2015-03-30 09:12:51 -07:00
|
|
|
// 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.
|
|
|
|
|
2020-02-03 08:23:44 -08:00
|
|
|
package parser
|
2015-03-30 09:12:51 -07:00
|
|
|
|
|
|
|
import (
|
2022-06-08 01:47:52 -07:00
|
|
|
"errors"
|
2015-03-30 09:12:51 -07:00
|
|
|
"fmt"
|
2021-01-20 02:57:39 -08:00
|
|
|
"math"
|
2017-08-11 11:45:52 -07:00
|
|
|
"os"
|
2015-03-30 09:12:51 -07:00
|
|
|
"runtime"
|
|
|
|
"strconv"
|
2015-04-29 02:36:41 -07:00
|
|
|
"strings"
|
2020-01-09 09:36:13 -08:00
|
|
|
"sync"
|
2015-03-30 09:12:51 -07:00
|
|
|
"time"
|
|
|
|
|
2015-08-20 08:18:46 -07:00
|
|
|
"github.com/prometheus/common/model"
|
2019-03-25 16:01:12 -07:00
|
|
|
|
2023-08-25 14:35:42 -07:00
|
|
|
"github.com/prometheus/prometheus/model/histogram"
|
2021-11-08 06:23:17 -08:00
|
|
|
"github.com/prometheus/prometheus/model/labels"
|
|
|
|
"github.com/prometheus/prometheus/model/timestamp"
|
2015-05-29 04:30:30 -07:00
|
|
|
"github.com/prometheus/prometheus/util/strutil"
|
2015-03-30 09:12:51 -07:00
|
|
|
)
|
|
|
|
|
2020-01-09 09:36:13 -08:00
|
|
|
var parserPool = sync.Pool{
|
|
|
|
New: func() interface{} {
|
|
|
|
return &parser{}
|
|
|
|
},
|
|
|
|
}
|
|
|
|
|
2023-03-22 02:02:10 -07:00
|
|
|
type Parser interface {
|
|
|
|
ParseExpr() (Expr, error)
|
|
|
|
Close()
|
|
|
|
}
|
|
|
|
|
2015-03-30 09:12:51 -07:00
|
|
|
type parser struct {
|
2020-01-09 09:36:13 -08:00
|
|
|
lex Lexer
|
2019-11-27 04:59:03 -08:00
|
|
|
|
2020-01-09 03:26:58 -08:00
|
|
|
inject ItemType
|
2019-11-27 04:59:03 -08:00
|
|
|
injecting bool
|
2019-12-05 08:16:12 -08:00
|
|
|
|
2023-03-22 02:02:10 -07:00
|
|
|
// functions contains all functions supported by the parser instance.
|
|
|
|
functions map[string]*Function
|
|
|
|
|
2020-01-14 08:12:15 -08:00
|
|
|
// Everytime an Item is lexed that could be the end
|
|
|
|
// of certain expressions its end position is stored here.
|
|
|
|
lastClosing Pos
|
|
|
|
|
2020-01-09 03:26:58 -08:00
|
|
|
yyParser yyParserImpl
|
|
|
|
|
2019-12-16 05:58:47 -08:00
|
|
|
generatedParserResult interface{}
|
2020-01-16 07:20:20 -08:00
|
|
|
parseErrors ParseErrors
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2023-03-22 02:02:10 -07:00
|
|
|
type Opt func(p *parser)
|
|
|
|
|
|
|
|
func WithFunctions(functions map[string]*Function) Opt {
|
|
|
|
return func(p *parser) {
|
|
|
|
p.functions = functions
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// NewParser returns a new parser.
|
2023-03-22 02:20:16 -07:00
|
|
|
// nolint:revive
|
2023-03-22 02:02:10 -07:00
|
|
|
func NewParser(input string, opts ...Opt) *parser {
|
|
|
|
p := parserPool.Get().(*parser)
|
|
|
|
|
|
|
|
p.functions = Functions
|
|
|
|
p.injecting = false
|
|
|
|
p.parseErrors = nil
|
|
|
|
p.generatedParserResult = nil
|
|
|
|
|
|
|
|
// Clear lexer struct before reusing.
|
|
|
|
p.lex = Lexer{
|
|
|
|
input: input,
|
|
|
|
state: lexStatements,
|
|
|
|
}
|
|
|
|
|
|
|
|
// Apply user define options.
|
|
|
|
for _, opt := range opts {
|
|
|
|
opt(p)
|
|
|
|
}
|
|
|
|
|
|
|
|
return p
|
|
|
|
}
|
|
|
|
|
|
|
|
func (p *parser) ParseExpr() (expr Expr, err error) {
|
|
|
|
defer p.recover(&err)
|
|
|
|
|
|
|
|
parseResult := p.parseGenerated(START_EXPRESSION)
|
|
|
|
|
|
|
|
if parseResult != nil {
|
|
|
|
expr = parseResult.(Expr)
|
|
|
|
}
|
|
|
|
|
|
|
|
// Only typecheck when there are no syntax errors.
|
|
|
|
if len(p.parseErrors) == 0 {
|
|
|
|
p.checkAST(expr)
|
|
|
|
}
|
|
|
|
|
|
|
|
if len(p.parseErrors) != 0 {
|
|
|
|
err = p.parseErrors
|
|
|
|
}
|
|
|
|
|
|
|
|
return expr, err
|
|
|
|
}
|
|
|
|
|
|
|
|
func (p *parser) Close() {
|
|
|
|
defer parserPool.Put(p)
|
|
|
|
}
|
|
|
|
|
2015-04-29 02:36:41 -07:00
|
|
|
// ParseErr wraps a parsing error with line and position context.
|
|
|
|
type ParseErr struct {
|
2020-01-15 12:01:49 -08:00
|
|
|
PositionRange PositionRange
|
|
|
|
Err error
|
|
|
|
Query string
|
|
|
|
|
2020-02-03 09:48:27 -08:00
|
|
|
// LineOffset is an additional line offset to be added. Only used inside unit tests.
|
|
|
|
LineOffset int
|
2015-04-29 02:36:41 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func (e *ParseErr) Error() string {
|
2020-01-15 12:01:49 -08:00
|
|
|
pos := int(e.PositionRange.Start)
|
|
|
|
lastLineBreak := -1
|
2020-02-03 09:48:27 -08:00
|
|
|
line := e.LineOffset + 1
|
2020-01-15 12:01:49 -08:00
|
|
|
|
|
|
|
var positionStr string
|
|
|
|
|
|
|
|
if pos < 0 || pos > len(e.Query) {
|
|
|
|
positionStr = "invalid position:"
|
|
|
|
} else {
|
|
|
|
|
2020-08-26 02:56:26 -07:00
|
|
|
for i, c := range e.Query[:pos] {
|
2020-01-15 12:01:49 -08:00
|
|
|
if c == '\n' {
|
|
|
|
lastLineBreak = i
|
|
|
|
line++
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
col := pos - lastLineBreak
|
|
|
|
positionStr = fmt.Sprintf("%d:%d:", line, col)
|
|
|
|
}
|
|
|
|
return fmt.Sprintf("%s parse error: %s", positionStr, e.Err)
|
2015-04-29 02:36:41 -07:00
|
|
|
}
|
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
type ParseErrors []ParseErr
|
|
|
|
|
|
|
|
// Since producing multiple error messages might look weird when combined with error wrapping,
|
|
|
|
// only the first error produced by the parser is included in the error string.
|
|
|
|
// If getting the full error list is desired, it is recommended to typecast the error returned
|
|
|
|
// by the parser to ParseErrors and work with the underlying slice.
|
|
|
|
func (errs ParseErrors) Error() string {
|
|
|
|
if len(errs) != 0 {
|
|
|
|
return errs[0].Error()
|
|
|
|
}
|
2020-03-23 07:47:11 -07:00
|
|
|
// Should never happen
|
|
|
|
// Panicking while printing an error seems like a bad idea, so the
|
|
|
|
// situation is explained in the error message instead.
|
|
|
|
return "error contains no error message"
|
2020-01-16 07:20:20 -08:00
|
|
|
}
|
|
|
|
|
2023-08-25 14:35:42 -07:00
|
|
|
// EnrichParseError enriches a single or list of parse errors (used for unit tests and promtool).
