blob: e4e56e28e0e807e84ff6828a669c7b55a3c8d06d [file] [log] [blame]
Zack Williamse940c7a2019-08-21 14:25:39 -07001package yaml
2
3import (
4 "encoding"
5 "encoding/base64"
6 "fmt"
7 "io"
8 "math"
9 "reflect"
10 "strconv"
11 "time"
12)
13
14const (
15 documentNode = 1 << iota
16 mappingNode
17 sequenceNode
18 scalarNode
19 aliasNode
20)
21
22type node struct {
23 kind int
24 line, column int
25 tag string
26 // For an alias node, alias holds the resolved alias.
27 alias *node
28 value string
29 implicit bool
30 children []*node
31 anchors map[string]*node
32}
33
34// ----------------------------------------------------------------------------
35// Parser, produces a node tree out of a libyaml event stream.
36
37type parser struct {
38 parser yaml_parser_t
39 event yaml_event_t
40 doc *node
41 doneInit bool
42}
43
44func newParser(b []byte) *parser {
45 p := parser{}
46 if !yaml_parser_initialize(&p.parser) {
47 panic("failed to initialize YAML emitter")
48 }
49 if len(b) == 0 {
50 b = []byte{'\n'}
51 }
52 yaml_parser_set_input_string(&p.parser, b)
53 return &p
54}
55
56func newParserFromReader(r io.Reader) *parser {
57 p := parser{}
58 if !yaml_parser_initialize(&p.parser) {
59 panic("failed to initialize YAML emitter")
60 }
61 yaml_parser_set_input_reader(&p.parser, r)
62 return &p
63}
64
65func (p *parser) init() {
66 if p.doneInit {
67 return
68 }
69 p.expect(yaml_STREAM_START_EVENT)
70 p.doneInit = true
71}
72
73func (p *parser) destroy() {
74 if p.event.typ != yaml_NO_EVENT {
75 yaml_event_delete(&p.event)
76 }
77 yaml_parser_delete(&p.parser)
78}
79
80// expect consumes an event from the event stream and
81// checks that it's of the expected type.
82func (p *parser) expect(e yaml_event_type_t) {
83 if p.event.typ == yaml_NO_EVENT {
84 if !yaml_parser_parse(&p.parser, &p.event) {
85 p.fail()
86 }
87 }
88 if p.event.typ == yaml_STREAM_END_EVENT {
89 failf("attempted to go past the end of stream; corrupted value?")
90 }
91 if p.event.typ != e {
92 p.parser.problem = fmt.Sprintf("expected %s event but got %s", e, p.event.typ)
93 p.fail()
94 }
95 yaml_event_delete(&p.event)
96 p.event.typ = yaml_NO_EVENT
97}
98
99// peek peeks at the next event in the event stream,
100// puts the results into p.event and returns the event type.
101func (p *parser) peek() yaml_event_type_t {
102 if p.event.typ != yaml_NO_EVENT {
103 return p.event.typ
104 }
105 if !yaml_parser_parse(&p.parser, &p.event) {
106 p.fail()
107 }
108 return p.event.typ
109}
110
111func (p *parser) fail() {
112 var where string
113 var line int
114 if p.parser.problem_mark.line != 0 {
115 line = p.parser.problem_mark.line
116 // Scanner errors don't iterate line before returning error
117 if p.parser.error == yaml_SCANNER_ERROR {
118 line++
119 }
120 } else if p.parser.context_mark.line != 0 {
121 line = p.parser.context_mark.line
122 }
123 if line != 0 {
124 where = "line " + strconv.Itoa(line) + ": "
125 }
126 var msg string
127 if len(p.parser.problem) > 0 {
128 msg = p.parser.problem
129 } else {
130 msg = "unknown problem parsing YAML content"
131 }
132 failf("%s%s", where, msg)
133}
134
135func (p *parser) anchor(n *node, anchor []byte) {
136 if anchor != nil {
137 p.doc.anchors[string(anchor)] = n
138 }
139}
140
141func (p *parser) parse() *node {
142 p.init()
143 switch p.peek() {
144 case yaml_SCALAR_EVENT:
145 return p.scalar()
146 case yaml_ALIAS_EVENT:
147 return p.alias()
148 case yaml_MAPPING_START_EVENT:
149 return p.mapping()
150 case yaml_SEQUENCE_START_EVENT:
151 return p.sequence()
152 case yaml_DOCUMENT_START_EVENT:
153 return p.document()
154 case yaml_STREAM_END_EVENT:
155 // Happens when attempting to decode an empty buffer.
