prometheus/model/labels/regexp_test.go

// Copyright 2020 The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

package labels

import (
	"bufio"
	"fmt"
	"math/rand"
	"os"
	"sort"
	"strconv"
	"strings"
	"testing"
	"time"

	"github.com/grafana/regexp"
	"github.com/grafana/regexp/syntax"
	"github.com/stretchr/testify/require"
)

var (
	asciiRunes = []rune("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789-_")
	regexes    = []string{
		"",
		"foo",
		"^foo",
		"(foo|bar)",
		"foo.*",
		".*foo",
		"^.*foo$",
		"^.+foo$",
		".*",
		".+",
		"foo.+",
		".+foo",
		"foo\n.+",
		"foo\n.*",
		".*foo.*",
		".+foo.+",
		"(?s:.*)",
		"(?s:.+)",
		"(?s:^.*foo$)",
		"(?i:foo)",
		"(?i:(foo|bar))",
		"(?i:(foo1|foo2|bar))",
		"^(?i:foo|oo)|(bar)$",
		"(?i:(foo1|foo2|aaa|bbb|ccc|ddd|eee|fff|ggg|hhh|iii|lll|mmm|nnn|ooo|ppp|qqq|rrr|sss|ttt|uuu|vvv|www|xxx|yyy|zzz))",
		"((.*)(bar|b|buzz)(.+)|foo)$",
		"^$",
		"(prometheus|api_prom)_api_v1_.+",
		"10\\.0\\.(1|2)\\.+",
		"10\\.0\\.(1|2).+",
		"((fo(bar))|.+foo)",
		// A long case sensitive alternation.
		"zQPbMkNO|NNSPdvMi|iWuuSoAl|qbvKMimS|IecrXtPa|seTckYqt|NxnyHkgB|fIDlOgKb|UhlWIygH|OtNoJxHG|cUTkFVIV|mTgFIHjr|jQkoIDtE|PPMKxRXl|AwMfwVkQ|CQyMrTQJ|BzrqxVSi|nTpcWuhF|PertdywG|ZZDgCtXN|WWdDPyyE|uVtNQsKk|BdeCHvPZ|wshRnFlH|aOUIitIp|RxZeCdXT|CFZMslCj|AVBZRDxl|IzIGCnhw|ythYuWiz|oztXVXhl|VbLkwqQx|qvaUgyVC|VawUjPWC|ecloYJuj|boCLTdSU|uPrKeAZx|hrMWLWBq|JOnUNHRM|rYnujkPq|dDEdZhIj|DRrfvugG|yEGfDxVV|YMYdJWuP|PHUQZNWM|AmKNrLis|zTxndVfn|FPsHoJnc|EIulZTua|KlAPhdzg|ScHJJCLt|NtTfMzME|eMCwuFdo|SEpJVJbR|cdhXZeCx|sAVtBwRh|kVFEVcMI|jzJrxraA|tGLHTell|NNWoeSaw|DcOKSetX|UXZAJyka|THpMphDP|rizheevl|kDCBRidd|pCZZRqyu|pSygkitl|SwZGkAaW|wILOrfNX|QkwVOerj|kHOMxPDr|EwOVycJv|AJvtzQFS|yEOjKYYB|LizIINLL|JBRSsfcG|YPiUqqNl|IsdEbvee|MjEpGcBm|OxXZVgEQ|xClXGuxa|UzRCGFEb|buJbvfvA|IPZQxRet|oFYShsMc|oBHffuHO|bzzKrcBR|KAjzrGCl|IPUsAVls|OGMUMbIU|gyDccHuR|bjlalnDd|ZLWjeMna|fdsuIlxQ|dVXtiomV|XxedTjNg|XWMHlNoA|nnyqArQX|opfkWGhb|wYtnhdYb",
		// An extremely long case sensitive alternation. This is a special
		// case because the values share common prefixes rather than being
		// entirely random. This is common in the real world. For example, the
		// values of a label like kubernetes pod will often include the
		// deployment name as a prefix.
		"jyyfj00j0061|jyyfj00j0062|jyyfj94j0093|jyyfj99j0093|jyyfm01j0021|jyyfm02j0021|jyefj00j0192|jyefj00j0193|jyefj00j0194|jyefj00j0195|jyefj00j0196|jyefj00j0197|jyefj00j0290|jyefj00j0291|jyefj00j0292|jyefj00j0293|jyefj00j0294|jyefj00j0295|jyefj00j0296|jyefj00j0297|jyefj89j0394|jyefj90j0394|jyefj91j0394|jyefj95j0347|jyefj96j0322|jyefj96j0347|jyefj97j0322|jyefj97j0347|jyefj98j0322|jyefj98j0347|jyefj99j0320|jyefj99j0322|jyefj99j0323|jyefj99j0335|jyefj99j0336|jyefj99j0344|jyefj99j0347|jyefj99j0349|jyefj99j0351|jyeff00j0117|lyyfm01j0025|lyyfm01j0028|lyyfm01j0041|lyyfm01j0133|lyyfm01j0701|lyyfm02j0025|lyyfm02j0028|lyyfm02j0041|lyyfm02j0133|lyyfm02j0701|lyyfm03j0701|lyefj00j0775|lyefj00j0776|lyefj00j0777|lyefj00j0778|lyefj00j0779|lyefj00j0780|lyefj00j0781|lyefj00j0782|lyefj50j3807|lyefj50j3852|lyefj51j3807|lyefj51j3852|lyefj52j3807|lyefj52j3852|lyefj53j3807|lyefj53j3852|lyefj54j3807|lyefj54j3852|lyefj54j3886|lyefj55j3807|lyefj55j3852|lyefj55j3886|lyefj56j3807|lyefj56j3852|lyefj56j3886|lyefj57j3807|lyefj57j3852|lyefj57j3886|lyefj58j3807|lyefj58j3852|lyefj58j3886|lyefj59j3807|lyefj59j3852|lyefj59j3886|lyefj60j3807|lyefj60j3852|lyefj60j3886|lyefj61j3807|lyefj61j3852|lyefj61j3886|lyefj62j3807|lyefj62j3852|lyefj62j3886|lyefj63j3807|lyefj63j3852|lyefj63j3886|lyefj64j3807|lyefj64j3852|lyefj64j3886|lyefj65j3807|lyefj65j3852|lyefj65j3886|lyefj66j3807|lyefj66j3852|lyefj66j3886|lyefj67j3807|lyefj67j3852|lyefj67j3886|lyefj68j3807|lyefj68j3852|lyefj68j3886|lyefj69j3807|lyefj69j3846|lyefj69j3852|lyefj69j3886|lyefj70j3807|lyefj70j3846|lyefj70j3852|lyefj70j3886|lyefj71j3807|lyefj71j3846|lyefj71j3852|lyefj71j3886|lyefj72j3807|lyefj72j3846|lyefj72j3852|lyefj72j3886|lyefj73j3807|lyefj73j3846|lyefj73j3852|lyefj73j3886|lyefj74j3807|lyefj74j3846|lyefj74j3852|lyefj74j3886|lyefj75j3807|lyefj75j3808|lyefj75j3846|lyefj75j3852|lyefj75j3886|lyefj76j3732|lyefj76j3807|lyefj76j3808|lyefj76j3846|lyefj76j3852|lyefj76j3886|lyefj77j3732|lyefj77j3807|lyefj77j3808|lyefj77j3846|lyefj77j3852|lyefj77j3886|lyefj78j3278|lyefj78j3732|lyefj78j3807|lyefj78j3808|lyefj78j3846|lyefj78j3852|lyefj78j3886|lyefj79j3732|lyefj79j3807|lyefj79j3808|lyefj79j3846|lyefj79j3852|lyefj79j3886|lyefj80j3732|lyefj80j3807|lyefj80j3808|lyefj80j3846|lyefj80j3852|lyefj80j3886|lyefj81j3732|lyefj81j3807|lyefj81j3808|lyefj81j3846|lyefj81j3852|lyefj81j3886|lyefj82j3732|lyefj82j3807|lyefj82j3808|lyefj82j3846|lyefj82j3852|lyefj82j3886|lyefj83j3732|lyefj83j3807|lyefj83j3808|lyefj83j3846|lyefj83j3852|lyefj83j3886|lyefj84j3732|lyefj84j3807|lyefj84j3808|lyefj84j3846|lyefj84j3852|lyefj84j3886|lyefj85j3732|lyefj85j3807|lyefj85j3808|lyefj85j3846|lyefj85j3852|lyefj85j3886|lyefj86j3278|lyefj86j3732|lyefj86j3807|lyefj86j3808|lyefj86j3846|lyefj86j3852|lyefj86j3886|lyefj87j3278|lyefj87j3732|lyefj87j3807|lyefj87j3808|lyefj87j3846|lyefj87j3852|lyefj87j3886|lyefj88j3732|lyefj88j3807|lyefj88j3808|lyefj88j3846|lyefj88j3852|lyefj88j3886|lyefj89j3732|lyefj89j3807|lyefj89j3808|lyefj89j3846|lyefj89j3852|lyefj89j3886|lyefj90j3732|lyefj90j3807|lyefj90j3808|lyefj90j3846|lyefj90j3852|lyefj90j3886|lyefj91j3732|lyefj91j3807|lyefj91j3808|lyefj91j3846|lyefj91j3852|lyefj91j3886|lyefj92j3732|lyefj92j3807|lyefj92j3808|lyefj92j3846|lyefj92j3852|lyefj92j3886|lyefj93j3732|lyefj93j3807|lyefj93j3808|lyefj93j3846|lyefj93j3852|lyefj93j3885|lyefj93j3886|lyefj94j3525|lyefj94j3732|lyefj94j3807|lyefj94j3808|lyefj94j3846|lyefj94j3852|lyefj94j3885|lyefj94j3886|lyefj95j3525|lyefj95j3732|lyefj95j3807|lyefj95j3808|lyefj95j3846|lyefj95j3852|lyefj95j3886|lyefj96j3732|lyefj96j3803|lyefj96j3807|lyefj96j3808|lyefj96j3846|lyefj96j3852|lyefj96j3886|lyefj97j3333|lyefj97j3732|lyefj97j3792|lyefj97j3803|lyefj97j3807|lyefj97j3808|lyefj97j3838|lyefj97j3843|lyefj97j3846|lyefj97j3852|lyefj97j3886|lyefj98j3083|lyefj98j3333|lyefj98j3732|lyefj98j3807|lyefj98j3808|lyefj98j3838|lyefj98j3843|lyefj98j3846|lyefj98j3852|lyefj98j3873|lyefj98j3877|lyefj98j3882|lyefj98j3886|lyefj99j2984|lyefj99j3083|lyefj99j3333|lyefj99j3732|lyefj99j3807|lyefj99j3808|lyefj99j3846|lyefj99j3849|lyefj99j3852|lyefj99j3873|lyefj99j3877|lyefj99j3882|lyefj99j3884|lyefj99j3886|lyeff00j0106|lyeff00j01
		// A long case insensitive alternation.
		"(?i:(zQPbMkNO|NNSPdvMi|iWuuSoAl|qbvKMimS|IecrXtPa|seTckYqt|NxnyHkgB|fIDlOgKb|UhlWIygH|OtNoJxHG|cUTkFVIV|mTgFIHjr|jQkoIDtE|PPMKxRXl|AwMfwVkQ|CQyMrTQJ|BzrqxVSi|nTpcWuhF|PertdywG|ZZDgCtXN|WWdDPyyE|uVtNQsKk|BdeCHvPZ|wshRnFlH|aOUIitIp|RxZeCdXT|CFZMslCj|AVBZRDxl|IzIGCnhw|ythYuWiz|oztXVXhl|VbLkwqQx|qvaUgyVC|VawUjPWC|ecloYJuj|boCLTdSU|uPrKeAZx|hrMWLWBq|JOnUNHRM|rYnujkPq|dDEdZhIj|DRrfvugG|yEGfDxVV|YMYdJWuP|PHUQZNWM|AmKNrLis|zTxndVfn|FPsHoJnc|EIulZTua|KlAPhdzg|ScHJJCLt|NtTfMzME|eMCwuFdo|SEpJVJbR|cdhXZeCx|sAVtBwRh|kVFEVcMI|jzJrxraA|tGLHTell|NNWoeSaw|DcOKSetX|UXZAJyka|THpMphDP|rizheevl|kDCBRidd|pCZZRqyu|pSygkitl|SwZGkAaW|wILOrfNX|QkwVOerj|kHOMxPDr|EwOVycJv|AJvtzQFS|yEOjKYYB|LizIINLL|JBRSsfcG|YPiUqqNl|IsdEbvee|MjEpGcBm|OxXZVgEQ|xClXGuxa|UzRCGFEb|buJbvfvA|IPZQxRet|oFYShsMc|oBHffuHO|bzzKrcBR|KAjzrGCl|IPUsAVls|OGMUMbIU|gyDccHuR|bjlalnDd|ZLWjeMna|fdsuIlxQ|dVXtiomV|XxedTjNg|XWMHlNoA|nnyqArQX|opfkWGhb|wYtnhdYb))",
	}
	values = []string{
		"foo", " foo bar", "bar", "buzz\nbar", "bar foo", "bfoo", "\n", "\nfoo", "foo\n", "hello foo world", "hello foo\n world", "",
		"FOO", "Foo", "OO", "Oo", "\nfoo\n", strings.Repeat("f", 20), "prometheus", "prometheus_api_v1", "prometheus_api_v1_foo",
		"10.0.1.20", "10.0.2.10", "10.0.3.30", "10.0.4.40",
		"foofoo0", "foofoo",
	}
)

