David K. Bainbridge | e05cf0c | 2021-08-19 03:16:50 +0000 | [diff] [blame] | 1 | // |
| 2 | // Copyright (c) 2011-2019 Canonical Ltd |
| 3 | // |
| 4 | // Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | // you may not use this file except in compliance with the License. |
| 6 | // You may obtain a copy of the License at |
| 7 | // |
| 8 | // http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | // |
| 10 | // Unless required by applicable law or agreed to in writing, software |
| 11 | // distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | // See the License for the specific language governing permissions and |
| 14 | // limitations under the License. |
| 15 | |
| 16 | package yaml |
| 17 | |
| 18 | import ( |
| 19 | "encoding" |
| 20 | "encoding/base64" |
| 21 | "fmt" |
| 22 | "io" |
| 23 | "math" |
| 24 | "reflect" |
| 25 | "strconv" |
| 26 | "time" |
| 27 | ) |
| 28 | |
| 29 | // ---------------------------------------------------------------------------- |
| 30 | // Parser, produces a node tree out of a libyaml event stream. |
| 31 | |
| 32 | type parser struct { |
| 33 | parser yaml_parser_t |
| 34 | event yaml_event_t |
| 35 | doc *Node |
| 36 | anchors map[string]*Node |
| 37 | doneInit bool |
| 38 | textless bool |
| 39 | } |
| 40 | |
| 41 | func newParser(b []byte) *parser { |
| 42 | p := parser{} |
| 43 | if !yaml_parser_initialize(&p.parser) { |
| 44 | panic("failed to initialize YAML emitter") |
| 45 | } |
| 46 | if len(b) == 0 { |
| 47 | b = []byte{'\n'} |
| 48 | } |
| 49 | yaml_parser_set_input_string(&p.parser, b) |
| 50 | return &p |
| 51 | } |
| 52 | |
| 53 | func newParserFromReader(r io.Reader) *parser { |
| 54 | p := parser{} |
| 55 | if !yaml_parser_initialize(&p.parser) { |
| 56 | panic("failed to initialize YAML emitter") |
| 57 | } |
| 58 | yaml_parser_set_input_reader(&p.parser, r) |
| 59 | return &p |
| 60 | } |
| 61 | |
| 62 | func (p *parser) init() { |
| 63 | if p.doneInit { |
| 64 | return |
| 65 | } |
| 66 | p.anchors = make(map[string]*Node) |
| 67 | p.expect(yaml_STREAM_START_EVENT) |
| 68 | p.doneInit = true |
| 69 | } |
| 70 | |
| 71 | func (p *parser) destroy() { |
| 72 | if p.event.typ != yaml_NO_EVENT { |
| 73 | yaml_event_delete(&p.event) |
| 74 | } |
| 75 | yaml_parser_delete(&p.parser) |
| 76 | } |
| 77 | |
| 78 | // expect consumes an event from the event stream and |
| 79 | // checks that it's of the expected type. |
| 80 | func (p *parser) expect(e yaml_event_type_t) { |
| 81 | if p.event.typ == yaml_NO_EVENT { |
| 82 | if !yaml_parser_parse(&p.parser, &p.event) { |
| 83 | p.fail() |
| 84 | } |
| 85 | } |
| 86 | if p.event.typ == yaml_STREAM_END_EVENT { |
| 87 | failf("attempted to go past the end of stream; corrupted value?") |
| 88 | } |
| 89 | if p.event.typ != e { |
| 90 | p.parser.problem = fmt.Sprintf("expected %s event but got %s", e, p.event.typ) |
| 91 | p.fail() |
| 92 | } |
| 93 | yaml_event_delete(&p.event) |
| 94 | p.event.typ = yaml_NO_EVENT |
| 95 | } |
| 96 | |
