Skip to main content

polyglot_sql/
lib.rs

1//! Polyglot Core - SQL parsing and dialect translation library
2//!
3//! This library provides the core functionality for parsing SQL statements,
4//! building an abstract syntax tree (AST), and generating SQL in different dialects.
5//!
6//! # Architecture
7//!
8//! The library follows a pipeline architecture:
9//! 1. **Tokenizer** - Converts SQL string to token stream
10//! 2. **Parser** - Builds AST from tokens
11//! 3. **Generator** - Converts AST back to SQL string
12//!
13//! Each stage can be customized per dialect.
14
15pub mod ast_transforms;
16pub mod builder;
17pub mod dialects;
18pub mod diff;
19pub mod error;
20pub mod expressions;
21pub mod function_catalog;
22mod function_registry;
23pub mod generator;
24pub mod helper;
25pub mod lineage;
26pub mod optimizer;
27pub mod parser;
28pub mod planner;
29pub mod resolver;
30pub mod schema;
31pub mod scope;
32pub mod time;
33pub mod tokens;
34pub mod transforms;
35pub mod traversal;
36pub mod trie;
37pub mod validation;
38
39use serde::{Deserialize, Serialize};
40
41pub use ast_transforms::{
42    add_select_columns, add_where, get_aggregate_functions, get_column_names, get_functions,
43    get_identifiers, get_literals, get_subqueries, get_table_names, get_window_functions,
44    node_count, qualify_columns, remove_limit_offset, remove_nodes, remove_select_columns,
45    remove_where, rename_columns, rename_tables, replace_by_type, replace_nodes, set_distinct,
46    set_limit, set_offset,
47};
48pub use dialects::{unregister_custom_dialect, CustomDialectBuilder, Dialect, DialectType};
49pub use error::{Error, Result, ValidationError, ValidationResult, ValidationSeverity};
50pub use expressions::Expression;
51pub use function_catalog::{
52    FunctionCatalog, FunctionNameCase, FunctionSignature, HashMapFunctionCatalog,
53};
54pub use generator::Generator;
55pub use helper::{
56    csv, find_new_name, is_date_unit, is_float, is_int, is_iso_date, is_iso_datetime, merge_ranges,
57    name_sequence, seq_get, split_num_words, tsort, while_changing, DATE_UNITS,
58};
59pub use optimizer::{annotate_types, TypeAnnotator, TypeCoercionClass};
60pub use parser::Parser;
61pub use resolver::{is_column_ambiguous, resolve_column, Resolver, ResolverError, ResolverResult};
62pub use schema::{
63    ensure_schema, from_simple_map, normalize_name, MappingSchema, Schema, SchemaError,
64};
65pub use scope::{
66    build_scope, find_all_in_scope, find_in_scope, traverse_scope, walk_in_scope, ColumnRef, Scope,
67    ScopeType, SourceInfo,
68};
69pub use time::{format_time, is_valid_timezone, subsecond_precision, TIMEZONES};
70pub use tokens::{Token, TokenType, Tokenizer};
71pub use traversal::{
72    contains_aggregate,
73    contains_subquery,
74    contains_window_function,
75    find_ancestor,
76    find_parent,
77    get_columns,
78    get_tables,
79    is_add,
80    is_aggregate,
81    is_alias,
82    is_alter_table,
83    is_and,
84    is_arithmetic,
85    is_avg,
86    is_between,
87    is_boolean,
88    is_case,
89    is_cast,
90    is_coalesce,
91    is_column,
92    is_comparison,
93    is_concat,
94    is_count,
95    is_create_index,
96    is_create_table,
97    is_create_view,
98    is_cte,
99    is_ddl,