func TestNewFastRegexMatcher(t *testing.T) {
	for _, r := range regexes {
		r := r
		for _, v := range values {
			v := v
			t.Run(r+` on "`+v+`"`, func(t *testing.T) {
				t.Parallel()
				m, err := NewFastRegexMatcher(r)
				require.NoError(t, err)
				re := regexp.MustCompile("^(?:" + r + ")$")
				require.Equal(t, re.MatchString(v), m.MatchString(v))
			})
		}

	}
}

func BenchmarkNewFastRegexMatcher(b *testing.B) {
	runBenchmark := func(newFunc func(v string) (*FastRegexMatcher, error)) func(b *testing.B) {
		return func(b *testing.B) {
			for _, r := range regexes {
				b.Run(getTestNameFromRegexp(r), func(b *testing.B) {
					for n := 0; n < b.N; n++ {
						_, err := newFunc(r)
						if err != nil {
							b.Fatal(err)
						}
					}
				})
			}
		}
	}

	b.Run("with cache", runBenchmark(NewFastRegexMatcher))
	b.Run("without cache", runBenchmark(newFastRegexMatcherWithoutCache))
}

func BenchmarkNewFastRegexMatcher_CacheMisses(b *testing.B) {
	// Init the random seed with a constant, so that it doesn't change between runs.
	randGenerator := rand.New(rand.NewSource(1))

	tests := map[string]string{
		"simple regexp":  randString(randGenerator, 10),
		"complex regexp": strings.Join(randStrings(randGenerator, 100, 10), "|"),
	}

	for testName, regexpPrefix := range tests {
		b.Run(testName, func(b *testing.B) {
			// Ensure the cache is empty.
			fastRegexMatcherCache.Purge()

			b.ResetTimer()

			for n := 0; n < b.N; n++ {
				// Unique regexp to emulate 100% cache misses.
				regexp := regexpPrefix + strconv.Itoa(n)

				_, err := NewFastRegexMatcher(regexp)
				if err != nil {
					b.Fatal(err)
				}
			}
		})
	}
}

func TestOptimizeConcatRegex(t *testing.T) {
	cases := []struct {
		regex    string
		prefix   string
		suffix   string
		contains string
	}{
		{regex: "foo(hello|bar)", prefix: "foo", suffix: "", contains: ""},
		{regex: "foo(hello|bar)world", prefix: "foo", suffix: "world", contains: ""},
		{regex: "foo.*", prefix: "foo", suffix: "", contains: ""},
		{regex: "foo.*hello.*bar", prefix: "foo", suffix: "bar", contains: "hello"},
		{regex: ".*foo", prefix: "", suffix: "foo", contains: ""},
		{regex: "^.*foo$", prefix: "", suffix: "foo", contains: ""},
		{regex: ".*foo.*", prefix: "", suffix: "", contains: "foo"},
		{regex: ".*foo.*bar.*", prefix: "", suffix: "", contains: "foo"},
		{regex: ".*(foo|bar).*", prefix: "", suffix: "", contains: ""},
		{regex: ".*[abc].*", prefix: "", suffix: "", contains: ""},
		{regex: ".*((?i)abc).*", prefix: "", suffix: "", contains: ""},
		{regex: ".*(?i:abc).*", prefix: "", suffix: "", contains: ""},
		{regex: "(?i:abc).*", prefix: "", suffix: "", contains: ""},
		{regex: ".*(?i:abc)", prefix: "", suffix: "", contains: ""},
		{regex: ".*(?i:abc)def.*", prefix: "", suffix: "", contains: "def"},
		{regex: "(?i).*(?-i:abc)def", prefix: "", suffix: "", contains: "abc"},
		{regex: ".*(?msU:abc).*", prefix: "", suffix: "", contains: "abc"},
		{regex: "[aA]bc.*", prefix: "", suffix: "", contains: "bc"},
		{regex: "^5..$", prefix: "5", suffix: "", contains: ""},
		{regex: "^release.*", prefix: "release", suffix: "", contains: ""},
		{regex: "^env-[0-9]+laio[1]?[^0-9].*", prefix: "env-", suffix: "", contains: "laio"},
	}

	for _, c := range cases {
		parsed, err := syntax.Parse(c.regex, syntax.Perl)
		require.NoError(t, err)

		prefix, suffix, contains := optimizeConcatRegex(parsed)
		require.Equal(t, c.prefix, prefix)
		require.Equal(t, c.suffix, suffix)
		require.Equal(t, c.contains, contains)
	}
}

