Module normalizer 

SQL normalization — rewrite the AST so structurally identical queries hash to the same string. See normalize as the entry point.

The base pass replaces every literal Value with a ? placeholder, so queries that differ only in their parameter values collapse to the same string.

“Every literal” is meant literally: it includes Values in structurally significant positions, not just bound-parameter slots. A JSON path (JSON_TABLE(data, '$.a'), JSON_EXTRACT(data, '$.a')), a CAST(x AS DATE FORMAT 'YYYY-MM-DD') format string, the TABLESAMPLE (BUCKET 3 OUT OF 10) / (10 PERCENT) counts, and LIMIT / OFFSET are all rewritten to ?. So two queries differing only in such a literal — e.g. selecting a different JSON field or sampling a different bucket — collapse to the same normalized string.

Three opt-in toggles (NormalizerOptions) further collapse repetitive shapes:

• unify_in_list: IN (1, 2, 3) → IN (...).
• unify_values: VALUES (1, 2, 3), (4, 5, 6) → VALUES (...).
• alphabetize_insert_columns: INSERT INTO t (c, b, a) VALUES (...) → INSERT INTO t (a, b, c) VALUES (...), only when VALUES is unified.

Output is one String per parsed statement, formatted by sqlparser’s Display after the rewrite.

Structs

Normalizer

VisitorMut impl that performs the normalization rewrite. Most callers go through normalize / normalize_with_options or Normalizer::normalize (which constructs and drives this visitor internally). Use the struct directly only when you want to integrate the rewrite into a larger AST traversal.

NormalizerOptions

Toggles for normalize_with_options. Defaults to all false (placeholder substitution only).

Functions

normalize

Parse sql under dialect and normalize each statement with default options (literal-to-? placeholder substitution only).

normalize_with_options

Parse sql under dialect and normalize each statement, applying any extra collapses enabled in options.