|
|
|
|
func EnrichParseError(err error, enrich func(parseErr *ParseErr)) {
|
|
|
|
var parseErr *ParseErr
|
|
|
|
if errors.As(err, &parseErr) {
|
|
|
|
enrich(parseErr)
|
|
|
|
}
|
|
|
|
var parseErrors ParseErrors
|
|
|
|
if errors.As(err, &parseErrors) {
|
|
|
|
for i, e := range parseErrors {
|
|
|
|
enrich(&e)
|
|
|
|
parseErrors[i] = e
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-03-30 09:12:51 -07:00
|
|
|
// ParseExpr returns the expression parsed from the input.
|
2020-01-08 03:04:47 -08:00
|
|
|
func ParseExpr(input string) (expr Expr, err error) {
|
2023-03-22 02:02:10 -07:00
|
|
|
p := NewParser(input)
|
|
|
|
defer p.Close()
|
|
|
|
return p.ParseExpr()
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2015-06-10 10:36:43 -07:00
|
|
|
// ParseMetric parses the input into a metric
|
2016-12-23 04:51:59 -08:00
|
|
|
func ParseMetric(input string) (m labels.Labels, err error) {
|
2023-03-22 02:02:10 -07:00
|
|
|
p := NewParser(input)
|
|
|
|
defer p.Close()
|
2015-06-10 10:36:43 -07:00
|
|
|
defer p.recover(&err)
|
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
parseResult := p.parseGenerated(START_METRIC)
|
|
|
|
if parseResult != nil {
|
|
|
|
m = parseResult.(labels.Labels)
|
|
|
|
}
|
|
|
|
|
|
|
|
if len(p.parseErrors) != 0 {
|
|
|
|
err = p.parseErrors
|
|
|
|
}
|
|
|
|
|
|
|
|
return m, err
|
2015-06-10 10:36:43 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// ParseMetricSelector parses the provided textual metric selector into a list of
|
|
|
|
// label matchers.
|
2016-12-25 02:34:22 -08:00
|
|
|
func ParseMetricSelector(input string) (m []*labels.Matcher, err error) {
|
2023-03-22 02:02:10 -07:00
|
|
|
p := NewParser(input)
|
|
|
|
defer p.Close()
|
2015-06-10 10:36:43 -07:00
|
|
|
defer p.recover(&err)
|
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
parseResult := p.parseGenerated(START_METRIC_SELECTOR)
|
|
|
|
if parseResult != nil {
|
|
|
|
m = parseResult.(*VectorSelector).LabelMatchers
|
|
|
|
}
|
|
|
|
|
|
|
|
if len(p.parseErrors) != 0 {
|
|
|
|
err = p.parseErrors
|
|
|
|
}
|
|
|
|
|
|
|
|
return m, err
|
2015-06-10 10:36:43 -07:00
|
|
|
}
|
|
|
|
|
2020-02-03 08:45:41 -08:00
|
|
|
// SequenceValue is an omittable value in a sequence of time series values.
|
|
|
|
type SequenceValue struct {
|
2023-08-25 14:35:42 -07:00
|
|
|
Value float64
|
|
|
|
Omitted bool
|
|
|
|
Histogram *histogram.FloatHistogram
|
2015-05-11 05:04:53 -07:00
|
|
|
}
|
|
|
|
|
2020-02-03 08:45:41 -08:00
|
|
|
func (v SequenceValue) String() string {
|
2020-02-03 09:48:27 -08:00
|
|
|
if v.Omitted {
|
2015-05-11 05:04:53 -07:00
|
|
|
return "_"
|
|
|
|
}
|
2023-08-25 14:35:42 -07:00
|
|
|
if v.Histogram != nil {
|
|
|
|
return v.Histogram.String()
|
|
|
|
}
|
2020-02-03 09:48:27 -08:00
|
|
|
return fmt.Sprintf("%f", v.Value)
|
2015-05-11 05:04:53 -07:00
|
|
|
}
|
|
|
|
|
2019-12-21 00:44:12 -08:00
|
|
|
type seriesDescription struct {
|
|
|
|
labels labels.Labels
|
2020-02-03 08:45:41 -08:00
|
|
|
values []SequenceValue
|
2019-12-21 00:44:12 -08:00
|
|
|
}
|
|
|
|
|
2020-02-03 09:48:27 -08:00
|
|
|
// ParseSeriesDesc parses the description of a time series.
|
|
|
|
func ParseSeriesDesc(input string) (labels labels.Labels, values []SequenceValue, err error) {
|
2023-03-22 02:02:10 -07:00
|
|
|
p := NewParser(input)
|
2018-03-20 07:30:52 -07:00
|
|
|
p.lex.seriesDesc = true
|
|
|
|
|
2023-03-22 02:02:10 -07:00
|
|
|
defer p.Close()
|
2015-05-11 05:04:53 -07:00
|
|
|
defer p.recover(&err)
|
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
parseResult := p.parseGenerated(START_SERIES_DESCRIPTION)
|
|
|
|
if parseResult != nil {
|
|
|
|
result := parseResult.(*seriesDescription)
|
2015-05-11 05:04:53 -07:00
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
labels = result.labels
|
|
|
|
values = result.values
|
2015-05-11 05:04:53 -07:00
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
}
|
2015-05-11 05:04:53 -07:00
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
if len(p.parseErrors) != 0 {
|
|
|
|
err = p.parseErrors
|
|
|
|
}
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2020-01-16 07:20:20 -08:00
|
|
|
return labels, values, err
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2020-02-26 22:08:31 -08:00
|
|
|
// addParseErrf formats the error and appends it to the list of parsing errors.
|
2020-01-17 07:16:58 -08:00
|
|
|
func (p *parser) addParseErrf(positionRange PositionRange, format string, args ...interface{}) {
|
2022-06-08 01:47:52 -07:00
|
|
|
p.addParseErr(positionRange, fmt.Errorf(format, args...))
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2020-01-17 07:16:58 -08:00
|
|
|
// addParseErr appends the provided error to the list of parsing errors.
|
|
|
|
func (p *parser) addParseErr(positionRange PositionRange, err error) {
|
2020-01-16 07:20:20 -08:00
|
|
|
perr := ParseErr{
|
2020-01-15 12:01:49 -08:00
|
|
|
PositionRange: positionRange,
|
|
|
|
Err: err,
|
|
|
|
Query: p.lex.input,
|
2015-04-29 02:36:41 -07:00
|
|
|
}
|
2020-01-16 07:20:20 -08:00
|
|
|
|
|
|
|
p.parseErrors = append(p.parseErrors, perr)
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2023-08-25 14:35:42 -07:00
|
|
|
func (p *parser) addSemanticError(err error) {
|
|
|
|
p.addParseErr(p.yyParser.lval.item.PositionRange(), err)
|
|
|
|
}
|
|
|
|
|
2019-12-18 09:36:43 -08:00
|
|
|
// unexpected creates a parser error complaining about an unexpected lexer item.
|
|
|
|
// The item that is presented as unexpected is always the last item produced
|
|
|
|
// by the lexer.
|
2021-10-22 01:06:44 -07:00
|
|
|
func (p *parser) unexpected(context, expected string) {
|
2019-12-18 09:36:43 -08:00
|
|
|
var errMsg strings.Builder
|
|
|
|
|
2020-03-16 07:47:47 -07:00
|
|
|
// Do not report lexer errors twice
|
|
|
|
if p.yyParser.lval.item.Typ == ERROR {
|
|
|
|
return
|
|
|
|
}
|
|
|
|
|
2019-12-18 09:36:43 -08:00
|
|
|
errMsg.WriteString("unexpected ")
|
2020-01-09 03:26:58 -08:00
|
|
|
errMsg.WriteString(p.yyParser.lval.item.desc())
|
2019-12-18 09:36:43 -08:00
|
|
|
|
|
|
|
if context != "" {
|
|
|
|
errMsg.WriteString(" in ")
|
|
|
|
errMsg.WriteString(context)
|
|
|
|
}
|
|
|
|
|
|
|
|
if expected != "" {
|
|
|
|
errMsg.WriteString(", expected ")
|
|
|
|
errMsg.WriteString(expected)
|
|
|
|
}
|
|
|
|
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErr(p.yyParser.lval.item.PositionRange(), errors.New(errMsg.String()))
|
2019-12-18 09:36:43 -08:00
|
|
|
}
|
|
|
|
|
2019-03-25 16:01:12 -07:00
|
|
|
var errUnexpected = errors.New("unexpected error")
|
2015-08-03 03:53:31 -07:00
|
|
|
|
2015-03-30 09:12:51 -07:00
|
|
|
// recover is the handler that turns panics into returns from the top level of Parse.
|
|
|
|
func (p *parser) recover(errp *error) {
|
|
|
|
e := recover()
|
style: Replace `else if` cascades with `switch`
Wiser coders than myself have come to the conclusion that a `switch`
statement is almost always superior to a statement that includes any
`else if`.