// Refer to https://github.com/prometheus/prometheus/issues/2651.
func TestFindSetMatches(t *testing.T) {
	for _, c := range []struct {
		pattern          string
		expMatches       []string
		expCaseSensitive bool
	}{
		// Single value, coming from a `bar=~"foo"` selector.
		{"foo", []string{"foo"}, true},
		{"^foo", []string{"foo"}, true},
		{"^foo$", []string{"foo"}, true},
		// Simple sets alternates.
		{"foo|bar|zz", []string{"foo", "bar", "zz"}, true},
		// Simple sets alternate and concat (bar|baz is parsed as "ba[rz]").
		{"foo|bar|baz", []string{"foo", "bar", "baz"}, true},
		// Simple sets alternate and concat and capture
		{"foo|bar|baz|(zz)", []string{"foo", "bar", "baz", "zz"}, true},
		// Simple sets alternate and concat and alternates with empty matches
		// parsed as  b(ar|(?:)|uzz) where b(?:) means literal b.
		{"bar|b|buzz", []string{"bar", "b", "buzz"}, true},
		// Skip outer anchors (it's enforced anyway at the root).
		{"^(bar|b|buzz)$", []string{"bar", "b", "buzz"}, true},
		{"^(?:prod|production)$", []string{"prod", "production"}, true},
		// Do not optimize regexp with inner anchors.
		{"(bar|b|b^uz$z)", nil, false},
		// Do not optimize regexp with empty string matcher.
		{"^$|Running", nil, false},
		// Simple sets containing escaped characters.
		{"fo\\.o|bar\\?|\\^baz", []string{"fo.o", "bar?", "^baz"}, true},
		// using charclass
		{"[abc]d", []string{"ad", "bd", "cd"}, true},
		// high low charset different => A(B[CD]|EF)|BC[XY]
		{"ABC|ABD|AEF|BCX|BCY", []string{"ABC", "ABD", "AEF", "BCX", "BCY"}, true},
		// triple concat
		{"api_(v1|prom)_push", []string{"api_v1_push", "api_prom_push"}, true},
		// triple concat with multiple alternates
		{"(api|rpc)_(v1|prom)_push", []string{"api_v1_push", "api_prom_push", "rpc_v1_push", "rpc_prom_push"}, true},
		{"(api|rpc)_(v1|prom)_(push|query)", []string{"api_v1_push", "api_v1_query", "api_prom_push", "api_prom_query", "rpc_v1_push", "rpc_v1_query", "rpc_prom_push", "rpc_prom_query"}, true},
		// class starting with "-"
		{"[-1-2][a-c]", []string{"-a", "-b", "-c", "1a", "1b", "1c", "2a", "2b", "2c"}, true},
		{"[1^3]", []string{"1", "3", "^"}, true},
		// OpPlus with concat
		{"(.+)/(foo|bar)", nil, false},
		// Simple sets containing special characters without escaping.
		{"fo.o|bar?|^baz", nil, false},
		// case sensitive wrapper.
		{"(?i)foo", []string{"FOO"}, false},
		// case sensitive wrapper on alternate.
		{"(?i)foo|bar|baz", []string{"FOO", "BAR", "BAZ", "BAr", "BAz"}, false},
		// mixed case sensitivity.
		{"(api|rpc)_(v1|prom)_((?i)push|query)", nil, false},
		// mixed case sensitivity concatenation only without capture group.
		{"api_v1_(?i)push", nil, false},
		// mixed case sensitivity alternation only without capture group.
		{"api|(?i)rpc", nil, false},
		// case sensitive after unsetting insensitivity.
		{"rpc|(?i)(?-i)api", []string{"rpc", "api"}, true},
		// case sensitive after unsetting insensitivity in all alternation options.
		{"(?i)((?-i)api|(?-i)rpc)", []string{"api", "rpc"}, true},
		// mixed case sensitivity after unsetting insensitivity.
		{"(?i)rpc|(?-i)api", nil, false},
		// too high charset combination
		{"(api|rpc)_[^0-9]", nil, false},
		// too many combinations
		{"[a-z][a-z]", nil, false},
	} {
		c := c
		t.Run(c.pattern, func(t *testing.T) {
			t.Parallel()
			parsed, err := syntax.Parse(c.pattern, syntax.Perl)
			require.NoError(t, err)
			matches, actualCaseSensitive := findSetMatches(parsed)
			require.Equal(t, c.expMatches, matches)
			require.Equal(t, c.expCaseSensitive, actualCaseSensitive)
		})
	}
}

func BenchmarkFastRegexMatcher(b *testing.B) {
	// Init the random seed with a constant, so that it doesn't change between runs.
	randGenerator := rand.New(rand.NewSource(1))

	// Generate variable lengths random texts to match against.
	texts := append([]string{}, randStrings(randGenerator, 10, 10)...)
	texts = append(texts, randStrings(randGenerator, 5, 30)...)
	texts = append(texts, randStrings(randGenerator, 1, 100)...)
	texts = append(texts, "foo"+randString(randGenerator, 50))
	texts = append(texts, randString(randGenerator, 50)+"foo")

	for _, r := range regexes {
		b.Run(getTestNameFromRegexp(r), func(b *testing.B) {
			m, err := NewFastRegexMatcher(r)
			require.NoError(b, err)
			b.ResetTimer()
			for i := 0; i < b.N; i++ {
				for _, text := range texts {
					_ = m.MatchString(text)
				}
			}
		})
	}
}

