1
0
mirror of https://github.com/Oxalide/vsphere-influxdb-go.git synced 2023-10-10 13:36:51 +02:00
vsphere-influxdb-go/vendor/github.com/influxdata/influxdb/influxql/parser.go

3015 lines
78 KiB
Go
Raw Normal View History

2017-10-25 22:52:40 +02:00
package influxql
import (
"bytes"
"errors"
"fmt"
"io"
"math"
"regexp"
"sort"
"strconv"
"strings"
"time"
)
const (
// DateFormat represents the format for date literals.
DateFormat = "2006-01-02"
// DateTimeFormat represents the format for date time literals.
DateTimeFormat = "2006-01-02 15:04:05.999999"
)
// Parser represents an InfluxQL parser.
type Parser struct {
s *bufScanner
params map[string]interface{}
}
// NewParser returns a new instance of Parser.
func NewParser(r io.Reader) *Parser {
return &Parser{s: newBufScanner(r)}
}
// SetParams sets the parameters that will be used for any bound parameter substitutions.
func (p *Parser) SetParams(params map[string]interface{}) {
p.params = params
}
// ParseQuery parses a query string and returns its AST representation.
func ParseQuery(s string) (*Query, error) { return NewParser(strings.NewReader(s)).ParseQuery() }
// ParseStatement parses a statement string and returns its AST representation.
func ParseStatement(s string) (Statement, error) {
return NewParser(strings.NewReader(s)).ParseStatement()
}
// MustParseStatement parses a statement string and returns its AST. Panic on error.
func MustParseStatement(s string) Statement {
stmt, err := ParseStatement(s)
if err != nil {
panic(err.Error())
}
return stmt
}
// ParseExpr parses an expression string and returns its AST representation.
func ParseExpr(s string) (Expr, error) { return NewParser(strings.NewReader(s)).ParseExpr() }
// MustParseExpr parses an expression string and returns its AST. Panic on error.
func MustParseExpr(s string) Expr {
expr, err := ParseExpr(s)
if err != nil {
panic(err.Error())
}
return expr
}
// ParseQuery parses an InfluxQL string and returns a Query AST object.
func (p *Parser) ParseQuery() (*Query, error) {
var statements Statements
semi := true
for {
if tok, pos, lit := p.scanIgnoreWhitespace(); tok == EOF {
return &Query{Statements: statements}, nil
} else if tok == SEMICOLON {
semi = true
} else {
if !semi {
return nil, newParseError(tokstr(tok, lit), []string{";"}, pos)
}
p.unscan()
s, err := p.ParseStatement()
if err != nil {
return nil, err
}
statements = append(statements, s)
semi = false
}
}
}
// ParseStatement parses an InfluxQL string and returns a Statement AST object.
func (p *Parser) ParseStatement() (Statement, error) {
// Inspect the first token.
tok, pos, lit := p.scanIgnoreWhitespace()
switch tok {
case SELECT:
return p.parseSelectStatement(targetNotRequired)
case DELETE:
return p.parseDeleteStatement()
case SHOW:
return p.parseShowStatement()
case CREATE:
return p.parseCreateStatement()
case DROP:
return p.parseDropStatement()
case GRANT:
return p.parseGrantStatement()
case REVOKE:
return p.parseRevokeStatement()
case ALTER:
return p.parseAlterStatement()
case SET:
return p.parseSetPasswordUserStatement()
case KILL:
return p.parseKillQueryStatement()
default:
return nil, newParseError(tokstr(tok, lit), []string{"SELECT", "DELETE", "SHOW", "CREATE", "DROP", "GRANT", "REVOKE", "ALTER", "SET", "KILL"}, pos)
}
}
// parseShowStatement parses a string and returns a list statement.
// This function assumes the SHOW token has already been consumed.
func (p *Parser) parseShowStatement() (Statement, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
switch tok {
case CONTINUOUS:
return p.parseShowContinuousQueriesStatement()
case GRANTS:
return p.parseGrantsForUserStatement()
case DATABASES:
return p.parseShowDatabasesStatement()
case FIELD:
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == KEYS {
return p.parseShowFieldKeysStatement()
}
return nil, newParseError(tokstr(tok, lit), []string{"KEYS"}, pos)
case MEASUREMENTS:
return p.parseShowMeasurementsStatement()
case QUERIES:
return p.parseShowQueriesStatement()
case RETENTION:
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == POLICIES {
return p.parseShowRetentionPoliciesStatement()
}
return nil, newParseError(tokstr(tok, lit), []string{"POLICIES"}, pos)
case SERIES:
return p.parseShowSeriesStatement()
case SHARD:
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == GROUPS {
return p.parseShowShardGroupsStatement()
}
return nil, newParseError(tokstr(tok, lit), []string{"GROUPS"}, pos)
case SHARDS:
return p.parseShowShardsStatement()
case STATS:
return p.parseShowStatsStatement()
case DIAGNOSTICS:
return p.parseShowDiagnosticsStatement()
case TAG:
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == KEYS {
return p.parseShowTagKeysStatement()
} else if tok == VALUES {
return p.parseShowTagValuesStatement()
}
return nil, newParseError(tokstr(tok, lit), []string{"KEYS", "VALUES"}, pos)
case USERS:
return p.parseShowUsersStatement()
case SUBSCRIPTIONS:
return p.parseShowSubscriptionsStatement()
}
showQueryKeywords := []string{
"CONTINUOUS",
"DATABASES",
"FIELD",
"GRANTS",
"MEASUREMENTS",
"QUERIES",
"RETENTION",
"SERIES",
"TAG",
"USERS",
"STATS",
"DIAGNOSTICS",
"SHARD",
"SHARDS",
"SUBSCRIPTIONS",
}
sort.Strings(showQueryKeywords)
return nil, newParseError(tokstr(tok, lit), showQueryKeywords, pos)
}
// parseCreateStatement parses a string and returns a create statement.
// This function assumes the CREATE token has already been consumed.
func (p *Parser) parseCreateStatement() (Statement, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == CONTINUOUS {
return p.parseCreateContinuousQueryStatement()
} else if tok == DATABASE {
return p.parseCreateDatabaseStatement()
} else if tok == USER {
return p.parseCreateUserStatement()
} else if tok == RETENTION {
tok, pos, lit = p.scanIgnoreWhitespace()
if tok != POLICY {
return nil, newParseError(tokstr(tok, lit), []string{"POLICY"}, pos)
}
return p.parseCreateRetentionPolicyStatement()
} else if tok == SUBSCRIPTION {
return p.parseCreateSubscriptionStatement()
}
return nil, newParseError(tokstr(tok, lit), []string{"CONTINUOUS", "DATABASE", "USER", "RETENTION", "SUBSCRIPTION"}, pos)
}
// parseDropStatement parses a string and returns a drop statement.
// This function assumes the DROP token has already been consumed.
func (p *Parser) parseDropStatement() (Statement, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
switch tok {
case CONTINUOUS:
return p.parseDropContinuousQueryStatement()
case DATABASE:
return p.parseDropDatabaseStatement()
case MEASUREMENT:
return p.parseDropMeasurementStatement()
case RETENTION:
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != POLICY {
return nil, newParseError(tokstr(tok, lit), []string{"POLICY"}, pos)
}
return p.parseDropRetentionPolicyStatement()
case SERIES:
return p.parseDropSeriesStatement()
case SHARD:
return p.parseDropShardStatement()
case SUBSCRIPTION:
return p.parseDropSubscriptionStatement()
case USER:
return p.parseDropUserStatement()
default:
return nil, newParseError(tokstr(tok, lit), []string{"CONTINUOUS", "MEASUREMENT", "RETENTION", "SERIES", "SHARD", "SUBSCRIPTION", "USER"}, pos)
}
}
// parseAlterStatement parses a string and returns an alter statement.
// This function assumes the ALTER token has already been consumed.
func (p *Parser) parseAlterStatement() (Statement, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == RETENTION {
if tok, pos, lit = p.scanIgnoreWhitespace(); tok != POLICY {
return nil, newParseError(tokstr(tok, lit), []string{"POLICY"}, pos)
}
return p.parseAlterRetentionPolicyStatement()
}
return nil, newParseError(tokstr(tok, lit), []string{"RETENTION"}, pos)
}
// parseSetPasswordUserStatement parses a string and returns a set statement.
// This function assumes the SET token has already been consumed.
func (p *Parser) parseSetPasswordUserStatement() (*SetPasswordUserStatement, error) {
stmt := &SetPasswordUserStatement{}
// Consume the required PASSWORD FOR tokens.
if err := p.parseTokens([]Token{PASSWORD, FOR}); err != nil {
return nil, err
}
// Parse username
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Consume the required = token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != EQ {
return nil, newParseError(tokstr(tok, lit), []string{"="}, pos)
}
// Parse new user's password
if ident, err = p.parseString(); err != nil {
return nil, err
}
stmt.Password = ident
return stmt, nil
}
// parseKillQueryStatement parses a string and returns a kill statement.
// This function assumes the KILL token has already been consumed.
func (p *Parser) parseKillQueryStatement() (*KillQueryStatement, error) {
if err := p.parseTokens([]Token{QUERY}); err != nil {
return nil, err
}
qid, err := p.parseUInt64()
if err != nil {
return nil, err
}
var host string
if tok, _, _ := p.scanIgnoreWhitespace(); tok == ON {
host, err = p.parseIdent()
if err != nil {
return nil, err
}
} else {
p.unscan()
}
return &KillQueryStatement{QueryID: qid, Host: host}, nil
}
// parseCreateSubscriptionStatement parses a string and returns a CreateSubscriptionStatement.
// This function assumes the "CREATE SUBSCRIPTION" tokens have already been consumed.
func (p *Parser) parseCreateSubscriptionStatement() (*CreateSubscriptionStatement, error) {
stmt := &CreateSubscriptionStatement{}
// Read the id of the subscription to create.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Expect an "ON" keyword.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != ON {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// Read the name of the database.