The exceptions that I have found in our codebase are just these two:
* The `if else` is followed by an additional statement before the next
condition (separated by a `;`).
* The whole thing is within a `for` loop and `break` statements are
used. In this case, using `switch` would require tagging the `for`
loop, which probably tips the balance.
Why are `switch` statements more readable?
For one, fewer curly braces. But more importantly, the conditions all
have the same alignment, so the whole thing follows the natural flow
of going down a list of conditions. With `else if`, in contrast, all
conditions but the first are "hidden" behind `} else if `, harder to
spot and (for no good reason) presented differently from the first
condition.
I'm sure the aforemention wise coders can list even more reasons.
In any case, I like it so much that I have found myself recommending
it in code reviews. I would like to make it a habit in our code base,
without making it a hard requirement that we would test on the CI. But
for that, there has to be a role model, so this commit eliminates all
`if else` occurrences, unless it is autogenerated code or fits one of
the exceptions above.
Signed-off-by: beorn7 <beorn@grafana.com>
2023-04-12 07:14:31 -07:00
|
|
|
switch _, ok := e.(runtime.Error); {
|
|
|
|
case ok:
|
2019-04-15 10:06:25 -07:00
|
|
|
// Print the stack trace but do not inhibit the running application.
|
|
|
|
buf := make([]byte, 64<<10)
|
|
|
|
buf = buf[:runtime.Stack(buf, false)]
|
|
|
|
|
|
|
|
fmt.Fprintf(os.Stderr, "parser panic: %v\n%s", e, buf)
|
|
|
|
*errp = errUnexpected
|
style: Replace `else if` cascades with `switch`
Wiser coders than myself have come to the conclusion that a `switch`
statement is almost always superior to a statement that includes any
`else if`.
The exceptions that I have found in our codebase are just these two:
* The `if else` is followed by an additional statement before the next
condition (separated by a `;`).
* The whole thing is within a `for` loop and `break` statements are
used. In this case, using `switch` would require tagging the `for`
loop, which probably tips the balance.
Why are `switch` statements more readable?
For one, fewer curly braces. But more importantly, the conditions all
have the same alignment, so the whole thing follows the natural flow
of going down a list of conditions. With `else if`, in contrast, all
conditions but the first are "hidden" behind `} else if `, harder to
spot and (for no good reason) presented differently from the first
condition.
I'm sure the aforemention wise coders can list even more reasons.
In any case, I like it so much that I have found myself recommending
it in code reviews. I would like to make it a habit in our code base,
without making it a hard requirement that we would test on the CI. But
for that, there has to be a role model, so this commit eliminates all
`if else` occurrences, unless it is autogenerated code or fits one of
the exceptions above.
Signed-off-by: beorn7 <beorn@grafana.com>
2023-04-12 07:14:31 -07:00
|
|
|
case e != nil:
|
2019-04-15 10:06:25 -07:00
|
|
|
*errp = e.(error)
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-11-26 05:28:36 -08:00
|
|
|
// Lex is expected by the yyLexer interface of the yacc generated parser.
|
2019-12-09 11:03:31 -08:00
|
|
|
// It writes the next Item provided by the lexer to the provided pointer address.
|
2019-11-26 05:28:36 -08:00
|
|
|
// Comments are skipped.
|
|
|
|
//
|
|
|
|
// The yyLexer interface is currently implemented by the parser to allow
|
|
|
|
// the generated and non-generated parts to work together with regards to lookahead
|
|
|
|
// and error handling.
|
|
|
|
//
|
2021-06-17 04:18:51 -07:00
|
|
|
// For more information, see https://pkg.go.dev/golang.org/x/tools/cmd/goyacc.
|
2019-11-26 05:28:36 -08:00
|
|
|
func (p *parser) Lex(lval *yySymType) int {
|
2020-01-09 03:26:58 -08:00
|
|
|
var typ ItemType
|
|
|
|
|
2019-11-27 04:59:03 -08:00
|
|
|
if p.injecting {
|
|
|
|
p.injecting = false
|
2020-01-09 03:26:58 -08:00
|
|
|
return int(p.inject)
|
2020-03-23 07:47:11 -07:00
|
|
|
}
|
|
|
|
// Skip comments.
|
|
|
|
for {
|
|
|
|
p.lex.NextItem(&lval.item)
|
|
|
|
typ = lval.item.Typ
|
|
|
|
if typ != COMMENT {
|
|
|
|
break
|
2020-01-09 03:26:58 -08:00
|
|
|
}
|
2019-11-27 04:59:03 -08:00
|
|
|
}
|
2019-11-26 05:28:36 -08:00
|
|
|
|
2020-01-14 08:12:15 -08:00
|
|
|
switch typ {
|
|
|
|
case ERROR:
|
2020-03-16 07:47:47 -07:00
|
|
|
pos := PositionRange{
|
|
|
|
Start: p.lex.start,
|
|
|
|
End: Pos(len(p.lex.input)),
|
|
|
|
}
|
|
|
|
p.addParseErr(pos, errors.New(p.yyParser.lval.item.Val))
|
|
|
|
|
|
|
|
// Tells yacc that this is the end of input.
|
|
|
|
return 0
|
2020-01-14 08:12:15 -08:00
|
|
|
case EOF:
|
2020-01-09 03:26:58 -08:00
|
|
|
lval.item.Typ = EOF
|
2020-01-08 06:59:25 -08:00
|
|
|
p.InjectItem(0)
|
2021-01-20 02:57:39 -08:00
|
|
|
case RIGHT_BRACE, RIGHT_PAREN, RIGHT_BRACKET, DURATION, NUMBER:
|
2020-01-14 08:12:15 -08:00
|
|
|
p.lastClosing = lval.item.Pos + Pos(len(lval.item.Val))
|
2019-12-05 08:16:12 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
return int(typ)
|
2019-11-26 05:28:36 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
// Error is expected by the yyLexer interface of the yacc generated parser.
|
|
|
|
//
|
2019-12-05 08:16:12 -08:00
|
|
|
// It is a no-op since the parsers error routines are triggered
|
2020-01-02 06:54:09 -08:00
|
|
|
// by mechanisms that allow more fine-grained control
|
2021-06-17 04:18:51 -07:00
|
|
|
// For more information, see https://pkg.go.dev/golang.org/x/tools/cmd/goyacc.
|
2023-04-12 04:05:41 -07:00
|
|
|
func (p *parser) Error(string) {
|
2019-11-26 05:28:36 -08:00
|
|
|
}
|
|
|
|
|
2019-12-09 11:03:31 -08:00
|
|
|
// InjectItem allows injecting a single Item at the beginning of the token stream
|
2019-11-27 04:59:03 -08:00
|
|
|
// consumed by the generated parser.
|
|
|
|
// This allows having multiple start symbols as described in
|
|
|
|
// https://www.gnu.org/software/bison/manual/html_node/Multiple-start_002dsymbols.html .
|
2019-12-09 11:03:31 -08:00
|
|
|
// Only the Lex function used by the generated parser is affected by this injected Item.
|
|
|
|
// Trying to inject when a previously injected Item has not yet been consumed will panic.
|
|
|
|
// Only Item types that are supposed to be used as start symbols are allowed as an argument.
