Skip to content
FlyQL

Changelog

2026.08.12

Section titled “2026.08.12”

Version: 1.1.0

The values allowlist on columns is now enforced consistently — and only where it makes sense — across the SQL generators, the in-memory matcher, and the validator. See Values Allowlist for the full semantics.

Behavioral changes:

  • in / not in lists are validated against the allowlist. Each list element on an allowlisted column is checked during SQL generation; an out-of-allowlist element now fails with unknown value instead of silently matching zero rows. Null elements and column references are exempt. Queries that previously generated SQL with typo’d list elements will now be rejected.
  • The in-memory matcher enforces the allowlist. When evaluating with a schema whose columns declare values, an out-of-allowlist = / != value or in-list element raises unknown value (previously it evaluated silently, in parity-breaking contrast to the generators). Schema-free evaluation is unchanged.
  • New validator diagnostic value_not_allowed. diagnose() now emits a positioned error when an equality value or in-list element falls outside the column’s allowlist. The dialect-to-core schema bridges (ToFlyQLSchema / toFlyQLSchema) now carry values, so bridged schemas participate in the diagnostic.

To opt out of enforcement for a column, remove its values list from the schema.

Bug fixes:

  • = null works on allowlisted columns. Null is a presence predicate, not a domain value: col = null / col != null on a column with a values allowlist now generate IS NULL / IS NOT NULL instead of failing with unknown value.
  • Patterns are no longer checked against the allowlist. like / ilike / ~ / !~ patterns on allowlisted columns generate normally; previously any pattern not literally present in the allowlist was rejected, making pattern matching on such columns impossible.
  • Go PostgreSQL generator resolves RHS column references before the allowlist check. col = other_column on an allowlisted column now generates a column-to-column comparison on all generators; previously the Go PostgreSQL generator rejected it with unknown value while ClickHouse and StarRocks accepted it.

Documentation:

2026.07.21

Section titled “2026.07.21”

Version: 1.0.2

Bug fixes:

  • Consistent character offsets for non-ASCII input. The parser now scans input by Unicode code point in all three languages (Go scans as []rune, JavaScript as Array.from(text)), so Range offsets advance one step per character regardless of byte or UTF-16 width. Previously Cyrillic and other multi-byte/astral characters desynced offsets across the Go, JavaScript, and Python ports.
  • Code-point token offsets in tokenize(). tokenize() now reports start/end as Unicode code-point offsets in every language, identical across Python, Go, and JavaScript for all input. Previously Go emitted byte-width token spans and JavaScript emitted UTF-16-code-unit spans for non-ASCII (and astral) characters, breaking the gap-free offset invariant.
  • Valid PostgreSQL string escaping. The PostgreSQL generator now emits escape-string literals (E'...') for values that contain backslash escapes such as quotes or newlines. A plain '...' literal treats backslashes literally under standard_conforming_strings (the default), which could produce invalid SQL; values that need no escaping still render as plain '...'.

2026.05.08

Section titled “2026.05.08”

Version: 1.0.0

Initial public release.