| 97 | // peek peeks at the next event in the event stream, |
| 98 | // puts the results into p.event and returns the event type. |
| 99 | func (p *parser) peek() yaml_event_type_t { |
| 100 | if p.event.typ != yaml_NO_EVENT { |
| 101 | return p.event.typ |
| 102 | } |
| 103 | if !yaml_parser_parse(&p.parser, &p.event) { |
| 104 | p.fail() |
| 105 | } |
| 106 | return p.event.typ |
| 107 | } |
| 108 | |
| 109 | func (p *parser) fail() { |
| 110 | var where string |
| 111 | var line int |
| 112 | if p.parser.context_mark.line != 0 { |
| 113 | line = p.parser.context_mark.line |
| 114 | // Scanner errors don't iterate line before returning error |
| 115 | if p.parser.error == yaml_SCANNER_ERROR { |
| 116 | line++ |
| 117 | } |
| 118 | } else if p.parser.problem_mark.line != 0 { |
| 119 | line = p.parser.problem_mark.line |
| 120 | // Scanner errors don't iterate line before returning error |
| 121 | if p.parser.error == yaml_SCANNER_ERROR { |
| 122 | line++ |
| 123 | } |
| 124 | } |
| 125 | if line != 0 { |
| 126 | where = "line " + strconv.Itoa(line) + ": " |
| 127 | } |
| 128 | var msg string |
| 129 | if len(p.parser.problem) > 0 { |
| 130 | msg = p.parser.problem |
| 131 | } else { |
| 132 | msg = "unknown problem parsing YAML content" |
| 133 | } |
| 134 | failf("%s%s", where, msg) |
| 135 | } |
| 136 | |
| 137 | func (p *parser) anchor(n *Node, anchor []byte) { |
| 138 | if anchor != nil { |
| 139 | n.Anchor = string(anchor) |
| 140 | p.anchors[n.Anchor] = n |
| 141 | } |
| 142 | } |
| 143 | |
| 144 | func (p *parser) parse() *Node { |
| 145 | p.init() |
| 146 | switch p.peek() { |
| 147 | case yaml_SCALAR_EVENT: |
| 148 | return p.scalar() |
| 149 | case yaml_ALIAS_EVENT: |
| 150 | return p.alias() |
| 151 | case yaml_MAPPING_START_EVENT: |
| 152 | return p.mapping() |
| 153 | case yaml_SEQUENCE_START_EVENT: |
| 154 | return p.sequence() |
| 155 | case yaml_DOCUMENT_START_EVENT: |
| 156 | return p.document() |
| 157 | case yaml_STREAM_END_EVENT: |
| 158 | // Happens when attempting to decode an empty buffer. |
| 159 | return nil |
| 160 | case yaml_TAIL_COMMENT_EVENT: |
| 161 | panic("internal error: unexpected tail comment event (please report)") |
| 162 | default: |
| 163 | panic("internal error: attempted to parse unknown event (please report): " + p.event.typ.String()) |
| 164 | } |
| 165 | } |
| 166 | |
| 167 | func (p *parser) node(kind Kind, defaultTag, tag, value string) *Node { |
| 168 | var style Style |
| 169 | if tag != "" && tag != "!" { |
| 170 | tag = shortTag(tag) |
| 171 | style = TaggedStyle |
| 172 | } else if defaultTag != "" { |
| 173 | tag = defaultTag |
| 174 | } else if kind == ScalarNode { |
| 175 | tag, _ = resolve("", value) |
| 176 | } |
| 177 | n := &Node{ |
| 178 | Kind: kind, |
| 179 | Tag: tag, |
| 180 | Value: value, |
| 181 | Style: style, |
| 182 | } |
| 183 | if !p.textless { |
| 184 | n.Line = p.event.start_mark.line + 1 |
| 185 | n.Column = p.event.start_mark.column + 1 |
| 186 | n.HeadComment = string(p.event.head_comment) |