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
stmt.Database = ident
if tok, pos, lit := p.scan(); tok != DOT {
return nil, newParseError(tokstr(tok, lit), []string{"."}, pos)
}
// Read the name of the retention policy.
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
stmt.RetentionPolicy = ident
// Expect a "DESTINATIONS" keyword.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != DESTINATIONS {
return nil, newParseError(tokstr(tok, lit), []string{"DESTINATIONS"}, pos)
}
// Expect one of "ANY ALL" keywords.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok == ALL || tok == ANY {
stmt.Mode = tokens[tok]
} else {
return nil, newParseError(tokstr(tok, lit), []string{"ALL", "ANY"}, pos)
}
// Read list of destinations.
var destinations []string
if destinations, err = p.parseStringList(); err != nil {
return nil, err
}
stmt.Destinations = destinations
return stmt, nil
}
// parseCreateRetentionPolicyStatement parses a string and returns a create retention policy statement.
// This function assumes the CREATE RETENTION POLICY tokens have already been consumed.
func (p *Parser) parseCreateRetentionPolicyStatement() (*CreateRetentionPolicyStatement, error) {
stmt := &CreateRetentionPolicyStatement{}
// Parse the retention policy name.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Consume the required ON token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != ON {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// Parse the database name.
ident, err = p.parseIdent()
if err != nil {
return nil, err
}
stmt.Database = ident
// Parse required DURATION token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != DURATION {
return nil, newParseError(tokstr(tok, lit), []string{"DURATION"}, pos)
}
// Parse duration value
d, err := p.parseDuration()
if err != nil {
return nil, err
}
stmt.Duration = d
// Parse required REPLICATION token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != REPLICATION {
return nil, newParseError(tokstr(tok, lit), []string{"REPLICATION"}, pos)
}
// Parse replication value.
n, err := p.parseInt(1, math.MaxInt32)
if err != nil {
return nil, err
}
stmt.Replication = n
// Parse optional SHARD token.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == SHARD {
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != DURATION {
return nil, newParseError(tokstr(tok, lit), []string{"DURATION"}, pos)
}
// Check to see if they used the INF keyword
tok, pos, _ := p.scanIgnoreWhitespace()
if tok == INF {
return nil, &ParseError{
Message: "invalid duration INF for shard duration",
Pos: pos,
}
}
p.unscan()
d, err := p.parseDuration()
if err != nil {
return nil, err
}
stmt.ShardGroupDuration = d
} else {
p.unscan()
}
// Parse optional DEFAULT token.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == DEFAULT {
stmt.Default = true
} else {
p.unscan()
}
return stmt, nil
}
// parseAlterRetentionPolicyStatement parses a string and returns an alter retention policy statement.
// This function assumes the ALTER RETENTION POLICY tokens have already been consumed.
func (p *Parser) parseAlterRetentionPolicyStatement() (*AlterRetentionPolicyStatement, error) {
stmt := &AlterRetentionPolicyStatement{}
// Parse the retention policy name.
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == DEFAULT {
stmt.Name = "default"
} else if tok == IDENT {
stmt.Name = lit
} else {
return nil, newParseError(tokstr(tok, lit), []string{"identifier"}, pos)
}
// Consume the required ON token.
if tok, pos, lit = p.scanIgnoreWhitespace(); tok != ON {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// Parse the database name.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Database = ident
// Loop through option tokens (DURATION, REPLICATION, SHARD DURATION, DEFAULT, etc.).
found := make(map[Token]struct{})
Loop:
for {
tok, pos, lit := p.scanIgnoreWhitespace()
if _, ok := found[tok]; ok {
return nil, &ParseError{
Message: fmt.Sprintf("found duplicate %s option", tok),
Pos: pos,
}
}
switch tok {
case DURATION:
d, err := p.parseDuration()
if err != nil {
return nil, err
}
stmt.Duration = &d
case REPLICATION:
n, err := p.parseInt(1, math.MaxInt32)
if err != nil {
return nil, err
}
stmt.Replication = &n
case SHARD:
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == DURATION {
// Check to see if they used the INF keyword
tok, pos, _ := p.scanIgnoreWhitespace()
if tok == INF {
return nil, &ParseError{
Message: "invalid duration INF for shard duration",
Pos: pos,
}
}
p.unscan()
d, err := p.parseDuration()
if err != nil {
return nil, err
}
stmt.ShardGroupDuration = &d
} else {
return nil, newParseError(tokstr(tok, lit), []string{"DURATION"}, pos)
}
case DEFAULT:
stmt.Default = true
default:
if len(found) == 0 {
return nil, newParseError(tokstr(tok, lit), []string{"DURATION", "REPLICATION", "SHARD", "DEFAULT"}, pos)
}
p.unscan()
break Loop
}
found[tok] = struct{}{}
}
return stmt, nil
}
// parseInt parses a string representing a base 10 integer and returns the number.
// It returns an error if the parsed number is outside the range [min, max].
func (p *Parser) parseInt(min, max int) (int, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != INTEGER {
return 0, newParseError(tokstr(tok, lit), []string{"integer"}, pos)
}
// Convert string to int.
n, err := strconv.Atoi(lit)
if err != nil {
return 0, &ParseError{Message: err.Error(), Pos: pos}
} else if min > n || n > max {
return 0, &ParseError{
Message: fmt.Sprintf("invalid value %d: must be %d <= n <= %d", n, min, max),
Pos: pos,
}
}
return n, nil
}
// parseUInt64 parses a string and returns a 64-bit unsigned integer literal.
func (p *Parser) parseUInt64() (uint64, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != INTEGER {
return 0, newParseError(tokstr(tok, lit), []string{"integer"}, pos)
}
// Convert string to unsigned 64-bit integer
n, err := strconv.ParseUint(lit, 10, 64)
if err != nil {
return 0, &ParseError{Message: err.Error(), Pos: pos}
}
return uint64(n), nil
}
// parseDuration parses a string and returns a duration literal.
// This function assumes the DURATION token has already been consumed.
func (p *Parser) parseDuration() (time.Duration, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != DURATIONVAL && tok != INF {
return 0, newParseError(tokstr(tok, lit), []string{"duration"}, pos)
}
if tok == INF {
return 0, nil
}
d, err := ParseDuration(lit)
if err != nil {
return 0, &ParseError{Message: err.Error(), Pos: pos}
}
return d, nil
}
// parseIdent parses an identifier.
func (p *Parser) parseIdent() (string, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != IDENT {
return "", newParseError(tokstr(tok, lit), []string{"identifier"}, pos)
}
return lit, nil
}
// parseIdentList parses a comma delimited list of identifiers.
func (p *Parser) parseIdentList() ([]string, error) {
// Parse first (required) identifier.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
idents := []string{ident}
// Parse remaining (optional) identifiers.
for {
if tok, _, _ := p.scanIgnoreWhitespace(); tok != COMMA {
p.unscan()
return idents, nil
}
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
idents = append(idents, ident)
}
}
// parseSegmentedIdents parses a segmented identifiers.
// e.g., "db"."rp".measurement or "db"..measurement
func (p *Parser) parseSegmentedIdents() ([]string, error) {
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
idents := []string{ident}
// Parse remaining (optional) identifiers.
for {
if tok, _, _ := p.scan(); tok != DOT {
// No more segments so we're done.
p.unscan()
break
}
if ch := p.peekRune(); ch == '/' {
// Next segment is a regex so we're done.
break
} else if ch == ':' {
// Next segment is context-specific so let caller handle it.
break
} else if ch == '.' {
// Add an empty identifier.
idents = append(idents, "")
continue
}
// Parse the next identifier.
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
idents = append(idents, ident)
}
if len(idents) > 3 {
msg := fmt.Sprintf("too many segments in %s", QuoteIdent(idents...))
return nil, &ParseError{Message: msg}
}
return idents, nil
}
// parseString parses a string.
func (p *Parser) parseString() (string, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != STRING {
return "", newParseError(tokstr(tok, lit), []string{"string"}, pos)
}
return lit, nil
}
// parseStringList parses a list of strings separated by commas.
func (p *Parser) parseStringList() ([]string, error) {
// Parse first (required) string.
str, err := p.parseString()
if err != nil {
return nil, err
}
strs := []string{str}
// Parse remaining (optional) strings.
for {
if tok, _, _ := p.scanIgnoreWhitespace(); tok != COMMA {
p.unscan()
return strs, nil
}
if str, err = p.parseString(); err != nil {
return nil, err
}
strs = append(strs, str)
}
}
// parseRevokeStatement parses a string and returns a revoke statement.
// This function assumes the REVOKE token has already been consumed.
func (p *Parser) parseRevokeStatement() (Statement, error) {
// Parse the privilege to be revoked.
priv, err := p.parsePrivilege()
if err != nil {
return nil, err
}
// Check for ON or FROM clauses.
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == ON {
stmt, err := p.parseRevokeOnStatement()
if err != nil {
return nil, err
}
stmt.Privilege = priv
return stmt, nil
} else if tok == FROM {
// Admin privilege is only revoked on ALL PRIVILEGES.
if priv != AllPrivileges {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
return p.parseRevokeAdminStatement()
}
// Only ON or FROM clauses are allowed after privilege.
if priv == AllPrivileges {
return nil, newParseError(tokstr(tok, lit), []string{"ON", "FROM"}, pos)
}
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// parseRevokeOnStatement parses a string and returns a revoke statement.
// This function assumes the [PRIVILEGE] ON tokens have already been consumed.
func (p *Parser) parseRevokeOnStatement() (*RevokeStatement, error) {
stmt := &RevokeStatement{}
// Parse the name of the database.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.On = lit
// Parse FROM clause.
tok, pos, lit := p.scanIgnoreWhitespace()
// Check for required FROM token.
if tok != FROM {
return nil, newParseError(tokstr(tok, lit), []string{"FROM"}, pos)
}
// Parse the name of the user.
lit, err = p.parseIdent()
if err != nil {
return nil, err
}
stmt.User = lit
return stmt, nil
}
// parseRevokeAdminStatement parses a string and returns a revoke admin statement.
