kinetica-query

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Kinetica SQL Dialect

Kinetica SQL 方言

Kinetica is a GPU-accelerated database with a PostgreSQL-compatible SQL dialect. It handles billions of rows, vector search, geospatial, time-series, and graph analytics.

Critical: Kinetica SQL has important deviations from standard PostgreSQL. Read references/kinetica-core-rules.md before writing any query.

Kinetica是一款GPU加速的数据库，兼容PostgreSQL SQL语法。它支持处理数十亿行数据、向量搜索、地理空间、时间序列以及图分析。

重要提示：Kinetica SQL与标准PostgreSQL存在重要差异。编写任何查询前请先阅读references/kinetica-core-rules.md。

Key Differences from PostgreSQL

与PostgreSQL的核心差异

No nested aggregate functions — use subqueries or CTEs instead
No backticks — use double quotes for identifiers
```
DATEDIFF
```
/
```
DATEADD
```
instead of timestamp arithmetic
Case-sensitive identifiers — always match exact column names
```
DECIMAL
```
not
```
NUMERIC
```

不支持嵌套聚合函数——请改用子查询或CTE
不支持反引号——标识符请使用双引号包裹
请使用
```
DATEDIFF
```
/
```
DATEADD
```
代替时间戳算术运算
标识符大小写敏感——请始终与列名的精确大小写匹配
使用
```
DECIMAL
```
而非
```
NUMERIC
```

Domain Capabilities

领域能力

SQL Core

SQL核心

See references/kinetica-core-rules.md — read this first. See references/sql-functions.md for Kinetica-specific functions. See references/sql-patterns.md for common query patterns.

请查看references/kinetica-core-rules.md——请优先阅读此文档。 Kinetica专属函数请查看references/sql-functions.md。常见查询模式请查看references/sql-patterns.md。

DDL & DML

CREATE TABLE with shard keys, partitioning, tier strategies, vector indexes. LOAD DATA for bulk ingestion, EXPORT for data extraction, upsert hints. See references/ddl-reference.md and references/dml-reference.md.

CREATE TABLE支持分片键、分区、分层策略、向量索引。 LOAD DATA用于批量导入，EXPORT用于数据导出，还有upsert提示。请查看references/ddl-reference.md和references/dml-reference.md。

Geospatial

地理空间

ST_* and accelerated STXY_* functions, H3 spatial indexing, SRID 4326 only, solution parameter (Euclidean/Haversine/Vincenty). See references/geospatial-functions.md and references/geospatial-examples.md.

支持ST_以及加速的STXY_函数、H3空间索引，仅支持SRID 4326，支持解决方案参数（欧几里得/半正矢/文森特公式）。请查看references/geospatial-functions.md和references/geospatial-examples.md。

Time-Series

时间序列

TIME_BUCKET, DATEDIFF/DATEADD, window functions, ASOF joins for temporal proximity. See references/timeseries-functions.md and references/timeseries-examples.md.

支持TIME_BUCKET、DATEDIFF/DATEADD、窗口函数、用于时序邻近匹配的ASOF连接。请查看references/timeseries-functions.md和references/timeseries-examples.md。

Vector Search & Embeddings

向量搜索与嵌入

Vector distance functions (L2, Cosine, Inner Product), GENERATE_EMBEDDINGS UDF, and semantic search patterns. See references/vector-patterns.md.

支持向量距离函数（L2、余弦、内积）、GENERATE_EMBEDDINGS UDF，以及语义搜索模式。请查看references/vector-patterns.md。

JSON, Arrays & Text Search

JSON、数组与文本搜索

JSON_EXTRACT_VALUE (must CAST!), ARRAY_CONTAINS, UNNEST, REGEXP_LIKE, fuzzy matching. See references/json-array-text.md.

支持JSON_EXTRACT_VALUE（必须显式CAST！）、ARRAY_CONTAINS、UNNEST、REGEXP_LIKE、模糊匹配。请查看references/json-array-text.md。

Graph Analytics

图分析

Build property graphs from existing tables, query with Cypher, run algorithms (shortest path, page rank, TSP). No separate graph database needed. See references/graph-functions.md and references/graph-examples.md.

可基于现有表构建属性图，使用Cypher查询，运行相关算法（最短路径、页面排名、TSP），无需单独的图数据库。请查看references/graph-functions.md和references/graph-examples.md。

UDFs, Procedures & ML

UDF、存储过程与ML

User-Defined Functions (Python), scheduled SQL procedures, built-in ML (PREDICT, OUTLIERS), and Docker model management. See references/udf-reference.md.

支持用户自定义函数（Python）、定时SQL存储过程、内置ML能力（PREDICT、OUTLIERS），以及Docker模型管理。请查看references/udf-reference.md。

Query Writing Guidelines

查询编写指南

Always check column names and types before writing SQL — Kinetica is case-sensitive
Quote schema-qualified table names:
```
"schema"."table"
```
Use LIMIT for exploration queries
Use CTEs instead of nested subqueries with aggregates
For date math, always use DATEDIFF/DATEADD — never subtract timestamps directly
When in doubt, consult the relevant reference file for the domain

编写SQL前请始终确认列名和类型——Kinetica大小写敏感
带schema的表名请加引号：
```
"schema"."table"
```
探索性查询请使用LIMIT
带聚合的查询请使用CTE代替嵌套子查询
日期运算请始终使用DATEDIFF/DATEADD——切勿直接对时间戳做减法
如有疑问，请查阅对应领域的相关参考文件