func Test_OptimizeRegex(t *testing.T) {
	for _, c := range []struct {
		pattern string
		exp     StringMatcher
	}{
		{".*", anyStringWithoutNewlineMatcher{}},
		{".*?", anyStringWithoutNewlineMatcher{}},
		{"(?s:.*)", trueMatcher{}},
		{"(.*)", anyStringWithoutNewlineMatcher{}},
		{"^.*$", anyStringWithoutNewlineMatcher{}},
		{".+", &anyNonEmptyStringMatcher{matchNL: false}},
		{"(?s:.+)", &anyNonEmptyStringMatcher{matchNL: true}},
		{"^.+$", &anyNonEmptyStringMatcher{matchNL: false}},
		{"(.+)", &anyNonEmptyStringMatcher{matchNL: false}},
		{"", emptyStringMatcher{}},
		{"^$", emptyStringMatcher{}},
		{"^foo$", &equalStringMatcher{s: "foo", caseSensitive: true}},
		{"^(?i:foo)$", &equalStringMatcher{s: "FOO", caseSensitive: false}},
		{"^((?i:foo)|(bar))$", orStringMatcher([]StringMatcher{&equalStringMatcher{s: "FOO", caseSensitive: false}, &equalStringMatcher{s: "bar", caseSensitive: true}})},
		{"^((?i:foo|oo)|(bar))$", orStringMatcher([]StringMatcher{&equalStringMatcher{s: "FOO", caseSensitive: false}, &equalStringMatcher{s: "OO", caseSensitive: false}, &equalStringMatcher{s: "bar", caseSensitive: true}})},
		{"(?i:(foo1|foo2|bar))", orStringMatcher([]StringMatcher{orStringMatcher([]StringMatcher{&equalStringMatcher{s: "FOO1", caseSensitive: false}, &equalStringMatcher{s: "FOO2", caseSensitive: false}}), &equalStringMatcher{s: "BAR", caseSensitive: false}})},
		{".*foo.*", &containsStringMatcher{substrings: []string{"foo"}, left: anyStringWithoutNewlineMatcher{}, right: anyStringWithoutNewlineMatcher{}}},
		{"(.*)foo.*", &containsStringMatcher{substrings: []string{"foo"}, left: anyStringWithoutNewlineMatcher{}, right: anyStringWithoutNewlineMatcher{}}},
		{"(.*)foo(.*)", &containsStringMatcher{substrings: []string{"foo"}, left: anyStringWithoutNewlineMatcher{}, right: anyStringWithoutNewlineMatcher{}}},
		{"(.+)foo(.*)", &containsStringMatcher{substrings: []string{"foo"}, left: &anyNonEmptyStringMatcher{matchNL: false}, right: anyStringWithoutNewlineMatcher{}}},
		{"^.+foo.+", &containsStringMatcher{substrings: []string{"foo"}, left: &anyNonEmptyStringMatcher{matchNL: false}, right: &anyNonEmptyStringMatcher{matchNL: false}}},
		{"^(.*)(foo)(.*)$", &containsStringMatcher{substrings: []string{"foo"}, left: anyStringWithoutNewlineMatcher{}, right: anyStringWithoutNewlineMatcher{}}},
		{"^(.*)(foo|foobar)(.*)$", &containsStringMatcher{substrings: []string{"foo", "foobar"}, left: anyStringWithoutNewlineMatcher{}, right: anyStringWithoutNewlineMatcher{}}},
		{"^(.*)(foo|foobar)(.+)$", &containsStringMatcher{substrings: []string{"foo", "foobar"}, left: anyStringWithoutNewlineMatcher{}, right: &anyNonEmptyStringMatcher{matchNL: false}}},
		{"^(.*)(bar|b|buzz)(.+)$", &containsStringMatcher{substrings: []string{"bar", "b", "buzz"}, left: anyStringWithoutNewlineMatcher{}, right: &anyNonEmptyStringMatcher{matchNL: false}}},
		{"10\\.0\\.(1|2)\\.+", nil},
		{"10\\.0\\.(1|2).+", &containsStringMatcher{substrings: []string{"10.0.1", "10.0.2"}, left: nil, right: &anyNonEmptyStringMatcher{matchNL: false}}},
		{"^.+foo", &containsStringMatcher{substrings: []string{"foo"}, left: &anyNonEmptyStringMatcher{matchNL: false}, right: nil}},
		{"foo-.*$", &containsStringMatcher{substrings: []string{"foo-"}, left: nil, right: anyStringWithoutNewlineMatcher{}}},
		{"(prometheus|api_prom)_api_v1_.+", &containsStringMatcher{substrings: []string{"prometheus_api_v1_", "api_prom_api_v1_"}, left: nil, right: &anyNonEmptyStringMatcher{matchNL: false}}},
		{"^((.*)(bar|b|buzz)(.+)|foo)$", orStringMatcher([]StringMatcher{&containsStringMatcher{substrings: []string{"bar", "b", "buzz"}, left: anyStringWithoutNewlineMatcher{}, right: &anyNonEmptyStringMatcher{matchNL: false}}, &equalStringMatcher{s: "foo", caseSensitive: true}})},
		{"((fo(bar))|.+foo)", orStringMatcher([]StringMatcher{orStringMatcher([]StringMatcher{&equalStringMatcher{s: "fobar", caseSensitive: true}}), &containsStringMatcher{substrings: []string{"foo"}, left: &anyNonEmptyStringMatcher{matchNL: false}, right: nil}})},
		{"(.+)/(gateway|cortex-gw|cortex-gw-internal)", &containsStringMatcher{substrings: []string{"/gateway", "/cortex-gw", "/cortex-gw-internal"}, left: &anyNonEmptyStringMatcher{matchNL: false}, right: nil}},
		// we don't support case insensitive matching for contains.
		// This is because there's no strings.IndexOfFold function.
		// We can revisit later if this is really popular by using strings.ToUpper.
		{"^(.*)((?i)foo|foobar)(.*)$", nil},
		{"(api|rpc)_(v1|prom)_((?i)push|query)", nil},
		{"[a-z][a-z]", nil},
		{"[1^3]", nil},
		{".*foo.*bar.*", nil},
		{`\d*`, nil},
		{".", nil},
		// This one is not supported because  `stringMatcherFromRegexp` is not reentrant for syntax.OpConcat.
		// It would make the code too complex to handle it.
		{"/|/bar.*", nil},
		{"(.+)/(foo.*|bar$)", nil},
	} {
		c := c
		t.Run(c.pattern, func(t *testing.T) {
			t.Parallel()
			parsed, err := syntax.Parse(c.pattern, syntax.Perl)
			require.NoError(t, err)
			matches := stringMatcherFromRegexp(parsed)
			require.Equal(t, c.exp, matches)
		})
	}
}