|
2019-11-27 04:59:03 -08:00
|
|
|
func (p *parser) InjectItem(typ ItemType) {
|
|
|
|
if p.injecting {
|
2019-12-09 11:03:31 -08:00
|
|
|
panic("cannot inject multiple Items into the token stream")
|
2019-11-27 04:59:03 -08:00
|
|
|
}
|
|
|
|
|
2019-12-05 08:16:12 -08:00
|
|
|
if typ != 0 && (typ <= startSymbolsStart || typ >= startSymbolsEnd) {
|
2019-11-27 04:59:03 -08:00
|
|
|
panic("cannot inject symbol that isn't start symbol")
|
|
|
|
}
|
|
|
|
|
2020-01-09 03:26:58 -08:00
|
|
|
p.inject = typ
|
2019-11-27 04:59:03 -08:00
|
|
|
p.injecting = true
|
|
|
|
}
|
2021-10-22 01:06:44 -07:00
|
|
|
|
|
|
|
func (p *parser) newBinaryExpression(lhs Node, op Item, modifiers, rhs Node) *BinaryExpr {
|
2020-01-08 03:04:47 -08:00
|
|
|
ret := modifiers.(*BinaryExpr)
|
2019-11-27 04:59:03 -08:00
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
ret.LHS = lhs.(Expr)
|
|
|
|
ret.RHS = rhs.(Expr)
|
|
|
|
ret.Op = op.Typ
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
return ret
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2020-01-14 08:12:15 -08:00
|
|
|
func (p *parser) assembleVectorSelector(vs *VectorSelector) {
|
|
|
|
if vs.Name != "" {
|
|
|
|
nameMatcher, err := labels.NewMatcher(labels.MatchEqual, labels.MetricName, vs.Name)
|
2018-12-22 05:47:13 -08:00
|
|
|
if err != nil {
|
2020-01-08 03:04:47 -08:00
|
|
|
panic(err) // Must not happen with labels.MatchEqual
|
2018-12-22 05:47:13 -08:00
|
|
|
}
|
2020-01-14 08:12:15 -08:00
|
|
|
vs.LabelMatchers = append(vs.LabelMatchers, nameMatcher)
|
2015-05-11 02:45:23 -07:00
|
|
|
}
|
2020-01-08 03:04:47 -08:00
|
|
|
}
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2021-10-22 01:06:44 -07:00
|
|
|
func (p *parser) newAggregateExpr(op Item, modifier, args Node) (ret *AggregateExpr) {
|
2020-01-08 03:04:47 -08:00
|
|
|
ret = modifier.(*AggregateExpr)
|
|
|
|
arguments := args.(Expressions)
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2020-01-17 07:16:58 -08:00
|
|
|
ret.PosRange = PositionRange{
|
|
|
|
Start: op.Pos,
|
|
|
|
End: p.lastClosing,
|
|
|
|
}
|
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
ret.Op = op.Typ
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
if len(arguments) == 0 {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(ret.PositionRange(), "no arguments for aggregate expression provided")
|
2016-02-07 10:03:16 -08:00
|
|
|
|
2020-01-17 07:16:58 -08:00
|
|
|
// Prevents invalid array accesses.
|
2020-01-08 03:04:47 -08:00
|
|
|
return
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
desiredArgs := 1
|
2020-02-03 09:48:27 -08:00
|
|
|
if ret.Op.IsAggregatorWithParam() {
|
2020-01-08 03:04:47 -08:00
|
|
|
desiredArgs = 2
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
ret.Param = arguments[0]
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
if len(arguments) != desiredArgs {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(ret.PositionRange(), "wrong number of arguments for aggregate expression provided, expected %d, got %d", desiredArgs, len(arguments))
|
2020-01-08 03:04:47 -08:00
|
|
|
return
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
ret.Expr = arguments[desiredArgs-1]
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
return ret
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
2023-08-25 14:35:42 -07:00
|
|
|
// newMap is used when building the FloatHistogram from a map.
|
|
|
|
func (p *parser) newMap() (ret map[string]interface{}) {
|
|
|
|
return map[string]interface{}{}
|
|
|
|
}
|
|
|
|
|
|
|
|
// mergeMaps is used to combine maps as they're used to later build the Float histogram.
|
|
|
|
// This will merge the right map into the left map.
|
|
|
|
func (p *parser) mergeMaps(left, right *map[string]interface{}) (ret *map[string]interface{}) {
|
|
|
|
for key, value := range *right {
|
|
|
|
if _, ok := (*left)[key]; ok {
|
|
|
|
p.addParseErrf(PositionRange{}, "duplicate key \"%s\" in histogram", key)
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
(*left)[key] = value
|
|
|
|
}
|
|
|
|
return left
|
|
|
|
}
|
|
|
|
|
|
|
|
func (p *parser) histogramsIncreaseSeries(base, inc *histogram.FloatHistogram, times uint64) ([]SequenceValue, error) {
|
|
|
|
return p.histogramsSeries(base, inc, times, func(a, b *histogram.FloatHistogram) *histogram.FloatHistogram {
|
|
|
|
return a.Add(b)
|
|
|
|
})
|
|
|
|
}
|
|
|
|
|
|
|
|
func (p *parser) histogramsDecreaseSeries(base, inc *histogram.FloatHistogram, times uint64) ([]SequenceValue, error) {
|
|
|
|
return p.histogramsSeries(base, inc, times, func(a, b *histogram.FloatHistogram) *histogram.FloatHistogram {
|
|
|
|
return a.Sub(b)
|
|
|
|
})
|
|
|
|
}
|
|
|
|
|
|
|
|
func (p *parser) histogramsSeries(base, inc *histogram.FloatHistogram, times uint64,
|
|
|
|
combine func(*histogram.FloatHistogram, *histogram.FloatHistogram) *histogram.FloatHistogram,
|
|
|
|
) ([]SequenceValue, error) {
|
|
|
|
ret := make([]SequenceValue, times+1)
|
|
|
|
// Add an additional value (the base) for time 0, which we ignore in tests.
|
|
|
|
ret[0] = SequenceValue{Histogram: base}
|
|
|
|
cur := base
|
|
|
|
for i := uint64(1); i <= times; i++ {
|
|
|
|
if cur.Schema > inc.Schema {
|
|
|
|
return nil, fmt.Errorf("error combining histograms: cannot merge from schema %d to %d", inc.Schema, cur.Schema)
|
|
|
|
}
|
|
|
|
|
|
|
|
cur = combine(cur.Copy(), inc)
|
|
|
|
ret[i] = SequenceValue{Histogram: cur}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret, nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// buildHistogramFromMap is used in the grammar to take then individual parts of the histogram and complete it.