100    is_delete,
101    is_div,
102    is_drop_index,
103    is_drop_table,
104    is_drop_view,
105    is_eq,
106    is_except,
107    is_exists,
108    is_from,
109    is_function,
110    is_group_by,
111    is_gt,
112    is_gte,
113    is_having,
114    is_identifier,
115    is_ilike,
116    is_in,
117    // Extended type predicates
118    is_insert,
119    is_intersect,
120    is_is_null,
121    is_join,
122    is_like,
123    is_limit,
124    is_literal,
125    is_logical,
126    is_lt,
127    is_lte,
128    is_max_func,
129    is_min_func,
130    is_mod,
131    is_mul,
132    is_neq,
133    is_not,
134    is_null_if,
135    is_null_literal,
136    is_offset,
137    is_or,
138    is_order_by,
139    is_ordered,
140    is_paren,
141    // Composite predicates
142    is_query,
143    is_safe_cast,
144    is_select,
145    is_set_operation,
146    is_star,
147    is_sub,
148    is_subquery,
149    is_sum,
150    is_table,
151    is_try_cast,
152    is_union,
153    is_update,
154    is_where,
155    is_window_function,
156    is_with,
157    transform,
158    transform_map,
159    BfsIter,
160    DfsIter,
161    ExpressionWalk,
162    ParentInfo,
163    TreeContext,
164};
165pub use trie::{new_trie, new_trie_from_keys, Trie, TrieResult};
166pub use validation::{
167    validate_with_schema, SchemaColumn, SchemaColumnReference, SchemaForeignKey, SchemaTable,
168    SchemaTableReference, SchemaValidationOptions, ValidationSchema,
169};
170
171const DEFAULT_FORMAT_MAX_INPUT_BYTES: usize = 16 * 1024 * 1024; // 16 MiB
172const DEFAULT_FORMAT_MAX_TOKENS: usize = 1_000_000;
173const DEFAULT_FORMAT_MAX_AST_NODES: usize = 1_000_000;
174const DEFAULT_FORMAT_MAX_SET_OP_CHAIN: usize = 256;
175
176fn default_format_max_input_bytes() -> Option<usize> {
177    Some(DEFAULT_FORMAT_MAX_INPUT_BYTES)
178}
179
180fn default_format_max_tokens() -> Option<usize> {
181    Some(DEFAULT_FORMAT_MAX_TOKENS)
182}
183
184fn default_format_max_ast_nodes() -> Option<usize> {
185    Some(DEFAULT_FORMAT_MAX_AST_NODES)
186}
187
188fn default_format_max_set_op_chain() -> Option<usize> {
189    Some(DEFAULT_FORMAT_MAX_SET_OP_CHAIN)
190}
191
192/// Guard options for SQL pretty-formatting.
193///
194/// These limits protect against extremely large/complex queries that can cause
195/// high memory pressure in constrained runtimes (for example browser WASM).
196#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
197#[serde(rename_all = "camelCase")]
198pub struct FormatGuardOptions {
199    /// Maximum allowed SQL input size in bytes.
200    /// `None` disables this check.
201    #[serde(default = "default_format_max_input_bytes")]
202    pub max_input_bytes: Option<usize>,
203    /// Maximum allowed number of tokens after tokenization.
204    /// `None` disables this check.
205    #[serde(default = "default_format_max_tokens")]
206    pub max_tokens: Option<usize>,
207    /// Maximum allowed AST node count after parsing.
208    /// `None` disables this check.
209    #[serde(default = "default_format_max_ast_nodes")]
210    pub max_ast_nodes: Option<usize>,
211    /// Maximum allowed count of set-operation operators (`UNION`/`INTERSECT`/`EXCEPT`)
212    /// observed in a statement before parsing.
213    ///
214    /// `None` disables this check.