| 187 | n.LineComment = string(p.event.line_comment) |
| 188 | n.FootComment = string(p.event.foot_comment) |
| 189 | } |
| 190 | return n |
| 191 | } |
| 192 | |
| 193 | func (p *parser) parseChild(parent *Node) *Node { |
| 194 | child := p.parse() |
| 195 | parent.Content = append(parent.Content, child) |
| 196 | return child |
| 197 | } |
| 198 | |
| 199 | func (p *parser) document() *Node { |
| 200 | n := p.node(DocumentNode, "", "", "") |
| 201 | p.doc = n |
| 202 | p.expect(yaml_DOCUMENT_START_EVENT) |
| 203 | p.parseChild(n) |
| 204 | if p.peek() == yaml_DOCUMENT_END_EVENT { |
| 205 | n.FootComment = string(p.event.foot_comment) |
| 206 | } |
| 207 | p.expect(yaml_DOCUMENT_END_EVENT) |
| 208 | return n |
| 209 | } |
| 210 | |
| 211 | func (p *parser) alias() *Node { |
| 212 | n := p.node(AliasNode, "", "", string(p.event.anchor)) |
| 213 | n.Alias = p.anchors[n.Value] |
| 214 | if n.Alias == nil { |
| 215 | failf("unknown anchor '%s' referenced", n.Value) |
| 216 | } |
| 217 | p.expect(yaml_ALIAS_EVENT) |
| 218 | return n |
| 219 | } |
| 220 | |
| 221 | func (p *parser) scalar() *Node { |
| 222 | var parsedStyle = p.event.scalar_style() |
| 223 | var nodeStyle Style |
| 224 | switch { |
| 225 | case parsedStyle&yaml_DOUBLE_QUOTED_SCALAR_STYLE != 0: |
| 226 | nodeStyle = DoubleQuotedStyle |
| 227 | case parsedStyle&yaml_SINGLE_QUOTED_SCALAR_STYLE != 0: |
| 228 | nodeStyle = SingleQuotedStyle |
| 229 | case parsedStyle&yaml_LITERAL_SCALAR_STYLE != 0: |
| 230 | nodeStyle = LiteralStyle |
| 231 | case parsedStyle&yaml_FOLDED_SCALAR_STYLE != 0: |
| 232 | nodeStyle = FoldedStyle |
| 233 | } |
| 234 | var nodeValue = string(p.event.value) |
| 235 | var nodeTag = string(p.event.tag) |
| 236 | var defaultTag string |
| 237 | if nodeStyle == 0 { |
| 238 | if nodeValue == "<<" { |
| 239 | defaultTag = mergeTag |
| 240 | } |
| 241 | } else { |
| 242 | defaultTag = strTag |
| 243 | } |
| 244 | n := p.node(ScalarNode, defaultTag, nodeTag, nodeValue) |
| 245 | n.Style |= nodeStyle |
| 246 | p.anchor(n, p.event.anchor) |
| 247 | p.expect(yaml_SCALAR_EVENT) |
| 248 | return n |
| 249 | } |
| 250 | |
| 251 | func (p *parser) sequence() *Node { |
| 252 | n := p.node(SequenceNode, seqTag, string(p.event.tag), "") |
| 253 | if p.event.sequence_style()&yaml_FLOW_SEQUENCE_STYLE != 0 { |
| 254 | n.Style |= FlowStyle |
| 255 | } |
| 256 | p.anchor(n, p.event.anchor) |
| 257 | p.expect(yaml_SEQUENCE_START_EVENT) |
| 258 | for p.peek() != yaml_SEQUENCE_END_EVENT { |
| 259 | p.parseChild(n) |
| 260 | } |
| 261 | n.LineComment = string(p.event.line_comment) |
| 262 | n.FootComment = string(p.event.foot_comment) |
| 263 | p.expect(yaml_SEQUENCE_END_EVENT) |
| 264 | return n |
| 265 | } |
| 266 | |
| 267 | func (p *parser) mapping() *Node { |
| 268 | n := p.node(MappingNode, mapTag, string(p.event.tag), "") |
| 269 | block := true |
| 270 | if p.event.mapping_style()&yaml_FLOW_MAPPING_STYLE != 0 { |
| 271 | block = false |
| 272 | n.Style |= FlowStyle |
| 273 | } |
| 274 | p.anchor(n, p.event.anchor) |