|
|
|
|
func (p *parser) buildHistogramFromMap(desc *map[string]interface{}) *histogram.FloatHistogram {
|
|
|
|
output := &histogram.FloatHistogram{}
|
|
|
|
|
|
|
|
val, ok := (*desc)["schema"]
|
|
|
|
if ok {
|
|
|
|
schema, ok := val.(int64)
|
|
|
|
if ok {
|
|
|
|
output.Schema = int32(schema)
|
|
|
|
} else {
|
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing schema number: %v", val)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
val, ok = (*desc)["sum"]
|
|
|
|
if ok {
|
|
|
|
sum, ok := val.(float64)
|
|
|
|
if ok {
|
|
|
|
output.Sum = sum
|
|
|
|
} else {
|
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing sum number: %v", val)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
val, ok = (*desc)["count"]
|
|
|
|
if ok {
|
|
|
|
count, ok := val.(float64)
|
|
|
|
if ok {
|
|
|
|
output.Count = count
|
|
|
|
} else {
|
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing count number: %v", val)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
val, ok = (*desc)["z_bucket"]
|
|
|
|
if ok {
|
|
|
|
bucket, ok := val.(float64)
|
|
|
|
if ok {
|
|
|
|
output.ZeroCount = bucket
|
|
|
|
} else {
|
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing z_bucket number: %v", val)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
val, ok = (*desc)["z_bucket_w"]
|
|
|
|
if ok {
|
|
|
|
bucketWidth, ok := val.(float64)
|
|
|
|
if ok {
|
|
|
|
output.ZeroThreshold = bucketWidth
|
|
|
|
} else {
|
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing z_bucket_w number: %v", val)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
buckets, spans := p.buildHistogramBucketsAndSpans(desc, "buckets", "offset")
|
|
|
|
output.PositiveBuckets = buckets
|
|
|
|
output.PositiveSpans = spans
|
|
|
|
|
|
|
|
buckets, spans = p.buildHistogramBucketsAndSpans(desc, "n_buckets", "n_offset")
|
|
|
|
output.NegativeBuckets = buckets
|
|
|
|
output.NegativeSpans = spans
|
|
|
|
|
|
|
|
return output
|
|
|
|
}
|
|
|
|
|
|
|
|
func (p *parser) buildHistogramBucketsAndSpans(desc *map[string]interface{}, bucketsKey, offsetKey string,
|
|
|
|
) (buckets []float64, spans []histogram.Span) {
|
|
|
|
bucketCount := 0
|
|
|
|
val, ok := (*desc)[bucketsKey]
|
|
|
|
if ok {
|
|
|
|
val, ok := val.([]float64)
|
|
|
|
if ok {
|
|
|
|
buckets = val
|
|
|
|
bucketCount = len(buckets)
|
|
|
|
} else {
|
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing %s float array: %v", bucketsKey, val)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
offset := int32(0)
|
|
|
|
val, ok = (*desc)[offsetKey]
|
|
|
|
if ok {
|
|
|
|
val, ok := val.(int64)
|
|
|
|
if ok {
|
|
|
|
offset = int32(val)
|
|
|
|
} else {
|
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing %s number: %v", offsetKey, val)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if bucketCount > 0 {
|
|
|
|
spans = []histogram.Span{{Offset: offset, Length: uint32(bucketCount)}}
|
|
|
|
}
|
|
|
|
return
|
|
|
|
}
|
|
|
|
|
2020-01-08 03:04:47 -08:00
|
|
|
// number parses a number.
|
|
|
|
func (p *parser) number(val string) float64 {
|
|
|
|
n, err := strconv.ParseInt(val, 0, 64)
|
|
|
|
f := float64(n)
|
2016-01-24 19:50:46 -08:00
|
|
|
if err != nil {
|
2020-01-08 03:04:47 -08:00
|
|
|
f, err = strconv.ParseFloat(val, 64)
|
2015-06-15 09:34:41 -07:00
|
|
|
}
|
2020-01-08 03:04:47 -08:00
|
|
|
if err != nil {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error parsing number: %s", err)
|
2015-06-15 09:34:41 -07:00
|
|
|
}
|
2020-01-08 03:04:47 -08:00
|
|
|
return f
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// expectType checks the type of the node and raises an error if it
|
|
|
|
// is not of the expected type.
|
2016-12-23 04:51:59 -08:00
|
|
|
func (p *parser) expectType(node Node, want ValueType, context string) {
|
2020-01-17 07:16:58 -08:00
|
|
|
t := p.checkAST(node)
|
2015-03-30 09:12:51 -07:00
|
|
|
if t != want {
|
2020-02-03 09:00:41 -08:00
|
|
|
p.addParseErrf(node.PositionRange(), "expected type %s in %s, got %s", DocumentedType(want), context, DocumentedType(t))
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-11-28 09:09:18 -08:00
|
|
|
// checkAST checks the validity of the provided AST. This includes type checking.
|
2020-01-17 07:16:58 -08:00
|
|
|
func (p *parser) checkAST(node Node) (typ ValueType) {
|
2015-03-30 09:12:51 -07:00
|
|
|
// For expressions the type is determined by their Type function.
|
2018-02-12 04:09:51 -08:00
|
|
|
// Lists do not have a type but are not invalid either.
|
2015-03-30 09:12:51 -07:00
|
|
|
switch n := node.(type) {
|
2018-02-12 04:09:51 -08:00
|
|
|
case Expressions:
|
2016-12-23 04:51:59 -08:00
|
|
|
typ = ValueTypeNone
|
2015-03-30 09:12:51 -07:00
|
|
|
case Expr:
|
|
|
|
typ = n.Type()
|
|
|
|
default:
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(node.PositionRange(), "unknown node type: %T", node)
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// Recursively check correct typing for child nodes and raise
|
|
|
|
// errors in case of bad typing.
|
|
|
|
switch n := node.(type) {
|
|
|
|
case *EvalStmt:
|
2020-01-17 07:16:58 -08:00
|
|
|
ty := p.checkAST(n.Expr)
|
2016-12-23 04:51:59 -08:00
|
|
|
if ty == ValueTypeNone {
|
2020-02-03 09:00:41 -08:00
|
|
|
p.addParseErrf(n.Expr.PositionRange(), "evaluation statement must have a valid expression type but got %s", DocumentedType(ty))
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
case Expressions:
|
|
|
|
for _, e := range n {
|
2020-01-17 07:16:58 -08:00
|
|
|
ty := p.checkAST(e)
|
2016-12-23 04:51:59 -08:00
|
|
|
if ty == ValueTypeNone {
|
2020-02-03 09:00:41 -08:00
|
|
|
p.addParseErrf(e.PositionRange(), "expression must have a valid expression type but got %s", DocumentedType(ty))
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
case *AggregateExpr:
|
2020-02-03 09:48:27 -08:00
|
|
|
if !n.Op.IsAggregator() {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "aggregation operator expected in aggregation expression but got %q", n.Op)
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
2016-12-23 04:51:59 -08:00
|
|
|
p.expectType(n.Expr, ValueTypeVector, "aggregation expression")
|
2019-11-26 05:29:42 -08:00
|
|
|
if n.Op == TOPK || n.Op == BOTTOMK || n.Op == QUANTILE {
|
2016-12-23 04:51:59 -08:00
|
|
|
p.expectType(n.Param, ValueTypeScalar, "aggregation parameter")
|
2016-07-04 10:03:05 -07:00
|
|
|
}
|
2019-11-26 05:29:42 -08:00
|
|
|
if n.Op == COUNT_VALUES {
|
2016-12-23 04:51:59 -08:00
|
|
|
p.expectType(n.Param, ValueTypeString, "aggregation parameter")
|
2016-07-05 09:12:19 -07:00
|
|
|
}
|
2015-03-30 09:12:51 -07:00
|
|
|
|
|
|
|
case *BinaryExpr:
|
2020-01-17 07:16:58 -08:00
|
|
|
lt := p.checkAST(n.LHS)
|
|
|
|
rt := p.checkAST(n.RHS)
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2020-01-17 05:06:27 -08:00
|
|
|
// opRange returns the PositionRange of the operator part of the BinaryExpr.
|
|
|
|
// This is made a function instead of a variable, so it is lazily evaluated on demand.
|
|
|
|
opRange := func() (r PositionRange) {
|
|
|
|
// Remove whitespace at the beginning and end of the range.
