bl/common/utils/cronex/parser.go

//Copyright (C) 2012 Rob Figueiredo
//2022 guonaihong,
// 1.修改内容，裁剪代码只保留一个干净的解析器
// 2.默认让parse支持到秒级别
//All Rights Reserved.
package cronex

import (
	"fmt"
	"math"
	"strconv"
	"strings"
	"time"
)

// Configuration options for creating a parser. Most options specify which
// fields should be included, while others enable features. If a field is not
// included the parser will assume a default value. These options do not change
// the order fields are parse in.
type ParseOption int

const (
	Second         ParseOption = 1 << iota // Seconds field, default 0
	SecondOptional                         // Optional seconds field, default 0
	Minute                                 // Minutes field, default 0
	Hour                                   // Hours field, default 0
	Dom                                    // Day of month field, default *
	Month                                  // Month field, default *
	Dow                                    // Day of week field, default *
	DowOptional                            // Optional day of week field, default *
	Descriptor                             // Allow descriptors such as @monthly, @weekly, etc.
)

var places = []ParseOption{
	Second,
	Minute,
	Hour,
	Dom,
	Month,
	Dow,
}

var defaults = []string{
	"0",
	"0",
	"0",
	"*",
	"*",
	"*",
}

// A custom Parser that can be configured.
type Parser struct {
	options ParseOption
}

// NewParser creates a Parser with custom options.
//
// It panics if more than one Optional is given, since it would be impossible to
// correctly infer which optional is provided or missing in general.
//
// Examples
//
//  // Standard parser without descriptors
//  specParser := NewParser(Minute | Hour | Dom | Month | Dow)
//  sched, err := specParser.Parse("0 0 15 */3 *")
//
//  // Same as above, just excludes time fields
//  specParser := NewParser(Dom | Month | Dow)
//  sched, err := specParser.Parse("15 */3 *")
//
//  // Same as above, just makes Dow optional
//  specParser := NewParser(Dom | Month | DowOptional)
//  sched, err := specParser.Parse("15 */3")
//
func NewParser(options ParseOption) Parser {
	optionals := 0
	if options&DowOptional > 0 {
		optionals++
	}
	if options&SecondOptional > 0 {
		optionals++
	}
	if optionals > 1 {
		panic("multiple optionals may not be configured")
	}
	return Parser{options}
}

// Parse returns a new crontab schedule representing the given spec.
// It returns a descriptive error if the spec is not valid.
// It accepts crontab specs and features configured by NewParser.
func (p Parser) Parse(spec string) (Schedule, error) {
	if len(spec) == 0 {
		return nil, fmt.Errorf("empty spec string")
	}

	// Extract timezone if present
	var loc = time.Local
	if strings.HasPrefix(spec, "TZ=") || strings.HasPrefix(spec, "CRON_TZ=") {
		var err error
		i := strings.Index(spec, " ")
		eq := strings.Index(spec, "=")
		if loc, err = time.LoadLocation(spec[eq+1 : i]); err != nil {
			return nil, fmt.Errorf("provided bad location %s: %v", spec[eq+1:i], err)
		}
		spec = strings.TrimSpace(spec[i:])
	}

	// Handle named schedules (descriptors), if configured
	if strings.HasPrefix(spec, "@") {
		if p.options&Descriptor == 0 {
			return nil, fmt.Errorf("parser does not accept descriptors: %v", spec)
		}
		return parseDescriptor(spec, loc)
	}

	// Split on whitespace.
	fields := strings.Fields(spec)

	// Validate & fill in any omitted or optional fields
	var err error
	fields, err = normalizeFields(fields, p.options)
	if err != nil {
		return nil, err
	}

	field := func(field string, r bounds) uint64 {
		if err != nil {
			return 0
		}
		var bits uint64
		bits, err = getField(field, r)
		return bits
	}

	var (
		second     = field(fields[0], seconds)
		minute     = field(fields[1], minutes)
		hour       = field(fields[2], hours)
		dayofmonth = field(fields[3], dom)
		month      = field(fields[4], months)
		dayofweek  = field(fields[5], dow)
	)
	if err != nil {
		return nil, err
	}

	return &SpecSchedule{
		Second:   second,
		Minute:   minute,
		Hour:     hour,
		Dom:      dayofmonth,
		Month:    month,
		Dow:      dayofweek,
		Location: loc,
	}, nil
}

// normalizeFields takes a subset set of the time fields and returns the full set
// with defaults (zeroes) populated for unset fields.
//
// As part of performing this function, it also validates that the provided
// fields are compatible with the configured options.
func normalizeFields(fields []string, options ParseOption) ([]string, error) {
	// Validate optionals & add their field to options
	optionals := 0
	if options&SecondOptional > 0 {
		options |= Second
		optionals++
	}
	if options&DowOptional > 0 {
		options |= Dow
		optionals++
	}
	if optionals > 1 {
		return nil, fmt.Errorf("multiple optionals may not be configured")
	}

