Support double-quoted identifiers in SQL parser

[fupelaqu]

Description

The SQL parser does not support double-quoted identifiers ("identifier"). This causes failures when SQL clients (Superset, DBeaver, etc.) generate standard SQL that quotes column aliases or table/column names.

Root Cause

The parser expects bare identifiers in ORDER BY, AS, and other clauses. When it encounters "Revenue" instead of Revenue, it fails with:

string matching regex '(?i)(FROM)\b' expected but 'A' found

Failing Queries (generated by Apache Superset)

-- Chart: Revenue by Country
SELECT country AS country, sum(total_price) AS "Revenue"
FROM ecommerce GROUP BY country ORDER BY "Revenue" DESC LIMIT 10

-- Chart: Orders by Category
SELECT category AS category, COUNT(*) AS "Orders"
FROM ecommerce GROUP BY category ORDER BY "Orders" DESC LIMIT 10

-- Chart: Avg Order Value by Payment Method
SELECT payment_method AS payment_method, AVG(total_price) AS "Avg Order Value"
FROM ecommerce GROUP BY payment_method ORDER BY "Avg Order Value" DESC LIMIT 10

-- Chart: Top Customers
SELECT customer_name AS customer_name, country AS country,
       sum(total_price) AS "Total Spend", COUNT(*) AS "Orders"
FROM ecommerce GROUP BY customer_name, country
ORDER BY "Total Spend" DESC LIMIT 10

All fail because:

1. AS "Revenue" — quoted alias in SELECT
2. ORDER BY "Revenue" — quoted identifier reference in ORDER BY

Expected Behaviour

Double-quoted identifiers should be treated as regular identifiers with the quotes stripped:

SELECT country, SUM(total_price) AS "Revenue"
FROM ecommerce GROUP BY country ORDER BY "Revenue" DESC

Should parse identically to:

SELECT country, SUM(total_price) AS Revenue
FROM ecommerce GROUP BY country ORDER BY Revenue DESC

For identifiers containing spaces or reserved words, the quotes are mandatory:

SELECT SUM(total_price) AS "Total Spend" FROM ecommerce

SQL Standard Reference

ANSI SQL uses double quotes for delimited identifiers (SQL-92, section 5.2). This is the standard quoting mechanism used by:

• PostgreSQL
• Apache Superset (SQLAlchemy default)
• DBeaver
• Most JDBC/ODBC tools

Scope

The parser should accept double-quoted identifiers in at least:

Context	Example
Column alias (AS)	SELECT col AS "My Alias"
ORDER BY	ORDER BY "My Alias" DESC
GROUP BY	GROUP BY "col name"
Column reference	SELECT "col" FROM t
WHERE clause	WHERE "col" > 10
HAVING clause	HAVING "Total" > 100

Impact

• Apache Superset charts fail entirely — all auto-generated SQL uses quoted identifiers
• DBeaver and other JDBC clients may quote identifiers depending on settings
• Blocks the Superset demo dashboard (all 6 charts return internal errors)

Resolution

Added two new parser combinators — quotedIdentifier and quotedAlias — that match "..." content and produce GenericIdentifier / Alias with the unquoted name. These are inserted as the first alternative in all identifier-expecting parser chains, before identifierWithArithmeticExpression (which would otherwise consume "..." via the literal → StringValue → functionAsIdentifier path).

Strategy: targeted insertion, not global replacement

The literal parser in type/package.scala was not modified — double-quoted strings remain StringValue in value positions (WHERE RHS, DDL OPTIONS). This preserves DDL round-trip safety.

Modified files

File	Change
Parser.scala	Added quotedIdentifier, quotedAlias; inserted quotedIdentifier in valueExpr
SelectParser.scala	quotedIdentifier first in field chain; quotedAlias | alias for alias parsing
OrderByParser.scala	quotedIdentifier first in fieldWithFunction
GroupByParser.scala	quotedIdentifier first in bucketWithFunction
WhereParser.scala	quotedIdentifier first in any_identifier; updated isNull/isNotNull

Tests

• Parser unit tests: ParserSpec — 6 new tests covering Superset queries, quoted aliases with spaces, quoted SELECT/WHERE/GROUP BY, reserved words as identifiers
• Integration test: ReplGatewayIntegrationSpec — end-to-end test with AS "City", AS "Total", AS "Avg Age" and ORDER BY "Total" DESC against real Elasticsearch

[fupelaqu]

Resolution

Added two new parser combinators — quotedIdentifier and quotedAlias — that match "..." content and produce GenericIdentifier / Alias with the unquoted name. These are inserted as the first alternative in all identifier-expecting parser chains, before identifierWithArithmeticExpression (which would otherwise consume "..." via the literal → StringValue → functionAsIdentifier path).

Strategy: targeted insertion, not global replacement

The literal parser in type/package.scala was not modified — double-quoted strings remain StringValue in value positions (WHERE RHS, DDL OPTIONS). This preserves DDL round-trip safety.

Modified files

File	Change
Parser.scala	Added quotedIdentifier, quotedAlias; inserted quotedIdentifier in valueExpr
SelectParser.scala	quotedIdentifier first in field chain; quotedAlias | alias for alias parsing
OrderByParser.scala	quotedIdentifier first in fieldWithFunction
GroupByParser.scala	quotedIdentifier first in bucketWithFunction
WhereParser.scala	quotedIdentifier first in any_identifier; updated isNull/isNotNull

Tests

• Parser unit tests: ParserSpec — 6 new tests covering Superset queries, quoted aliases with spaces, quoted SELECT/WHERE/GROUP BY, reserved words as identifiers
• Integration test: ReplGatewayIntegrationSpec — end-to-end test with AS "City", AS "Total", AS "Avg Age" and ORDER BY "Total" DESC against real Elasticsearch