| 275 | p.expect(yaml_MAPPING_START_EVENT) |
| 276 | for p.peek() != yaml_MAPPING_END_EVENT { |
| 277 | k := p.parseChild(n) |
| 278 | if block && k.FootComment != "" { |
| 279 | // Must be a foot comment for the prior value when being dedented. |
| 280 | if len(n.Content) > 2 { |
| 281 | n.Content[len(n.Content)-3].FootComment = k.FootComment |
| 282 | k.FootComment = "" |
| 283 | } |
| 284 | } |
| 285 | v := p.parseChild(n) |
| 286 | if k.FootComment == "" && v.FootComment != "" { |
| 287 | k.FootComment = v.FootComment |
| 288 | v.FootComment = "" |
| 289 | } |
| 290 | if p.peek() == yaml_TAIL_COMMENT_EVENT { |
| 291 | if k.FootComment == "" { |
| 292 | k.FootComment = string(p.event.foot_comment) |
| 293 | } |
| 294 | p.expect(yaml_TAIL_COMMENT_EVENT) |
| 295 | } |
| 296 | } |
| 297 | n.LineComment = string(p.event.line_comment) |
| 298 | n.FootComment = string(p.event.foot_comment) |
| 299 | if n.Style&FlowStyle == 0 && n.FootComment != "" && len(n.Content) > 1 { |
| 300 | n.Content[len(n.Content)-2].FootComment = n.FootComment |
| 301 | n.FootComment = "" |
| 302 | } |
| 303 | p.expect(yaml_MAPPING_END_EVENT) |
| 304 | return n |
| 305 | } |
| 306 | |
| 307 | // ---------------------------------------------------------------------------- |
| 308 | // Decoder, unmarshals a node into a provided value. |
| 309 | |
| 310 | type decoder struct { |
| 311 | doc *Node |
| 312 | aliases map[*Node]bool |
| 313 | terrors []string |
| 314 | |
| 315 | stringMapType reflect.Type |
| 316 | generalMapType reflect.Type |
| 317 | |
| 318 | knownFields bool |
| 319 | uniqueKeys bool |
| 320 | decodeCount int |
| 321 | aliasCount int |
| 322 | aliasDepth int |
| 323 | } |
| 324 | |
| 325 | var ( |
| 326 | nodeType = reflect.TypeOf(Node{}) |
| 327 | durationType = reflect.TypeOf(time.Duration(0)) |
| 328 | stringMapType = reflect.TypeOf(map[string]interface{}{}) |
| 329 | generalMapType = reflect.TypeOf(map[interface{}]interface{}{}) |
| 330 | ifaceType = generalMapType.Elem() |
| 331 | timeType = reflect.TypeOf(time.Time{}) |
| 332 | ptrTimeType = reflect.TypeOf(&time.Time{}) |
| 333 | ) |
| 334 | |
| 335 | func newDecoder() *decoder { |
| 336 | d := &decoder{ |
| 337 | stringMapType: stringMapType, |
| 338 | generalMapType: generalMapType, |
| 339 | uniqueKeys: true, |
| 340 | } |
| 341 | d.aliases = make(map[*Node]bool) |
| 342 | return d |
| 343 | } |
| 344 | |
| 345 | func (d *decoder) terror(n *Node, tag string, out reflect.Value) { |
| 346 | if n.Tag != "" { |
| 347 | tag = n.Tag |
| 348 | } |
| 349 | value := n.Value |
| 350 | if tag != seqTag && tag != mapTag { |
| 351 | if len(value) > 10 { |
| 352 | value = " `" + value[:7] + "...`" |
| 353 | } else { |
| 354 | value = " `" + value + "`" |
| 355 | } |
| 356 | } |
| 357 | d.terrors = append(d.terrors, fmt.Sprintf("line %d: cannot unmarshal %s%s into %s", n.Line, shortTag(tag), value, out.Type())) |
| 358 | } |
| 359 | |
| 360 | func (d *decoder) callUnmarshaler(n *Node, u Unmarshaler) (good bool) { |