215    #[serde(default = "default_format_max_set_op_chain")]
216    pub max_set_op_chain: Option<usize>,
217}
218
219impl Default for FormatGuardOptions {
220    fn default() -> Self {
221        Self {
222            max_input_bytes: default_format_max_input_bytes(),
223            max_tokens: default_format_max_tokens(),
224            max_ast_nodes: default_format_max_ast_nodes(),
225            max_set_op_chain: default_format_max_set_op_chain(),
226        }
227    }
228}
229
230fn format_guard_error(code: &str, actual: usize, limit: usize) -> Error {
231    Error::generate(format!(
232        "{code}: value {actual} exceeds configured limit {limit}"
233    ))
234}
235
236fn enforce_input_guard(sql: &str, options: &FormatGuardOptions) -> Result<()> {
237    if let Some(max) = options.max_input_bytes {
238        let input_bytes = sql.len();
239        if input_bytes > max {
240            return Err(format_guard_error(
241                "E_GUARD_INPUT_TOO_LARGE",
242                input_bytes,
243                max,
244            ));
245        }
246    }
247    Ok(())
248}
249
250fn parse_with_token_guard(
251    sql: &str,
252    dialect: &Dialect,
253    options: &FormatGuardOptions,
254) -> Result<Vec<Expression>> {
255    let tokens = dialect.tokenize(sql)?;
256    if let Some(max) = options.max_tokens {
257        let token_count = tokens.len();
258        if token_count > max {
259            return Err(format_guard_error(
260                "E_GUARD_TOKEN_BUDGET_EXCEEDED",
261                token_count,
262                max,
263            ));
264        }
265    }
266    enforce_set_op_chain_guard(&tokens, options)?;
267
268    let config = crate::parser::ParserConfig {
269        dialect: Some(dialect.dialect_type()),
270        ..Default::default()
271    };
272    let mut parser = Parser::with_source(tokens, config, sql.to_string());
273    parser.parse()
274}
275
276fn is_trivia_token(token_type: TokenType) -> bool {
277    matches!(
278        token_type,
279        TokenType::Space | TokenType::Break | TokenType::LineComment | TokenType::BlockComment
280    )
281}
282
283fn next_significant_token(tokens: &[Token], start: usize) -> Option<&Token> {
284    tokens
285        .iter()
286        .skip(start)
287        .find(|token| !is_trivia_token(token.token_type))
288}
289
290fn is_set_operation_token(tokens: &[Token], idx: usize) -> bool {
291    let token = &tokens[idx];
292    match token.token_type {
293        TokenType::Union | TokenType::Intersect => true,
294        TokenType::Except => {
295            // MINUS is aliased to EXCEPT in the tokenizer, but in ClickHouse minus(...)
296            // is a function call rather than a set operation.
297            if token.text.eq_ignore_ascii_case("minus")
298                && matches!(
299                    next_significant_token(tokens, idx + 1).map(|t| t.token_type),
300                    Some(TokenType::LParen)
301                )
302            {
303                return false;
304            }
305            true
306        }
307        _ => false,
308    }
309}
310
311fn enforce_set_op_chain_guard(tokens: &[Token], options: &FormatGuardOptions) -> Result<()> {
312    let Some(max) = options.max_set_op_chain else {
313        return Ok(());
314    };
315
316    let mut set_op_count = 0usize;
317    for (idx, token) in tokens.iter().enumerate() {
318        if token.token_type == TokenType::Semicolon {
319            set_op_count = 0;
320            continue;
321        }
322
323        if is_set_operation_token(tokens, idx) {
324            set_op_count += 1;
325            if set_op_count > max {
326                return Err(format_guard_error(
327                    "E_GUARD_SET_OP_CHAIN_EXCEEDED",
328                    set_op_count,
329                    max,
330                ));
331            }
332        }
333    }
334
335    Ok(())
336}
337
338fn enforce_ast_guard(expressions: &[Expression], options: &FormatGuardOptions) -> Result<()> {
339    if let Some(max) = options.max_ast_nodes {
340        let ast_nodes: usize = expressions.iter().map(node_count).sum();
341        if ast_nodes > max {
342            return Err(format_guard_error(
343                "E_GUARD_AST_BUDGET_EXCEEDED",
344                ast_nodes,
345                max,
346            ));
347        }
348    }
349    Ok(())
350}
351
352fn format_with_dialect(
353    sql: &str,
354    dialect: &Dialect,
355    options: &FormatGuardOptions,
356) -> Result<Vec<String>> {
357    enforce_input_guard(sql, options)?;
358    let expressions = parse_with_token_guard(sql, dialect, options)?;
359    enforce_ast_guard(&expressions, options)?;
360
361    expressions
362        .iter()
363        .map(|expr| dialect.generate_pretty(expr))
364        .collect()
365}
366
367/// Transpile SQL from one dialect to another.