	// Figure out how many fields we need
	max := 0
	for _, place := range places {
		if options&place > 0 {
			max++
		}
	}
	min := max - optionals

	// Validate number of fields
	if count := len(fields); count < min || count > max {
		if min == max {
			return nil, fmt.Errorf("expected exactly %d fields, found %d: %s", min, count, fields)
		}
		return nil, fmt.Errorf("expected %d to %d fields, found %d: %s", min, max, count, fields)
	}

	// Populate the optional field if not provided
	if min < max && len(fields) == min {
		switch {
		case options&DowOptional > 0:
			fields = append(fields, defaults[5]) // TODO: improve access to default
		case options&SecondOptional > 0:
			fields = append([]string{defaults[0]}, fields...)
		default:
			return nil, fmt.Errorf("unknown optional field")
		}
	}

	// Populate all fields not part of options with their defaults
	n := 0
	expandedFields := make([]string, len(places))
	copy(expandedFields, defaults)
	for i, place := range places {
		if options&place > 0 {
			expandedFields[i] = fields[n]
			n++
		}
	}
	return expandedFields, nil
}

var standardParser = NewParser(
	SecondOptional | Minute | Hour | Dom | Month | Dow | Descriptor,
)

// ParseStandard returns a new crontab schedule representing the given
// standardSpec (https://en.wikipedia.org/wiki/Cron). It requires 5 entries
// representing: minute, hour, day of month, month and day of week, in that
// order. It returns a descriptive error if the spec is not valid.
//
// It accepts
//   - Standard crontab specs, e.g. "* * * * ?"
//   - Descriptors, e.g. "@midnight", "@every 1h30m"
func ParseStandard(standardSpec string) (Schedule, error) {
	return standardParser.Parse(standardSpec)
}

// getField returns an Int with the bits set representing all of the times that
// the field represents or error parsing field value.  A "field" is a comma-separated
// list of "ranges".
func getField(field string, r bounds) (uint64, error) {
	var bits uint64
	ranges := strings.FieldsFunc(field, func(r rune) bool { return r == ',' })
	for _, expr := range ranges {
		bit, err := getRange(expr, r)
		if err != nil {
			return bits, err
		}
		bits |= bit
	}
	return bits, nil
}

// getRange returns the bits indicated by the given expression:
//   number | number "-" number [ "/" number ]
// or error parsing range.
func getRange(expr string, r bounds) (uint64, error) {
	var (
		start, end, step uint
		rangeAndStep     = strings.Split(expr, "/")
		lowAndHigh       = strings.Split(rangeAndStep[0], "-")
		singleDigit      = len(lowAndHigh) == 1
		err              error
	)

	var extra uint64
	if lowAndHigh[0] == "*" || lowAndHigh[0] == "?" {
		start = r.min
		end = r.max
		extra = starBit
	} else {
		start, err = parseIntOrName(lowAndHigh[0], r.names)
		if err != nil {
			return 0, err
		}
		switch len(lowAndHigh) {
		case 1:
			end = start
		case 2:
			end, err = parseIntOrName(lowAndHigh[1], r.names)
			if err != nil {
				return 0, err
			}
		default:
			return 0, fmt.Errorf("too many hyphens: %s", expr)
		}
	}

	switch len(rangeAndStep) {
	case 1:
		step = 1
	case 2:
		step, err = mustParseInt(rangeAndStep[1])
		if err != nil {
			return 0, err
		}

		// Special handling: "N/step" means "N-max/step".
		if singleDigit {
			end = r.max
		}
		if step > 1 {
			extra = 0
		}
	default:
		return 0, fmt.Errorf("too many slashes: %s", expr)
	}

	if start < r.min {
		return 0, fmt.Errorf("beginning of range (%d) below minimum (%d): %s", start, r.min, expr)
	}
	if end > r.max {
		return 0, fmt.Errorf("end of range (%d) above maximum (%d): %s", end, r.max, expr)
	}
	if start > end {
		return 0, fmt.Errorf("beginning of range (%d) beyond end of range (%d): %s", start, end, expr)
	}
	if step == 0 {
		return 0, fmt.Errorf("step of range should be a positive number: %s", expr)
	}

	return getBits(start, end, step) | extra, nil
}

// parseIntOrName returns the (possibly-named) integer contained in expr.
func parseIntOrName(expr string, names map[string]uint) (uint, error) {
	if names != nil {
		if namedInt, ok := names[strings.ToLower(expr)]; ok {
			return namedInt, nil
		}
	}
	return mustParseInt(expr)
}