| 361 | err := u.UnmarshalYAML(n) |
| 362 | if e, ok := err.(*TypeError); ok { |
| 363 | d.terrors = append(d.terrors, e.Errors...) |
| 364 | return false |
| 365 | } |
| 366 | if err != nil { |
| 367 | fail(err) |
| 368 | } |
| 369 | return true |
| 370 | } |
| 371 | |
| 372 | func (d *decoder) callObsoleteUnmarshaler(n *Node, u obsoleteUnmarshaler) (good bool) { |
| 373 | terrlen := len(d.terrors) |
| 374 | err := u.UnmarshalYAML(func(v interface{}) (err error) { |
| 375 | defer handleErr(&err) |
| 376 | d.unmarshal(n, reflect.ValueOf(v)) |
| 377 | if len(d.terrors) > terrlen { |
| 378 | issues := d.terrors[terrlen:] |
| 379 | d.terrors = d.terrors[:terrlen] |
| 380 | return &TypeError{issues} |
| 381 | } |
| 382 | return nil |
| 383 | }) |
| 384 | if e, ok := err.(*TypeError); ok { |
| 385 | d.terrors = append(d.terrors, e.Errors...) |
| 386 | return false |
| 387 | } |
| 388 | if err != nil { |
| 389 | fail(err) |
| 390 | } |
| 391 | return true |
| 392 | } |
| 393 | |
| 394 | // d.prepare initializes and dereferences pointers and calls UnmarshalYAML |
| 395 | // if a value is found to implement it. |
| 396 | // It returns the initialized and dereferenced out value, whether |
| 397 | // unmarshalling was already done by UnmarshalYAML, and if so whether |
| 398 | // its types unmarshalled appropriately. |
| 399 | // |
| 400 | // If n holds a null value, prepare returns before doing anything. |
| 401 | func (d *decoder) prepare(n *Node, out reflect.Value) (newout reflect.Value, unmarshaled, good bool) { |
| 402 | if n.ShortTag() == nullTag { |
| 403 | return out, false, false |
| 404 | } |
| 405 | again := true |
| 406 | for again { |
| 407 | again = false |
| 408 | if out.Kind() == reflect.Ptr { |
| 409 | if out.IsNil() { |
| 410 | out.Set(reflect.New(out.Type().Elem())) |
| 411 | } |
| 412 | out = out.Elem() |
| 413 | again = true |
| 414 | } |
| 415 | if out.CanAddr() { |
| 416 | outi := out.Addr().Interface() |
| 417 | if u, ok := outi.(Unmarshaler); ok { |
| 418 | good = d.callUnmarshaler(n, u) |
| 419 | return out, true, good |
| 420 | } |
| 421 | if u, ok := outi.(obsoleteUnmarshaler); ok { |
| 422 | good = d.callObsoleteUnmarshaler(n, u) |
| 423 | return out, true, good |
| 424 | } |
| 425 | } |
| 426 | } |
| 427 | return out, false, false |
| 428 | } |
| 429 | |
| 430 | func (d *decoder) fieldByIndex(n *Node, v reflect.Value, index []int) (field reflect.Value) { |
| 431 | if n.ShortTag() == nullTag { |
| 432 | return reflect.Value{} |
| 433 | } |
| 434 | for _, num := range index { |
| 435 | for { |
| 436 | if v.Kind() == reflect.Ptr { |
| 437 | if v.IsNil() { |
| 438 | v.Set(reflect.New(v.Type().Elem())) |
| 439 | } |
| 440 | v = v.Elem() |
| 441 | continue |
| 442 | } |
| 443 | break |
| 444 | } |
| 445 | v = v.Field(num) |
| 446 | } |
| 447 | return v |
| 448 | } |
| 449 | |
| 450 | const ( |
| 451 | // 400,000 decode operations is ~500kb of dense object declarations, or |
| 452 | // ~5kb of dense object declarations with 10000% alias expansion |
