1
0
mirror of https://github.com/danog/liquid.git synced 2024-12-02 14:17:45 +01:00
liquid/expressions/scanner.go
2017-07-13 20:18:23 -04:00

647 lines
13 KiB
Go

//line scanner.rl:1
package expressions
import "strconv"
//line scanner.go:9
var _expression_actions []byte = []byte{
0, 1, 0, 1, 1, 1, 2, 1, 10,
1, 11, 1, 12, 1, 13, 1, 14,
1, 15, 1, 16, 1, 17, 1, 18,
1, 19, 1, 20, 1, 21, 1, 22,
1, 23, 1, 24, 1, 25, 1, 26,
1, 27, 1, 28, 1, 29, 2, 2,
3, 2, 2, 4, 2, 2, 5, 2,
2, 6, 2, 2, 7, 2, 2, 8,
2, 2, 9,
}
var _expression_key_offsets []int16 = []int16{
0, 1, 2, 3, 4, 5, 6, 7,
8, 9, 10, 11, 12, 14, 16, 17,
18, 19, 20, 21, 22, 23, 24, 25,
52, 55, 56, 57, 59, 60, 63, 65,
68, 76, 85, 94, 95, 96, 97, 107,
108, 119, 130, 141, 152, 163, 174, 185,
196, 207, 218, 229, 240, 251, 262, 273,
284, 295, 306, 317,
}
var _expression_trans_keys []byte = []byte{
34, 115, 115, 105, 103, 110, 32, 111,
111, 112, 32, 39, 48, 57, 99, 119,
121, 99, 108, 101, 32, 104, 101, 110,
32, 32, 33, 34, 37, 39, 45, 46,
60, 61, 62, 95, 97, 99, 102, 105,
110, 111, 116, 123, 9, 13, 48, 57,
65, 90, 98, 122, 32, 9, 13, 61,
34, 97, 108, 39, 46, 48, 57, 48,
57, 46, 48, 57, 45, 95, 48, 57,
65, 90, 97, 122, 45, 63, 95, 48,
57, 65, 90, 97, 122, 45, 63, 95,
48, 57, 65, 90, 97, 122, 61, 61,
61, 45, 58, 63, 95, 48, 57, 65,
90, 97, 122, 58, 45, 58, 63, 95,
110, 48, 57, 65, 90, 97, 122, 45,
58, 63, 95, 100, 48, 57, 65, 90,
97, 122, 45, 58, 63, 95, 111, 48,
57, 65, 90, 97, 122, 45, 58, 63,
95, 110, 48, 57, 65, 90, 97, 122,
45, 58, 63, 95, 116, 48, 57, 65,
90, 97, 122, 45, 58, 63, 95, 97,
48, 57, 65, 90, 98, 122, 45, 58,
63, 95, 105, 48, 57, 65, 90, 97,
122, 45, 58, 63, 95, 110, 48, 57,
65, 90, 97, 122, 45, 58, 63, 95,
115, 48, 57, 65, 90, 97, 122, 45,
58, 63, 95, 97, 48, 57, 65, 90,
98, 122, 45, 58, 63, 95, 108, 48,
57, 65, 90, 97, 122, 45, 58, 63,
95, 115, 48, 57, 65, 90, 97, 122,
45, 58, 63, 95, 101, 48, 57, 65,
90, 97, 122, 45, 58, 63, 95, 110,
48, 57, 65, 90, 97, 122, 45, 58,
63, 95, 105, 48, 57, 65, 90, 97,
122, 45, 58, 63, 95, 108, 48, 57,
65, 90, 97, 122, 45, 58, 63, 95,
114, 48, 57, 65, 90, 97, 122, 45,
58, 63, 95, 114, 48, 57, 65, 90,
97, 122, 45, 58, 63, 95, 117, 48,
57, 65, 90, 97, 122, 37,
}
var _expression_single_lengths []byte = []byte{
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 0, 2, 1, 1,
1, 1, 1, 1, 1, 1, 1, 19,
1, 1, 1, 2, 1, 1, 0, 1,
2, 3, 3, 1, 1, 1, 4, 1,
5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 1,
}
var _expression_range_lengths []byte = []byte{
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 4,
1, 0, 0, 0, 0, 1, 1, 1,
3, 3, 3, 0, 0, 0, 3, 0,
3, 3, 3, 3, 3, 3, 3, 3,
3, 3, 3, 3, 3, 3, 3, 3,
3, 3, 3, 0,
}
var _expression_index_offsets []int16 = []int16{
0, 2, 4, 6, 8, 10, 12, 14,
16, 18, 20, 22, 24, 26, 29, 31,
33, 35, 37, 39, 41, 43, 45, 47,
71, 74, 76, 78, 81, 83, 86, 88,
91, 97, 104, 111, 113, 115, 117, 125,
