mirror of
https://github.com/ceph/ceph-csi.git
synced 2024-11-25 07:40:19 +00:00
484 lines
16 KiB
Go
484 lines
16 KiB
Go
|
// Copyright 2020 Google LLC
|
||
|
//
|
||
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
||
|
// you may not use this file except in compliance with the License.
|
||
|
// You may obtain a copy of the License at
|
||
|
//
|
||
|
// http://www.apache.org/licenses/LICENSE-2.0
|
||
|
//
|
||
|
// Unless required by applicable law or agreed to in writing, software
|
||
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
||
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||
|
// See the License for the specific language governing permissions and
|
||
|
// limitations under the License.
|
||
|
|
||
|
// Package ext contains CEL extension libraries where each library defines a related set of
|
||
|
// constants, functions, macros, or other configuration settings which may not be covered by
|
||
|
// the core CEL spec.
|
||
|
package ext
|
||
|
|
||
|
import (
|
||
|
"fmt"
|
||
|
"reflect"
|
||
|
"strings"
|
||
|
"unicode"
|
||
|
|
||
|
"github.com/google/cel-go/cel"
|
||
|
"github.com/google/cel-go/common/types"
|
||
|
"github.com/google/cel-go/common/types/ref"
|
||
|
)
|
||
|
|
||
|
// Strings returns a cel.EnvOption to configure extended functions for string manipulation.
|
||
|
// As a general note, all indices are zero-based.
|
||
|
//
|
||
|
// CharAt
|
||
|
//
|
||
|
// Returns the character at the given position. If the position is negative, or greater than
|
||
|
// the length of the string, the function will produce an error:
|
||
|
//
|
||
|
// <string>.charAt(<int>) -> <string>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'hello'.charAt(4) // return 'o'
|
||
|
// 'hello'.charAt(5) // return ''
|
||
|
// 'hello'.charAt(-1) // error
|
||
|
//
|
||
|
// IndexOf
|
||
|
//
|
||
|
// Returns the integer index of the first occurrence of the search string. If the search string is
|
||
|
// not found the function returns -1.
|
||
|
//
|
||
|
// The function also accepts an optional position from which to begin the substring search. If the
|
||
|
// substring is the empty string, the index where the search starts is returned (zero or custom).
|
||
|
//
|
||
|
// <string>.indexOf(<string>) -> <int>
|
||
|
// <string>.indexOf(<string>, <int>) -> <int>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'hello mellow'.indexOf('') // returns 0
|
||
|
// 'hello mellow'.indexOf('ello') // returns 1
|
||
|
// 'hello mellow'.indexOf('jello') // returns -1
|
||
|
// 'hello mellow'.indexOf('', 2) // returns 2
|
||
|
// 'hello mellow'.indexOf('ello', 2) // returns 7
|
||
|
// 'hello mellow'.indexOf('ello', 20) // error
|
||
|
//
|
||
|
// Join
|
||
|
//
|
||
|
// Returns a new string where the elements of string list are concatenated.
|
||
|
//
|
||
|
// The function also accepts an optional separator which is placed between elements in the resulting string.
|
||
|
//
|
||
|
// <list<string>>.join() -> <string>
|
||
|
// <list<string>>.join(<string>) -> <string>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// ['hello', 'mellow'].join() // returns 'hellomellow'
|
||
|
// ['hello', 'mellow'].join(' ') // returns 'hello mellow'
|
||
|
// [].join() // returns ''
|
||
|
// [].join('/') // returns ''
|
||
|
//
|
||
|
// LastIndexOf
|
||
|
//
|
||
|
// Returns the integer index at the start of the last occurrence of the search string. If the
|
||
|
// search string is not found the function returns -1.
|
||
|
//
|
||
|
// The function also accepts an optional position which represents the last index to be
|
||
|
// considered as the beginning of the substring match. If the substring is the empty string,
|
||
|
// the index where the search starts is returned (string length or custom).
|
||
|
//
|
||
|
// <string>.lastIndexOf(<string>) -> <int>
|
||
|
// <string>.lastIndexOf(<string>, <int>) -> <int>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'hello mellow'.lastIndexOf('') // returns 12
|
||
|
// 'hello mellow'.lastIndexOf('ello') // returns 7
|
||
|
// 'hello mellow'.lastIndexOf('jello') // returns -1
|
||
|
// 'hello mellow'.lastIndexOf('ello', 6) // returns 1
|
||
|
// 'hello mellow'.lastIndexOf('ello', -1) // error
|
||
|
//
|
||
|
// LowerAscii
|
||
|
//
|
||
|
// Returns a new string where all ASCII characters are lower-cased.
