prometheus/promql/parser/lex_test.go
Harkishen Singh fc8e769d71
Use ASSIGN when using = inside braces (#7911)
* Fix EQL when using = inside braces.

Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>

* EQL => EQLC and ASSIGN => EQL

Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>

* Aligned yacc code.

Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>
2020-09-09 15:40:02 +05:30

742 lines
16 KiB
Go

// Copyright 2015 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 parser
import (
"testing"
"github.com/prometheus/prometheus/util/testutil"
)
type testCase struct {
input string
expected []Item
fail bool
seriesDesc bool // Whether to lex a series description.
}
var tests = []struct {
name string
tests []testCase
}{
{
name: "common",
tests: []testCase{
{
input: ",",
expected: []Item{{COMMA, 0, ","}},
}, {
input: "()",
expected: []Item{{LEFT_PAREN, 0, `(`}, {RIGHT_PAREN, 1, `)`}},
}, {
input: "{}",
expected: []Item{{LEFT_BRACE, 0, `{`}, {RIGHT_BRACE, 1, `}`}},
}, {
input: "[5m]",
expected: []Item{
{LEFT_BRACKET, 0, `[`},
{DURATION, 1, `5m`},
{RIGHT_BRACKET, 3, `]`},
},
}, {
input: "[ 5m]",
expected: []Item{
{LEFT_BRACKET, 0, `[`},
{DURATION, 2, `5m`},
{RIGHT_BRACKET, 4, `]`},
},
}, {
input: "[ 5m]",
expected: []Item{
{LEFT_BRACKET, 0, `[`},
{DURATION, 3, `5m`},
{RIGHT_BRACKET, 5, `]`},
},
}, {
input: "[ 5m ]",
expected: []Item{
{LEFT_BRACKET, 0, `[`},
{DURATION, 3, `5m`},
{RIGHT_BRACKET, 6, `]`},
},
}, {
input: "\r\n\r",
expected: []Item{},
},
},
},
{
name: "numbers",
tests: []testCase{
{
input: "1",
expected: []Item{{NUMBER, 0, "1"}},
}, {
input: "4.23",
expected: []Item{{NUMBER, 0, "4.23"}},
}, {
input: ".3",
expected: []Item{{NUMBER, 0, ".3"}},
}, {
input: "5.",
expected: []Item{{NUMBER, 0, "5."}},
}, {
input: "NaN",
expected: []Item{{NUMBER, 0, "NaN"}},
}, {
input: "nAN",
expected: []Item{{NUMBER, 0, "nAN"}},
}, {
input: "NaN 123",
expected: []Item{{NUMBER, 0, "NaN"}, {NUMBER, 4, "123"}},
}, {
input: "NaN123",
expected: []Item{{IDENTIFIER, 0, "NaN123"}},
}, {
input: "iNf",
expected: []Item{{NUMBER, 0, "iNf"}},
}, {
input: "Inf",
expected: []Item{{NUMBER, 0, "Inf"}},
}, {
input: "+Inf",
expected: []Item{{ADD, 0, "+"}, {NUMBER, 1, "Inf"}},
}, {
input: "+Inf 123",
expected: []Item{{ADD, 0, "+"}, {NUMBER, 1, "Inf"}, {NUMBER, 5, "123"}},
}, {
input: "-Inf",
expected: []Item{{SUB, 0, "-"}, {NUMBER, 1, "Inf"}},
}, {
input: "Infoo",
expected: []Item{{IDENTIFIER, 0, "Infoo"}},
}, {
input: "-Infoo",
expected: []Item{{SUB, 0, "-"}, {IDENTIFIER, 1, "Infoo"}},
}, {
input: "-Inf 123",
expected: []Item{{SUB, 0, "-"}, {NUMBER, 1, "Inf"}, {NUMBER, 5, "123"}},
}, {
input: "0x123",
expected: []Item{{NUMBER, 0, "0x123"}},
},
},
},
{
name: "strings",
tests: []testCase{
{
input: "\"test\\tsequence\"",
expected: []Item{{STRING, 0, `"test\tsequence"`}},
},
{
input: "\"test\\\\.expression\"",
expected: []Item{{STRING, 0, `"test\\.expression"`}},
},
{
input: "\"test\\.expression\"",
expected: []Item{
{ERROR, 0, "unknown escape sequence U+002E '.'"},
{STRING, 0, `"test\.expression"`},
},
},
{
input: "`test\\.expression`",
expected: []Item{{STRING, 0, "`test\\.expression`"}},
},
{
// See https://github.com/prometheus/prometheus/issues/939.