// mustParseInt parses the given expression as an int or returns an error.
func mustParseInt(expr string) (uint, error) {
	num, err := strconv.Atoi(expr)
	if err != nil {
		return 0, fmt.Errorf("failed to parse int from %s: %s", expr, err)
	}
	if num < 0 {
		return 0, fmt.Errorf("negative number (%d) not allowed: %s", num, expr)
	}

	return uint(num), nil
}

// getBits sets all bits in the range [min, max], modulo the given step size.
func getBits(min, max, step uint) uint64 {
	var bits uint64

	// If step is 1, use shifts.
	if step == 1 {
		return ^(math.MaxUint64 << (max + 1)) & (math.MaxUint64 << min)
	}

	// Else, use a simple loop.
	for i := min; i <= max; i += step {
		bits |= 1 << i
	}
	return bits
}

// all returns all bits within the given bounds.  (plus the star bit)
func all(r bounds) uint64 {
	return getBits(r.min, r.max, 1) | starBit
}

// parseDescriptor returns a predefined schedule for the expression, or error if none matches.
func parseDescriptor(descriptor string, loc *time.Location) (Schedule, error) {
	switch descriptor {
	case "@yearly", "@annually":
		return &SpecSchedule{
			Second:   1 << seconds.min,
			Minute:   1 << minutes.min,
			Hour:     1 << hours.min,
			Dom:      1 << dom.min,
			Month:    1 << months.min,
			Dow:      all(dow),
			Location: loc,
		}, nil

	case "@monthly":
		return &SpecSchedule{
			Second:   1 << seconds.min,
			Minute:   1 << minutes.min,
			Hour:     1 << hours.min,
			Dom:      1 << dom.min,
			Month:    all(months),
			Dow:      all(dow),
			Location: loc,
		}, nil

	case "@weekly":
		return &SpecSchedule{
			Second:   1 << seconds.min,
			Minute:   1 << minutes.min,
			Hour:     1 << hours.min,
			Dom:      all(dom),
			Month:    all(months),
			Dow:      1 << dow.min,
			Location: loc,
		}, nil

	case "@daily", "@midnight":
		return &SpecSchedule{
			Second:   1 << seconds.min,
			Minute:   1 << minutes.min,
			Hour:     1 << hours.min,
			Dom:      all(dom),
			Month:    all(months),
			Dow:      all(dow),
			Location: loc,
		}, nil

	case "@hourly":
		return &SpecSchedule{
			Second:   1 << seconds.min,
			Minute:   1 << minutes.min,
			Hour:     all(hours),
			Dom:      all(dom),
			Month:    all(months),
			Dow:      all(dow),
			Location: loc,
		}, nil

	}

	const every = "@every "
	if strings.HasPrefix(descriptor, every) {
		duration, err := time.ParseDuration(descriptor[len(every):])
		if err != nil {
			return nil, fmt.Errorf("failed to parse duration %s: %s", descriptor, err)
		}
		return Every(duration), nil
	}

	return nil, fmt.Errorf("unrecognized descriptor: %s", descriptor)
}

// SpecSchedule specifies a duty cycle (to the second granularity), based on a
// traditional crontab specification. It is computed initially and stored as bit sets.
type SpecSchedule struct {
	Second, Minute, Hour, Dom, Month, Dow uint64

	// Override location for this schedule.
	Location *time.Location
}

// Next returns the next time this schedule is activated, greater than the given
// time.  If no time can be found to satisfy the schedule, return the zero time.
func (s *SpecSchedule) Next(t time.Time) time.Time {
	// General approach
	//
	// For Month, Day, Hour, Minute, Second:
	// Check if the time value matches.  If yes, continue to the next field.
	// If the field doesn't match the schedule, then increment the field until it matches.
	// While incrementing the field, a wrap-around brings it back to the beginning
	// of the field list (since it is necessary to re-verify previous field
	// values)

	// Convert the given time into the schedule's timezone, if one is specified.
	// Save the original timezone so we can convert back after we find a time.
	// Note that schedules without a time zone specified (time.Local) are treated
	// as local to the time provided.
	origLocation := t.Location()
	loc := s.Location
	if loc == time.Local {
		loc = t.Location()
	}
	if s.Location != time.Local {
		t = t.In(s.Location)
	}

	// Start at the earliest possible time (the upcoming second).
	t = t.Add(1*time.Second - time.Duration(t.Nanosecond())*time.Nanosecond)

	// This flag indicates whether a field has been incremented.
	added := false

	// If no time is found within five years, return zero.
	yearLimit := t.Year() + 5