// This function assumes the ALL [PRVILEGES] FROM token has already been consumed.
func (p *Parser) parseRevokeAdminStatement() (*RevokeAdminStatement, error) {
// Admin privilege is always false when revoke admin clause is called.
stmt := &RevokeAdminStatement{}
// Parse the name of the user.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.User = lit
return stmt, nil
}
// parseGrantStatement parses a string and returns a grant statement.
// This function assumes the GRANT token has already been consumed.
func (p *Parser) parseGrantStatement() (Statement, error) {
// Parse the privilege to be granted.
priv, err := p.parsePrivilege()
if err != nil {
return nil, err
}
// Check for ON or TO clauses.
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == ON {
stmt, err := p.parseGrantOnStatement()
if err != nil {
return nil, err
}
stmt.Privilege = priv
return stmt, nil
} else if tok == TO {
// Admin privilege is only granted on ALL PRIVILEGES.
if priv != AllPrivileges {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
return p.parseGrantAdminStatement()
}
// Only ON or TO clauses are allowed after privilege.
if priv == AllPrivileges {
return nil, newParseError(tokstr(tok, lit), []string{"ON", "TO"}, pos)
}
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// parseGrantOnStatement parses a string and returns a grant statement.
// This function assumes the [PRIVILEGE] ON tokens have already been consumed.
func (p *Parser) parseGrantOnStatement() (*GrantStatement, error) {
stmt := &GrantStatement{}
// Parse the name of the database.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.On = lit
// Parse TO clause.
tok, pos, lit := p.scanIgnoreWhitespace()
// Check for required TO token.
if tok != TO {
return nil, newParseError(tokstr(tok, lit), []string{"TO"}, pos)
}
// Parse the name of the user.
lit, err = p.parseIdent()
if err != nil {
return nil, err
}
stmt.User = lit
return stmt, nil
}
// parseGrantAdminStatement parses a string and returns a grant admin statement.
// This function assumes the ALL [PRVILEGES] TO tokens have already been consumed.
func (p *Parser) parseGrantAdminStatement() (*GrantAdminStatement, error) {
// Admin privilege is always true when grant admin clause is called.
stmt := &GrantAdminStatement{}
// Parse the name of the user.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.User = lit
return stmt, nil
}
// parsePrivilege parses a string and returns a Privilege.
func (p *Parser) parsePrivilege() (Privilege, error) {
tok, pos, lit := p.scanIgnoreWhitespace()
switch tok {
case READ:
return ReadPrivilege, nil
case WRITE:
return WritePrivilege, nil
case ALL:
// Consume optional PRIVILEGES token
tok, pos, lit = p.scanIgnoreWhitespace()
if tok != PRIVILEGES {
p.unscan()
}
return AllPrivileges, nil
}
return 0, newParseError(tokstr(tok, lit), []string{"READ", "WRITE", "ALL [PRIVILEGES]"}, pos)
}
// parseSelectStatement parses a select string and returns a Statement AST object.
// This function assumes the SELECT token has already been consumed.
func (p *Parser) parseSelectStatement(tr targetRequirement) (*SelectStatement, error) {
stmt := &SelectStatement{}
var err error
// Parse fields: "FIELD+".
if stmt.Fields, err = p.parseFields(); err != nil {
return nil, err
}
// Parse target: "INTO"
if stmt.Target, err = p.parseTarget(tr); err != nil {
return nil, err
}
// Parse source: "FROM".
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != FROM {
return nil, newParseError(tokstr(tok, lit), []string{"FROM"}, pos)
}
if stmt.Sources, err = p.parseSources(true); err != nil {
return nil, err
}
// Parse condition: "WHERE EXPR".
if stmt.Condition, err = p.parseCondition(); err != nil {
return nil, err
}
// Parse dimensions: "GROUP BY DIMENSION+".
if stmt.Dimensions, err = p.parseDimensions(); err != nil {
return nil, err
}
// Parse fill options: "fill(<option>)"
if stmt.Fill, stmt.FillValue, err = p.parseFill(); err != nil {
return nil, err
}
// Parse sort: "ORDER BY FIELD+".
if stmt.SortFields, err = p.parseOrderBy(); err != nil {
return nil, err
}
// Parse limit: "LIMIT <n>".
if stmt.Limit, err = p.parseOptionalTokenAndInt(LIMIT); err != nil {
return nil, err
}
// Parse offset: "OFFSET <n>".
if stmt.Offset, err = p.parseOptionalTokenAndInt(OFFSET); err != nil {
return nil, err
}
// Parse series limit: "SLIMIT <n>".
if stmt.SLimit, err = p.parseOptionalTokenAndInt(SLIMIT); err != nil {
return nil, err
}
// Parse series offset: "SOFFSET <n>".
if stmt.SOffset, err = p.parseOptionalTokenAndInt(SOFFSET); err != nil {
return nil, err
}
// Parse timezone: "TZ(<timezone>)".
if stmt.Location, err = p.parseLocation(); err != nil {
return nil, err
}
// Set if the query is a raw data query or one with an aggregate
stmt.IsRawQuery = true
WalkFunc(stmt.Fields, func(n Node) {
if _, ok := n.(*Call); ok {
stmt.IsRawQuery = false
}
})
if err := stmt.validate(tr); err != nil {
return nil, err
}
return stmt, nil
}
// targetRequirement specifies whether or not a target clause is required.
type targetRequirement int
const (
targetRequired targetRequirement = iota
targetNotRequired
targetSubquery
)
// parseTarget parses a string and returns a Target.
func (p *Parser) parseTarget(tr targetRequirement) (*Target, error) {
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != INTO {
if tr == targetRequired {
return nil, newParseError(tokstr(tok, lit), []string{"INTO"}, pos)
}
p.unscan()
return nil, nil
}
// db, rp, and / or measurement
idents, err := p.parseSegmentedIdents()
if err != nil {
return nil, err
}
if len(idents) < 3 {
// Check for source measurement reference.
if ch := p.peekRune(); ch == ':' {
if err := p.parseTokens([]Token{COLON, MEASUREMENT}); err != nil {
return nil, err
}
// Append empty measurement name.
idents = append(idents, "")
}
}
t := &Target{Measurement: &Measurement{IsTarget: true}}
switch len(idents) {
case 1:
t.Measurement.Name = idents[0]
case 2:
t.Measurement.RetentionPolicy = idents[0]
t.Measurement.Name = idents[1]
case 3:
t.Measurement.Database = idents[0]
t.Measurement.RetentionPolicy = idents[1]
t.Measurement.Name = idents[2]
}
return t, nil
}
// parseDeleteStatement parses a string and returns a delete statement.
// This function assumes the DELETE token has already been consumed.
func (p *Parser) parseDeleteStatement() (Statement, error) {
stmt := &DeleteSeriesStatement{}
var err error
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == FROM {
// Parse source.
if stmt.Sources, err = p.parseSources(false); err != nil {
return nil, err
}
var err error
WalkFunc(stmt.Sources, func(n Node) {
if t, ok := n.(*Measurement); ok {
// Don't allow database or retention policy in from clause for delete
// statement. They apply to the selected database across all retention
// policies.
if t.Database != "" {
err = &ParseError{Message: "database not supported"}
}
if t.RetentionPolicy != "" {
err = &ParseError{Message: "retention policy not supported"}
}
}
})
if err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse condition: "WHERE EXPR".
if stmt.Condition, err = p.parseCondition(); err != nil {
return nil, err
}
// If they didn't provide a FROM or a WHERE, this query is invalid
if stmt.Condition == nil && stmt.Sources == nil {
return nil, newParseError(tokstr(tok, lit), []string{"FROM", "WHERE"}, pos)
}
return stmt, nil
}
// parseShowSeriesStatement parses a string and returns a ShowSeriesStatement.
// This function assumes the "SHOW SERIES" tokens have already been consumed.
func (p *Parser) parseShowSeriesStatement() (*ShowSeriesStatement, error) {
stmt := &ShowSeriesStatement{}
var err error
// Parse optional ON clause.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == ON {
// Parse the database.
stmt.Database, err = p.parseIdent()
if err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse optional FROM.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == FROM {
if stmt.Sources, err = p.parseSources(false); err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse condition: "WHERE EXPR".
if stmt.Condition, err = p.parseCondition(); err != nil {
return nil, err
}
// Parse sort: "ORDER BY FIELD+".
if stmt.SortFields, err = p.parseOrderBy(); err != nil {
return nil, err
}
// Parse limit: "LIMIT <n>".
if stmt.Limit, err = p.parseOptionalTokenAndInt(LIMIT); err != nil {
return nil, err
}
// Parse offset: "OFFSET <n>".
if stmt.Offset, err = p.parseOptionalTokenAndInt(OFFSET); err != nil {
return nil, err
}
return stmt, nil
}
// parseShowMeasurementsStatement parses a string and returns a ShowSeriesStatement.
// This function assumes the "SHOW MEASUREMENTS" tokens have already been consumed.
func (p *Parser) parseShowMeasurementsStatement() (*ShowMeasurementsStatement, error) {
stmt := &ShowMeasurementsStatement{}
var err error
// Parse optional ON clause.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == ON {
// Parse the database.
stmt.Database, err = p.parseIdent()
if err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse optional WITH clause.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == WITH {
// Parse required MEASUREMENT token.
if err := p.parseTokens([]Token{MEASUREMENT}); err != nil {
return nil, err
}
// Parse required operator: = or =~.
tok, pos, lit := p.scanIgnoreWhitespace()
switch tok {
case EQ, EQREGEX:
// Parse required source (measurement name or regex).
if stmt.Source, err = p.parseSource(false); err != nil {
return nil, err
}
default:
return nil, newParseError(tokstr(tok, lit), []string{"=", "=~"}, pos)
}
} else {
// Not a WITH clause so put the token back.
p.unscan()
}
// Parse condition: "WHERE EXPR".
if stmt.Condition, err = p.parseCondition(); err != nil {
return nil, err
}
// Parse sort: "ORDER BY FIELD+".
if stmt.SortFields, err = p.parseOrderBy(); err != nil {
return nil, err
}
// Parse limit: "LIMIT <n>".
if stmt.Limit, err = p.parseOptionalTokenAndInt(LIMIT); err != nil {
return nil, err
}
// Parse offset: "OFFSET <n>".
if stmt.Offset, err = p.parseOptionalTokenAndInt(OFFSET); err != nil {
return nil, err
}
return stmt, nil
}
// parseShowQueriesStatement parses a string and returns a ShowQueriesStatement.