156 return nil
157 default:
158 panic("attempted to parse unknown event: " + p.event.typ.String())
159 }
160}
161
162func (p *parser) node(kind int) *node {
163 return &node{
164 kind: kind,
165 line: p.event.start_mark.line,
166 column: p.event.start_mark.column,
167 }
168}
169
170func (p *parser) document() *node {
171 n := p.node(documentNode)
172 n.anchors = make(map[string]*node)
173 p.doc = n
174 p.expect(yaml_DOCUMENT_START_EVENT)
175 n.children = append(n.children, p.parse())
176 p.expect(yaml_DOCUMENT_END_EVENT)
177 return n
178}
179
180func (p *parser) alias() *node {
181 n := p.node(aliasNode)
182 n.value = string(p.event.anchor)
183 n.alias = p.doc.anchors[n.value]
184 if n.alias == nil {
185 failf("unknown anchor '%s' referenced", n.value)
186 }
187 p.expect(yaml_ALIAS_EVENT)
188 return n
189}
190
191func (p *parser) scalar() *node {
192 n := p.node(scalarNode)
193 n.value = string(p.event.value)
194 n.tag = string(p.event.tag)
195 n.implicit = p.event.implicit
196 p.anchor(n, p.event.anchor)
197 p.expect(yaml_SCALAR_EVENT)
198 return n
199}
200
201func (p *parser) sequence() *node {
202 n := p.node(sequenceNode)
203 p.anchor(n, p.event.anchor)
204 p.expect(yaml_SEQUENCE_START_EVENT)
205 for p.peek() != yaml_SEQUENCE_END_EVENT {
206 n.children = append(n.children, p.parse())
207 }
208 p.expect(yaml_SEQUENCE_END_EVENT)
209 return n
210}
211
212func (p *parser) mapping() *node {
213 n := p.node(mappingNode)
214 p.anchor(n, p.event.anchor)
215 p.expect(yaml_MAPPING_START_EVENT)
216 for p.peek() != yaml_MAPPING_END_EVENT {
217 n.children = append(n.children, p.parse(), p.parse())
218 }
219 p.expect(yaml_MAPPING_END_EVENT)
220 return n
221}
222
223// ----------------------------------------------------------------------------
224// Decoder, unmarshals a node into a provided value.
225
226type decoder struct {
227 doc *node
228 aliases map[*node]bool
229 mapType reflect.Type
230 terrors []string
231 strict bool
232}
233
234var (
235 mapItemType = reflect.TypeOf(MapItem{})
236 durationType = reflect.TypeOf(time.Duration(0))
237 defaultMapType = reflect.TypeOf(map[interface{}]interface{}{})
238 ifaceType = defaultMapType.Elem()
239 timeType = reflect.TypeOf(time.Time{})
240 ptrTimeType = reflect.TypeOf(&time.Time{})
241)
242
243func newDecoder(strict bool) *decoder {
244 d := &decoder{mapType: defaultMapType, strict: strict}
245 d.aliases = make(map[*node]bool)
246 return d
247}
248
249func (d *decoder) terror(n *node, tag string, out reflect.Value) {
250 if n.tag != "" {
251 tag = n.tag
252 }
253 value := n.value
254 if tag != yaml_SEQ_TAG && tag != yaml_MAP_TAG {
255 if len(value) > 10 {
256 value = " `" + value[:7] + "...`"
257 } else {
258 value = " `" + value + "`"
259 }
260 }
261 d.terrors = append(d.terrors, fmt.Sprintf("line %d: cannot unmarshal %s%s into %s", n.line+1, shortTag(tag), value, out.Type()))
262}
263
264func (d *decoder) callUnmarshaler(n *node, u Unmarshaler) (good bool) {
265 terrlen := len(d.terrors)
266 err := u.UnmarshalYAML(func(v interface{}) (err error) {
267 defer handleErr(&err)
268 d.unmarshal(n, reflect.ValueOf(v))
269 if len(d.terrors) > terrlen {
270 issues := d.terrors[terrlen:]
271 d.terrors = d.terrors[:terrlen]
272 return &TypeError{issues}
273 }
274 return nil
275 })
276 if e, ok := err.(*TypeError); ok {
277 d.terrors = append(d.terrors, e.Errors...)