127, 136, 145, 154, 163, 172, 181, 190,
199, 208, 217, 226, 235, 244, 253, 262,
271, 280, 289, 298,
}
var _expression_indicies []byte = []byte{
2, 1, 3, 0, 4, 0, 5, 0,
6, 0, 7, 0, 8, 0, 9, 0,
10, 0, 11, 0, 12, 0, 2, 13,
14, 0, 15, 16, 0, 17, 0, 18,
0, 19, 0, 20, 0, 21, 0, 22,
0, 23, 0, 24, 0, 25, 0, 27,
28, 29, 30, 31, 32, 33, 35, 36,
37, 38, 39, 40, 41, 42, 43, 44,
45, 46, 27, 34, 38, 38, 26, 27,
27, 47, 49, 48, 2, 1, 50, 51,
48, 2, 13, 52, 34, 48, 14, 53,
14, 34, 54, 55, 55, 56, 55, 55,
48, 55, 58, 55, 55, 55, 55, 57,
55, 58, 55, 56, 55, 55, 53, 59,
48, 60, 48, 61, 48, 38, 63, 64,
38, 38, 38, 38, 62, 63, 65, 38,
63, 64, 38, 66, 38, 38, 38, 65,
38, 63, 64, 38, 67, 38, 38, 38,
65, 38, 63, 64, 38, 68, 38, 38,
38, 65, 38, 63, 64, 38, 69, 38,
38, 38, 65, 38, 63, 64, 38, 70,
38, 38, 38, 65, 38, 63, 64, 38,
71, 38, 38, 38, 65, 38, 63, 64,
38, 72, 38, 38, 38, 65, 38, 63,
64, 38, 73, 38, 38, 38, 65, 38,
63, 64, 38, 74, 38, 38, 38, 65,
38, 63, 64, 38, 75, 38, 38, 38,
65, 38, 63, 64, 38, 76, 38, 38,
38, 65, 38, 63, 64, 38, 77, 38,
38, 38, 65, 38, 63, 64, 38, 78,
38, 38, 38, 65, 38, 63, 64, 38,
79, 38, 38, 38, 65, 38, 63, 64,
38, 80, 38, 38, 38, 65, 38, 63,
64, 38, 81, 38, 38, 38, 65, 38,
63, 64, 38, 82, 38, 38, 38, 65,
38, 63, 64, 38, 83, 38, 38, 38,
65, 38, 63, 64, 38, 77, 38, 38,
38, 65, 84, 48,
}
var _expression_trans_targs []byte = []byte{
23, 0, 23, 2, 3, 4, 5, 6,
23, 8, 9, 10, 23, 11, 30, 14,
19, 15, 16, 17, 18, 23, 20, 21,
22, 23, 23, 24, 25, 26, 27, 28,
29, 32, 31, 35, 36, 37, 38, 40,
42, 49, 53, 54, 56, 57, 59, 23,
23, 23, 1, 7, 12, 23, 23, 33,
34, 23, 23, 23, 23, 23, 23, 23,
39, 23, 41, 38, 43, 44, 45, 46,
47, 48, 38, 50, 51, 52, 38, 38,
55, 38, 38, 58, 13,
}
var _expression_trans_actions []byte = []byte{
43, 0, 15, 0, 0, 0, 0, 0,
7, 0, 0, 0, 11, 0, 0, 0,
0, 0, 0, 0, 0, 9, 0, 0,
0, 13, 29, 0, 0, 5, 5, 5,
5, 0, 0, 0, 0, 0, 65, 0,
0, 0, 0, 0, 0, 0, 5, 39,
41, 19, 0, 0, 0, 33, 31, 0,
0, 37, 27, 23, 17, 21, 45, 25,
0, 35, 0, 53, 0, 0, 0, 0,
0, 0, 59, 0, 0, 0, 47, 62,
0, 50, 56, 0, 0,
}
var _expression_to_state_actions []byte = []byte{
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 1,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0,
}
var _expression_from_state_actions []byte = []byte{
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 3,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0,
}
var _expression_eof_trans []int16 = []int16{
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 0,
48, 49, 49, 49, 49, 49, 54, 55,
49, 58, 54, 49, 49, 49, 63, 66,
66, 66, 66, 66, 66, 66, 66, 66,
66, 66, 66, 66, 66, 66, 66, 66,
66, 66, 66, 49,
}
const expression_start int = 23
const expression_first_final int = 23
const expression_error int = -1
const expression_en_main int = 23
//line scanner.rl:11
type lexer struct {
parseValue
data []byte
p, pe, cs int
ts, te, act int
}
