Skip to content

Lexical structure

SQL is a domain-specific language for managing and manipulating data. It’s used primarily to work with structured data, where the types and relationships across entities are well-defined. Originally adopted for relational databases, SQL is rapidly becoming the language of choice for stream processing. It’s declarative, expressive, and ubiquitous.

The American National Standards Institute (ANSI) maintains a standard for the specification of SQL. SQL-92, the third revision to the standard, is generally the most recognized form of the specification. Beyond the standard, there are many flavors and extensions to SQL so that it can express programs beyond what's possible with the SQL-92 grammar.

ksqlDB’s SQL grammar was built initially around Presto's grammar and has been extended judiciously. ksqlDB goes beyond SQL-92, because the standard currently has no constructs for streaming queries, which are a core aspect of this project.

Syntax

SQL inputs are made up of a series of statements. Each statements is made up of a series of tokens and ends in a semicolon (;). The tokens that apply depend on the statement being invoked.

A token is any keyword, identifier, backticked identifier, literal, or special character. By convention, tokens are separated by whitespace, unless there is no ambiguity in the grammar. This happens when tokens flank a special character.

The following example statements are syntactically valid ksqlDB SQL input:

1
2
3
4
5
6
7
8
INSERT INTO s1 (a, b) VALUES ('k1', 'v1');

CREATE STREAM s2 AS
    SELECT a, b
    FROM s1
    EMIT CHANGES;

SELECT * FROM t1 WHERE k1='foo' EMIT CHANGES;

Keywords

Some tokens, such as SELECT, INSERT, and CREATE, are keywords. Keywords are reserved tokens that have a specific meaning in ksqlDB's syntax. They control their surrounding allowable tokens and execution semantics. Keywords are case insensitive, meaning SELECT and select are equivalent. You can't create an identifier that is already a reserved word, unless you use backticked identifiers.

A complete list of keywords can be found in the appendix.

Identifiers

Identifiers are symbols that represent user-defined entities, like streams, tables, columns, and other objects. For example, if you have a stream named s1, s1 is an identifier for that stream. By default, identifiers are case-insensitive, meaning s1 and S1 refer to the same stream. Under the hood, ksqlDB capitalizes all of the characters in the identifier for all future display purposes.

Unless an identifier is backticked, it may be composed only of characters that are a letter, number, or underscore. There is no imposed limit on the number of characters.

To make it possible to use any character in an identifier, you can enclose it in backticks (`) when you declare and use it. A backticked identifier is useful when you don't control the data, so it might have special characters, or even keywords. When you use backticked identifers, ksqlDB captures the case exactly, and any future references to the identifer become case-sensitive. For example, if you declare the following stream:

1
2
3
4
5
6
7
8
CREATE STREAM `s1` (
    k VARCHAR KEY,
    `@MY-identifier-stream-column!` INT
) WITH (
    kafka_topic = 's1',
    partitions = 3,
    value_format = 'json'
);

You must select from it by backticking the stream name and column name and using the original casing:

1
SELECT `@MY-identifier-stream-column!` FROM `s1` EMIT CHANGES;

Constants

There are three implicitly typed constants, or literals, in ksqlDB: strings, numbers, and booleans.

String constants

A string constant is an arbitrary series of characters surrounded by single quotes ('), like 'Hello world'. To include a quote inside of a string literal, escape the quote by prefixing it with another quote, for example 'You can call me ''Stuart'', or Stu.'

Numeric constants

Numeric constants are accepted in the following forms:

  1. digits
  2. digits.[digits][e[+-]digits]
  3. [digits].digits[e[+-]digits]
  4. digitse[+-]digits

where digits is one or more single-digit integers (0 through 9).

  • At least one digit must be present before or after the decimal point, if there is one.
  • At least one digit must follow the exponent symbol e, if there is one.
  • No spaces, underscores, or any other characters are allowed in the constant.
  • Numeric constants may also have a + or - prefix, but this is considered to be a function applied to the constant, not the constant itself.

Here are some examples of valid numeric constants:

  • 5
  • 7.2
  • 0.0087
  • 1.
  • .5
  • 1e-3
  • 1.332434e+2
  • +100
  • -250

Boolean constants

A boolean constant is represented as either the identifer true or false. Boolean constants are not case-sensitive, meaning true evaluates to the same value as TRUE.

Operators

Operators are infix functions composed of special characters. A complete list of operators can be found in the appendix. ksqlDB doesn't allow you to add user-space operators.

Special characters

Some characters have a particular meaning that doesn't correspond to an operator. The following list describes the special characters and their purpose.

  • Parentheses (()) retain their usual meaning in programming languages for grouping expressions and controlling the order of evaluation.
  • Brackets ([]) are used to work with arrays, both in their construction and subscript access. They also allow you to key into maps.
  • Commas (,) delineate a discrete list of entities.
  • The semi-colons (;) terminates a SQL command.
  • The asterisk (*), when used in particular syntax, is used as an "all" qualifier. This is seen most commonly in a SELECT command to retrieve all columns.
  • The period (.) accesses a column in a stream or table.
  • The arrow (->) accesses a field in a struct data type.
  • The dollar sign/brace combination (${...}) combination references a defined variable.

Comments

A comment is a string beginning with twos dashes. It includes all of the content from the dashes to the end of the line:

1
-- Here is a comment.

You can also span a comment over multiple lines by using C-style syntax:

1
2
3
/* Here is
   another comment.
 */

Lexical precedence

Operators are evaluated using the following order of precedence:

  1. *, /, %
  2. +, -
  3. =, >, <, >=, <=, <>, !=
  4. NOT
  5. AND
  6. BETWEEN, LIKE, OR

In an expression, when two operators have the same precedence level, they're evaluated left-to-right based on their position.

You can enclose an expression in parentheses to force precedence or clarify precedence, for example, (5 + 2) * 3.

Last update: 2020-12-15