// This function assumes the "SHOW QUERIES" tokens have been consumed.
func (p *Parser) parseShowQueriesStatement() (*ShowQueriesStatement, error) {
return &ShowQueriesStatement{}, nil
}
// parseShowRetentionPoliciesStatement parses a string and returns a ShowRetentionPoliciesStatement.
// This function assumes the "SHOW RETENTION POLICIES" tokens have been consumed.
func (p *Parser) parseShowRetentionPoliciesStatement() (*ShowRetentionPoliciesStatement, error) {
stmt := &ShowRetentionPoliciesStatement{}
// Expect an "ON" keyword.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == ON {
// Parse the database.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Database = ident
} else {
p.unscan()
}
return stmt, nil
}
// parseShowTagKeysStatement parses a string and returns a ShowSeriesStatement.
// This function assumes the "SHOW TAG KEYS" tokens have already been consumed.
func (p *Parser) parseShowTagKeysStatement() (*ShowTagKeysStatement, error) {
stmt := &ShowTagKeysStatement{}
var err error
// Parse optional ON clause.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == ON {
// Parse the database.
stmt.Database, err = p.parseIdent()
if err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse optional source.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == FROM {
if stmt.Sources, err = p.parseSources(false); err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse condition: "WHERE EXPR".
if stmt.Condition, err = p.parseCondition(); err != nil {
return nil, err
}
// Parse sort: "ORDER BY FIELD+".
if stmt.SortFields, err = p.parseOrderBy(); err != nil {
return nil, err
}
// Parse limit: "LIMIT <n>".
if stmt.Limit, err = p.parseOptionalTokenAndInt(LIMIT); err != nil {
return nil, err
}
// Parse offset: "OFFSET <n>".
if stmt.Offset, err = p.parseOptionalTokenAndInt(OFFSET); err != nil {
return nil, err
}
// Parse series limit: "SLIMIT <n>".
if stmt.SLimit, err = p.parseOptionalTokenAndInt(SLIMIT); err != nil {
return nil, err
}
// Parse series offset: "SOFFSET <n>".
if stmt.SOffset, err = p.parseOptionalTokenAndInt(SOFFSET); err != nil {
return nil, err
}
return stmt, nil
}
// parseShowTagValuesStatement parses a string and returns a ShowSeriesStatement.
// This function assumes the "SHOW TAG VALUES" tokens have already been consumed.
func (p *Parser) parseShowTagValuesStatement() (*ShowTagValuesStatement, error) {
stmt := &ShowTagValuesStatement{}
var err error
// Parse optional ON clause.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == ON {
// Parse the database.
stmt.Database, err = p.parseIdent()
if err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse optional source.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == FROM {
if stmt.Sources, err = p.parseSources(false); err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse required WITH KEY.
if stmt.Op, stmt.TagKeyExpr, err = p.parseTagKeyExpr(); err != nil {
return nil, err
}
// Parse condition: "WHERE EXPR".
if stmt.Condition, err = p.parseCondition(); err != nil {
return nil, err
}
// Parse sort: "ORDER BY FIELD+".
if stmt.SortFields, err = p.parseOrderBy(); err != nil {
return nil, err
}
// Parse limit: "LIMIT <n>".
if stmt.Limit, err = p.parseOptionalTokenAndInt(LIMIT); err != nil {
return nil, err
}
// Parse offset: "OFFSET <n>".
if stmt.Offset, err = p.parseOptionalTokenAndInt(OFFSET); err != nil {
return nil, err
}
return stmt, nil
}
// parseTagKeys parses a string and returns a list of tag keys.
func (p *Parser) parseTagKeyExpr() (Token, Literal, error) {
var err error
// Parse required WITH KEY tokens.
if err := p.parseTokens([]Token{WITH, KEY}); err != nil {
return 0, nil, err
}
// Parse required IN, EQ, or EQREGEX token.
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == IN {
// Parse required ( token.
if tok, pos, lit = p.scanIgnoreWhitespace(); tok != LPAREN {
return 0, nil, newParseError(tokstr(tok, lit), []string{"("}, pos)
}
// Parse tag key list.
var tagKeys []string
if tagKeys, err = p.parseIdentList(); err != nil {
return 0, nil, err
}
// Parse required ) token.
if tok, pos, lit = p.scanIgnoreWhitespace(); tok != RPAREN {
return 0, nil, newParseError(tokstr(tok, lit), []string{")"}, pos)
}
return IN, &ListLiteral{Vals: tagKeys}, nil
} else if tok == EQ || tok == NEQ {
// Parse required tag key.
ident, err := p.parseIdent()
if err != nil {
return 0, nil, err
}
return tok, &StringLiteral{Val: ident}, nil
} else if tok == EQREGEX || tok == NEQREGEX {
re, err := p.parseRegex()
if err != nil {
return 0, nil, err
} else if re == nil {
// parseRegex can return an empty type, but we need it to be present
tok, pos, lit := p.scanIgnoreWhitespace()
return 0, nil, newParseError(tokstr(tok, lit), []string{"regex"}, pos)
}
return tok, re, nil
}
return 0, nil, newParseError(tokstr(tok, lit), []string{"IN", "=", "=~"}, pos)
}
// parseShowUsersStatement parses a string and returns a ShowUsersStatement.
// This function assumes the "SHOW USERS" tokens have been consumed.
func (p *Parser) parseShowUsersStatement() (*ShowUsersStatement, error) {
return &ShowUsersStatement{}, nil
}
// parseShowSubscriptionsStatement parses a string and returns a ShowSubscriptionsStatement
// This function assumes the "SHOW SUBSCRIPTIONS" tokens have been consumed.
func (p *Parser) parseShowSubscriptionsStatement() (*ShowSubscriptionsStatement, error) {
stmt := &ShowSubscriptionsStatement{}
return stmt, nil
}
// parseShowFieldKeysStatement parses a string and returns a ShowSeriesStatement.
// This function assumes the "SHOW FIELD KEYS" tokens have already been consumed.
func (p *Parser) parseShowFieldKeysStatement() (*ShowFieldKeysStatement, error) {
stmt := &ShowFieldKeysStatement{}
var err error
// Parse optional ON clause.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == ON {
// Parse the database.
stmt.Database, err = p.parseIdent()
if err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse optional source.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == FROM {
if stmt.Sources, err = p.parseSources(false); err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse sort: "ORDER BY FIELD+".
if stmt.SortFields, err = p.parseOrderBy(); err != nil {
return nil, err
}
// Parse limit: "LIMIT <n>".
if stmt.Limit, err = p.parseOptionalTokenAndInt(LIMIT); err != nil {
return nil, err
}
// Parse offset: "OFFSET <n>".
if stmt.Offset, err = p.parseOptionalTokenAndInt(OFFSET); err != nil {
return nil, err
}
return stmt, nil
}
// parseDropMeasurementStatement parses a string and returns a DropMeasurementStatement.
// This function assumes the "DROP MEASUREMENT" tokens have already been consumed.
func (p *Parser) parseDropMeasurementStatement() (*DropMeasurementStatement, error) {
stmt := &DropMeasurementStatement{}
// Parse the name of the measurement to be dropped.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = lit
return stmt, nil
}
// parseDropSeriesStatement parses a string and returns a DropSeriesStatement.
// This function assumes the "DROP SERIES" tokens have already been consumed.
func (p *Parser) parseDropSeriesStatement() (*DropSeriesStatement, error) {
stmt := &DropSeriesStatement{}
var err error
tok, pos, lit := p.scanIgnoreWhitespace()
if tok == FROM {
// Parse source.
if stmt.Sources, err = p.parseSources(false); err != nil {
return nil, err
}
var err error
WalkFunc(stmt.Sources, func(n Node) {
if t, ok := n.(*Measurement); ok {
// Don't allow database or retention policy in from clause for delete
// statement. They apply to the selected database across all retention
// policies.
if t.Database != "" {
err = &ParseError{Message: "database not supported"}
}
if t.RetentionPolicy != "" {
err = &ParseError{Message: "retention policy not supported"}
}
}
})
if err != nil {
return nil, err
}
} else {
p.unscan()
}
// Parse condition: "WHERE EXPR".
if stmt.Condition, err = p.parseCondition(); err != nil {
return nil, err
}
// If they didn't provide a FROM or a WHERE, this query is invalid
if stmt.Condition == nil && stmt.Sources == nil {
return nil, newParseError(tokstr(tok, lit), []string{"FROM", "WHERE"}, pos)
}
return stmt, nil
}
// parseDropShardStatement parses a string and returns a
// DropShardStatement. This function assumes the "DROP SHARD" tokens
// have already been consumed.
func (p *Parser) parseDropShardStatement() (*DropShardStatement, error) {
var err error
stmt := &DropShardStatement{}
// Parse the ID of the shard to be dropped.
if stmt.ID, err = p.parseUInt64(); err != nil {
return nil, err
}
return stmt, nil
}
// parseShowContinuousQueriesStatement parses a string and returns a ShowContinuousQueriesStatement.
// This function assumes the "SHOW CONTINUOUS" tokens have already been consumed.
func (p *Parser) parseShowContinuousQueriesStatement() (*ShowContinuousQueriesStatement, error) {
stmt := &ShowContinuousQueriesStatement{}
// Expect a "QUERIES" token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != QUERIES {
return nil, newParseError(tokstr(tok, lit), []string{"QUERIES"}, pos)
}
return stmt, nil
}
// parseGrantsForUserStatement parses a string and returns a ShowGrantsForUserStatement.
