API Documentation

From your editors, initialize the tokenizer like this:

var t = new SqlTokenizer(sql);

Then call any of these methods — no need to touch .sqlTokenized or know internal token positions:

• t.getTables() → ["orders", "customers"]
  Deduplicated, CTE-aware, digs into CTE bodies.
• t.getTables({withAliases: true}) → [{name, alias, schema, statementIndex}]
  Use this when you need the full picture of table usage.
• t.getColumns() → [{column, alias, table, role, statementIndex, statementType}]
  Works across SELECT, INSERT, UPDATE, and CREATE TABLE.
  • role: "target" (INSERT/UPDATE/CREATE columns being written), "source" (INSERT...SELECT source columns), or "select" (SELECT output columns).
  • For INSERT and UPDATE, each target entry also gets a value field. For CREATE TABLE, each entry gets a datatype.
• t.getStatements() → [{index, type, sql, tables, columns, ctes, where, targetTable, targetColumns, sourceColumns}]
  One-call overview. (targetColumns/sourceColumns are pre-filtered from columns by role.)
• t.getStatement(index) → Returns the same detailed object as above for a single statement by its index.
• t.getStatementTypes() → ["SELECT", "INSERT", "UPDATE", "CREATE TABLE", "DELETE"]
• t.getStatementCount() → Returns the total integer count of statements parsed.
• t.getCTEs() → [{name, recursive, columns, body, statementIndex}]
• t.getSQL() → Returns the reconstructed, normalized SQL string.

Every method that returns per-statement data accepts an optional {statement: n} parameter to filter the results to just one statement. All return shapes are flat objects with string/array fields — no internal token positions are leaked through the API.