|
||
|
//
|
||
|
// This function does not perform Unicode case-mapping for characters outside the ASCII range.
|
||
|
//
|
||
|
// <string>.lowerAscii() -> <string>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'TacoCat'.lowerAscii() // returns 'tacocat'
|
||
|
// 'TacoCÆt Xii'.lowerAscii() // returns 'tacocÆt xii'
|
||
|
//
|
||
|
// Replace
|
||
|
//
|
||
|
// Returns a new string based on the target, which replaces the occurrences of a search string
|
||
|
// with a replacement string if present. The function accepts an optional limit on the number of
|
||
|
// substring replacements to be made.
|
||
|
//
|
||
|
// When the replacement limit is 0, the result is the original string. When the limit is a negative
|
||
|
// number, the function behaves the same as replace all.
|
||
|
//
|
||
|
// <string>.replace(<string>, <string>) -> <string>
|
||
|
// <string>.replace(<string>, <string>, <int>) -> <string>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'hello hello'.replace('he', 'we') // returns 'wello wello'
|
||
|
// 'hello hello'.replace('he', 'we', -1) // returns 'wello wello'
|
||
|
// 'hello hello'.replace('he', 'we', 1) // returns 'wello hello'
|
||
|
// 'hello hello'.replace('he', 'we', 0) // returns 'hello hello'
|
||
|
//
|
||
|
// Split
|
||
|
//
|
||
|
// Returns a list of strings split from the input by the given separator. The function accepts
|
||
|
// an optional argument specifying a limit on the number of substrings produced by the split.
|
||
|
//
|
||
|
// When the split limit is 0, the result is an empty list. When the limit is 1, the result is the
|
||
|
// target string to split. When the limit is a negative number, the function behaves the same as
|
||
|
// split all.
|
||
|
//
|
||
|
// <string>.split(<string>) -> <list<string>>
|
||
|
// <string>.split(<string>, <int>) -> <list<string>>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'hello hello hello'.split(' ') // returns ['hello', 'hello', 'hello']
|
||
|
// 'hello hello hello'.split(' ', 0) // returns []
|
||
|
// 'hello hello hello'.split(' ', 1) // returns ['hello hello hello']
|
||
|
// 'hello hello hello'.split(' ', 2) // returns ['hello', 'hello hello']
|
||
|
// 'hello hello hello'.split(' ', -1) // returns ['hello', 'hello', 'hello']
|
||
|
//
|
||
|
// Substring
|
||
|
//
|
||
|
// Returns the substring given a numeric range corresponding to character positions. Optionally
|
||
|
// may omit the trailing range for a substring from a given character position until the end of
|
||
|
// a string.
|
||
|
//
|
||
|
// Character offsets are 0-based with an inclusive start range and exclusive end range. It is an
|
||
|
// error to specify an end range that is lower than the start range, or for either the start or end
|
||
|
// index to be negative or exceed the string length.
|
||
|
//
|
||
|
// <string>.substring(<int>) -> <string>
|
||
|
// <string>.substring(<int>, <int>) -> <string>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'tacocat'.substring(4) // returns 'cat'
|
||
|
// 'tacocat'.substring(0, 4) // returns 'taco'
|
||
|
// 'tacocat'.substring(-1) // error
|
||
|
// 'tacocat'.substring(2, 1) // error
|
||
|
//
|
||
|
// Trim
|
||
|
//
|
||
|
// Returns a new string which removes the leading and trailing whitespace in the target string.
|
||
|
// The trim function uses the Unicode definition of whitespace which does not include the
|
||
|
// zero-width spaces. See: https://en.wikipedia.org/wiki/Whitespace_character#Unicode
|
||
|
//
|
||
|
// <string>.trim() -> <string>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// ' \ttrim\n '.trim() // returns 'trim'
|
||
|
//
|
||
|
// UpperAscii
|
||
|
//
|
||
|
// Returns a new string where all ASCII characters are upper-cased.
|
||
|
//
|
||
|
// This function does not perform Unicode case-mapping for characters outside the ASCII range.