// This function assumes the "SHOW GRANTS" tokens have already been consumed.
func (p *Parser) parseGrantsForUserStatement() (*ShowGrantsForUserStatement, error) {
stmt := &ShowGrantsForUserStatement{}
// Expect a "FOR" token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != FOR {
return nil, newParseError(tokstr(tok, lit), []string{"FOR"}, pos)
}
// Parse the name of the user to be displayed.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = lit
return stmt, nil
}
// parseShowDatabasesStatement parses a string and returns a ShowDatabasesStatement.
// This function assumes the "SHOW DATABASE" tokens have already been consumed.
func (p *Parser) parseShowDatabasesStatement() (*ShowDatabasesStatement, error) {
stmt := &ShowDatabasesStatement{}
return stmt, nil
}
// parseCreateContinuousQueriesStatement parses a string and returns a CreateContinuousQueryStatement.
// This function assumes the "CREATE CONTINUOUS" tokens have already been consumed.
func (p *Parser) parseCreateContinuousQueryStatement() (*CreateContinuousQueryStatement, error) {
stmt := &CreateContinuousQueryStatement{}
// Expect a "QUERY" token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != QUERY {
return nil, newParseError(tokstr(tok, lit), []string{"QUERY"}, pos)
}
// Read the id of the query to create.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Expect an "ON" keyword.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != ON {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// Read the name of the database to create the query on.
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
stmt.Database = ident
if p.parseTokenMaybe(RESAMPLE) {
stmt.ResampleEvery, stmt.ResampleFor, err = p.parseResample()
if err != nil {
return nil, err
}
}
// Expect a "BEGIN SELECT" tokens.
if err := p.parseTokens([]Token{BEGIN, SELECT}); err != nil {
return nil, err
}
// Read the select statement to be used as the source.
source, err := p.parseSelectStatement(targetRequired)
if err != nil {
return nil, err
}
stmt.Source = source
// validate that the statement has a non-zero group by interval if it is aggregated
if !source.IsRawQuery {
d, err := source.GroupByInterval()
if d == 0 || err != nil {
// rewind so we can output an error with some info
p.unscan() // unscan the whitespace
p.unscan() // unscan the last token
tok, pos, lit := p.scanIgnoreWhitespace()
expected := []string{"GROUP BY time(...)"}
if err != nil {
expected = append(expected, err.Error())
}
return nil, newParseError(tokstr(tok, lit), expected, pos)
}
}
// Expect a "END" keyword.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != END {
return nil, newParseError(tokstr(tok, lit), []string{"END"}, pos)
}
if err := stmt.validate(); err != nil {
return nil, err
}
return stmt, nil
}
// parseCreateDatabaseStatement parses a string and returns a CreateDatabaseStatement.
// This function assumes the "CREATE DATABASE" tokens have already been consumed.
func (p *Parser) parseCreateDatabaseStatement() (*CreateDatabaseStatement, error) {
stmt := &CreateDatabaseStatement{}
// Parse the name of the database to be created.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = lit
// Look for "WITH"
if tok, _, _ := p.scanIgnoreWhitespace(); tok == WITH {
// validate that at least one of DURATION, NAME, REPLICATION or SHARD is provided
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != DURATION && tok != NAME && tok != REPLICATION && tok != SHARD {
return nil, newParseError(tokstr(tok, lit), []string{"DURATION", "NAME", "REPLICATION", "SHARD"}, pos)
}
// rewind
p.unscan()
// mark statement as having a RetentionPolicyInfo defined
stmt.RetentionPolicyCreate = true
// Look for "DURATION"
if err := p.parseTokens([]Token{DURATION}); err != nil {
p.unscan()
} else {
rpDuration, err := p.parseDuration()
if err != nil {
return nil, err
}
stmt.RetentionPolicyDuration = &rpDuration
}
// Look for "REPLICATION"
if err := p.parseTokens([]Token{REPLICATION}); err != nil {
p.unscan()
} else {
rpReplication, err := p.parseInt(1, math.MaxInt32)
if err != nil {
return nil, err
}
stmt.RetentionPolicyReplication = &rpReplication
}
// Look for "SHARD"
if err := p.parseTokens([]Token{SHARD}); err != nil {
p.unscan()
} else {
// Look for "DURATION"
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != DURATION {
return nil, newParseError(tokstr(tok, lit), []string{"DURATION"}, pos)
}
stmt.RetentionPolicyShardGroupDuration, err = p.parseDuration()
if err != nil {
return nil, err
}
}
// Look for "NAME"
if err := p.parseTokens([]Token{NAME}); err != nil {
p.unscan()
} else {
stmt.RetentionPolicyName, err = p.parseIdent()
if err != nil {
return nil, err
}
}
} else {
p.unscan()
}
return stmt, nil
}
// parseDropDatabaseStatement parses a string and returns a DropDatabaseStatement.
// This function assumes the DROP DATABASE tokens have already been consumed.
func (p *Parser) parseDropDatabaseStatement() (*DropDatabaseStatement, error) {
stmt := &DropDatabaseStatement{}
// Parse the name of the database to be dropped.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = lit
return stmt, nil
}
// parseDropSubscriptionStatement parses a string and returns a DropSubscriptionStatement.
// This function assumes the "DROP SUBSCRIPTION" tokens have already been consumed.
func (p *Parser) parseDropSubscriptionStatement() (*DropSubscriptionStatement, error) {
stmt := &DropSubscriptionStatement{}
// Read the id of the subscription to drop.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Expect an "ON" keyword.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != ON {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// Read the name of the database.
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
stmt.Database = ident
if tok, pos, lit := p.scan(); tok != DOT {
return nil, newParseError(tokstr(tok, lit), []string{"."}, pos)
}
// Read the name of the retention policy.
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
stmt.RetentionPolicy = ident
return stmt, nil
}
// parseDropRetentionPolicyStatement parses a string and returns a DropRetentionPolicyStatement.
// This function assumes the DROP RETENTION POLICY tokens have been consumed.
func (p *Parser) parseDropRetentionPolicyStatement() (*DropRetentionPolicyStatement, error) {
stmt := &DropRetentionPolicyStatement{}
// Parse the policy name.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Consume the required ON token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != ON {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// Parse the database name.
if stmt.Database, err = p.parseIdent(); err != nil {
return nil, err
}
return stmt, nil
}
// parseCreateUserStatement parses a string and returns a CreateUserStatement.
// This function assumes the "CREATE USER" tokens have already been consumed.
func (p *Parser) parseCreateUserStatement() (*CreateUserStatement, error) {
stmt := &CreateUserStatement{}
// Parse name of the user to be created.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Consume "WITH PASSWORD" tokens
if err := p.parseTokens([]Token{WITH, PASSWORD}); err != nil {
return nil, err
}
// Parse new user's password
if ident, err = p.parseString(); err != nil {
return nil, err
}
stmt.Password = ident
// Check for option WITH clause.
if tok, _, _ := p.scanIgnoreWhitespace(); tok != WITH {
p.unscan()
return stmt, nil
}
// "WITH ALL PRIVILEGES" grants the new user admin privilege.
// Only admin privilege can be set on user creation.
if err := p.parseTokens([]Token{ALL, PRIVILEGES}); err != nil {
return nil, err
}
stmt.Admin = true
return stmt, nil
}
// parseDropUserStatement parses a string and returns a DropUserStatement.
// This function assumes the DROP USER tokens have already been consumed.
func (p *Parser) parseDropUserStatement() (*DropUserStatement, error) {
stmt := &DropUserStatement{}
// Parse the name of the user to be dropped.
lit, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = lit
return stmt, nil
}
// parseShowShardGroupsStatement parses a string for "SHOW SHARD GROUPS" statement.
// This function assumes the "SHOW SHARD GROUPS" tokens have already been consumed.
func (p *Parser) parseShowShardGroupsStatement() (*ShowShardGroupsStatement, error) {
return &ShowShardGroupsStatement{}, nil
}
// parseShowShardsStatement parses a string for "SHOW SHARDS" statement.
// This function assumes the "SHOW SHARDS" tokens have already been consumed.
func (p *Parser) parseShowShardsStatement() (*ShowShardsStatement, error) {
return &ShowShardsStatement{}, nil
}
// parseShowStatsStatement parses a string and returns a ShowStatsStatement.
// This function assumes the "SHOW STATS" tokens have already been consumed.
func (p *Parser) parseShowStatsStatement() (*ShowStatsStatement, error) {
stmt := &ShowStatsStatement{}
var err error
if tok, _, _ := p.scanIgnoreWhitespace(); tok == FOR {
stmt.Module, err = p.parseString()
} else {
p.unscan()
}
return stmt, err
}
// parseShowDiagnostics parses a string and returns a ShowDiagnosticsStatement.
func (p *Parser) parseShowDiagnosticsStatement() (*ShowDiagnosticsStatement, error) {
stmt := &ShowDiagnosticsStatement{}
var err error
if tok, _, _ := p.scanIgnoreWhitespace(); tok == FOR {
stmt.Module, err = p.parseString()
} else {
p.unscan()
}
return stmt, err
}
// parseDropContinuousQueriesStatement parses a string and returns a DropContinuousQueryStatement.