368///
369/// # Arguments
370/// * `sql` - The SQL string to transpile
371/// * `read` - The source dialect to parse with
372/// * `write` - The target dialect to generate
373///
374/// # Returns
375/// A vector of transpiled SQL statements
376///
377/// # Example
378/// ```
379/// use polyglot_sql::{transpile, DialectType};
380///
381/// let result = transpile(
382///     "SELECT EPOCH_MS(1618088028295)",
383///     DialectType::DuckDB,
384///     DialectType::Hive
385/// );
386/// ```
387pub fn transpile(sql: &str, read: DialectType, write: DialectType) -> Result<Vec<String>> {
388    let read_dialect = Dialect::get(read);
389    let write_dialect = Dialect::get(write);
390    let generic_identity = read == DialectType::Generic && write == DialectType::Generic;
391
392    let expressions = read_dialect.parse(sql)?;
393
394    expressions
395        .into_iter()
396        .map(|expr| {
397            if generic_identity {
398                write_dialect.generate_with_source(&expr, read)
399            } else {
400                let transformed = write_dialect.transform(expr)?;
401                write_dialect.generate_with_source(&transformed, read)
402            }
403        })
404        .collect()
405}
406
407/// Parse SQL into an AST.
408///
409/// # Arguments
410/// * `sql` - The SQL string to parse
411/// * `dialect` - The dialect to use for parsing
412///
413/// # Returns
414/// A vector of parsed expressions
415pub fn parse(sql: &str, dialect: DialectType) -> Result<Vec<Expression>> {
416    let d = Dialect::get(dialect);
417    d.parse(sql)
418}
419
420/// Parse a single SQL statement.
421///
422/// # Arguments
423/// * `sql` - The SQL string containing a single statement
424/// * `dialect` - The dialect to use for parsing
425///
426/// # Returns
427/// The parsed expression, or an error if multiple statements found
428pub fn parse_one(sql: &str, dialect: DialectType) -> Result<Expression> {
429    let mut expressions = parse(sql, dialect)?;
430
431    if expressions.len() != 1 {
432        return Err(Error::parse(
433            format!("Expected 1 statement, found {}", expressions.len()),
434            0,
435            0,
436        ));
437    }
438
439    Ok(expressions.remove(0))
440}
441
442/// Generate SQL from an AST.
443///
444/// # Arguments
445/// * `expression` - The expression to generate SQL from
446/// * `dialect` - The target dialect
447///
448/// # Returns
449/// The generated SQL string
450pub fn generate(expression: &Expression, dialect: DialectType) -> Result<String> {
451    let d = Dialect::get(dialect);
452    d.generate(expression)
453}
454
455/// Format/pretty-print SQL statements.
456///
457/// Uses [`FormatGuardOptions::default`] guards.
458pub fn format(sql: &str, dialect: DialectType) -> Result<Vec<String>> {
459    format_with_options(sql, dialect, &FormatGuardOptions::default())
460}
461
462/// Format/pretty-print SQL statements with configurable guard limits.
463pub fn format_with_options(
464    sql: &str,
465    dialect: DialectType,
466    options: &FormatGuardOptions,
467) -> Result<Vec<String>> {
468    let d = Dialect::get(dialect);
469    format_with_dialect(sql, &d, options)
470}
471
472/// Validate SQL syntax.
473///
474/// # Arguments
475/// * `sql` - The SQL string to validate
476/// * `dialect` - The dialect to use for validation
477///
478/// # Returns
479/// A validation result with any errors found
480pub fn validate(sql: &str, dialect: DialectType) -> ValidationResult {
481    validate_with_options(sql, dialect, &ValidationOptions::default())
482}
483
484/// Options for syntax validation behavior.