278 return false
279 }
280 if err != nil {
281 fail(err)
282 }
283 return true
284}
285
286// d.prepare initializes and dereferences pointers and calls UnmarshalYAML
287// if a value is found to implement it.
288// It returns the initialized and dereferenced out value, whether
289// unmarshalling was already done by UnmarshalYAML, and if so whether
290// its types unmarshalled appropriately.
291//
292// If n holds a null value, prepare returns before doing anything.
293func (d *decoder) prepare(n *node, out reflect.Value) (newout reflect.Value, unmarshaled, good bool) {
294 if n.tag == yaml_NULL_TAG || n.kind == scalarNode && n.tag == "" && (n.value == "null" || n.value == "~" || n.value == "" && n.implicit) {
295 return out, false, false
296 }
297 again := true
298 for again {
299 again = false
300 if out.Kind() == reflect.Ptr {
301 if out.IsNil() {
302 out.Set(reflect.New(out.Type().Elem()))
303 }
304 out = out.Elem()
305 again = true
306 }
307 if out.CanAddr() {
308 if u, ok := out.Addr().Interface().(Unmarshaler); ok {
309 good = d.callUnmarshaler(n, u)
310 return out, true, good
311 }
312 }
313 }
314 return out, false, false
315}
316
317func (d *decoder) unmarshal(n *node, out reflect.Value) (good bool) {
318 switch n.kind {
319 case documentNode:
320 return d.document(n, out)
321 case aliasNode:
322 return d.alias(n, out)
323 }
324 out, unmarshaled, good := d.prepare(n, out)
325 if unmarshaled {
326 return good
327 }
328 switch n.kind {
329 case scalarNode:
330 good = d.scalar(n, out)
331 case mappingNode:
332 good = d.mapping(n, out)
333 case sequenceNode:
334 good = d.sequence(n, out)
335 default:
336 panic("internal error: unknown node kind: " + strconv.Itoa(n.kind))
337 }
338 return good
339}
340
341func (d *decoder) document(n *node, out reflect.Value) (good bool) {
342 if len(n.children) == 1 {
343 d.doc = n
344 d.unmarshal(n.children[0], out)
345 return true
346 }
347 return false
348}
349
350func (d *decoder) alias(n *node, out reflect.Value) (good bool) {
351 if d.aliases[n] {
352 // TODO this could actually be allowed in some circumstances.
353 failf("anchor '%s' value contains itself", n.value)
354 }
355 d.aliases[n] = true
356 good = d.unmarshal(n.alias, out)
357 delete(d.aliases, n)
358 return good
359}
360
361var zeroValue reflect.Value
362
363func resetMap(out reflect.Value) {
364 for _, k := range out.MapKeys() {
365 out.SetMapIndex(k, zeroValue)
366 }
367}
368
369func (d *decoder) scalar(n *node, out reflect.Value) bool {
370 var tag string
371 var resolved interface{}
372 if n.tag == "" && !n.implicit {
373 tag = yaml_STR_TAG
374 resolved = n.value
375 } else {
376 tag, resolved = resolve(n.tag, n.value)
377 if tag == yaml_BINARY_TAG {
378 data, err := base64.StdEncoding.DecodeString(resolved.(string))
379 if err != nil {
380 failf("!!binary value contains invalid base64 data")
381 }
382 resolved = string(data)
383 }
384 }
385 if resolved == nil {
386 if out.Kind() == reflect.Map && !out.CanAddr() {
387 resetMap(out)
388 } else {
389 out.Set(reflect.Zero(out.Type()))
390 }
391 return true
392 }
393 if resolvedv := reflect.ValueOf(resolved); out.Type() == resolvedv.Type() {
394 // We've resolved to exactly the type we want, so use that.