// This function assumes the "DROP CONTINUOUS" tokens have already been consumed.
func (p *Parser) parseDropContinuousQueryStatement() (*DropContinuousQueryStatement, error) {
stmt := &DropContinuousQueryStatement{}
// Expect a "QUERY" token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != QUERY {
return nil, newParseError(tokstr(tok, lit), []string{"QUERY"}, pos)
}
// Read the id of the query to drop.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
stmt.Name = ident
// Expect an "ON" keyword.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != ON {
return nil, newParseError(tokstr(tok, lit), []string{"ON"}, pos)
}
// Read the name of the database to remove the query from.
if ident, err = p.parseIdent(); err != nil {
return nil, err
}
stmt.Database = ident
return stmt, nil
}
// parseFields parses a list of one or more fields.
func (p *Parser) parseFields() (Fields, error) {
var fields Fields
for {
// Parse the field.
f, err := p.parseField()
if err != nil {
return nil, err
}
// Add new field.
fields = append(fields, f)
// If there's not a comma next then stop parsing fields.
if tok, _, _ := p.scan(); tok != COMMA {
p.unscan()
break
}
}
return fields, nil
}
// parseField parses a single field.
func (p *Parser) parseField() (*Field, error) {
f := &Field{}
// Attempt to parse a regex.
re, err := p.parseRegex()
if err != nil {
return nil, err
} else if re != nil {
f.Expr = re
} else {
_, pos, _ := p.scanIgnoreWhitespace()
p.unscan()
// Parse the expression first.
expr, err := p.ParseExpr()
if err != nil {
return nil, err
}
var c validateField
Walk(&c, expr)
if c.foundInvalid {
return nil, fmt.Errorf("invalid operator %s in SELECT clause at line %d, char %d; operator is intended for WHERE clause", c.badToken, pos.Line+1, pos.Char+1)
}
f.Expr = expr
}
// Parse the alias if the current and next tokens are "WS AS".
alias, err := p.parseAlias()
if err != nil {
return nil, err
}
f.Alias = alias
// Consume all trailing whitespace.
p.consumeWhitespace()
return f, nil
}
// validateField checks if the Expr is a valid field. We disallow all binary expression
// that return a boolean.
type validateField struct {
foundInvalid bool
badToken Token
}
func (c *validateField) Visit(n Node) Visitor {
e, ok := n.(*BinaryExpr)
if !ok {
return c
}
switch e.Op {
case EQ, NEQ, EQREGEX,
NEQREGEX, LT, LTE, GT, GTE,
AND, OR:
c.foundInvalid = true
c.badToken = e.Op
return nil
}
return c
}
// parseAlias parses the "AS IDENT" alias for fields and dimensions.
func (p *Parser) parseAlias() (string, error) {
// Check if the next token is "AS". If not, then unscan and exit.
if tok, _, _ := p.scanIgnoreWhitespace(); tok != AS {
p.unscan()
return "", nil
}
// Then we should have the alias identifier.
lit, err := p.parseIdent()
if err != nil {
return "", err
}
return lit, nil
}
// parseSources parses a comma delimited list of sources.
func (p *Parser) parseSources(subqueries bool) (Sources, error) {
var sources Sources
for {
s, err := p.parseSource(subqueries)
if err != nil {
return nil, err
}
sources = append(sources, s)
if tok, _, _ := p.scanIgnoreWhitespace(); tok != COMMA {
p.unscan()
break
}
}
return sources, nil
}
// peekRune returns the next rune that would be read by the scanner.
func (p *Parser) peekRune() rune {
r, _, _ := p.s.s.r.ReadRune()
if r != eof {
_ = p.s.s.r.UnreadRune()
}
return r
}
func (p *Parser) parseSource(subqueries bool) (Source, error) {
m := &Measurement{}
// Attempt to parse a regex.
re, err := p.parseRegex()
if err != nil {
return nil, err
} else if re != nil {
m.Regex = re
// Regex is always last so we're done.
return m, nil
}
// If there is no regular expression, this might be a subquery.
// Parse the subquery if we are in a query that allows them as a source.
if m.Regex == nil && subqueries {
if tok, _, _ := p.scanIgnoreWhitespace(); tok == LPAREN {
if err := p.parseTokens([]Token{SELECT}); err != nil {
return nil, err
}
stmt, err := p.parseSelectStatement(targetSubquery)
if err != nil {
return nil, err
}
if err := p.parseTokens([]Token{RPAREN}); err != nil {
return nil, err
}
return &SubQuery{Statement: stmt}, nil
} else {
p.unscan()
}
}
// Didn't find a regex so parse segmented identifiers.
idents, err := p.parseSegmentedIdents()
if err != nil {
return nil, err
}
// If we already have the max allowed idents, we're done.
if len(idents) == 3 {
m.Database, m.RetentionPolicy, m.Name = idents[0], idents[1], idents[2]
return m, nil
}
// Check again for regex.
re, err = p.parseRegex()
if err != nil {
return nil, err
} else if re != nil {
m.Regex = re
}
// Assign identifiers to their proper locations.
switch len(idents) {
case 1:
if re != nil {
m.RetentionPolicy = idents[0]
} else {
m.Name = idents[0]
}
case 2:
if re != nil {
m.Database, m.RetentionPolicy = idents[0], idents[1]
} else {
m.RetentionPolicy, m.Name = idents[0], idents[1]
}
}
return m, nil
}
// parseCondition parses the "WHERE" clause of the query, if it exists.
func (p *Parser) parseCondition() (Expr, error) {
// Check if the WHERE token exists.
if tok, _, _ := p.scanIgnoreWhitespace(); tok != WHERE {
p.unscan()
return nil, nil
}
// Scan the identifier for the source.
expr, err := p.ParseExpr()
if err != nil {
return nil, err
}
return expr, nil
}
// parseDimensions parses the "GROUP BY" clause of the query, if it exists.
func (p *Parser) parseDimensions() (Dimensions, error) {
// If the next token is not GROUP then exit.
if tok, _, _ := p.scanIgnoreWhitespace(); tok != GROUP {
p.unscan()
return nil, nil
}
// Now the next token should be "BY".
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != BY {
return nil, newParseError(tokstr(tok, lit), []string{"BY"}, pos)
}
var dimensions Dimensions
for {
// Parse the dimension.
d, err := p.parseDimension()
if err != nil {
return nil, err
}
// Add new dimension.
dimensions = append(dimensions, d)
// If there's not a comma next then stop parsing dimensions.
if tok, _, _ := p.scan(); tok != COMMA {
p.unscan()
break
}
}
return dimensions, nil
}
// parseDimension parses a single dimension.
func (p *Parser) parseDimension() (*Dimension, error) {
re, err := p.parseRegex()
if err != nil {
return nil, err
} else if re != nil {
return &Dimension{Expr: re}, nil
}
// Parse the expression first.
expr, err := p.ParseExpr()
if err != nil {
return nil, err
}
// Consume all trailing whitespace.
p.consumeWhitespace()
return &Dimension{Expr: expr}, nil
}
// parseFill parses the fill call and its options.
func (p *Parser) parseFill() (FillOption, interface{}, error) {
// Parse the expression first.
tok, _, lit := p.scanIgnoreWhitespace()
p.unscan()
if tok != IDENT || strings.ToLower(lit) != "fill" {
return NullFill, nil, nil
}
expr, err := p.ParseExpr()
if err != nil {
return NullFill, nil, err
}
fill, ok := expr.(*Call)
if !ok {
return NullFill, nil, errors.New("fill must be a function call")
} else if len(fill.Args) != 1 {
return NullFill, nil, errors.New("fill requires an argument, e.g.: 0, null, none, previous, linear")
}
switch fill.Args[0].String() {
case "null":
return NullFill, nil, nil
case "none":
return NoFill, nil, nil
case "previous":
return PreviousFill, nil, nil
case "linear":
return LinearFill, nil, nil
default:
switch num := fill.Args[0].(type) {
case *IntegerLiteral:
return NumberFill, num.Val, nil
case *NumberLiteral:
return NumberFill, num.Val, nil
default:
return NullFill, nil, fmt.Errorf("expected number argument in fill()")
}
}
}
// parseLocation parses the timezone call and its arguments.
func (p *Parser) parseLocation() (*time.Location, error) {
// Parse the expression first.
tok, _, lit := p.scanIgnoreWhitespace()
p.unscan()
if tok != IDENT || strings.ToLower(lit) != "tz" {
return nil, nil
}
expr, err := p.ParseExpr()
if err != nil {
return nil, err
}
tz, ok := expr.(*Call)
if !ok {
return nil, errors.New("tz must be a function call")
} else if len(tz.Args) != 1 {
return nil, errors.New("tz requires exactly one argument")
}
tzname, ok := tz.Args[0].(*StringLiteral)
if !ok {
return nil, errors.New("expected string argument in tz()")
}
loc, err := time.LoadLocation(tzname.Val)
if err != nil {
// Do not pass the same error message as the error may contain sensitive pathnames.
return nil, fmt.Errorf("unable to find time zone %s", tzname.Val)
}
return loc, nil
}
// parseOptionalTokenAndInt parses the specified token followed
// by an int, if it exists.
func (p *Parser) parseOptionalTokenAndInt(t Token) (int, error) {
// Check if the token exists.
if tok, _, _ := p.scanIgnoreWhitespace(); tok != t {
p.unscan()
return 0, nil
}
// Scan the number.