func randString(randGenerator *rand.Rand, length int) string {
	b := make([]rune, length)
	for i := range b {
		b[i] = asciiRunes[randGenerator.Intn(len(asciiRunes))]
	}
	return string(b)
}

func randStrings(randGenerator *rand.Rand, many, length int) []string {
	out := make([]string, 0, many)
	for i := 0; i < many; i++ {
		out = append(out, randString(randGenerator, length))
	}
	return out
}

func FuzzFastRegexMatcher_WithStaticallyDefinedRegularExpressions(f *testing.F) {
	// Create all matchers.
	matchers := make([]*FastRegexMatcher, 0, len(regexes))
	res := make([]*regexp.Regexp, 0, len(regexes))
	for _, re := range regexes {
		m, err := NewFastRegexMatcher(re)
		require.NoError(f, err)
		r := regexp.MustCompile("^(?:" + re + ")$")
		matchers = append(matchers, m)
		res = append(res, r)
	}

	// Add known values to seed corpus.
	for _, v := range values {
		f.Add(v)
	}

	f.Fuzz(func(t *testing.T, text string) {
		for i, m := range matchers {
			require.Equalf(t, res[i].MatchString(text), m.MatchString(text), "regexp: %s text: %s", res[i].String(), text)
		}
	})
}

func FuzzFastRegexMatcher_WithFuzzyRegularExpressions(f *testing.F) {
	for _, re := range regexes {
		for _, text := range values {
			f.Add(re, text)
		}
	}

	f.Fuzz(func(t *testing.T, re, text string) {
		m, err := NewFastRegexMatcher(re)
		if err != nil {
			// Ignore invalid regexes.
			return
		}

		reg, err := regexp.Compile("^(?:" + re + ")$")
		if err != nil {
			// Ignore invalid regexes.
			return
		}

		require.Equalf(t, reg.MatchString(text), m.MatchString(text), "regexp: %s text: %s", reg.String(), text)
	})
}

// This test can be used to analyze real queries from Mimir logs. You can extract real queries with a regexp matcher
// running the following command:
//
// logcli --addr=XXX --username=YYY --password=ZZZ query '{namespace=~"(cortex|mimir).*",name="query-frontend"} |= "query stats" |= "=~" --limit=100000 > logs.txt
func TestAnalyzeRealQueries(t *testing.T) {
	t.Skip("Decomment this test only to manually analyze real queries")

	type labelValueInfo struct {
		numMatchingQueries       int     //nolint:unused
		numShardedQueries        int     //nolint:unused
		numSplitQueries          int     //nolint:unused
		optimized                bool    //nolint:unused
		averageParsingTimeMillis float64 //nolint:unused

		// Sorted list of timestamps when the queries have been received.
		queryStartTimes []time.Time
	}

	labelValueRE := regexp.MustCompile(`=~\\"([^"]+)\\"`)
	tsRE := regexp.MustCompile(`ts=([^ ]+)`)
	shardedQueriesRE := regexp.MustCompile(`sharded_queries=(\d+)`)
	splitQueriesRE := regexp.MustCompile(`split_queries=(\d+)`)

	labelValues := make(map[string]*labelValueInfo)

	// Read the logs file line-by-line, and find all values for regex label matchers.
	readFile, err := os.Open("logs.txt")
	require.NoError(t, err)

	fileScanner := bufio.NewScanner(readFile)
	fileScanner.Split(bufio.ScanLines)

	numQueries := 0

	for fileScanner.Scan() {
		line := fileScanner.Text()
		matches := labelValueRE.FindAllStringSubmatch(line, -1)
		if len(matches) == 0 {
			continue
		}

		// Look up query stats.
		tsRaw := tsRE.FindStringSubmatch(line)
		shardedQueriesRaw := shardedQueriesRE.FindStringSubmatch(line)
		splitQueriesRaw := splitQueriesRE.FindStringSubmatch(line)
		shardedQueries := 0
		splitQueries := 0
		var ts time.Time

		if len(tsRaw) > 0 {
			ts, _ = time.Parse(time.RFC3339Nano, tsRaw[1])
		}
		if len(shardedQueriesRaw) > 0 {
			shardedQueries, _ = strconv.Atoi(shardedQueriesRaw[1])
		}
		if len(splitQueriesRaw) > 0 {
			splitQueries, _ = strconv.Atoi(splitQueriesRaw[1])
		}

		numQueries++

		for _, match := range matches {
			info := labelValues[match[1]]
			if info == nil {
				info = &labelValueInfo{}
				labelValues[match[1]] = info
			}

			info.numMatchingQueries++
			info.numShardedQueries += shardedQueries
			info.numSplitQueries += splitQueries

			if !ts.IsZero() {
				info.queryStartTimes = append(info.queryStartTimes, ts)
			}
		}
	}

	// Sort query start times.
	for _, info := range labelValues {
		sort.Slice(info.queryStartTimes, func(i, j int) bool {
			return info.queryStartTimes[i].Before(info.queryStartTimes[j])
		})
	}

	require.NoError(t, readFile.Close())
	t.Logf("Found %d unique regexp matchers out of %d queries", len(labelValues), numQueries)

	// Analyze each regexp matcher found.
	numChecked := 0
	numOptimized := 0

	for re, info := range labelValues {
		m, err := NewFastRegexMatcher(re)
		if err != nil {
			// Ignore it, because we may have failed to extract the label matcher.
			continue
		}

		numChecked++

		// Check if each regexp matcher is supported by our optimization.
		if m.isOptimized() {
			numOptimized++
			info.optimized = true
		}