485#[derive(Debug, Clone, Copy, Default, Serialize, Deserialize)]
486#[serde(rename_all = "camelCase")]
487pub struct ValidationOptions {
488    /// When enabled, validation rejects non-canonical trailing commas that the parser
489    /// would otherwise accept for compatibility (e.g. `SELECT a, FROM t`).
490    #[serde(default)]
491    pub strict_syntax: bool,
492}
493
494/// Validate SQL syntax with additional validation options.
495pub fn validate_with_options(
496    sql: &str,
497    dialect: DialectType,
498    options: &ValidationOptions,
499) -> ValidationResult {
500    let d = Dialect::get(dialect);
501    match d.parse(sql) {
502        Ok(expressions) => {
503            // Reject bare expressions that aren't valid SQL statements.
504            // The parser accepts any expression at the top level, but bare identifiers,
505            // literals, function calls, etc. are not valid statements.
506            for expr in &expressions {
507                if !expr.is_statement() {
508                    let msg = format!("Invalid expression / Unexpected token");
509                    return ValidationResult::with_errors(vec![ValidationError::error(
510                        msg, "E004",
511                    )]);
512                }
513            }
514            if options.strict_syntax {
515                if let Some(error) = strict_syntax_error(sql, &d) {
516                    return ValidationResult::with_errors(vec![error]);
517                }
518            }
519            ValidationResult::success()
520        }
521        Err(e) => {
522            let error = match &e {
523                Error::Syntax {
524                    message,
525                    line,
526                    column,
527                } => ValidationError::error(message.clone(), "E001").with_location(*line, *column),
528                Error::Tokenize {
529                    message,
530                    line,
531                    column,
532                } => ValidationError::error(message.clone(), "E002").with_location(*line, *column),
533                Error::Parse {
534                    message,
535                    line,
536                    column,
537                } => ValidationError::error(message.clone(), "E003").with_location(*line, *column),
538                _ => ValidationError::error(e.to_string(), "E000"),
539            };
540            ValidationResult::with_errors(vec![error])
541        }
542    }
543}
544
545fn strict_syntax_error(sql: &str, dialect: &Dialect) -> Option<ValidationError> {
546    let tokens = dialect.tokenize(sql).ok()?;
547
548    for (idx, token) in tokens.iter().enumerate() {
549        if token.token_type != TokenType::Comma {
550            continue;
551        }
552
553        let next = tokens.get(idx + 1);
554        let (is_boundary, boundary_name) = match next.map(|t| t.token_type) {
555            Some(TokenType::From) => (true, "FROM"),
556            Some(TokenType::Where) => (true, "WHERE"),
557            Some(TokenType::GroupBy) => (true, "GROUP BY"),
558            Some(TokenType::Having) => (true, "HAVING"),
559            Some(TokenType::Order) | Some(TokenType::OrderBy) => (true, "ORDER BY"),
560            Some(TokenType::Limit) => (true, "LIMIT"),
561            Some(TokenType::Offset) => (true, "OFFSET"),
562            Some(TokenType::Union) => (true, "UNION"),
563            Some(TokenType::Intersect) => (true, "INTERSECT"),
564            Some(TokenType::Except) => (true, "EXCEPT"),
565            Some(TokenType::Qualify) => (true, "QUALIFY"),
566            Some(TokenType::Window) => (true, "WINDOW"),
567            Some(TokenType::Semicolon) | None => (true, "end of statement"),
568            _ => (false, ""),
569        };
570
571        if is_boundary {
572            let message = format!(
573                "Trailing comma before {} is not allowed in strict syntax mode",
574                boundary_name
575            );
576            return Some(
577                ValidationError::error(message, "E005")
578                    .with_location(token.span.line, token.span.column),
579            );
580        }
581    }
582
583    None
584}
585
586/// Transpile SQL from one dialect to another, using string dialect names.