input: ".٩",
fail: true,
},
},
},
{
name: "durations",
tests: []testCase{
{
input: "5s",
expected: []Item{{DURATION, 0, "5s"}},
}, {
input: "123m",
expected: []Item{{DURATION, 0, "123m"}},
}, {
input: "1h",
expected: []Item{{DURATION, 0, "1h"}},
}, {
input: "3w",
expected: []Item{{DURATION, 0, "3w"}},
}, {
input: "1y",
expected: []Item{{DURATION, 0, "1y"}},
},
},
},
{
name: "identifiers",
tests: []testCase{
{
input: "abc",
expected: []Item{{IDENTIFIER, 0, "abc"}},
}, {
input: "a:bc",
expected: []Item{{METRIC_IDENTIFIER, 0, "a:bc"}},
}, {
input: "abc d",
expected: []Item{{IDENTIFIER, 0, "abc"}, {IDENTIFIER, 4, "d"}},
}, {
input: ":bc",
expected: []Item{{METRIC_IDENTIFIER, 0, ":bc"}},
}, {
input: "0a:bc",
fail: true,
},
},
},
{
name: "comments",
tests: []testCase{
{
input: "# some comment",
expected: []Item{{COMMENT, 0, "# some comment"}},
}, {
input: "5 # 1+1\n5",
expected: []Item{
{NUMBER, 0, "5"},
{COMMENT, 2, "# 1+1"},
{NUMBER, 8, "5"},
},
},
},
},
{
name: "operators",
tests: []testCase{
{
input: `=`,
expected: []Item{{EQL, 0, `=`}},
}, {
// Inside braces equality is a single '=' character but in terms of a token
// it should be treated as ASSIGN.
input: `{=}`,
expected: []Item{{LEFT_BRACE, 0, `{`}, {EQL, 1, `=`}, {RIGHT_BRACE, 2, `}`}},
}, {
input: `==`,
expected: []Item{{EQLC, 0, `==`}},
}, {
input: `!=`,
expected: []Item{{NEQ, 0, `!=`}},
}, {
input: `<`,
expected: []Item{{LSS, 0, `<`}},
}, {
input: `>`,
expected: []Item{{GTR, 0, `>`}},
}, {
input: `>=`,
expected: []Item{{GTE, 0, `>=`}},
}, {
input: `<=`,
expected: []Item{{LTE, 0, `<=`}},
}, {
input: `+`,
expected: []Item{{ADD, 0, `+`}},
}, {
input: `-`,
expected: []Item{{SUB, 0, `-`}},
}, {
input: `*`,
expected: []Item{{MUL, 0, `*`}},
}, {
input: `/`,
expected: []Item{{DIV, 0, `/`}},
}, {
input: `^`,
expected: []Item{{POW, 0, `^`}},
}, {
input: `%`,
expected: []Item{{MOD, 0, `%`}},
}, {
input: `AND`,
expected: []Item{{LAND, 0, `AND`}},
}, {
input: `or`,
expected: []Item{{LOR, 0, `or`}},
}, {
input: `unless`,
expected: []Item{{LUNLESS, 0, `unless`}},
},
},
},
{
name: "aggregators",
tests: []testCase{
{
input: `sum`,
expected: []Item{{SUM, 0, `sum`}},
}, {
input: `AVG`,
expected: []Item{{AVG, 0, `AVG`}},
}, {
input: `GROUP`,
expected: []Item{{GROUP, 0, `GROUP`}},
}, {
input: `MAX`,
expected: []Item{{MAX, 0, `MAX`}},
}, {
input: `min`,
expected: []Item{{MIN, 0, `min`}},
}, {
input: `count`,
expected: []Item{{COUNT, 0, `count`}},
}, {
input: `stdvar`,
expected: []Item{{STDVAR, 0, `stdvar`}},
}, {
input: `stddev`,
expected: []Item{{STDDEV, 0, `stddev`}},
},
},
},
{
name: "keywords",
tests: []testCase{
