1--
2-- Regular expression tests
3--
4
5-- Don't want to have to double backslashes in regexes
6set standard_conforming_strings = on;
7
8-- Test simple quantified backrefs
9select 'bbbbb' ~ '^([bc])\1*$' as t;
10select 'ccc' ~ '^([bc])\1*$' as t;
11select 'xxx' ~ '^([bc])\1*$' as f;
12select 'bbc' ~ '^([bc])\1*$' as f;
13select 'b' ~ '^([bc])\1*$' as t;
14
15-- Test quantified backref within a larger expression
16select 'abc abc abc' ~ '^(\w+)( \1)+$' as t;
17select 'abc abd abc' ~ '^(\w+)( \1)+$' as f;
18select 'abc abc abd' ~ '^(\w+)( \1)+$' as f;
19select 'abc abc abc' ~ '^(.+)( \1)+$' as t;
20select 'abc abd abc' ~ '^(.+)( \1)+$' as f;
21select 'abc abc abd' ~ '^(.+)( \1)+$' as f;
22
23-- Test some cases that crashed in 9.2beta1 due to pmatch[] array overrun
24select substring('asd TO foo' from ' TO (([a-z0-9._]+|"([^"]+|"")+")+)');
25select substring('a' from '((a))+');
26select substring('a' from '((a)+)');
27
28-- Test lookahead constraints
29select regexp_matches('ab', 'a(?=b)b*');
30select regexp_matches('a', 'a(?=b)b*');
31select regexp_matches('abc', 'a(?=b)b*(?=c)c*');
32select regexp_matches('ab', 'a(?=b)b*(?=c)c*');
33select regexp_matches('ab', 'a(?!b)b*');
34select regexp_matches('a', 'a(?!b)b*');
35select regexp_matches('b', '(?=b)b');
36select regexp_matches('a', '(?=b)b');
37
38-- Test lookbehind constraints
39select regexp_matches('abb', '(?<=a)b*');
40select regexp_matches('a', 'a(?<=a)b*');
41select regexp_matches('abc', 'a(?<=a)b*(?<=b)c*');
42select regexp_matches('ab', 'a(?<=a)b*(?<=b)c*');
43select regexp_matches('ab', 'a*(?<!a)b*');
44select regexp_matches('ab', 'a*(?<!a)b+');
45select regexp_matches('b', 'a*(?<!a)b+');
46select regexp_matches('a', 'a(?<!a)b*');
47select regexp_matches('b', '(?<=b)b');
48select regexp_matches('foobar', '(?<=f)b+');
49select regexp_matches('foobar', '(?<=foo)b+');
50select regexp_matches('foobar', '(?<=oo)b+');
51
52-- Test optimization of single-chr-or-bracket-expression lookaround constraints
53select 'xz' ~ 'x(?=[xy])';
54select 'xy' ~ 'x(?=[xy])';
55select 'xz' ~ 'x(?![xy])';
56select 'xy' ~ 'x(?![xy])';
57select 'x'  ~ 'x(?![xy])';
58select 'xyy' ~ '(?<=[xy])yy+';
59select 'zyy' ~ '(?<=[xy])yy+';
60select 'xyy' ~ '(?<![xy])yy+';
61select 'zyy' ~ '(?<![xy])yy+';
62
63-- Test conversion of regex patterns to indexable conditions
64explain (costs off) select * from pg_proc where proname ~ 'abc';
65explain (costs off) select * from pg_proc where proname ~ '^abc';
66explain (costs off) select * from pg_proc where proname ~ '^abc$';
67explain (costs off) select * from pg_proc where proname ~ '^abcd*e';
68explain (costs off) select * from pg_proc where proname ~ '^abc+d';
69explain (costs off) select * from pg_proc where proname ~ '^(abc)(def)';
70explain (costs off) select * from pg_proc where proname ~ '^(abc)$';
71explain (costs off) select * from pg_proc where proname ~ '^(abc)?d';