|
||
|
//
|
||
|
// <string>.upperAscii() -> <string>
|
||
|
//
|
||
|
// Examples:
|
||
|
//
|
||
|
// 'TacoCat'.upperAscii() // returns 'TACOCAT'
|
||
|
// 'TacoCÆt Xii'.upperAscii() // returns 'TACOCÆT XII'
|
||
|
func Strings() cel.EnvOption {
|
||
|
return cel.Lib(stringLib{})
|
||
|
}
|
||
|
|
||
|
type stringLib struct{}
|
||
|
|
||
|
func (stringLib) CompileOptions() []cel.EnvOption {
|
||
|
return []cel.EnvOption{
|
||
|
cel.Function("charAt",
|
||
|
cel.MemberOverload("string_char_at_int", []*cel.Type{cel.StringType, cel.IntType}, cel.StringType,
|
||
|
cel.BinaryBinding(func(str, ind ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
i := ind.(types.Int)
|
||
|
return stringOrError(charAt(string(s), int64(i)))
|
||
|
}))),
|
||
|
cel.Function("indexOf",
|
||
|
cel.MemberOverload("string_index_of_string", []*cel.Type{cel.StringType, cel.StringType}, cel.IntType,
|
||
|
cel.BinaryBinding(func(str, substr ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
sub := substr.(types.String)
|
||
|
return intOrError(indexOf(string(s), string(sub)))
|
||
|
})),
|
||
|
cel.MemberOverload("string_index_of_string_int", []*cel.Type{cel.StringType, cel.StringType, cel.IntType}, cel.IntType,
|
||
|
cel.FunctionBinding(func(args ...ref.Val) ref.Val {
|
||
|
s := args[0].(types.String)
|
||
|
sub := args[1].(types.String)
|
||
|
offset := args[2].(types.Int)
|
||
|
return intOrError(indexOfOffset(string(s), string(sub), int64(offset)))
|
||
|
}))),
|
||
|
cel.Function("lastIndexOf",
|
||
|
cel.MemberOverload("string_last_index_of_string", []*cel.Type{cel.StringType, cel.StringType}, cel.IntType,
|
||
|
cel.BinaryBinding(func(str, substr ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
sub := substr.(types.String)
|
||
|
return intOrError(lastIndexOf(string(s), string(sub)))
|
||
|
})),
|
||
|
cel.MemberOverload("string_last_index_of_string_int", []*cel.Type{cel.StringType, cel.StringType, cel.IntType}, cel.IntType,
|
||
|
cel.FunctionBinding(func(args ...ref.Val) ref.Val {
|
||
|
s := args[0].(types.String)
|
||
|
sub := args[1].(types.String)
|
||
|
offset := args[2].(types.Int)
|
||
|
return intOrError(lastIndexOfOffset(string(s), string(sub), int64(offset)))
|
||
|
}))),
|
||
|
cel.Function("lowerAscii",
|
||
|
cel.MemberOverload("string_lower_ascii", []*cel.Type{cel.StringType}, cel.StringType,
|
||
|
cel.UnaryBinding(func(str ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
return stringOrError(lowerASCII(string(s)))
|
||
|
}))),
|
||
|
cel.Function("replace",
|
||
|
cel.MemberOverload(
|
||
|
"string_replace_string_string", []*cel.Type{cel.StringType, cel.StringType, cel.StringType}, cel.StringType,
|
||
|
cel.FunctionBinding(func(args ...ref.Val) ref.Val {
|
||
|
str := args[0].(types.String)
|
||
|
old := args[1].(types.String)
|
||
|
new := args[2].(types.String)
|
||
|
return stringOrError(replace(string(str), string(old), string(new)))
|
||
|
})),
|
||
|
cel.MemberOverload(
|
||
|
"string_replace_string_string_int", []*cel.Type{cel.StringType, cel.StringType, cel.StringType, cel.IntType}, cel.StringType,
|
||
|
cel.FunctionBinding(func(args ...ref.Val) ref.Val {
|
||
|
str := args[0].(types.String)
|
||
|
old := args[1].(types.String)
|
||
|
new := args[2].(types.String)
|
||
|
n := args[3].(types.Int)
|
||
|
return stringOrError(replaceN(string(str), string(old), string(new), int64(n)))
|
||
|
}))),
|
||
|
cel.Function("split",
|
||
|
cel.MemberOverload("string_split_string", []*cel.Type{cel.StringType, cel.StringType}, cel.ListType(cel.StringType),
|
||
|