{
input: "offset",
expected: []Item{{OFFSET, 0, "offset"}},
}, {
input: "by",
expected: []Item{{BY, 0, "by"}},
}, {
input: "without",
expected: []Item{{WITHOUT, 0, "without"}},
}, {
input: "on",
expected: []Item{{ON, 0, "on"}},
}, {
input: "ignoring",
expected: []Item{{IGNORING, 0, "ignoring"}},
}, {
input: "group_left",
expected: []Item{{GROUP_LEFT, 0, "group_left"}},
}, {
input: "group_right",
expected: []Item{{GROUP_RIGHT, 0, "group_right"}},
}, {
input: "bool",
expected: []Item{{BOOL, 0, "bool"}},
},
},
},
{
name: "selectors",
tests: []testCase{
{
input: `台北`,
fail: true,
}, {
input: `{台北='a'}`,
fail: true,
}, {
input: `{0a='a'}`,
fail: true,
}, {
input: `{foo='bar'}`,
expected: []Item{
{LEFT_BRACE, 0, `{`},
{IDENTIFIER, 1, `foo`},
{EQL, 4, `=`},
{STRING, 5, `'bar'`},
{RIGHT_BRACE, 10, `}`},
},
}, {
input: `{foo="bar"}`,
expected: []Item{
{LEFT_BRACE, 0, `{`},
{IDENTIFIER, 1, `foo`},
{EQL, 4, `=`},
{STRING, 5, `"bar"`},
{RIGHT_BRACE, 10, `}`},
},
}, {
input: `{foo="bar\"bar"}`,
expected: []Item{
{LEFT_BRACE, 0, `{`},
{IDENTIFIER, 1, `foo`},
{EQL, 4, `=`},
{STRING, 5, `"bar\"bar"`},
{RIGHT_BRACE, 15, `}`},
},
}, {
input: `{NaN != "bar" }`,
expected: []Item{
{LEFT_BRACE, 0, `{`},
{IDENTIFIER, 1, `NaN`},
{NEQ, 5, `!=`},
{STRING, 8, `"bar"`},
{RIGHT_BRACE, 14, `}`},
},
}, {
input: `{alert=~"bar" }`,
expected: []Item{
{LEFT_BRACE, 0, `{`},
{IDENTIFIER, 1, `alert`},
{EQL_REGEX, 6, `=~`},
{STRING, 8, `"bar"`},
{RIGHT_BRACE, 14, `}`},
},
}, {
input: `{on!~"bar"}`,
expected: []Item{
{LEFT_BRACE, 0, `{`},
{IDENTIFIER, 1, `on`},
{NEQ_REGEX, 3, `!~`},
{STRING, 5, `"bar"`},
{RIGHT_BRACE, 10, `}`},
},
}, {
input: `{alert!#"bar"}`, fail: true,
}, {
input: `{foo:a="bar"}`, fail: true,
},
},
},
{
name: "common errors",
tests: []testCase{
{
input: `=~`, fail: true,
}, {
input: `!~`, fail: true,
}, {
input: `!(`, fail: true,
}, {
input: "1a", fail: true,
},
},
},
{
name: "mismatched parentheses",
tests: []testCase{
{
input: `(`, fail: true,
}, {
input: `())`, fail: true,
}, {
input: `(()`, fail: true,
}, {
input: `{`, fail: true,
}, {
input: `}`, fail: true,
}, {
input: "{{", fail: true,
}, {
input: "{{}}", fail: true,
}, {
input: `[`, fail: true,
}, {
input: `[[`, fail: true,
}, {
input: `[]]`, fail: true,
}, {
input: `[[]]`, fail: true,
}, {
input: `]`, fail: true,
},
},
},
{
name: "encoding issues",
tests: []testCase{
{
input: "\"\xff\"", fail: true,
},
{
input: "`\xff`", fail: true,
},
},
},
{
name: "series descriptions",
tests: []testCase{
{
input: `{} _ 1 x .3`,
expected: []Item{
{LEFT_BRACE, 0, `{`},
{RIGHT_BRACE, 1, `}`},
{SPACE, 2, ` `},
{BLANK, 3, `_`},