|
2023-04-12 04:05:41 -07:00
|
|
|
for r.Start = n.LHS.PositionRange().End; isSpace(rune(p.lex.input[r.Start])); r.Start++ { // nolint:revive
|
2020-01-17 05:06:27 -08:00
|
|
|
}
|
2023-04-12 04:05:41 -07:00
|
|
|
for r.End = n.RHS.PositionRange().Start - 1; isSpace(rune(p.lex.input[r.End])); r.End-- { // nolint:revive
|
2020-01-17 05:06:27 -08:00
|
|
|
}
|
|
|
|
return
|
|
|
|
}
|
|
|
|
|
2020-02-03 09:48:27 -08:00
|
|
|
if n.ReturnBool && !n.Op.IsComparisonOperator() {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(opRange(), "bool modifier can only be used on comparison operators")
|
2020-01-17 05:06:27 -08:00
|
|
|
}
|
|
|
|
|
2020-02-03 09:48:27 -08:00
|
|
|
if n.Op.IsComparisonOperator() && !n.ReturnBool && n.RHS.Type() == ValueTypeScalar && n.LHS.Type() == ValueTypeScalar {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(opRange(), "comparisons between scalars must use BOOL modifier")
|
2020-01-17 05:06:27 -08:00
|
|
|
}
|
|
|
|
|
2020-02-03 09:48:27 -08:00
|
|
|
if n.Op.IsSetOperator() && n.VectorMatching.Card == CardOneToOne {
|
2020-01-17 05:06:27 -08:00
|
|
|
n.VectorMatching.Card = CardManyToMany
|
|
|
|
}
|
|
|
|
|
|
|
|
for _, l1 := range n.VectorMatching.MatchingLabels {
|
|
|
|
for _, l2 := range n.VectorMatching.Include {
|
|
|
|
if l1 == l2 && n.VectorMatching.On {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(opRange(), "label %q must not occur in ON and GROUP clause at once", l1)
|
2020-01-17 05:06:27 -08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-02-03 09:48:27 -08:00
|
|
|
if !n.Op.IsOperator() {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "binary expression does not support operator %q", n.Op)
|
2020-01-15 12:01:49 -08:00
|
|
|
}
|
|
|
|
if lt != ValueTypeScalar && lt != ValueTypeVector {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.LHS.PositionRange(), "binary expression must contain only scalar and instant vector types")
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
2020-01-15 12:01:49 -08:00
|
|
|
if rt != ValueTypeScalar && rt != ValueTypeVector {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.RHS.PositionRange(), "binary expression must contain only scalar and instant vector types")
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
style: Replace `else if` cascades with `switch`
Wiser coders than myself have come to the conclusion that a `switch`
statement is almost always superior to a statement that includes any
`else if`.
The exceptions that I have found in our codebase are just these two:
* The `if else` is followed by an additional statement before the next
condition (separated by a `;`).
* The whole thing is within a `for` loop and `break` statements are
used. In this case, using `switch` would require tagging the `for`
loop, which probably tips the balance.
Why are `switch` statements more readable?
For one, fewer curly braces. But more importantly, the conditions all
have the same alignment, so the whole thing follows the natural flow
of going down a list of conditions. With `else if`, in contrast, all
conditions but the first are "hidden" behind `} else if `, harder to
spot and (for no good reason) presented differently from the first
condition.
I'm sure the aforemention wise coders can list even more reasons.
In any case, I like it so much that I have found myself recommending
it in code reviews. I would like to make it a habit in our code base,
without making it a hard requirement that we would test on the CI. But
for that, there has to be a role model, so this commit eliminates all
`if else` occurrences, unless it is autogenerated code or fits one of
the exceptions above.
Signed-off-by: beorn7 <beorn@grafana.com>
2023-04-12 07:14:31 -07:00
|
|
|
switch {
|
|
|
|
case (lt != ValueTypeVector || rt != ValueTypeVector) && n.VectorMatching != nil:
|
2016-04-26 06:28:36 -07:00
|
|
|
if len(n.VectorMatching.MatchingLabels) > 0 {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "vector matching only allowed between instant vectors")
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
n.VectorMatching = nil
|
style: Replace `else if` cascades with `switch`
Wiser coders than myself have come to the conclusion that a `switch`
statement is almost always superior to a statement that includes any
`else if`.
The exceptions that I have found in our codebase are just these two:
* The `if else` is followed by an additional statement before the next
condition (separated by a `;`).
* The whole thing is within a `for` loop and `break` statements are
used. In this case, using `switch` would require tagging the `for`
loop, which probably tips the balance.
Why are `switch` statements more readable?
For one, fewer curly braces. But more importantly, the conditions all
have the same alignment, so the whole thing follows the natural flow
of going down a list of conditions. With `else if`, in contrast, all
conditions but the first are "hidden" behind `} else if `, harder to
spot and (for no good reason) presented differently from the first
condition.
I'm sure the aforemention wise coders can list even more reasons.
In any case, I like it so much that I have found myself recommending
it in code reviews. I would like to make it a habit in our code base,
without making it a hard requirement that we would test on the CI. But
for that, there has to be a role model, so this commit eliminates all
`if else` occurrences, unless it is autogenerated code or fits one of
the exceptions above.
Signed-off-by: beorn7 <beorn@grafana.com>
2023-04-12 07:14:31 -07:00
|
|
|
case n.Op.IsSetOperator(): // Both operands are Vectors.
|
|
|
|
if n.VectorMatching.Card == CardOneToMany || n.VectorMatching.Card == CardManyToOne {
|
|
|
|
p.addParseErrf(n.PositionRange(), "no grouping allowed for %q operation", n.Op)
|
|
|
|
}
|
|
|
|
if n.VectorMatching.Card != CardManyToMany {
|
|
|
|
p.addParseErrf(n.PositionRange(), "set operations must always be many-to-many")
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-02-03 09:48:27 -08:00
|
|
|
if (lt == ValueTypeScalar || rt == ValueTypeScalar) && n.Op.IsSetOperator() {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "set operator %q not allowed in binary scalar expression", n.Op)
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
case *Call:
|
|
|
|
nargs := len(n.Func.ArgTypes)
|
2017-06-16 06:51:22 -07:00
|
|
|
if n.Func.Variadic == 0 {
|
|
|
|
if nargs != len(n.Args) {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "expected %d argument(s) in call to %q, got %d", nargs, n.Func.Name, len(n.Args))
|
2017-06-16 06:51:22 -07:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
na := nargs - 1
|
|
|
|
if na > len(n.Args) {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "expected at least %d argument(s) in call to %q, got %d", na, n.Func.Name, len(n.Args))
|
2017-06-16 06:51:22 -07:00
|
|
|
} else if nargsmax := na + n.Func.Variadic; n.Func.Variadic > 0 && nargsmax < len(n.Args) {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "expected at most %d argument(s) in call to %q, got %d", nargsmax, n.Func.Name, len(n.Args))
|
2017-06-16 06:51:22 -07:00
|
|
|
}
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
2017-06-16 06:51:22 -07:00
|
|
|
|
2015-03-30 09:12:51 -07:00
|
|
|
for i, arg := range n.Args {
|
2017-06-16 06:51:22 -07:00
|
|
|
if i >= len(n.Func.ArgTypes) {
|
2020-03-08 05:09:24 -07:00
|
|
|
if n.Func.Variadic == 0 {
|
|
|
|
// This is not a vararg function so we should not check the
|
|
|
|
// type of the extra arguments.
|
|
|
|
break
|
|
|
|
}
|
2017-06-16 06:51:22 -07:00
|
|
|
i = len(n.Func.ArgTypes) - 1
|
|
|
|
}
|
2015-03-30 09:12:51 -07:00
|
|
|
p.expectType(arg, n.Func.ArgTypes[i], fmt.Sprintf("call to function %q", n.Func.Name))
|
|
|
|
}
|
|
|
|
|
|
|
|
case *ParenExpr:
|
2020-01-17 07:16:58 -08:00
|
|
|
p.checkAST(n.Expr)
|
2015-03-30 09:12:51 -07:00
|
|
|
|
|
|
|
case *UnaryExpr:
|
2019-11-26 05:29:42 -08:00
|
|
|
if n.Op != ADD && n.Op != SUB {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "only + and - operators allowed for unary expressions")
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
2020-01-17 07:16:58 -08:00
|
|
|
if t := p.checkAST(n.Expr); t != ValueTypeScalar && t != ValueTypeVector {
|
2020-02-03 09:00:41 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "unary expression only allowed on expressions of type scalar or instant vector, got %q", DocumentedType(t))
|
2015-08-04 05:57:34 -07:00
|
|
|
}
|
2015-03-30 09:12:51 -07:00
|
|
|
|
2018-12-22 05:47:13 -08:00
|
|
|
case *SubqueryExpr:
|
2020-01-17 07:16:58 -08:00
|
|
|
ty := p.checkAST(n.Expr)
|
2018-12-22 05:47:13 -08:00
|
|
|
if ty != ValueTypeVector {
|
2020-11-12 06:25:52 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "subquery is only allowed on instant vector, got %s instead", ty)
|
2018-12-22 05:47:13 -08:00
|
|
|
}
|
2020-01-10 06:25:41 -08:00
|
|
|
case *MatrixSelector:
|
2020-01-17 07:16:58 -08:00
|
|
|
p.checkAST(n.VectorSelector)
|
2018-12-22 05:47:13 -08:00
|
|
|
|
2020-01-17 07:16:58 -08:00
|
|
|
case *VectorSelector:
|
2021-01-07 02:38:38 -08:00
|
|
|
if n.Name != "" {
|
|
|
|
// In this case the last LabelMatcher is checking for the metric name
|
|
|
|
// set outside the braces. This checks if the name has already been set
|
|
|
|
// previously.