cel.BinaryBinding(func(str, separator ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
sep := separator.(types.String)
|
||
|
return listStringOrError(split(string(s), string(sep)))
|
||
|
})),
|
||
|
cel.MemberOverload("string_split_string_int", []*cel.Type{cel.StringType, cel.StringType, cel.IntType}, cel.ListType(cel.StringType),
|
||
|
cel.FunctionBinding(func(args ...ref.Val) ref.Val {
|
||
|
s := args[0].(types.String)
|
||
|
sep := args[1].(types.String)
|
||
|
n := args[2].(types.Int)
|
||
|
return listStringOrError(splitN(string(s), string(sep), int64(n)))
|
||
|
}))),
|
||
|
cel.Function("substring",
|
||
|
cel.MemberOverload("string_substring_int", []*cel.Type{cel.StringType, cel.IntType}, cel.StringType,
|
||
|
cel.BinaryBinding(func(str, offset ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
off := offset.(types.Int)
|
||
|
return stringOrError(substr(string(s), int64(off)))
|
||
|
})),
|
||
|
cel.MemberOverload("string_substring_int_int", []*cel.Type{cel.StringType, cel.IntType, cel.IntType}, cel.StringType,
|
||
|
cel.FunctionBinding(func(args ...ref.Val) ref.Val {
|
||
|
s := args[0].(types.String)
|
||
|
start := args[1].(types.Int)
|
||
|
end := args[2].(types.Int)
|
||
|
return stringOrError(substrRange(string(s), int64(start), int64(end)))
|
||
|
}))),
|
||
|
cel.Function("trim",
|
||
|
cel.MemberOverload("string_trim", []*cel.Type{cel.StringType}, cel.StringType,
|
||
|
cel.UnaryBinding(func(str ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
return stringOrError(trimSpace(string(s)))
|
||
|
}))),
|
||
|
cel.Function("upperAscii",
|
||
|
cel.MemberOverload("string_upper_ascii", []*cel.Type{cel.StringType}, cel.StringType,
|
||
|
cel.UnaryBinding(func(str ref.Val) ref.Val {
|
||
|
s := str.(types.String)
|
||
|
return stringOrError(upperASCII(string(s)))
|
||
|
}))),
|
||
|
cel.Function("join",
|
||
|
cel.MemberOverload("list_join", []*cel.Type{cel.ListType(cel.StringType)}, cel.StringType,
|
||
|
cel.UnaryBinding(func(list ref.Val) ref.Val {
|
||
|
l, err := list.ConvertToNative(stringListType)
|
||
|
if err != nil {
|
||
|
return types.NewErr(err.Error())
|
||
|
}
|
||
|
return stringOrError(join(l.([]string)))
|
||
|
})),
|
||
|
cel.MemberOverload("list_join_string", []*cel.Type{cel.ListType(cel.StringType), cel.StringType}, cel.StringType,
|
||
|
cel.BinaryBinding(func(list, delim ref.Val) ref.Val {
|
||
|
l, err := list.ConvertToNative(stringListType)
|
||
|
if err != nil {
|
||
|
return types.NewErr(err.Error())
|
||
|
}
|
||
|
d := delim.(types.String)
|
||
|
return stringOrError(joinSeparator(l.([]string), string(d)))
|
||
|
}))),
|
||
|
}
|
||
|
}
|
||
|
|
||
|
func (stringLib) ProgramOptions() []cel.ProgramOption {
|
||
|
return []cel.ProgramOption{}
|
||
|
}
|
||
|
|
||
|
func charAt(str string, ind int64) (string, error) {
|
||
|
i := int(ind)
|
||
|
runes := []rune(str)
|
||
|
if i < 0 || i > len(runes) {
|
||
|
return "", fmt.Errorf("index out of range: %d", ind)
|
||
|
}
|
||
|
if i == len(runes) {
|
||
|
return "", nil
|
||
|
}
|
||
|
return string(runes[i]), nil
|
||
|
}
|
||
|
|
||
|
func indexOf(str, substr string) (int64, error) {
|
||
|
return indexOfOffset(str, substr, int64(0))
|
||
|
}
|
||
|
|
||
|
func indexOfOffset(str, substr string, offset int64) (int64, error) {
|
||
|
if substr == "" {
|
||
|
return offset, nil
|
||
|
}
|
||
|
off := int(offset)
|
||
|
runes := []rune(str)
|
||
|
subrunes := []rune(substr)
|
||
|
if off < 0 || off >= len(runes) {
|
||
|
return -1, fmt.Errorf("index out of range: %d", off)
|
||
|
}
|
||
|
