KSML Type Inference Rules

This document describes KSML attribute type inference rules.
It explains how KSML infers field types based on explicit declarations and literal values, and how these types map to Java, SQL, and GraphQL.

---

1. Overview

In KSML, attribute types can be:

1. Explicitly declared using type functions such as string(), id(), date()
2. Implicitly inferred from literal values using regex-based matchers
3. Inherited via parent type relationships

Type inference is deterministic and rule-ordered.

---

2. Base Type Flags

Every type carries a set of semantic flags that guide downstream code generation.

Flag	Meaning
isString	String-like value
isText	Large text (CLOB / TEXT)
isNumber	Numeric value
isInt	Integer number
isBool	Boolean
isDate	Date component
isTime	Time component
isDateTime	Date + time
isId	Primary identifier
isVersion	Version field
isBaseEntityField	System-level base field
isPassword	Password field

---

3. Core Built-in Types

3.1 Identifier & Version

Type	Java Type	SQL Type	Special Flags
id	Long	BIGINT	isId, isBaseEntityField
version	Long	BIGINT	isVersion, isBaseEntityField

Example

id="id()"
version="version()"

---

3.2 String Types

Type	Parent	Java Type	SQL Type	Notes
string	—	String	VARCHAR(100)	Default type
text	string	String	TEXT / CLOB	Large text
jsonMe	text	String	TEXT	JSON, zipped
password	string	String	VARCHAR	Encrypted
uniStr	string	String	VARCHAR	Auto-generated

Example

name="string()"
description="text()"
password="password()"

---

3.3 Numeric Types

Type	Parent	Java Type	SQL Type	Notes
number	—	BigDecimal	NUMERIC(19,7)	Decimal
integer	number	Integer	INTEGER	Whole number
long	number	Long	BIGINT	Large integer

Example

age="integer()"
amount="number()"
count="long()"

---

3.4 Boolean

Type	Java Type	SQL Type
bool	Boolean	BOOLEAN

Example

enabled="bool()"

---

3.5 Date & Time

Type	Java Type	SQL Type	Notes
date	LocalDate	DATE	Date only
time	LocalTime	TIME	Time only
dateTime	LocalDateTime	TIMESTAMP	Date + time
createTime	LocalDateTime	TIMESTAMP	Auto now()
updateTime	LocalDateTime	TIMESTAMP	Auto now()

Example

birth_date="date()"
create_time="createTime()"
update_time="updateTime()"

---

4. Literal-Based Type Inference (Type Matchers)

When no explicit type is declared, KSML infers the type using regex matchers in priority order.

4.1 Type Matcher Table

Order	Regex Pattern	Inferred Type	Example Value
1	^\d+$	integer	123
2	`^\d+(l	L)$`	long
3	^\d+\.\d+$	number	12.34
4	`^(true	false	on
5	^\d{4}-\d{1,2}-\d{1,2}T\d{1,2}:\d{1,2}:\d{1,2}$	dateTime	2024-01-01T10:00:00
6	^\d{4}-\d{1,2}-\d{1,2}$	date	2024-01-01

---

4.2 Inference Examples

age="18"

➡ inferred as integer

total_amount="99.95"

➡ inferred as number

active="true"

➡ inferred as bool

created_at="2024-06-01T12:30:00"

➡ inferred as dateTime

---

5. Parent Type Inheritance

Types can inherit from parent types.

Child Type	Parent	Inherited Behavior
text	string	String semantics + large storage
integer	number	Numeric semantics
password	string	String + security
createTime	dateTime	Auto creation
updateTime	dateTime	Auto create & update

---

6. Type Resolution Priority

1. Explicit type declaration (e.g. string())
2. Built-in semantic type (id(), createTime())
3. Regex matcher inference
4. Default type (string)

---

7. Summary

• KSML type inference is rule-based, ordered, and deterministic
• Explicit types always override inferred types
• Parent-child type inheritance reduces duplication
• Type flags drive multi-target generation (Java / SQL / GraphQL)

Result: a single KSML model can reliably generate consistent schemas across systems.