func (l *lexer) token() string {
return string(l.data[l.ts:l.te])
}
func newLexer(data []byte) *lexer {
lex := &lexer{
data: data,
pe: len(data),
}
//line scanner.go:238
{
lex.cs = expression_start
lex.ts = 0
lex.te = 0
lex.act = 0
}
//line scanner.rl:30
return lex
}
func (lex *lexer) Lex(out *yySymType) int {
eof := lex.pe
tok := 0
//line scanner.go:255
{
var _klen int
var _trans int
var _acts int
var _nacts uint
var _keys int
if (lex.p) == (lex.pe) {
goto _test_eof
}
_resume:
_acts = int(_expression_from_state_actions[lex.cs])
_nacts = uint(_expression_actions[_acts])
_acts++
for ; _nacts > 0; _nacts-- {
_acts++
switch _expression_actions[_acts-1] {
case 1:
//line NONE:1
lex.ts = (lex.p)
//line scanner.go:275
}
}
_keys = int(_expression_key_offsets[lex.cs])
_trans = int(_expression_index_offsets[lex.cs])
_klen = int(_expression_single_lengths[lex.cs])
if _klen > 0 {
_lower := int(_keys)
var _mid int
_upper := int(_keys + _klen - 1)
for {
if _upper < _lower {
break
}
_mid = _lower + ((_upper - _lower) >> 1)
switch {
case lex.data[(lex.p)] < _expression_trans_keys[_mid]:
_upper = _mid - 1
case lex.data[(lex.p)] > _expression_trans_keys[_mid]:
_lower = _mid + 1
default:
_trans += int(_mid - int(_keys))
goto _match
}
}
_keys += _klen
_trans += _klen
}
_klen = int(_expression_range_lengths[lex.cs])
if _klen > 0 {
_lower := int(_keys)
var _mid int
_upper := int(_keys + (_klen << 1) - 2)
for {
if _upper < _lower {
break
}
_mid = _lower + (((_upper - _lower) >> 1) & ^1)
switch {
case lex.data[(lex.p)] < _expression_trans_keys[_mid]:
_upper = _mid - 2
case lex.data[(lex.p)] > _expression_trans_keys[_mid+1]:
_lower = _mid + 2
default:
_trans += int((_mid - int(_keys)) >> 1)
goto _match
}
}
_trans += _klen
}
_match:
_trans = int(_expression_indicies[_trans])
_eof_trans:
lex.cs = int(_expression_trans_targs[_trans])
if _expression_trans_actions[_trans] == 0 {
goto _again
}
_acts = int(_expression_trans_actions[_trans])
_nacts = uint(_expression_actions[_acts])
_acts++
for ; _nacts > 0; _nacts-- {
_acts++
switch _expression_actions[_acts-1] {
case 2:
//line NONE:1
lex.te = (lex.p) + 1
case 3:
//line scanner.rl:38
lex.act = 8
case 4:
//line scanner.rl:95
lex.act = 9
case 5:
//line scanner.rl:102
lex.act = 14
case 6:
//line scanner.rl:103
lex.act = 15
case 7:
//line scanner.rl:104
lex.act = 16
case 8:
//line scanner.rl:107
lex.act = 17
case 9:
//line scanner.rl:43
lex.act = 19
case 10:
//line scanner.rl:83
lex.te = (lex.p) + 1
{
tok = ASSIGN
(lex.p)++
goto _out
}
case 11:
//line scanner.rl:84
lex.te = (lex.p) + 1
{
tok = CYCLE
(lex.p)++
goto _out
}
case 12:
//line scanner.rl:85
lex.te = (lex.p) + 1
{
tok = LOOP
(lex.p)++
goto _out
}
case 13:
//line scanner.rl:86
lex.te = (lex.p) + 1
{
tok = WHEN
(lex.p)++
goto _out
}
case 14:
//line scanner.rl:66
lex.te = (lex.p) + 1
{
tok = LITERAL
// TODO unescape \x