587///
588/// This supports both built-in dialect names (e.g., "postgresql", "mysql") and
589/// custom dialects registered via [`CustomDialectBuilder`].
590///
591/// # Arguments
592/// * `sql` - The SQL string to transpile
593/// * `read` - The source dialect name
594/// * `write` - The target dialect name
595///
596/// # Returns
597/// A vector of transpiled SQL statements, or an error if a dialect name is unknown.
598pub fn transpile_by_name(sql: &str, read: &str, write: &str) -> Result<Vec<String>> {
599    let read_dialect = Dialect::get_by_name(read)
600        .ok_or_else(|| Error::parse(format!("Unknown dialect: {}", read), 0, 0))?;
601    let write_dialect = Dialect::get_by_name(write)
602        .ok_or_else(|| Error::parse(format!("Unknown dialect: {}", write), 0, 0))?;
603    let generic_identity = read_dialect.dialect_type() == DialectType::Generic
604        && write_dialect.dialect_type() == DialectType::Generic;
605
606    let expressions = read_dialect.parse(sql)?;
607
608    expressions
609        .into_iter()
610        .map(|expr| {
611            if generic_identity {
612                write_dialect.generate_with_source(&expr, read_dialect.dialect_type())
613            } else {
614                let transformed = write_dialect.transform(expr)?;
615                write_dialect.generate_with_source(&transformed, read_dialect.dialect_type())
616            }
617        })
618        .collect()
619}
620
621/// Parse SQL into an AST using a string dialect name.
622///
623/// Supports both built-in and custom dialect names.
624pub fn parse_by_name(sql: &str, dialect: &str) -> Result<Vec<Expression>> {
625    let d = Dialect::get_by_name(dialect)
626        .ok_or_else(|| Error::parse(format!("Unknown dialect: {}", dialect), 0, 0))?;
627    d.parse(sql)
628}
629
630/// Generate SQL from an AST using a string dialect name.
631///
632/// Supports both built-in and custom dialect names.
633pub fn generate_by_name(expression: &Expression, dialect: &str) -> Result<String> {
634    let d = Dialect::get_by_name(dialect)
635        .ok_or_else(|| Error::parse(format!("Unknown dialect: {}", dialect), 0, 0))?;
636    d.generate(expression)
637}
638
639/// Format SQL using a string dialect name.
640///
641/// Uses [`FormatGuardOptions::default`] guards.
642pub fn format_by_name(sql: &str, dialect: &str) -> Result<Vec<String>> {
643    format_with_options_by_name(sql, dialect, &FormatGuardOptions::default())
644}
645
646/// Format SQL using a string dialect name with configurable guard limits.
647pub fn format_with_options_by_name(
648    sql: &str,
649    dialect: &str,
650    options: &FormatGuardOptions,
651) -> Result<Vec<String>> {
652    let d = Dialect::get_by_name(dialect)
653        .ok_or_else(|| Error::parse(format!("Unknown dialect: {}", dialect), 0, 0))?;
654    format_with_dialect(sql, &d, options)
655}
656
657#[cfg(test)]
658mod validation_tests {
659    use super::*;
660
661    #[test]
662    fn validate_is_permissive_by_default_for_trailing_commas() {
663        let result = validate("SELECT name, FROM employees", DialectType::Generic);
664        assert!(result.valid, "Result: {:?}", result.errors);
665    }
666
667    #[test]
668    fn validate_with_options_rejects_trailing_comma_before_from() {
669        let options = ValidationOptions {
670            strict_syntax: true,
671        };
672        let result = validate_with_options(
673            "SELECT name, FROM employees",
674            DialectType::Generic,
675            &options,
676        );
677        assert!(!result.valid, "Result should be invalid");
678        assert!(
679            result.errors.iter().any(|e| e.code == "E005"),
680            "Expected E005, got: {:?}",
681            result.errors
682        );
683    }
684
685    #[test]
686    fn validate_with_options_rejects_trailing_comma_before_where() {
