Apache Flink SQL Keyword Compatibility Reference

Generated for GSP Java version 4.1.0.8 on 2026-03-15

This page was generated using hybrid static extraction from parser source files combined with runtime validation against the actual GSP parser. Re-run the extraction script after parser updates to keep this page current.

Keyword-as-Column-Name Support

As of version 4.1.0.8, the GSP Apache Flink parser includes a lexer lookahead mechanism that allows 29 vendor-unreserved keywords to be used as unquoted column names in SELECT statements.

The lookahead pre-scans the token list before parsing and converts context-specific keywords to identifiers when they appear in column-name position:

• After: SELECT, ,, DISTINCT, or ALL
• Before: FROM, AS, WHERE, GROUP, ORDER, HAVING, LIMIT, UNION, INTERSECT, EXCEPT, INTO, ,, ), or ;

-- Works: keyword as column name
SELECT between FROM t;

-- Works: keyword as column name
SELECT case FROM t;

-- Works: keyword as column name
SELECT coalesce FROM t;

-- Original keyword syntax also still works

---

Full Classification Overview

Out of 312 keywords recognized by the GSP Apache Flink parser:

Classification	Count	Description
Allowed	280	Can be used as an unquoted column name in both canonical contexts
Context-specific	31	Fails as SELECT keyword FROM t but works as SELECT t.keyword FROM t
Blocked	1	Cannot be used as an unquoted column name in either context

Context-Specific Keywords (31)

These keywords fail when used as bare column names (SELECT keyword FROM t) but succeed when table-qualified (SELECT t.keyword FROM t).

Keyword	Reason
ALL	SELECT qualifier
BETWEEN	Operator keyword
CASE	Expression keyword
COALESCE	Expression keyword
CUMULATE	Grammar keyword
CURRENT_ROW_TIMESTAMP	Grammar keyword
CURRENT_WATERMARK	Grammar keyword
DATE_FORMAT	Grammar keyword
DESCRIPTOR	Grammar keyword
DISTINCT	SELECT qualifier
EXISTS	Operator keyword
HOP	Grammar keyword
IF	Expression keyword
IFNULL	Grammar keyword
INSERT	Grammar keyword
INTERVAL	Type keyword
IS	Operator keyword
JOIN	JOIN keyword
LIKE	Operator keyword
NOT	Operator keyword
NULLIF	Expression keyword
PROCTIME	Grammar keyword
RIGHT	JOIN keyword
ROW	Grammar keyword
ROWTIME	Grammar keyword
SESSION	Grammar keyword
TO_DATE	Grammar keyword
TO_TIMESTAMP	Grammar keyword
TO_TIMESTAMP_LTZ	Grammar keyword
TRY_CAST	Grammar keyword
TUMBLE	Grammar keyword

Blocked Keywords (1)

These keywords cannot be used as unquoted column names in either context.

Keyword	Workaround
FROM	SELECT "from" FROM t

Workaround: Double-Quoted Identifiers

For any keyword that fails as an unquoted column name, you can use double-quoted identifiers:

-- Blocked or context-specific keyword as column name
SELECT "from" FROM t;

-- Or use table qualification for context-specific keywords
SELECT t.all FROM t;

Scope and Limitations

• Tested contexts: SELECT keyword FROM t and SELECT t.keyword FROM t. Other contexts (DDL column definitions, INSERT column lists, aliases) may behave differently.
• Version-specific: This report reflects GSP Java version 4.1.0.8.
• Case sensitivity: Keywords are case-insensitive. select, SELECT, and Select are all treated the same.

How to Report Discrepancies

If you encounter a keyword that behaves differently from what this page describes, please report it through your support channel. Include:

1. The exact SQL statement
2. The GSP parser version
3. Whether the same SQL works in Apache Flink

Methodology

1. Static extraction: A Python script parses the lexer (.cod) and grammar (.y) source files to identify all 312 keywords and their grammar classifications.
2. Runtime validation: A Java test harness validates every classification against actual TGSqlParser runtime behavior.
3. JSON dataset: The authoritative data is stored in docs/generated/flink_keyword_compatibility.json.