WRAP:
	if t.Year() > yearLimit {
		return time.Time{}
	}

	// Find the first applicable month.
	// If it's this month, then do nothing.
	for 1<<uint(t.Month())&s.Month == 0 {
		// If we have to add a month, reset the other parts to 0.
		if !added {
			added = true
			// Otherwise, set the date at the beginning (since the current time is irrelevant).
			t = time.Date(t.Year(), t.Month(), 1, 0, 0, 0, 0, loc)
		}
		t = t.AddDate(0, 1, 0)

		// Wrapped around.
		if t.Month() == time.January {
			goto WRAP
		}
	}

	// Now get a day in that month.
	//
	// NOTE: This causes issues for daylight savings regimes where midnight does
	// not exist.  For example: Sao Paulo has DST that transforms midnight on
	// 11/3 into 1am. Handle that by noticing when the Hour ends up != 0.
	for !dayMatches(s, t) {
		if !added {
			added = true
			t = time.Date(t.Year(), t.Month(), t.Day(), 0, 0, 0, 0, loc)
		}
		t = t.AddDate(0, 0, 1)
		// Notice if the hour is no longer midnight due to DST.
		// Add an hour if it's 23, subtract an hour if it's 1.
		if t.Hour() != 0 {
			if t.Hour() > 12 {
				t = t.Add(time.Duration(24-t.Hour()) * time.Hour)
			} else {
				t = t.Add(time.Duration(-t.Hour()) * time.Hour)
			}
		}

		if t.Day() == 1 {
			goto WRAP
		}
	}

	for 1<<uint(t.Hour())&s.Hour == 0 {
		if !added {
			added = true
			t = time.Date(t.Year(), t.Month(), t.Day(), t.Hour(), 0, 0, 0, loc)
		}
		t = t.Add(1 * time.Hour)

		if t.Hour() == 0 {
			goto WRAP
		}
	}

	for 1<<uint(t.Minute())&s.Minute == 0 {
		if !added {
			added = true
			t = t.Truncate(time.Minute)
		}
		t = t.Add(1 * time.Minute)

		if t.Minute() == 0 {
			goto WRAP
		}
	}

	for 1<<uint(t.Second())&s.Second == 0 {
		if !added {
			added = true
			t = t.Truncate(time.Second)
		}
		t = t.Add(1 * time.Second)

		if t.Second() == 0 {
			goto WRAP
		}
	}

	return t.In(origLocation)
}

// dayMatches returns true if the schedule's day-of-week and day-of-month
// restrictions are satisfied by the given time.
func dayMatches(s *SpecSchedule, t time.Time) bool {
	var (
		domMatch bool = 1<<uint(t.Day())&s.Dom > 0
		dowMatch bool = 1<<uint(t.Weekday())&s.Dow > 0
	)
	if s.Dom&starBit > 0 || s.Dow&starBit > 0 {
		return domMatch && dowMatch
	}
	return domMatch || dowMatch
}

// bounds provides a range of acceptable values (plus a map of name to value).
type bounds struct {
	min, max uint
	names    map[string]uint
}

// The bounds for each field.
var (
	seconds = bounds{0, 59, nil}
	minutes = bounds{0, 59, nil}
	hours   = bounds{0, 23, nil}
	dom     = bounds{1, 31, nil}
	months  = bounds{1, 12, map[string]uint{
		"jan": 1,
		"feb": 2,
		"mar": 3,
		"apr": 4,
		"may": 5,
		"jun": 6,
		"jul": 7,
		"aug": 8,
		"sep": 9,
		"oct": 10,
		"nov": 11,
		"dec": 12,
	}}
	dow = bounds{0, 6, map[string]uint{
		"sun": 0,
		"mon": 1,
		"tue": 2,
		"wed": 3,
		"thu": 4,
		"fri": 5,
		"sat": 6,
	}}
)

const (
	// Set the top bit if a star was included in the expression.
	starBit = 1 << 63
)

// ConstantDelaySchedule represents a simple recurring duty cycle, e.g. "Every 5 minutes".
// It does not support jobs more frequent than once a second.
type ConstantDelaySchedule struct {
	Delay time.Duration
}

// Every returns a crontab Schedule that activates once every duration.
// Delays of less than a second are not supported (will round up to 1 second).
// Any fields less than a Second are truncated.
func Every(duration time.Duration) ConstantDelaySchedule {
	if duration < time.Second {
		duration = time.Second
	}
	return ConstantDelaySchedule{
		Delay: duration - time.Duration(duration.Nanoseconds())%time.Second,
	}
}

// Next returns the next time this should be run.
// This rounds so that the next activation time will be on the second.
func (schedule ConstantDelaySchedule) Next(t time.Time) time.Time {
	return t.Add(schedule.Delay - time.Duration(t.Nanosecond())*time.Nanosecond)
}

// Schedule describes a job's duty cycle.
type Schedule interface {
	// Next returns the next activation time, later than the given time.
	// Next is invoked initially, and then each time the job is run.
	Next(time.Time) time.Time
}