sqruff_lib/rules/ambiguous/
am02.rs

1use ahash::AHashMap;
2use sqruff_lib_core::dialects::init::DialectKind;
3use sqruff_lib_core::dialects::syntax::{SyntaxKind, SyntaxSet};
4use sqruff_lib_core::lint_fix::LintFix;
5use sqruff_lib_core::parser::segments::base::SegmentBuilder;
6
7use crate::core::config::Value;
8use crate::core::rules::base::{CloneRule, ErasedRule, LintResult, Rule, RuleGroups};
9use crate::core::rules::context::RuleContext;
10use crate::core::rules::crawlers::{Crawler, SegmentSeekerCrawler};
11
12#[derive(Clone, Debug, Default)]
13pub struct RuleAM02;
14
15impl Rule for RuleAM02 {
16    fn load_from_config(&self, _config: &AHashMap<String, Value>) -> Result<ErasedRule, String> {
17        Ok(RuleAM02.erased())
18    }
19
20    fn name(&self) -> &'static str {
21        "ambiguous.union"
22    }
23
24    fn description(&self) -> &'static str {
25        "Look for UNION keyword not immediately followed by DISTINCT or ALL"
26    }
27
28    fn long_description(&self) -> &'static str {
29        r#"
30**Anti-pattern**
31
32In this example, `UNION DISTINCT` should be preferred over `UNION`, because explicit is better than implicit.
33
34
35```sql
36SELECT a, b FROM table_1
37UNION
38SELECT a, b FROM table_2
39```
40
41**Best practice**
42
43Specify `DISTINCT` or `ALL` after `UNION` (note that `DISTINCT` is the default behavior).
44
45```sql
46SELECT a, b FROM table_1
47UNION DISTINCT
48SELECT a, b FROM table_2
49```
50"#
51    }
52
53    fn groups(&self) -> &'static [RuleGroups] {
54        &[RuleGroups::All, RuleGroups::Core, RuleGroups::Ambiguous]
55    }
56
57    fn dialect_skip(&self) -> &'static [DialectKind] {
58        // TODO: add ansi, hive, mysql
59        // TODO This feels wrong and should bneed fixing
60        &[
61            DialectKind::Bigquery,
62            DialectKind::Postgres,
63            DialectKind::Snowflake,
64            DialectKind::Clickhouse,
65            DialectKind::Sparksql,
66            DialectKind::Duckdb,
67        ]
68    }
69
70    fn eval(&self, rule_cx: &RuleContext) -> Vec<LintResult> {
71        let raw = rule_cx.segment.raw();
72        let raw_upper = raw.to_uppercase();
73
74        if rule_cx.segment.raw().contains("union")
75            && !(raw_upper.contains("ALL") || raw_upper.contains("DISTINCT"))
76        {
77            let edits = vec![
78                SegmentBuilder::keyword(rule_cx.tables.next_id(), "union"),
79                SegmentBuilder::whitespace(rule_cx.tables.next_id(), " "),
80                SegmentBuilder::keyword(rule_cx.tables.next_id(), "distinct"),
81            ];
82
83            let segments = rule_cx.segment.clone();
84            let fixes = vec![LintFix::replace(
85                rule_cx.segment.segments()[0].clone(),
86                edits,
87                None,
88            )];
89
90            return vec![LintResult::new(Some(segments), fixes, None, None)];
91        } else if raw_upper.contains("UNION")
92            && !(raw_upper.contains("ALL") || raw_upper.contains("DISTINCT"))
93        {
94            let edits = vec![
95                SegmentBuilder::keyword(rule_cx.tables.next_id(), "UNION"),
96                SegmentBuilder::whitespace(rule_cx.tables.next_id(), " "),
97                SegmentBuilder::keyword(rule_cx.tables.next_id(), "DISTINCT"),
98            ];
99
100            let segments = rule_cx.segment.clone();
101            let fixes = vec![LintFix::replace(
102                rule_cx.segment.segments()[0].clone(),
103                edits,
104                None,
105            )];
106
107            return vec![LintResult::new(Some(segments), fixes, None, None)];
108        }
109
110        Vec::new()
111    }
112
113    fn is_fix_compatible(&self) -> bool {
114        true
115    }
116
117    fn crawl_behaviour(&self) -> Crawler {
118        SegmentSeekerCrawler::new(const { SyntaxSet::new(&[SyntaxKind::SetOperator]) }).into()
119    }
120}