数会报错：
+
+```sql
+SELECT INNER_PRODUCT(
+ MAP(1, CAST(NULL AS FLOAT)),
+ MAP(1, CAST(2 AS FLOAT))
+);
+```
+
+```text
+ERROR 1105 (HY000): errCode = 2, detailMessage =
(127.0.0.1)[INVALID_ARGUMENT]First argument for function inner_product cannot
have null
+```
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