for i := off; i < len(runes)-(len(subrunes)-1); i++ {
|
||
|
found := true
|
||
|
for j := 0; j < len(subrunes); j++ {
|
||
|
if runes[i+j] != subrunes[j] {
|
||
|
found = false
|
||
|
break
|
||
|
}
|
||
|
}
|
||
|
if found {
|
||
|
return int64(i), nil
|
||
|
}
|
||
|
}
|
||
|
return -1, nil
|
||
|
}
|
||
|
|
||
|
func lastIndexOf(str, substr string) (int64, error) {
|
||
|
runes := []rune(str)
|
||
|
if substr == "" {
|
||
|
return int64(len(runes)), nil
|
||
|
}
|
||
|
return lastIndexOfOffset(str, substr, int64(len(runes)-1))
|
||
|
}
|
||
|
|
||
|
func lastIndexOfOffset(str, substr string, offset int64) (int64, error) {
|
||
|
if substr == "" {
|
||
|
return offset, nil
|
||
|
}
|
||
|
off := int(offset)
|
||
|
runes := []rune(str)
|
||
|
subrunes := []rune(substr)
|
||
|
if off < 0 || off >= len(runes) {
|
||
|
return -1, fmt.Errorf("index out of range: %d", off)
|
||
|
}
|
||
|
if off > len(runes)-len(subrunes) {
|
||
|
off = len(runes) - len(subrunes)
|
||
|
}
|
||
|
for i := off; i >= 0; i-- {
|
||
|
found := true
|
||
|
for j := 0; j < len(subrunes); j++ {
|
||
|
if runes[i+j] != subrunes[j] {
|
||
|
found = false
|
||
|
break
|
||
|
}
|
||
|
}
|
||
|
if found {
|
||
|
return int64(i), nil
|
||
|
}
|
||
|
}
|
||
|
return -1, nil
|
||
|
}
|
||
|
|
||
|
func lowerASCII(str string) (string, error) {
|
||
|
runes := []rune(str)
|
||
|
for i, r := range runes {
|
||
|
if r <= unicode.MaxASCII {
|
||
|
r = unicode.ToLower(r)
|
||
|
runes[i] = r
|
||
|
}
|
||
|
}
|
||
|
return string(runes), nil
|
||
|
}
|
||
|
|
||
|
func replace(str, old, new string) (string, error) {
|
||
|
return strings.ReplaceAll(str, old, new), nil
|
||
|
}
|
||
|
|
||
|
func replaceN(str, old, new string, n int64) (string, error) {
|
||
|
return strings.Replace(str, old, new, int(n)), nil
|
||
|
}
|
||
|
|
||
|
func split(str, sep string) ([]string, error) {
|
||
|
return strings.Split(str, sep), nil
|
||
|
}
|
||
|
|
||
|
func splitN(str, sep string, n int64) ([]string, error) {
|
||
|
return strings.SplitN(str, sep, int(n)), nil
|
||
|
}
|
||
|
|
||
|
func substr(str string, start int64) (string, error) {
|
||
|
runes := []rune(str)
|
||
|
if int(start) < 0 || int(start) > len(runes) {
|
||
|
return "", fmt.Errorf("index out of range: %d", start)
|
||
|
}
|
||
|
return string(runes[start:]), nil
|
||
|
}
|
||
|
|
||
|
func substrRange(str string, start, end int64) (string, error) {
|
||
|
runes := []rune(str)
|
||
|
l := len(runes)
|
||
|
if start > end {
|
||
|
return "", fmt.Errorf("invalid substring range. start: %d, end: %d", start, end)
|
||
|
}
|
||
|
if int(start) < 0 || int(start) > l {
|
||
|
return "", fmt.Errorf("index out of range: %d", start)
|
||
|
}
|
||
|
if int(end) < 0 || int(end) > l {
|
||
|
return "", fmt.Errorf("index out of range: %d", end)
|
||
|
}
|
||
|
return string(runes[int(start):int(end)]), nil
|
||
|
}
|
||
|
|
||
|
func trimSpace(str string) (string, error) {
|
||
|
return strings.TrimSpace(str), nil
|
||
|
}
|
||
|
|
||
|
func upperASCII(str string) (string, error) {
|
||
|
runes := []rune(str)
|
||
|
for i, r := range runes {
|
||
|
if r <= unicode.MaxASCII {
|
||
|
r = unicode.ToUpper(r)
|
||
|
runes[i] = r
|
||
|
}
|
||
|
}
|
||
|
return string(runes), nil
|
||
|
}
|
||
|
|
||
|
func joinSeparator(strs []string, separator string) (string, error) {
|
||
|
return strings.Join(strs, separator), nil
|
||
|
}
|
||
|
|
||
|
func join(strs []string) (string, error) {
|
||
|
return strings.Join(strs, ""), nil
|
||
|
}
|
||
|
|
||
|
var (
|
||
|
stringListType = reflect.TypeOf([]string{})
|
||
|
)
|