{SPACE, 4, ` `},
{NUMBER, 5, `1`},
{SPACE, 6, ` `},
{TIMES, 7, `x`},
{SPACE, 8, ` `},
{NUMBER, 9, `.3`},
},
seriesDesc: true,
},
{
input: `metric +Inf Inf NaN`,
expected: []Item{
{IDENTIFIER, 0, `metric`},
{SPACE, 6, ` `},
{ADD, 7, `+`},
{NUMBER, 8, `Inf`},
{SPACE, 11, ` `},
{NUMBER, 12, `Inf`},
{SPACE, 15, ` `},
{NUMBER, 16, `NaN`},
},
seriesDesc: true,
},
{
input: `metric 1+1x4`,
expected: []Item{
{IDENTIFIER, 0, `metric`},
{SPACE, 6, ` `},
{NUMBER, 7, `1`},
{ADD, 8, `+`},
{NUMBER, 9, `1`},
{TIMES, 10, `x`},
{NUMBER, 11, `4`},
},
seriesDesc: true,
},
},
},
{
name: "subqueries",
tests: []testCase{
{
input: `test_name{on!~"bar"}[4m:4s]`,
expected: []Item{
{IDENTIFIER, 0, `test_name`},
{LEFT_BRACE, 9, `{`},
{IDENTIFIER, 10, `on`},
{NEQ_REGEX, 12, `!~`},
{STRING, 14, `"bar"`},
{RIGHT_BRACE, 19, `}`},
{LEFT_BRACKET, 20, `[`},
{DURATION, 21, `4m`},
{COLON, 23, `:`},
{DURATION, 24, `4s`},
{RIGHT_BRACKET, 26, `]`},
},
},
{
input: `test:name{on!~"bar"}[4m:4s]`,
expected: []Item{
{METRIC_IDENTIFIER, 0, `test:name`},
{LEFT_BRACE, 9, `{`},
{IDENTIFIER, 10, `on`},
{NEQ_REGEX, 12, `!~`},
{STRING, 14, `"bar"`},
{RIGHT_BRACE, 19, `}`},
{LEFT_BRACKET, 20, `[`},
{DURATION, 21, `4m`},
{COLON, 23, `:`},
{DURATION, 24, `4s`},
{RIGHT_BRACKET, 26, `]`},
},
}, {
input: `test:name{on!~"b:ar"}[4m:4s]`,
expected: []Item{
{METRIC_IDENTIFIER, 0, `test:name`},
{LEFT_BRACE, 9, `{`},
{IDENTIFIER, 10, `on`},
{NEQ_REGEX, 12, `!~`},
{STRING, 14, `"b:ar"`},
{RIGHT_BRACE, 20, `}`},
{LEFT_BRACKET, 21, `[`},
{DURATION, 22, `4m`},
{COLON, 24, `:`},
{DURATION, 25, `4s`},
{RIGHT_BRACKET, 27, `]`},
},
}, {
input: `test:name{on!~"b:ar"}[4m:]`,
expected: []Item{
{METRIC_IDENTIFIER, 0, `test:name`},
{LEFT_BRACE, 9, `{`},
{IDENTIFIER, 10, `on`},
{NEQ_REGEX, 12, `!~`},
{STRING, 14, `"b:ar"`},
{RIGHT_BRACE, 20, `}`},
{LEFT_BRACKET, 21, `[`},
{DURATION, 22, `4m`},
{COLON, 24, `:`},
{RIGHT_BRACKET, 25, `]`},
},
}, { // Nested Subquery.
input: `min_over_time(rate(foo{bar="baz"}[2s])[5m:])[4m:3s]`,
expected: []Item{
{IDENTIFIER, 0, `min_over_time`},
{LEFT_PAREN, 13, `(`},
{IDENTIFIER, 14, `rate`},
{LEFT_PAREN, 18, `(`},
{IDENTIFIER, 19, `foo`},
{LEFT_BRACE, 22, `{`},
{IDENTIFIER, 23, `bar`},
{EQL, 26, `=`},
{STRING, 27, `"baz"`},
{RIGHT_BRACE, 32, `}`},
{LEFT_BRACKET, 33, `[`},
{DURATION, 34, `2s`},
{RIGHT_BRACKET, 36, `]`},
{RIGHT_PAREN, 37, `)`},
{LEFT_BRACKET, 38, `[`},
{DURATION, 39, `5m`},
{COLON, 41, `:`},
{RIGHT_BRACKET, 42, `]`},
{RIGHT_PAREN, 43, `)`},
{LEFT_BRACKET, 44, `[`},
{DURATION, 45, `4m`},
{COLON, 47, `:`},
{DURATION, 48, `3s`},
{RIGHT_BRACKET, 50, `]`},
},
},
// Subquery with offset.