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != INTEGER {
return 0, newParseError(tokstr(tok, lit), []string{"integer"}, pos)
}
// Parse number.
n, _ := strconv.ParseInt(lit, 10, 64)
if n < 0 {
msg := fmt.Sprintf("%s must be >= 0", t.String())
return 0, &ParseError{Message: msg, Pos: pos}
}
return int(n), nil
}
// parseOrderBy parses the "ORDER BY" clause of a query, if it exists.
func (p *Parser) parseOrderBy() (SortFields, error) {
// Return nil result and nil error if no ORDER token at this position.
if tok, _, _ := p.scanIgnoreWhitespace(); tok != ORDER {
p.unscan()
return nil, nil
}
// Parse the required BY token.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != BY {
return nil, newParseError(tokstr(tok, lit), []string{"BY"}, pos)
}
// Parse the ORDER BY fields.
fields, err := p.parseSortFields()
if err != nil {
return nil, err
}
return fields, nil
}
// parseSortFields parses the sort fields for an ORDER BY clause.
func (p *Parser) parseSortFields() (SortFields, error) {
var fields SortFields
tok, pos, lit := p.scanIgnoreWhitespace()
switch tok {
// The first field after an order by may not have a field name (e.g. ORDER BY ASC)
case ASC, DESC:
fields = append(fields, &SortField{Ascending: (tok == ASC)})
// If it's a token, parse it as a sort field. At least one is required.
case IDENT:
p.unscan()
field, err := p.parseSortField()
if err != nil {
return nil, err
}
if lit != "time" {
return nil, errors.New("only ORDER BY time supported at this time")
}
fields = append(fields, field)
// Parse error...
default:
return nil, newParseError(tokstr(tok, lit), []string{"identifier", "ASC", "DESC"}, pos)
}
// Parse additional fields.
for {
tok, _, _ := p.scanIgnoreWhitespace()
if tok != COMMA {
p.unscan()
break
}
field, err := p.parseSortField()
if err != nil {
return nil, err
}
fields = append(fields, field)
}
if len(fields) > 1 {
return nil, errors.New("only ORDER BY time supported at this time")
}
return fields, nil
}
// parseSortField parses one field of an ORDER BY clause.
func (p *Parser) parseSortField() (*SortField, error) {
field := &SortField{}
// Parse sort field name.
ident, err := p.parseIdent()
if err != nil {
return nil, err
}
field.Name = ident
// Check for optional ASC or DESC clause. Default is ASC.
tok, _, _ := p.scanIgnoreWhitespace()
if tok != ASC && tok != DESC {
p.unscan()
tok = ASC
}
field.Ascending = (tok == ASC)
return field, nil
}
// parseVarRef parses a reference to a measurement or field.
func (p *Parser) parseVarRef() (*VarRef, error) {
// Parse the segments of the variable ref.
segments, err := p.parseSegmentedIdents()
if err != nil {
return nil, err
}
var dtype DataType
if tok, _, _ := p.scan(); tok == DOUBLECOLON {
tok, pos, lit := p.scan()
switch tok {
case IDENT:
switch strings.ToLower(lit) {
case "float":
dtype = Float
case "integer":
dtype = Integer
case "string":
dtype = String
case "boolean":
dtype = Boolean
default:
return nil, newParseError(tokstr(tok, lit), []string{"float", "integer", "string", "boolean", "field", "tag"}, pos)
}
case FIELD:
dtype = AnyField
case TAG:
dtype = Tag
default:
return nil, newParseError(tokstr(tok, lit), []string{"float", "integer", "string", "boolean", "field", "tag"}, pos)
}
} else {
p.unscan()
}
vr := &VarRef{Val: strings.Join(segments, "."), Type: dtype}
return vr, nil
}
// ParseExpr parses an expression.
func (p *Parser) ParseExpr() (Expr, error) {
var err error
// Dummy root node.
root := &BinaryExpr{}
// Parse a non-binary expression type to start.
// This variable will always be the root of the expression tree.
root.RHS, err = p.parseUnaryExpr()
if err != nil {
return nil, err
}
// Loop over operations and unary exprs and build a tree based on precendence.
for {
// If the next token is NOT an operator then return the expression.
op, _, _ := p.scanIgnoreWhitespace()
if !op.isOperator() {
p.unscan()
return root.RHS, nil
}
// Otherwise parse the next expression.
var rhs Expr
if IsRegexOp(op) {
// RHS of a regex operator must be a regular expression.
p.consumeWhitespace()
if rhs, err = p.parseRegex(); err != nil {
return nil, err
}
// parseRegex can return an empty type, but we need it to be present
if rhs.(*RegexLiteral) == nil {
tok, pos, lit := p.scanIgnoreWhitespace()
return nil, newParseError(tokstr(tok, lit), []string{"regex"}, pos)
}
} else {
if rhs, err = p.parseUnaryExpr(); err != nil {
return nil, err
}
}
// Find the right spot in the tree to add the new expression by
// descending the RHS of the expression tree until we reach the last
// BinaryExpr or a BinaryExpr whose RHS has an operator with
// precedence >= the operator being added.
for node := root; ; {
r, ok := node.RHS.(*BinaryExpr)
if !ok || r.Op.Precedence() >= op.Precedence() {
// Add the new expression here and break.
node.RHS = &BinaryExpr{LHS: node.RHS, RHS: rhs, Op: op}
break
}
node = r
}
}
}
// parseUnaryExpr parses an non-binary expression.
func (p *Parser) parseUnaryExpr() (Expr, error) {
// If the first token is a LPAREN then parse it as its own grouped expression.
if tok, _, _ := p.scanIgnoreWhitespace(); tok == LPAREN {
expr, err := p.ParseExpr()
if err != nil {
return nil, err
}
// Expect an RPAREN at the end.
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != RPAREN {
return nil, newParseError(tokstr(tok, lit), []string{")"}, pos)
}
return &ParenExpr{Expr: expr}, nil
}
p.unscan()
// Read next token.
tok, pos, lit := p.scanIgnoreWhitespace()
switch tok {
case IDENT:
// If the next immediate token is a left parentheses, parse as function call.
// Otherwise parse as a variable reference.
if tok0, _, _ := p.scan(); tok0 == LPAREN {
return p.parseCall(lit)
}
p.unscan() // unscan the last token (wasn't an LPAREN)
p.unscan() // unscan the IDENT token
// Parse it as a VarRef.
return p.parseVarRef()
case DISTINCT:
// If the next immediate token is a left parentheses, parse as function call.
// Otherwise parse as a Distinct expression.
tok0, pos, lit := p.scan()
if tok0 == LPAREN {
return p.parseCall("distinct")
} else if tok0 == WS {
tok1, pos, lit := p.scanIgnoreWhitespace()
if tok1 != IDENT {
return nil, newParseError(tokstr(tok1, lit), []string{"identifier"}, pos)
}
return &Distinct{Val: lit}, nil
}
return nil, newParseError(tokstr(tok0, lit), []string{"(", "identifier"}, pos)
case STRING:
return &StringLiteral{Val: lit}, nil
case NUMBER:
v, err := strconv.ParseFloat(lit, 64)
if err != nil {
return nil, &ParseError{Message: "unable to parse number", Pos: pos}
}
return &NumberLiteral{Val: v}, nil
case INTEGER:
v, err := strconv.ParseInt(lit, 10, 64)
if err != nil {
return nil, &ParseError{Message: "unable to parse integer", Pos: pos}
}
return &IntegerLiteral{Val: v}, nil
case TRUE, FALSE:
return &BooleanLiteral{Val: (tok == TRUE)}, nil
case DURATIONVAL:
v, err := ParseDuration(lit)
if err != nil {
return nil, err
}
return &DurationLiteral{Val: v}, nil
case MUL:
wc := &Wildcard{}
if tok, _, _ := p.scan(); tok == DOUBLECOLON {
tok, pos, lit := p.scan()
switch tok {
case FIELD, TAG:
wc.Type = tok
default:
return nil, newParseError(tokstr(tok, lit), []string{"field", "tag"}, pos)
}
} else {
p.unscan()
}
return wc, nil
case REGEX:
re, err := regexp.Compile(lit)
if err != nil {
return nil, &ParseError{Message: err.Error(), Pos: pos}
}
return &RegexLiteral{Val: re}, nil
case BOUNDPARAM:
k := strings.TrimPrefix(lit, "$")
if len(k) == 0 {
return nil, errors.New("empty bound parameter")
}
v, ok := p.params[k]
if !ok {
return nil, fmt.Errorf("missing parameter: %s", k)
}
switch v := v.(type) {
case float64:
return &NumberLiteral{Val: v}, nil
case int64:
return &IntegerLiteral{Val: v}, nil
case string:
return &StringLiteral{Val: v}, nil
case bool:
return &BooleanLiteral{Val: v}, nil
default:
return nil, fmt.Errorf("unable to bind parameter with type %T", v)
}
case ADD, SUB:
mul := 1
if tok == SUB {
mul = -1
}
tok0, pos0, lit0 := p.scanIgnoreWhitespace()
switch tok0 {
case NUMBER, INTEGER, DURATIONVAL, LPAREN, IDENT:
// Unscan the token and use parseUnaryExpr.
p.unscan()
lit, err := p.parseUnaryExpr()
if err != nil {
return nil, err
}
switch lit := lit.(type) {
case *NumberLiteral:
lit.Val *= float64(mul)
case *IntegerLiteral:
lit.Val *= int64(mul)
case *DurationLiteral:
lit.Val *= time.Duration(mul)
case *VarRef, *Call, *ParenExpr:
// Multiply the variable.
return &BinaryExpr{
Op: MUL,
LHS: &IntegerLiteral{Val: int64(mul)},
RHS: lit,
}, nil
default:
panic(fmt.Sprintf("unexpected literal: %T", lit))
}
return lit, nil
default:
return nil, newParseError(tokstr(tok0, lit0), []string{"identifier", "number", "duration", "("}, pos0)
}
default:
return nil, newParseError(tokstr(tok, lit), []string{"identifier", "string", "number", "bool"}, pos)
}
}
// parseRegex parses a regular expression.
func (p *Parser) parseRegex() (*RegexLiteral, error) {
nextRune := p.peekRune()
if isWhitespace(nextRune) {
p.consumeWhitespace()
}
// If the next character is not a '/', then return nils.
nextRune = p.peekRune()
if nextRune != '/' {
return nil, nil
}
tok, pos, lit := p.s.ScanRegex()
if tok == BADESCAPE {
msg := fmt.Sprintf("bad escape: %s", lit)
return nil, &ParseError{Message: msg, Pos: pos}
} else if tok == BADREGEX {
msg := fmt.Sprintf("bad regex: %s", lit)
return nil, &ParseError{Message: msg, Pos: pos}
} else if tok != REGEX {
return nil, newParseError(tokstr(tok, lit), []string{"regex"}, pos)
}
re, err := regexp.Compile(lit)
if err != nil {
return nil, &ParseError{Message: err.Error(), Pos: pos}
}
return &RegexLiteral{Val: re}, nil
}
// parseCall parses a function call.