		// Estimate the parsing complexity.
		startTime := time.Now()
		const numParsingRuns = 1000

		for i := 0; i < numParsingRuns; i++ {
			NewFastRegexMatcher(re)
		}

		info.averageParsingTimeMillis = float64(time.Since(startTime).Milliseconds()) / float64(numParsingRuns)
	}

	t.Logf("Found %d out of %d (%.2f%%) regexp matchers optimized by FastRegexMatcher", numOptimized, numChecked, (float64(numOptimized)/float64(numChecked))*100)

	// Print some statistics.
	for labelValue, info := range labelValues {
		// Find the min/avg/max difference between query start times.
		var (
			minQueryStartTimeDiff time.Duration
			maxQueryStartTimeDiff time.Duration
			avgQueryStartTimeDiff time.Duration
			sumQueryStartTime     time.Duration
			countQueryStartTime   int
		)

		for i := 1; i < len(info.queryStartTimes); i++ {
			diff := info.queryStartTimes[i].Sub(info.queryStartTimes[i-1])

			sumQueryStartTime += diff
			countQueryStartTime++

			if minQueryStartTimeDiff == 0 || diff < minQueryStartTimeDiff {
				minQueryStartTimeDiff = diff
			}
			if diff > maxQueryStartTimeDiff {
				maxQueryStartTimeDiff = diff
			}
		}

		if countQueryStartTime > 0 {
			avgQueryStartTimeDiff = sumQueryStartTime / time.Duration(countQueryStartTime)
		}

		t.Logf("num queries: %d\t num split queries: %d\t num sharded queries: %d\t optimized: %t\t parsing time: %.0fms\t min/avg/max query start time diff (sec): %.2f/%.2f/%.2f regexp: %s",
			info.numMatchingQueries, info.numSplitQueries, info.numShardedQueries, info.optimized, info.averageParsingTimeMillis,
			minQueryStartTimeDiff.Seconds(), avgQueryStartTimeDiff.Seconds(), maxQueryStartTimeDiff.Seconds(), labelValue)
	}
}

func TestOptimizeEqualStringMatchers(t *testing.T) {
	tests := map[string]struct {
		input                 StringMatcher
		expectedValues        []string
		expectedCaseSensitive bool
	}{
		"should skip optimization on orStringMatcher with containsStringMatcher": {
			input: orStringMatcher{
				&equalStringMatcher{s: "FOO", caseSensitive: true},
				&containsStringMatcher{substrings: []string{"a", "b", "c"}},
			},
			expectedValues: nil,
		},
		"should run optimization on orStringMatcher with equalStringMatcher and same case sensitivity": {
			input: orStringMatcher{
				&equalStringMatcher{s: "FOO", caseSensitive: true},
				&equalStringMatcher{s: "bar", caseSensitive: true},
				&equalStringMatcher{s: "baz", caseSensitive: true},
			},
			expectedValues:        []string{"FOO", "bar", "baz"},
			expectedCaseSensitive: true,
		},
		"should skip optimization on orStringMatcher with equalStringMatcher but different case sensitivity": {
			input: orStringMatcher{
				&equalStringMatcher{s: "FOO", caseSensitive: true},
				&equalStringMatcher{s: "bar", caseSensitive: false},
				&equalStringMatcher{s: "baz", caseSensitive: true},
			},
			expectedValues: nil,
		},
		"should run optimization on orStringMatcher with nested orStringMatcher and equalStringMatcher, and same case sensitivity": {
			input: orStringMatcher{
				&equalStringMatcher{s: "FOO", caseSensitive: true},
				orStringMatcher{
					&equalStringMatcher{s: "bar", caseSensitive: true},
					&equalStringMatcher{s: "xxx", caseSensitive: true},
				},
				&equalStringMatcher{s: "baz", caseSensitive: true},
			},
			expectedValues:        []string{"FOO", "bar", "xxx", "baz"},
			expectedCaseSensitive: true,
		},
		"should skip optimization on orStringMatcher with nested orStringMatcher and equalStringMatcher, but different case sensitivity": {
			input: orStringMatcher{
				&equalStringMatcher{s: "FOO", caseSensitive: true},
				orStringMatcher{
					// Case sensitivity is different within items at the same level.
					&equalStringMatcher{s: "bar", caseSensitive: true},
					&equalStringMatcher{s: "xxx", caseSensitive: false},
				},
				&equalStringMatcher{s: "baz", caseSensitive: true},
			},
			expectedValues: nil,
		},
		"should skip optimization on orStringMatcher with nested orStringMatcher and equalStringMatcher, but different case sensitivity in the nested one": {
			input: orStringMatcher{
				&equalStringMatcher{s: "FOO", caseSensitive: true},
				// Case sensitivity is different between the parent and child.
				orStringMatcher{
					&equalStringMatcher{s: "bar", caseSensitive: false},
					&equalStringMatcher{s: "xxx", caseSensitive: false},
				},
				&equalStringMatcher{s: "baz", caseSensitive: true},
			},
			expectedValues: nil,
		},
		"should return unchanged values on few case insensitive matchers": {
			input: orStringMatcher{
				&equalStringMatcher{s: "FOO", caseSensitive: false},
				orStringMatcher{
					&equalStringMatcher{s: "bAr", caseSensitive: false},
				},
				&equalStringMatcher{s: "baZ", caseSensitive: false},
			},
			expectedValues:        []string{"FOO", "bAr", "baZ"},
			expectedCaseSensitive: false,
		},
	}

	for testName, testData := range tests {
		t.Run(testName, func(t *testing.T) {
			actualMatcher := optimizeEqualStringMatchers(testData.input, 0)

			if testData.expectedValues == nil {
				require.IsType(t, testData.input, actualMatcher)
			} else {
				require.IsType(t, &equalMultiStringSliceMatcher{}, actualMatcher)
				require.Equal(t, testData.expectedValues, actualMatcher.(*equalMultiStringSliceMatcher).values)
				require.Equal(t, testData.expectedCaseSensitive, actualMatcher.(*equalMultiStringSliceMatcher).caseSensitive)
			}
		})
	}
}