687        let options = ValidationOptions {
688            strict_syntax: true,
689        };
690        let result = validate_with_options(
691            "SELECT name FROM employees, WHERE salary > 10",
692            DialectType::Generic,
693            &options,
694        );
695        assert!(!result.valid, "Result should be invalid");
696        assert!(
697            result.errors.iter().any(|e| e.code == "E005"),
698            "Expected E005, got: {:?}",
699            result.errors
700        );
701    }
702}
703
704#[cfg(test)]
705mod format_tests {
706    use super::*;
707
708    #[test]
709    fn format_basic_query() {
710        let result = format("SELECT a,b FROM t", DialectType::Generic).expect("format failed");
711        assert_eq!(result.len(), 1);
712        assert!(result[0].contains('\n'));
713    }
714
715    #[test]
716    fn format_guard_rejects_large_input() {
717        let options = FormatGuardOptions {
718            max_input_bytes: Some(7),
719            max_tokens: None,
720            max_ast_nodes: None,
721            max_set_op_chain: None,
722        };
723        let err = format_with_options("SELECT 1", DialectType::Generic, &options)
724            .expect_err("expected guard error");
725        assert!(err.to_string().contains("E_GUARD_INPUT_TOO_LARGE"));
726    }
727
728    #[test]
729    fn format_guard_rejects_token_budget() {
730        let options = FormatGuardOptions {
731            max_input_bytes: None,
732            max_tokens: Some(1),
733            max_ast_nodes: None,
734            max_set_op_chain: None,
735        };
736        let err = format_with_options("SELECT 1", DialectType::Generic, &options)
737            .expect_err("expected guard error");
738        assert!(err.to_string().contains("E_GUARD_TOKEN_BUDGET_EXCEEDED"));
739    }
740
741    #[test]
742    fn format_guard_rejects_ast_budget() {
743        let options = FormatGuardOptions {
744            max_input_bytes: None,
745            max_tokens: None,
746            max_ast_nodes: Some(1),
747            max_set_op_chain: None,
748        };
749        let err = format_with_options("SELECT 1", DialectType::Generic, &options)
750            .expect_err("expected guard error");
751        assert!(err.to_string().contains("E_GUARD_AST_BUDGET_EXCEEDED"));
752    }
753
754    #[test]
755    fn format_guard_rejects_set_op_chain_budget() {
756        let options = FormatGuardOptions {
757            max_input_bytes: None,
758            max_tokens: None,
759            max_ast_nodes: None,
760            max_set_op_chain: Some(1),
761        };
762        let err = format_with_options(
763            "SELECT 1 UNION ALL SELECT 2 UNION ALL SELECT 3",
764            DialectType::Generic,
765            &options,
766        )
767        .expect_err("expected guard error");
768        assert!(err.to_string().contains("E_GUARD_SET_OP_CHAIN_EXCEEDED"));
769    }
770
771    #[test]
772    fn format_guard_does_not_treat_clickhouse_minus_function_as_set_op() {
773        let options = FormatGuardOptions {
774            max_input_bytes: None,
775            max_tokens: None,
776            max_ast_nodes: None,
777            max_set_op_chain: Some(0),
778        };
779        let result = format_with_options("SELECT minus(3, 2)", DialectType::ClickHouse, &options);
780        assert!(result.is_ok(), "Result: {:?}", result);
781    }
782
783    #[test]
784    fn format_default_guard_rejects_deep_union_chain_before_parse() {
785        let base = "SELECT col0, col1 FROM t";
786        let mut sql = base.to_string();
787        for _ in 0..1100 {
788            sql.push_str(" UNION ALL ");
789            sql.push_str(base);
790        }
791
792        let err = format(&sql, DialectType::Athena).expect_err("expected guard error");
793        assert!(err.to_string().contains("E_GUARD_SET_OP_CHAIN_EXCEEDED"));
794    }
795}