395 out.Set(resolvedv)
396 return true
397 }
398 // Perhaps we can use the value as a TextUnmarshaler to
399 // set its value.
400 if out.CanAddr() {
401 u, ok := out.Addr().Interface().(encoding.TextUnmarshaler)
402 if ok {
403 var text []byte
404 if tag == yaml_BINARY_TAG {
405 text = []byte(resolved.(string))
406 } else {
407 // We let any value be unmarshaled into TextUnmarshaler.
408 // That might be more lax than we'd like, but the
409 // TextUnmarshaler itself should bowl out any dubious values.
410 text = []byte(n.value)
411 }
412 err := u.UnmarshalText(text)
413 if err != nil {
414 fail(err)
415 }
416 return true
417 }
418 }
419 switch out.Kind() {
420 case reflect.String:
421 if tag == yaml_BINARY_TAG {
422 out.SetString(resolved.(string))
423 return true
424 }
425 if resolved != nil {
426 out.SetString(n.value)
427 return true
428 }
429 case reflect.Interface:
430 if resolved == nil {
431 out.Set(reflect.Zero(out.Type()))
432 } else if tag == yaml_TIMESTAMP_TAG {
433 // It looks like a timestamp but for backward compatibility
434 // reasons we set it as a string, so that code that unmarshals
435 // timestamp-like values into interface{} will continue to
436 // see a string and not a time.Time.
437 // TODO(v3) Drop this.
438 out.Set(reflect.ValueOf(n.value))
439 } else {
440 out.Set(reflect.ValueOf(resolved))
441 }
442 return true
443 case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
444 switch resolved := resolved.(type) {
445 case int:
446 if !out.OverflowInt(int64(resolved)) {
447 out.SetInt(int64(resolved))
448 return true
449 }
450 case int64:
451 if !out.OverflowInt(resolved) {
452 out.SetInt(resolved)
453 return true
454 }
455 case uint64:
456 if resolved <= math.MaxInt64 && !out.OverflowInt(int64(resolved)) {
457 out.SetInt(int64(resolved))
458 return true
459 }
460 case float64:
461 if resolved <= math.MaxInt64 && !out.OverflowInt(int64(resolved)) {
462 out.SetInt(int64(resolved))
463 return true
464 }
465 case string:
466 if out.Type() == durationType {
467 d, err := time.ParseDuration(resolved)
468 if err == nil {
469 out.SetInt(int64(d))
470 return true
471 }
472 }
473 }
474 case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
475 switch resolved := resolved.(type) {
476 case int:
477 if resolved >= 0 && !out.OverflowUint(uint64(resolved)) {
478 out.SetUint(uint64(resolved))
479 return true
480 }
481 case int64:
482 if resolved >= 0 && !out.OverflowUint(uint64(resolved)) {
483 out.SetUint(uint64(resolved))
484 return true
485 }
486 case uint64:
487 if !out.OverflowUint(uint64(resolved)) {
488 out.SetUint(uint64(resolved))
489 return true
490 }
491 case float64:
492 if resolved <= math.MaxUint64 && !out.OverflowUint(uint64(resolved)) {
493 out.SetUint(uint64(resolved))
494 return true
495 }
496 }
497 case reflect.Bool:
498 switch resolved := resolved.(type) {
499 case bool:
500 out.SetBool(resolved)
501 return true
502 }
503 case reflect.Float32, reflect.Float64:
504 switch resolved := resolved.(type) {
505 case int:
506 out.SetFloat(float64(resolved))
507 return true
508 case int64:
509 out.SetFloat(float64(resolved))
510 return true
511 case uint64:
512 out.SetFloat(float64(resolved))
513 return true
514 case float64:
515 out.SetFloat(resolved)
516 return true
517 }
518 case reflect.Struct:
519 if resolvedv := reflect.ValueOf(resolved); out.Type() == resolvedv.Type() {
520 out.Set(resolvedv)
521 return true
522 }
523 case reflect.Ptr:
524 if out.Type().Elem() == reflect.TypeOf(resolved) {
525 // TODO DOes this make sense? When is out a Ptr except when decoding a nil value?