72explain (costs off) select * from pg_proc where proname ~ '^abcd(x|(?=\w\w)q)';
73
74-- Test for infinite loop in pullback() (CVE-2007-4772)
75select 'a' ~ '($|^)*';
76
77-- These cases expose a bug in the original fix for CVE-2007-4772
78select 'a' ~ '(^)+^';
79select 'a' ~ '$($$)+';
80
81-- More cases of infinite loop in pullback(), not fixed by CVE-2007-4772 fix
82select 'a' ~ '($^)+';
83select 'a' ~ '(^$)*';
84select 'aa bb cc' ~ '(^(?!aa))+';
85select 'aa x' ~ '(^(?!aa)(?!bb)(?!cc))+';
86select 'bb x' ~ '(^(?!aa)(?!bb)(?!cc))+';
87select 'cc x' ~ '(^(?!aa)(?!bb)(?!cc))+';
88select 'dd x' ~ '(^(?!aa)(?!bb)(?!cc))+';
89
90-- Test for infinite loop in fixempties() (Tcl bugs 3604074, 3606683)
91select 'a' ~ '((((((a)*)*)*)*)*)*';
92select 'a' ~ '((((((a+|)+|)+|)+|)+|)+|)';
93
94-- These cases used to give too-many-states failures
95select 'x' ~ 'abcd(\m)+xyz';
96select 'a' ~ '^abcd*(((((^(a c(e?d)a+|)+|)+|)+|)+|a)+|)';
97select 'x' ~ 'a^(^)bcd*xy(((((($a+|)+|)+|)+$|)+|)+|)^$';
98select 'x' ~ 'xyz(\Y\Y)+';
99select 'x' ~ 'x|(?:\M)+';
100
101-- This generates O(N) states but O(N^2) arcs, so it causes problems
102-- if arc count is not constrained
103select 'x' ~ repeat('x*y*z*', 1000);
104
105-- Test backref in combination with non-greedy quantifier
106-- https://core.tcl.tk/tcl/tktview/6585b21ca8fa6f3678d442b97241fdd43dba2ec0
107select 'Programmer' ~ '(\w).*?\1' as t;
108select regexp_matches('Programmer', '(\w)(.*?\1)', 'g');
109
110-- Test for proper matching of non-greedy iteration (bug #11478)
111select regexp_matches('foo/bar/baz',
112                      '^([^/]+?)(?:/([^/]+?))(?:/([^/]+?))?$', '');
113
114-- Test that greediness can be overridden by outer quantifier
115select regexp_matches('llmmmfff', '^(l*)(.*)(f*)$');
116select regexp_matches('llmmmfff', '^(l*){1,1}(.*)(f*)$');
117select regexp_matches('llmmmfff', '^(l*){1,1}?(.*)(f*)$');
118select regexp_matches('llmmmfff', '^(l*){1,1}?(.*){1,1}?(f*)$');
119select regexp_matches('llmmmfff', '^(l*?)(.*)(f*)$');
120select regexp_matches('llmmmfff', '^(l*?){1,1}(.*)(f*)$');
121select regexp_matches('llmmmfff', '^(l*?){1,1}?(.*)(f*)$');
122select regexp_matches('llmmmfff', '^(l*?){1,1}?(.*){1,1}?(f*)$');
123
124-- Test for infinite loop in cfindloop with zero-length possible match
125-- but no actual match (can only happen in the presence of backrefs)
126select 'a' ~ '$()|^\1';
127select 'a' ~ '.. ()|\1';
128select 'a' ~ '()*\1';
129select 'a' ~ '()+\1';
130
131-- Test incorrect removal of capture groups within {0}
132select 'xxx' ~ '(.){0}(\1)' as f;
133select 'xxx' ~ '((.)){0}(\2)' as f;
134select 'xyz' ~ '((.)){0}(\2){0}' as t;
135
136-- Test ancient oversight in when to apply zaptreesubs
137select 'abcdef' ~ '^(.)\1|\1.' as f;
138select 'abadef' ~ '^((.)\2|..)\2' as f;
139
140-- Error conditions
141select 'xyz' ~ 'x(\w)(?=\1)';  -- no backrefs in LACONs
142select 'xyz' ~ 'x(\w)(?=(\1))';
143select 'a' ~ '\x7fffffff';  -- invalid chr code
144