{
input: `test:name{on!~"b:ar"}[4m:4s] offset 10m`,
expected: []Item{
{METRIC_IDENTIFIER, 0, `test:name`},
{LEFT_BRACE, 9, `{`},
{IDENTIFIER, 10, `on`},
{NEQ_REGEX, 12, `!~`},
{STRING, 14, `"b:ar"`},
{RIGHT_BRACE, 20, `}`},
{LEFT_BRACKET, 21, `[`},
{DURATION, 22, `4m`},
{COLON, 24, `:`},
{DURATION, 25, `4s`},
{RIGHT_BRACKET, 27, `]`},
{OFFSET, 29, "offset"},
{DURATION, 36, "10m"},
},
}, {
input: `min_over_time(rate(foo{bar="baz"}[2s])[5m:] offset 6m)[4m:3s]`,
expected: []Item{
{IDENTIFIER, 0, `min_over_time`},
{LEFT_PAREN, 13, `(`},
{IDENTIFIER, 14, `rate`},
{LEFT_PAREN, 18, `(`},
{IDENTIFIER, 19, `foo`},
{LEFT_BRACE, 22, `{`},
{IDENTIFIER, 23, `bar`},
{EQL, 26, `=`},
{STRING, 27, `"baz"`},
{RIGHT_BRACE, 32, `}`},
{LEFT_BRACKET, 33, `[`},
{DURATION, 34, `2s`},
{RIGHT_BRACKET, 36, `]`},
{RIGHT_PAREN, 37, `)`},
{LEFT_BRACKET, 38, `[`},
{DURATION, 39, `5m`},
{COLON, 41, `:`},
{RIGHT_BRACKET, 42, `]`},
{OFFSET, 44, `offset`},
{DURATION, 51, `6m`},
{RIGHT_PAREN, 53, `)`},
{LEFT_BRACKET, 54, `[`},
{DURATION, 55, `4m`},
{COLON, 57, `:`},
{DURATION, 58, `3s`},
{RIGHT_BRACKET, 60, `]`},
},
},
{
input: `test:name[ 5m]`,
expected: []Item{
{METRIC_IDENTIFIER, 0, `test:name`},
{LEFT_BRACKET, 9, `[`},
{DURATION, 11, `5m`},
{RIGHT_BRACKET, 13, `]`},
},
},
{
input: `test:name{o:n!~"bar"}[4m:4s]`,
fail: true,
},
{
input: `test:name{on!~"bar"}[4m:4s:4h]`,
fail: true,
},
{
input: `test:name{on!~"bar"}[4m:4s:]`,
fail: true,
},
{
input: `test:name{on!~"bar"}[4m::]`,
fail: true,
},
{
input: `test:name{on!~"bar"}[:4s]`,
fail: true,
},
},
},
}
// TestLexer tests basic functionality of the lexer. More elaborate tests are implemented
// for the parser to avoid duplicated effort.
func TestLexer(t *testing.T) {
for _, typ := range tests {
t.Run(typ.name, func(t *testing.T) {
for i, test := range typ.tests {
l := &Lexer{
input: test.input,
seriesDesc: test.seriesDesc,
}
var out []Item
for l.state = lexStatements; l.state != nil; {
out = append(out, Item{})
l.NextItem(&out[len(out)-1])
}
lastItem := out[len(out)-1]
if test.fail {
hasError := false
for _, item := range out {
if item.Typ == ERROR {
hasError = true
}
}
if !hasError {
t.Logf("%d: input %q", i, test.input)
t.Fatalf("expected lexing error but did not fail")
}
continue
}
if lastItem.Typ == ERROR {
t.Logf("%d: input %q", i, test.input)
t.Fatalf("unexpected lexing error at position %d: %s", lastItem.Pos, lastItem)
}
eofItem := Item{EOF, Pos(len(test.input)), ""}
testutil.Equals(t, lastItem, eofItem, "%d: input %q", i, test.input)
out = out[:len(out)-1]
testutil.Equals(t, test.expected, out, "%d: input %q", i, test.input)
}
})
}
}