526 elem := reflect.New(out.Type().Elem())
527 elem.Elem().Set(reflect.ValueOf(resolved))
528 out.Set(elem)
529 return true
530 }
531 }
532 d.terror(n, tag, out)
533 return false
534}
535
536func settableValueOf(i interface{}) reflect.Value {
537 v := reflect.ValueOf(i)
538 sv := reflect.New(v.Type()).Elem()
539 sv.Set(v)
540 return sv
541}
542
543func (d *decoder) sequence(n *node, out reflect.Value) (good bool) {
544 l := len(n.children)
545
546 var iface reflect.Value
547 switch out.Kind() {
548 case reflect.Slice:
549 out.Set(reflect.MakeSlice(out.Type(), l, l))
550 case reflect.Array:
551 if l != out.Len() {
552 failf("invalid array: want %d elements but got %d", out.Len(), l)
553 }
554 case reflect.Interface:
555 // No type hints. Will have to use a generic sequence.
556 iface = out
557 out = settableValueOf(make([]interface{}, l))
558 default:
559 d.terror(n, yaml_SEQ_TAG, out)
560 return false
561 }
562 et := out.Type().Elem()
563
564 j := 0
565 for i := 0; i < l; i++ {
566 e := reflect.New(et).Elem()
567 if ok := d.unmarshal(n.children[i], e); ok {
568 out.Index(j).Set(e)
569 j++
570 }
571 }
572 if out.Kind() != reflect.Array {
573 out.Set(out.Slice(0, j))
574 }
575 if iface.IsValid() {
576 iface.Set(out)
577 }
578 return true
579}
580
581func (d *decoder) mapping(n *node, out reflect.Value) (good bool) {
582 switch out.Kind() {
583 case reflect.Struct:
584 return d.mappingStruct(n, out)
585 case reflect.Slice:
586 return d.mappingSlice(n, out)
587 case reflect.Map:
588 // okay
589 case reflect.Interface:
590 if d.mapType.Kind() == reflect.Map {
591 iface := out
592 out = reflect.MakeMap(d.mapType)
593 iface.Set(out)
594 } else {
595 slicev := reflect.New(d.mapType).Elem()
596 if !d.mappingSlice(n, slicev) {
597 return false
598 }
599 out.Set(slicev)
600 return true
601 }
602 default:
603 d.terror(n, yaml_MAP_TAG, out)
604 return false
605 }
606 outt := out.Type()
607 kt := outt.Key()
608 et := outt.Elem()
609
610 mapType := d.mapType
611 if outt.Key() == ifaceType && outt.Elem() == ifaceType {
612 d.mapType = outt
613 }
614
615 if out.IsNil() {
616 out.Set(reflect.MakeMap(outt))
617 }
618 l := len(n.children)
619 for i := 0; i < l; i += 2 {
620 if isMerge(n.children[i]) {
621 d.merge(n.children[i+1], out)
622 continue
623 }
624 k := reflect.New(kt).Elem()
625 if d.unmarshal(n.children[i], k) {
626 kkind := k.Kind()
627 if kkind == reflect.Interface {
628 kkind = k.Elem().Kind()
629 }
630 if kkind == reflect.Map || kkind == reflect.Slice {
631 failf("invalid map key: %#v", k.Interface())
632 }
633 e := reflect.New(et).Elem()
634 if d.unmarshal(n.children[i+1], e) {
635 d.setMapIndex(n.children[i+1], out, k, e)
636 }
637 }
638 }
639 d.mapType = mapType
640 return true
641}
642
643func (d *decoder) setMapIndex(n *node, out, k, v reflect.Value) {
644 if d.strict && out.MapIndex(k) != zeroValue {
645 d.terrors = append(d.terrors, fmt.Sprintf("line %d: key %#v already set in map", n.line+1, k.Interface()))
646 return
647 }
648 out.SetMapIndex(k, v)
649}
650
651func (d *decoder) mappingSlice(n *node, out reflect.Value) (good bool) {
652 outt := out.Type()
653 if outt.Elem() != mapItemType {
654 d.terror(n, yaml_MAP_TAG, out)