out.val = string(lex.data[lex.ts+1 : lex.te-1])
(lex.p)++
goto _out
}
case 15:
//line scanner.rl:98
lex.te = (lex.p) + 1
{
tok = EQ
(lex.p)++
goto _out
}
case 16:
//line scanner.rl:99
lex.te = (lex.p) + 1
{
tok = NEQ
(lex.p)++
goto _out
}
case 17:
//line scanner.rl:100
lex.te = (lex.p) + 1
{
tok = GE
(lex.p)++
goto _out
}
case 18:
//line scanner.rl:101
lex.te = (lex.p) + 1
{
tok = LE
(lex.p)++
goto _out
}
case 19:
//line scanner.rl:109
lex.te = (lex.p) + 1
{
tok = KEYWORD
out.name = string(lex.data[lex.ts : lex.te-1])
(lex.p)++
goto _out
}
case 20:
//line scanner.rl:111
lex.te = (lex.p) + 1
{
tok = PROPERTY
out.name = string(lex.data[lex.ts+1 : lex.te])
(lex.p)++
goto _out
}
case 21:
//line scanner.rl:114
lex.te = (lex.p) + 1
{
tok = int(lex.data[lex.ts])
(lex.p)++
goto _out
}
case 22:
//line scanner.rl:48
lex.te = (lex.p)
(lex.p)--
{
tok = LITERAL
n, err := strconv.ParseInt(lex.token(), 10, 64)
if err != nil {
panic(err)
}
out.val = int(n)
(lex.p)++
goto _out
}
case 23:
//line scanner.rl:57
lex.te = (lex.p)
(lex.p)--
{
tok = LITERAL
n, err := strconv.ParseFloat(lex.token(), 64)
if err != nil {
panic(err)
}
out.val = n
(lex.p)++
goto _out
}
case 24:
//line scanner.rl:43
lex.te = (lex.p)
(lex.p)--
{
tok = IDENTIFIER
out.name = lex.token()
(lex.p)++
goto _out
}
case 25:
//line scanner.rl:111
lex.te = (lex.p)
(lex.p)--
{
tok = PROPERTY
out.name = string(lex.data[lex.ts+1 : lex.te])
(lex.p)++
goto _out
}
case 26:
//line scanner.rl:113
lex.te = (lex.p)
(lex.p)--
case 27:
//line scanner.rl:114
lex.te = (lex.p)
(lex.p)--
{
tok = int(lex.data[lex.ts])
(lex.p)++
goto _out
}
case 28:
//line scanner.rl:114
(lex.p) = (lex.te) - 1
{
tok = int(lex.data[lex.ts])
(lex.p)++
goto _out
}
case 29:
//line NONE:1
switch lex.act {
case 8:
{
(lex.p) = (lex.te) - 1
tok = LITERAL
out.val = lex.token() == "true"
(lex.p)++
goto _out
}
case 9:
{
(lex.p) = (lex.te) - 1
tok = LITERAL
out.val = nil
(lex.p)++
goto _out
}
case 14:
{
(lex.p) = (lex.te) - 1
tok = AND
(lex.p)++
goto _out
}
case 15:
{
(lex.p) = (lex.te) - 1
tok = OR
(lex.p)++
goto _out
}
case 16:
{
(lex.p) = (lex.te) - 1
tok = CONTAINS
(lex.p)++
goto _out
}
case 17:
{
(lex.p) = (lex.te) - 1
tok = IN
(lex.p)++
goto _out
}
case 19:
{
(lex.p) = (lex.te) - 1
tok = IDENTIFIER
out.name = lex.token()
(lex.p)++
goto _out
}
}
//line scanner.go:536
}
}
_again:
_acts = int(_expression_to_state_actions[lex.cs])
_nacts = uint(_expression_actions[_acts])
_acts++
for ; _nacts > 0; _nacts-- {
_acts++
switch _expression_actions[_acts-1] {
case 0:
//line NONE:1
lex.ts = 0
//line scanner.go:550
}
}
(lex.p)++
if (lex.p) != (lex.pe) {
goto _resume
}
_test_eof:
{
}
if (lex.p) == eof {
if _expression_eof_trans[lex.cs] > 0 {
_trans = int(_expression_eof_trans[lex.cs] - 1)
goto _eof_trans
}
}
_out:
{
}
}
//line scanner.rl:118
return tok
}
func (lex *lexer) Error(e string) {
// fmt.Println("scan error:", e)
}