|
|
|
|
for _, m := range n.LabelMatchers[0 : len(n.LabelMatchers)-1] {
|
|
|
|
if m != nil && m.Name == labels.MetricName {
|
|
|
|
p.addParseErrf(n.PositionRange(), "metric name must not be set twice: %q or %q", n.Name, m.Value)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Skip the check for non-empty matchers because an explicit
|
|
|
|
// metric name is a non-empty matcher.
|
|
|
|
break
|
|
|
|
}
|
|
|
|
|
2020-01-17 07:16:58 -08:00
|
|
|
// A Vector selector must contain at least one non-empty matcher to prevent
|
|
|
|
// implicit selection of all metrics (e.g. by a typo).
|
|
|
|
notEmpty := false
|
|
|
|
for _, lm := range n.LabelMatchers {
|
|
|
|
if lm != nil && !lm.Matches("") {
|
|
|
|
notEmpty = true
|
|
|
|
break
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if !notEmpty {
|
|
|
|
p.addParseErrf(n.PositionRange(), "vector selector must contain at least one non-empty matcher")
|
|
|
|
}
|
|
|
|
|
|
|
|
case *NumberLiteral, *StringLiteral:
|
2015-03-30 09:12:51 -07:00
|
|
|
// Nothing to do for terminals.
|
|
|
|
|
|
|
|
default:
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(n.PositionRange(), "unknown node type: %T", node)
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
|
|
|
return
|
|
|
|
}
|
|
|
|
|
2015-09-30 12:27:08 -07:00
|
|
|
func (p *parser) unquoteString(s string) string {
|
|
|
|
unquoted, err := strutil.Unquote(s)
|
|
|
|
if err != nil {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(p.yyParser.lval.item.PositionRange(), "error unquoting string %q: %s", s, err)
|
2015-09-30 12:27:08 -07:00
|
|
|
}
|
|
|
|
return unquoted
|
|
|
|
}
|
|
|
|
|
2015-03-30 09:12:51 -07:00
|
|
|
func parseDuration(ds string) (time.Duration, error) {
|
2016-01-29 06:23:11 -08:00
|
|
|
dur, err := model.ParseDuration(ds)
|
2015-03-30 09:12:51 -07:00
|
|
|
if err != nil {
|
|
|
|
return 0, err
|
|
|
|
}
|
|
|
|
if dur == 0 {
|
2019-03-25 16:01:12 -07:00
|
|
|
return 0, errors.New("duration must be greater than 0")
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
2016-01-29 06:23:11 -08:00
|
|
|
return time.Duration(dur), nil
|
2015-03-30 09:12:51 -07:00
|
|
|
}
|
2019-12-05 08:16:12 -08:00
|
|
|
|
|
|
|
// parseGenerated invokes the yacc generated parser.
|
|
|
|
// The generated parser gets the provided startSymbol injected into
|
|
|
|
// the lexer stream, based on which grammar will be used.
|
2020-01-08 06:59:25 -08:00
|
|
|
func (p *parser) parseGenerated(startSymbol ItemType) interface{} {
|
2019-12-05 08:16:12 -08:00
|
|
|
p.InjectItem(startSymbol)
|
|
|
|
|
2020-01-09 03:26:58 -08:00
|
|
|
p.yyParser.Parse(p)
|
2019-12-05 08:16:12 -08:00
|
|
|
|
|
|
|
return p.generatedParserResult
|
|
|
|
}
|
|
|
|
|
2021-10-22 01:06:44 -07:00
|
|
|
func (p *parser) newLabelMatcher(label, operator, value Item) *labels.Matcher {
|
2019-12-09 11:03:31 -08:00
|
|
|
op := operator.Typ
|
|
|
|
val := p.unquoteString(value.Val)
|
2019-12-05 08:16:12 -08:00
|
|
|
|
2019-12-09 11:03:31 -08:00
|
|
|
// Map the Item to the respective match type.
|
2019-12-05 08:16:12 -08:00
|
|
|
var matchType labels.MatchType
|
|
|
|
switch op {
|
|
|
|
case EQL:
|
|
|
|
matchType = labels.MatchEqual
|
|
|
|
case NEQ:
|
|
|
|
matchType = labels.MatchNotEqual
|
|
|
|
case EQL_REGEX:
|
|
|
|
matchType = labels.MatchRegexp
|
|
|
|
case NEQ_REGEX:
|
|
|
|
matchType = labels.MatchNotRegexp
|
|
|
|
default:
|
2020-01-02 06:54:09 -08:00
|
|
|
// This should never happen, since the error should have been caught
|
2019-12-05 08:16:12 -08:00
|
|
|
// by the generated parser.
|
|
|
|
panic("invalid operator")
|
|
|
|
}
|
|
|
|
|
2019-12-09 11:03:31 -08:00
|
|
|
m, err := labels.NewMatcher(matchType, label.Val, val)
|
2019-12-05 08:16:12 -08:00
|
|
|
if err != nil {
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErr(mergeRanges(&label, &value), err)
|
2019-12-05 08:16:12 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
return m
|
|
|
|
}
|
2020-01-08 03:04:47 -08:00
|
|
|
|
2021-01-20 02:57:39 -08:00
|
|
|
// addOffset is used to set the offset in the generated parser.
|
2020-01-08 03:04:47 -08:00
|
|
|
func (p *parser) addOffset(e Node, offset time.Duration) {
|
2021-01-20 02:57:39 -08:00
|
|
|
var orgoffsetp *time.Duration
|
2020-01-14 08:12:15 -08:00
|
|
|
var endPosp *Pos
|
2020-01-08 03:04:47 -08:00
|
|
|
|
|
|
|
switch s := e.(type) {
|
|
|
|
case *VectorSelector:
|
2021-01-20 02:57:39 -08:00
|
|
|
orgoffsetp = &s.OriginalOffset
|
2020-01-14 08:12:15 -08:00
|
|
|
endPosp = &s.PosRange.End
|
2020-01-08 03:04:47 -08:00
|
|
|
case *MatrixSelector:
|
2021-01-20 02:57:39 -08:00
|
|
|
vs, ok := s.VectorSelector.(*VectorSelector)
|
|
|
|
if !ok {
|
|
|
|
p.addParseErrf(e.PositionRange(), "ranges only allowed for vector selectors")
|
|
|
|
return
|
2020-01-17 07:16:58 -08:00
|
|
|
}
|
2021-01-20 02:57:39 -08:00
|
|
|
orgoffsetp = &vs.OriginalOffset
|
2020-01-14 08:12:15 -08:00
|
|
|
endPosp = &s.EndPos
|
2020-01-08 03:04:47 -08:00
|
|
|
case *SubqueryExpr:
|
2021-01-20 02:57:39 -08:00
|
|
|
orgoffsetp = &s.OriginalOffset
|
2020-01-14 08:12:15 -08:00
|
|
|
endPosp = &s.EndPos
|
2020-01-08 03:04:47 -08:00
|
|
|
default:
|
2021-05-10 14:33:26 -07:00
|
|
|
p.addParseErrf(e.PositionRange(), "offset modifier must be preceded by an instant vector selector or range vector selector or a subquery")
|
2020-01-08 03:04:47 -08:00
|
|
|
return
|
|
|
|
}
|
|
|
|
|
|
|
|
// it is already ensured by parseDuration func that there never will be a zero offset modifier
|
style: Replace `else if` cascades with `switch`
Wiser coders than myself have come to the conclusion that a `switch`
statement is almost always superior to a statement that includes any
`else if`.
The exceptions that I have found in our codebase are just these two:
* The `if else` is followed by an additional statement before the next
condition (separated by a `;`).
* The whole thing is within a `for` loop and `break` statements are
used. In this case, using `switch` would require tagging the `for`
loop, which probably tips the balance.