655 return false
656 }
657
658 mapType := d.mapType
659 d.mapType = outt
660
661 var slice []MapItem
662 var l = len(n.children)
663 for i := 0; i < l; i += 2 {
664 if isMerge(n.children[i]) {
665 d.merge(n.children[i+1], out)
666 continue
667 }
668 item := MapItem{}
669 k := reflect.ValueOf(&item.Key).Elem()
670 if d.unmarshal(n.children[i], k) {
671 v := reflect.ValueOf(&item.Value).Elem()
672 if d.unmarshal(n.children[i+1], v) {
673 slice = append(slice, item)
674 }
675 }
676 }
677 out.Set(reflect.ValueOf(slice))
678 d.mapType = mapType
679 return true
680}
681
682func (d *decoder) mappingStruct(n *node, out reflect.Value) (good bool) {
683 sinfo, err := getStructInfo(out.Type())
684 if err != nil {
685 panic(err)
686 }
687 name := settableValueOf("")
688 l := len(n.children)
689
690 var inlineMap reflect.Value
691 var elemType reflect.Type
692 if sinfo.InlineMap != -1 {
693 inlineMap = out.Field(sinfo.InlineMap)
694 inlineMap.Set(reflect.New(inlineMap.Type()).Elem())
695 elemType = inlineMap.Type().Elem()
696 }
697
698 var doneFields []bool
699 if d.strict {
700 doneFields = make([]bool, len(sinfo.FieldsList))
701 }
702 for i := 0; i < l; i += 2 {
703 ni := n.children[i]
704 if isMerge(ni) {
705 d.merge(n.children[i+1], out)
706 continue
707 }
708 if !d.unmarshal(ni, name) {
709 continue
710 }
711 if info, ok := sinfo.FieldsMap[name.String()]; ok {
712 if d.strict {
713 if doneFields[info.Id] {
714 d.terrors = append(d.terrors, fmt.Sprintf("line %d: field %s already set in type %s", ni.line+1, name.String(), out.Type()))
715 continue
716 }
717 doneFields[info.Id] = true
718 }
719 var field reflect.Value
720 if info.Inline == nil {
721 field = out.Field(info.Num)
722 } else {
723 field = out.FieldByIndex(info.Inline)
724 }
725 d.unmarshal(n.children[i+1], field)
726 } else if sinfo.InlineMap != -1 {
727 if inlineMap.IsNil() {
728 inlineMap.Set(reflect.MakeMap(inlineMap.Type()))
729 }
730 value := reflect.New(elemType).Elem()
731 d.unmarshal(n.children[i+1], value)
732 d.setMapIndex(n.children[i+1], inlineMap, name, value)
733 } else if d.strict {
734 d.terrors = append(d.terrors, fmt.Sprintf("line %d: field %s not found in type %s", ni.line+1, name.String(), out.Type()))
735 }
736 }
737 return true
738}
739
740func failWantMap() {
741 failf("map merge requires map or sequence of maps as the value")
742}
743
744func (d *decoder) merge(n *node, out reflect.Value) {
745 switch n.kind {
746 case mappingNode:
747 d.unmarshal(n, out)
748 case aliasNode:
749 an, ok := d.doc.anchors[n.value]
750 if ok && an.kind != mappingNode {
751 failWantMap()
752 }
753 d.unmarshal(n, out)
754 case sequenceNode:
755 // Step backwards as earlier nodes take precedence.
756 for i := len(n.children) - 1; i >= 0; i-- {
757 ni := n.children[i]
758 if ni.kind == aliasNode {
759 an, ok := d.doc.anchors[ni.value]
760 if ok && an.kind != mappingNode {
761 failWantMap()
762 }
763 } else if ni.kind != mappingNode {
764 failWantMap()
765 }
766 d.unmarshal(ni, out)
767 }
768 default:
769 failWantMap()
770 }
771}
772
773func isMerge(n *node) bool {
774 return n.kind == scalarNode && n.value == "<<" && (n.implicit == true || n.tag == yaml_MERGE_TAG)
775}