// This function assumes the function name and LPAREN have been consumed.
func (p *Parser) parseCall(name string) (*Call, error) {
name = strings.ToLower(name)
// Parse first function argument if one exists.
var args []Expr
re, err := p.parseRegex()
if err != nil {
return nil, err
} else if re != nil {
args = append(args, re)
} else {
// If there's a right paren then just return immediately.
if tok, _, _ := p.scan(); tok == RPAREN {
return &Call{Name: name}, nil
}
p.unscan()
arg, err := p.ParseExpr()
if err != nil {
return nil, err
}
args = append(args, arg)
}
// Parse additional function arguments if there is a comma.
for {
// If there's not a comma, stop parsing arguments.
if tok, _, _ := p.scanIgnoreWhitespace(); tok != COMMA {
p.unscan()
break
}
re, err := p.parseRegex()
if err != nil {
return nil, err
} else if re != nil {
args = append(args, re)
continue
}
// Parse an expression argument.
arg, err := p.ParseExpr()
if err != nil {
return nil, err
}
args = append(args, arg)
}
// There should be a right parentheses at the end.
if tok, pos, lit := p.scan(); tok != RPAREN {
return nil, newParseError(tokstr(tok, lit), []string{")"}, pos)
}
return &Call{Name: name, Args: args}, nil
}
// parseResample parses a RESAMPLE [EVERY <duration>] [FOR <duration>].
// This function assumes RESAMPLE has already been consumed.
// EVERY and FOR are optional, but at least one of the two has to be used.
func (p *Parser) parseResample() (time.Duration, time.Duration, error) {
var interval time.Duration
if p.parseTokenMaybe(EVERY) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != DURATIONVAL {
return 0, 0, newParseError(tokstr(tok, lit), []string{"duration"}, pos)
}
d, err := ParseDuration(lit)
if err != nil {
return 0, 0, &ParseError{Message: err.Error(), Pos: pos}
}
interval = d
}
var maxDuration time.Duration
if p.parseTokenMaybe(FOR) {
tok, pos, lit := p.scanIgnoreWhitespace()
if tok != DURATIONVAL {
return 0, 0, newParseError(tokstr(tok, lit), []string{"duration"}, pos)
}
d, err := ParseDuration(lit)
if err != nil {
return 0, 0, &ParseError{Message: err.Error(), Pos: pos}
}
maxDuration = d
}
// Neither EVERY or FOR were read, so read the next token again
// so we can return a suitable error message.
if interval == 0 && maxDuration == 0 {
tok, pos, lit := p.scanIgnoreWhitespace()
return 0, 0, newParseError(tokstr(tok, lit), []string{"EVERY", "FOR"}, pos)
}
return interval, maxDuration, nil
}
// scan returns the next token from the underlying scanner.
func (p *Parser) scan() (tok Token, pos Pos, lit string) { return p.s.Scan() }
// scanIgnoreWhitespace scans the next non-whitespace and non-comment token.
func (p *Parser) scanIgnoreWhitespace() (tok Token, pos Pos, lit string) {
for {
tok, pos, lit = p.scan()
if tok == WS || tok == COMMENT {
continue
}
return
}
}
// consumeWhitespace scans the next token if it's whitespace.
func (p *Parser) consumeWhitespace() {
if tok, _, _ := p.scan(); tok != WS {
p.unscan()
}
}
// unscan pushes the previously read token back onto the buffer.
func (p *Parser) unscan() { p.s.Unscan() }
// ParseDuration parses a time duration from a string.
// This is needed instead of time.ParseDuration because this will support
// the full syntax that InfluxQL supports for specifying durations
// including weeks and days.
func ParseDuration(s string) (time.Duration, error) {
// Return an error if the string is blank or one character
if len(s) < 2 {
return 0, ErrInvalidDuration
}
// Split string into individual runes.
a := []rune(s)
// Start with a zero duration.
var d time.Duration
i := 0
// Check for a negative.
isNegative := false
if a[i] == '-' {
isNegative = true
i++
}
var measure int64
var unit string
// Parsing loop.
for i < len(a) {
// Find the number portion.
start := i
for ; i < len(a) && isDigit(a[i]); i++ {
// Scan for the digits.
}
// Check if we reached the end of the string prematurely.
if i >= len(a) || i == start {
return 0, ErrInvalidDuration
}
// Parse the numeric part.
n, err := strconv.ParseInt(string(a[start:i]), 10, 64)
if err != nil {
return 0, ErrInvalidDuration
}
measure = n
// Extract the unit of measure.
// If the last two characters are "ms" then parse as milliseconds.
// Otherwise just use the last character as the unit of measure.
unit = string(a[i])
switch a[i] {
case 'n':
if i+1 < len(a) && a[i+1] == 's' {
unit = string(a[i : i+2])
d += time.Duration(n)
i += 2
continue
}
return 0, ErrInvalidDuration
case 'u', 'µ':
d += time.Duration(n) * time.Microsecond
case 'm':
if i+1 < len(a) && a[i+1] == 's' {
unit = string(a[i : i+2])
d += time.Duration(n) * time.Millisecond
i += 2
continue
}
d += time.Duration(n) * time.Minute
case 's':
d += time.Duration(n) * time.Second
case 'h':
d += time.Duration(n) * time.Hour
case 'd':
d += time.Duration(n) * 24 * time.Hour
case 'w':
d += time.Duration(n) * 7 * 24 * time.Hour
default:
return 0, ErrInvalidDuration
}
i++
}
// Check to see if we overflowed a duration
if d < 0 && !isNegative {
return 0, fmt.Errorf("overflowed duration %d%s: choose a smaller duration or INF", measure, unit)
}
if isNegative {
d = -d
}
return d, nil
}
// FormatDuration formats a duration to a string.
func FormatDuration(d time.Duration) string {
if d == 0 {
return "0s"
} else if d%(7*24*time.Hour) == 0 {
return fmt.Sprintf("%dw", d/(7*24*time.Hour))
} else if d%(24*time.Hour) == 0 {
return fmt.Sprintf("%dd", d/(24*time.Hour))
} else if d%time.Hour == 0 {
return fmt.Sprintf("%dh", d/time.Hour)
} else if d%time.Minute == 0 {
return fmt.Sprintf("%dm", d/time.Minute)
} else if d%time.Second == 0 {
return fmt.Sprintf("%ds", d/time.Second)
} else if d%time.Millisecond == 0 {
return fmt.Sprintf("%dms", d/time.Millisecond)
}
// Although we accept both "u" and "µ" when reading microsecond durations,
// we output with "u", which can be represented in 1 byte,
// instead of "µ", which requires 2 bytes.
return fmt.Sprintf("%du", d/time.Microsecond)
}
// parseTokens consumes an expected sequence of tokens.
func (p *Parser) parseTokens(toks []Token) error {
for _, expected := range toks {
if tok, pos, lit := p.scanIgnoreWhitespace(); tok != expected {
return newParseError(tokstr(tok, lit), []string{tokens[expected]}, pos)
}
}
return nil
}
// parseTokenMaybe consumes the next token if it matches the expected one and
// does nothing if the next token is not the next one.
func (p *Parser) parseTokenMaybe(expected Token) bool {
tok, _, _ := p.scanIgnoreWhitespace()
if tok != expected {
p.unscan()
return false
}
return true
}
var (
// Quote String replacer.
qsReplacer = strings.NewReplacer("\n", `\n`, `\`, `\\`, `'`, `\'`)
// Quote Ident replacer.
qiReplacer = strings.NewReplacer("\n", `\n`, `\`, `\\`, `"`, `\"`)
)
// QuoteString returns a quoted string.
func QuoteString(s string) string {
return `'` + qsReplacer.Replace(s) + `'`
}
// QuoteIdent returns a quoted identifier from multiple bare identifiers.
func QuoteIdent(segments ...string) string {
var buf bytes.Buffer
for i, segment := range segments {
needQuote := IdentNeedsQuotes(segment) ||
((i < len(segments)-1) && segment != "") || // not last segment && not ""
((i == 0 || i == len(segments)-1) && segment == "") // the first or last segment and an empty string
if needQuote {
_ = buf.WriteByte('"')
}
_, _ = buf.WriteString(qiReplacer.Replace(segment))
if needQuote {
_ = buf.WriteByte('"')
}
if i < len(segments)-1 {
_ = buf.WriteByte('.')
}
}
return buf.String()
}
// IdentNeedsQuotes returns true if the ident string given would require quotes.
func IdentNeedsQuotes(ident string) bool {
// check if this identifier is a keyword
tok := Lookup(ident)
if tok != IDENT {
return true
}
for i, r := range ident {
if i == 0 && !isIdentFirstChar(r) {
return true
} else if i > 0 && !isIdentChar(r) {
return true
}
}
return false
}
// isDateString returns true if the string looks like a date-only time literal.
func isDateString(s string) bool { return dateStringRegexp.MatchString(s) }
// isDateTimeString returns true if the string looks like a date+time time literal.
func isDateTimeString(s string) bool { return dateTimeStringRegexp.MatchString(s) }
var dateStringRegexp = regexp.MustCompile(`^\d{4}-\d{2}-\d{2}$`)
var dateTimeStringRegexp = regexp.MustCompile(`^\d{4}-\d{2}-\d{2}.+`)
// ErrInvalidDuration is returned when parsing a malformed duration.
var ErrInvalidDuration = errors.New("invalid duration")
// ParseError represents an error that occurred during parsing.
type ParseError struct {
Message string
Found string
Expected []string
Pos Pos
}
// newParseError returns a new instance of ParseError.
func newParseError(found string, expected []string, pos Pos) *ParseError {
return &ParseError{Found: found, Expected: expected, Pos: pos}
}
// Error returns the string representation of the error.
func (e *ParseError) Error() string {
if e.Message != "" {
return fmt.Sprintf("%s at line %d, char %d", e.Message, e.Pos.Line+1, e.Pos.Char+1)
}
return fmt.Sprintf("found %s, expected %s at line %d, char %d", e.Found, strings.Join(e.Expected, ", "), e.Pos.Line+1, e.Pos.Char+1)
}