Why are `switch` statements more readable?
For one, fewer curly braces. But more importantly, the conditions all
have the same alignment, so the whole thing follows the natural flow
of going down a list of conditions. With `else if`, in contrast, all
conditions but the first are "hidden" behind `} else if `, harder to
spot and (for no good reason) presented differently from the first
condition.
I'm sure the aforemention wise coders can list even more reasons.
In any case, I like it so much that I have found myself recommending
it in code reviews. I would like to make it a habit in our code base,
without making it a hard requirement that we would test on the CI. But
for that, there has to be a role model, so this commit eliminates all
`if else` occurrences, unless it is autogenerated code or fits one of
the exceptions above.
Signed-off-by: beorn7 <beorn@grafana.com>
2023-04-12 07:14:31 -07:00
|
|
|
switch {
|
|
|
|
case *orgoffsetp != 0:
|
2020-01-17 07:16:58 -08:00
|
|
|
p.addParseErrf(e.PositionRange(), "offset may not be set multiple times")
|
style: Replace `else if` cascades with `switch`
Wiser coders than myself have come to the conclusion that a `switch`
statement is almost always superior to a statement that includes any
`else if`.
The exceptions that I have found in our codebase are just these two:
* The `if else` is followed by an additional statement before the next
condition (separated by a `;`).
* The whole thing is within a `for` loop and `break` statements are
used. In this case, using `switch` would require tagging the `for`
loop, which probably tips the balance.
Why are `switch` statements more readable?
For one, fewer curly braces. But more importantly, the conditions all
have the same alignment, so the whole thing follows the natural flow
of going down a list of conditions. With `else if`, in contrast, all
conditions but the first are "hidden" behind `} else if `, harder to
spot and (for no good reason) presented differently from the first
condition.
I'm sure the aforemention wise coders can list even more reasons.
In any case, I like it so much that I have found myself recommending
it in code reviews. I would like to make it a habit in our code base,
without making it a hard requirement that we would test on the CI. But
for that, there has to be a role model, so this commit eliminates all
`if else` occurrences, unless it is autogenerated code or fits one of
the exceptions above.
Signed-off-by: beorn7 <beorn@grafana.com>
2023-04-12 07:14:31 -07:00
|
|
|
case orgoffsetp != nil:
|
2021-01-20 02:57:39 -08:00
|
|
|
*orgoffsetp = offset
|
|
|
|
}
|
|
|
|
|
|
|
|
*endPosp = p.lastClosing
|
|
|
|
}
|
|
|
|
|
|
|
|
// setTimestamp is used to set the timestamp from the @ modifier in the generated parser.
|
|
|
|
func (p *parser) setTimestamp(e Node, ts float64) {
|
|
|
|
if math.IsInf(ts, -1) || math.IsInf(ts, 1) || math.IsNaN(ts) ||
|
|
|
|
ts >= float64(math.MaxInt64) || ts <= float64(math.MinInt64) {
|
|
|
|
p.addParseErrf(e.PositionRange(), "timestamp out of bounds for @ modifier: %f", ts)
|
|
|
|
}
|
|
|
|
var timestampp **int64
|
|
|
|
var endPosp *Pos
|
|
|
|
|
2021-02-09 08:03:16 -08:00
|
|
|
timestampp, _, endPosp, ok := p.getAtModifierVars(e)
|
|
|
|
if !ok {
|
|
|
|
return
|
|
|
|
}
|
|
|
|
|
|
|
|
if timestampp != nil {
|
|
|
|
*timestampp = new(int64)
|
|
|
|
**timestampp = timestamp.FromFloatSeconds(ts)
|
|
|
|
}
|
|
|
|
|
|
|
|
*endPosp = p.lastClosing
|
|
|
|
}
|
|
|
|
|
|
|
|
// setAtModifierPreprocessor is used to set the preprocessor for the @ modifier.
|
|
|
|
func (p *parser) setAtModifierPreprocessor(e Node, op Item) {
|
|
|
|
_, preprocp, endPosp, ok := p.getAtModifierVars(e)
|
|
|
|
if !ok {
|
|
|
|
return
|
|
|
|
}
|
|
|
|
|
|
|
|
if preprocp != nil {
|
|
|
|
*preprocp = op.Typ
|
|
|
|
}
|
|
|
|
|
|
|
|
*endPosp = p.lastClosing
|
|
|
|
}
|
|
|
|
|
|
|
|
func (p *parser) getAtModifierVars(e Node) (**int64, *ItemType, *Pos, bool) {
|
|
|
|
var (
|
|
|
|
timestampp **int64
|
|
|
|
preprocp *ItemType
|
|
|
|
endPosp *Pos
|
|
|
|
)
|
2021-01-20 02:57:39 -08:00
|
|
|
switch s := e.(type) {
|
|
|
|
case *VectorSelector:
|
|
|
|
timestampp = &s.Timestamp
|
2021-02-09 08:03:16 -08:00
|
|
|
preprocp = &s.StartOrEnd
|
2021-01-20 02:57:39 -08:00
|
|
|
endPosp = &s.PosRange.End
|
|
|
|
case *MatrixSelector:
|
|
|
|
vs, ok := s.VectorSelector.(*VectorSelector)
|
|
|
|
if !ok {
|
|
|
|
p.addParseErrf(e.PositionRange(), "ranges only allowed for vector selectors")
|
2021-02-09 08:03:16 -08:00
|
|
|
return nil, nil, nil, false
|
2021-01-20 02:57:39 -08:00
|
|
|
}
|
2021-02-09 08:03:16 -08:00
|
|
|
preprocp = &vs.StartOrEnd
|
2021-01-20 02:57:39 -08:00
|
|
|
timestampp = &vs.Timestamp
|
|
|
|
endPosp = &s.EndPos
|
|
|
|
case *SubqueryExpr:
|
2021-02-09 08:03:16 -08:00
|
|
|
preprocp = &s.StartOrEnd
|
2021-01-20 02:57:39 -08:00
|
|
|
timestampp = &s.Timestamp
|
|
|
|
endPosp = &s.EndPos
|
|
|
|
default:
|
2021-05-10 14:33:26 -07:00
|
|
|
p.addParseErrf(e.PositionRange(), "@ modifier must be preceded by an instant vector selector or range vector selector or a subquery")
|
2021-02-09 08:03:16 -08:00
|
|
|
return nil, nil, nil, false
|
2021-01-20 02:57:39 -08:00
|
|
|
}
|
|
|
|
|
2021-02-09 08:03:16 -08:00
|
|
|
if *timestampp != nil || (*preprocp) == START || (*preprocp) == END {
|
2021-01-20 02:57:39 -08:00
|
|
|
p.addParseErrf(e.PositionRange(), "@ <timestamp> may not be set multiple times")
|
2021-02-09 08:03:16 -08:00
|
|
|
return nil, nil, nil, false
|
2020-01-08 03:04:47 -08:00
|
|
|
}
|
|
|
|
|
2021-02-09 08:03:16 -08:00
|
|
|
return timestampp, preprocp, endPosp, true
|
2021-01-20 02:57:39 -08:00
|
|
|
}
|
2020-01-14 08:12:15 -08:00
|
|
|
|
2021-01-20 02:57:39 -08:00
|
|
|
func MustLabelMatcher(mt labels.MatchType, name, val string) *labels.Matcher {
|
|
|
|
m, err := labels.NewMatcher(mt, name, val)
|
|
|
|
if err != nil {
|
|
|
|
panic(err)
|
|
|
|
}
|
|
|
|
return m
|
|
|
|
}
|
|
|
|
|
|
|
|
func MustGetFunction(name string) *Function {
|
2023-03-22 02:02:10 -07:00
|
|
|
f, ok := getFunction(name, Functions)
|
2021-01-20 02:57:39 -08:00
|
|
|
if !ok {
|
2022-06-08 01:47:52 -07:00
|
|
|
panic(fmt.Errorf("function %q does not exist", name))
|
2021-01-20 02:57:39 -08:00
|
|
|
}
|
|
|
|
return f
|
2020-01-08 03:04:47 -08:00
|
|
|
}
|