func TestNewEqualMultiStringMatcher(t *testing.T) {
	tests := map[string]struct {
		values             []string
		caseSensitive      bool
		expectedValuesMap  map[string]struct{}
		expectedValuesList []string
	}{
		"few case sensitive values": {
			values:             []string{"a", "B"},
			caseSensitive:      true,
			expectedValuesList: []string{"a", "B"},
		},
		"few case insensitive values": {
			values:             []string{"a", "B"},
			caseSensitive:      false,
			expectedValuesList: []string{"a", "B"},
		},
		"many case sensitive values": {
			values:            []string{"a", "B", "c", "D", "e", "F", "g", "H", "i", "L", "m", "N", "o", "P", "q", "r"},
			caseSensitive:     true,
			expectedValuesMap: map[string]struct{}{"a": {}, "B": {}, "c": {}, "D": {}, "e": {}, "F": {}, "g": {}, "H": {}, "i": {}, "L": {}, "m": {}, "N": {}, "o": {}, "P": {}, "q": {}, "r": {}},
		},
		"many case insensitive values": {
			values:            []string{"a", "B", "c", "D", "e", "F", "g", "H", "i", "L", "m", "N", "o", "P", "q", "r"},
			caseSensitive:     false,
			expectedValuesMap: map[string]struct{}{"a": {}, "b": {}, "c": {}, "d": {}, "e": {}, "f": {}, "g": {}, "h": {}, "i": {}, "l": {}, "m": {}, "n": {}, "o": {}, "p": {}, "q": {}, "r": {}},
		},
	}

	for testName, testData := range tests {
		t.Run(testName, func(t *testing.T) {
			matcher := newEqualMultiStringMatcher(testData.caseSensitive, len(testData.values))
			for _, v := range testData.values {
				matcher.add(v)
			}
			if testData.expectedValuesMap != nil {
				require.IsType(t, &equalMultiStringMapMatcher{}, matcher)
				require.Equal(t, testData.expectedValuesMap, matcher.(*equalMultiStringMapMatcher).values)
				require.Equal(t, testData.caseSensitive, matcher.(*equalMultiStringMapMatcher).caseSensitive)
			}
			if testData.expectedValuesList != nil {
				require.IsType(t, &equalMultiStringSliceMatcher{}, matcher)
				require.Equal(t, testData.expectedValuesList, matcher.(*equalMultiStringSliceMatcher).values)
				require.Equal(t, testData.caseSensitive, matcher.(*equalMultiStringSliceMatcher).caseSensitive)
			}
		})
	}
}

func TestEqualMultiStringMatcher_Matches(t *testing.T) {
	tests := map[string]struct {
		values             []string
		caseSensitive      bool
		expectedMatches    []string
		expectedNotMatches []string
	}{
		"few case sensitive values": {
			values:             []string{"a", "B"},
			caseSensitive:      true,
			expectedMatches:    []string{"a", "B"},
			expectedNotMatches: []string{"A", "b"},
		},
		"few case insensitive values": {
			values:             []string{"a", "B"},
			caseSensitive:      false,
			expectedMatches:    []string{"a", "A", "b", "B"},
			expectedNotMatches: []string{"c", "C"},
		},
		"many case sensitive values": {
			values:             []string{"a", "B", "c", "D", "e", "F", "g", "H", "i", "L", "m", "N", "o", "P", "q", "r"},
			caseSensitive:      true,
			expectedMatches:    []string{"a", "B"},
			expectedNotMatches: []string{"A", "b"},
		},
		"many case insensitive values": {
			values:             []string{"a", "B", "c", "D", "e", "F", "g", "H", "i", "L", "m", "N", "o", "P", "q", "r"},
			caseSensitive:      false,
			expectedMatches:    []string{"a", "A", "b", "B"},
			expectedNotMatches: []string{"x", "X"},
		},
	}

	for testName, testData := range tests {
		t.Run(testName, func(t *testing.T) {
			matcher := newEqualMultiStringMatcher(testData.caseSensitive, len(testData.values))
			for _, v := range testData.values {
				matcher.add(v)
			}

			for _, v := range testData.expectedMatches {
				require.True(t, matcher.Matches(v), "value: %s", v)
			}
			for _, v := range testData.expectedNotMatches {
				require.False(t, matcher.Matches(v), "value: %s", v)
			}
		})
	}
}

// This benchmark is used to find a good threshold to use to apply the optimization
// done by optimizeEqualStringMatchers()
func BenchmarkOptimizeEqualStringMatchers(b *testing.B) {
	randGenerator := rand.New(rand.NewSource(time.Now().UnixNano()))

	// Generate variable lengths random texts to match against.
	texts := append([]string{}, randStrings(randGenerator, 10, 10)...)
	texts = append(texts, randStrings(randGenerator, 5, 30)...)
	texts = append(texts, randStrings(randGenerator, 1, 100)...)

	for numAlternations := 2; numAlternations <= 256; numAlternations *= 2 {
		for _, caseSensitive := range []bool{true, false} {
			b.Run(fmt.Sprintf("alternations: %d case sensitive: %t", numAlternations, caseSensitive), func(b *testing.B) {
				// Generate a regex with the expected number of alternations.
				re := strings.Join(randStrings(randGenerator, numAlternations, 10), "|")
				if !caseSensitive {
					re = "(?i:(" + re + "))"
				}

				parsed, err := syntax.Parse(re, syntax.Perl)
				require.NoError(b, err)

				unoptimized := stringMatcherFromRegexpInternal(parsed)
				require.IsType(b, orStringMatcher{}, unoptimized)

				optimized := optimizeEqualStringMatchers(unoptimized, 0)
				require.IsType(b, &equalMultiStringMapMatcher{}, optimized)

				b.Run("without optimizeEqualStringMatchers()", func(b *testing.B) {
					for n := 0; n < b.N; n++ {
						for _, t := range texts {
							unoptimized.Matches(t)
						}
					}
				})

				b.Run("with optimizeEqualStringMatchers()", func(b *testing.B) {
					for n := 0; n < b.N; n++ {
						for _, t := range texts {
							optimized.Matches(t)
						}
					}
				})
			})
		}
	}
}

func getTestNameFromRegexp(re string) string {
	if len(re) > 32 {
		return re[:32]
	}
	return re
}