| 453 | alias_ratio_range_low = 400000 |
| 454 | |
| 455 | // 4,000,000 decode operations is ~5MB of dense object declarations, or |
| 456 | // ~4.5MB of dense object declarations with 10% alias expansion |
| 457 | alias_ratio_range_high = 4000000 |
| 458 | |
| 459 | // alias_ratio_range is the range over which we scale allowed alias ratios |
| 460 | alias_ratio_range = float64(alias_ratio_range_high - alias_ratio_range_low) |
| 461 | ) |
| 462 | |
| 463 | func allowedAliasRatio(decodeCount int) float64 { |
| 464 | switch { |
| 465 | case decodeCount <= alias_ratio_range_low: |
| 466 | // allow 99% to come from alias expansion for small-to-medium documents |
| 467 | return 0.99 |
| 468 | case decodeCount >= alias_ratio_range_high: |
| 469 | // allow 10% to come from alias expansion for very large documents |
| 470 | return 0.10 |
| 471 | default: |
| 472 | // scale smoothly from 99% down to 10% over the range. |
| 473 | // this maps to 396,000 - 400,000 allowed alias-driven decodes over the range. |
| 474 | // 400,000 decode operations is ~100MB of allocations in worst-case scenarios (single-item maps). |
| 475 | return 0.99 - 0.89*(float64(decodeCount-alias_ratio_range_low)/alias_ratio_range) |
| 476 | } |
| 477 | } |
| 478 | |
| 479 | func (d *decoder) unmarshal(n *Node, out reflect.Value) (good bool) { |
| 480 | d.decodeCount++ |
| 481 | if d.aliasDepth > 0 { |
| 482 | d.aliasCount++ |
| 483 | } |
| 484 | if d.aliasCount > 100 && d.decodeCount > 1000 && float64(d.aliasCount)/float64(d.decodeCount) > allowedAliasRatio(d.decodeCount) { |
| 485 | failf("document contains excessive aliasing") |
| 486 | } |
| 487 | if out.Type() == nodeType { |
| 488 | out.Set(reflect.ValueOf(n).Elem()) |
| 489 | return true |
| 490 | } |
| 491 | switch n.Kind { |
| 492 | case DocumentNode: |
| 493 | return d.document(n, out) |
| 494 | case AliasNode: |
| 495 | return d.alias(n, out) |
| 496 | } |
| 497 | out, unmarshaled, good := d.prepare(n, out) |
| 498 | if unmarshaled { |
| 499 | return good |
| 500 | } |
| 501 | switch n.Kind { |
| 502 | case ScalarNode: |
| 503 | good = d.scalar(n, out) |
| 504 | case MappingNode: |
| 505 | good = d.mapping(n, out) |
| 506 | case SequenceNode: |
| 507 | good = d.sequence(n, out) |
| 508 | case 0: |
| 509 | if n.IsZero() { |
| 510 | return d.null(out) |
| 511 | } |
| 512 | fallthrough |
| 513 | default: |
| 514 | failf("cannot decode node with unknown kind %d", n.Kind) |
| 515 | } |
| 516 | return good |
| 517 | } |
| 518 | |
| 519 | func (d *decoder) document(n *Node, out reflect.Value) (good bool) { |
| 520 | if len(n.Content) == 1 { |
| 521 | d.doc = n |
| 522 | d.unmarshal(n.Content[0], out) |
| 523 | return true |
| 524 | } |
| 525 | return false |
| 526 | } |
| 527 | |
| 528 | func (d *decoder) alias(n *Node, out reflect.Value) (good bool) { |
| 529 | if d.aliases[n] { |
| 530 | // TODO this could actually be allowed in some circumstances. |
| 531 | failf("anchor '%s' value contains itself", n.Value) |
| 532 | } |
| 533 | d.aliases[n] = true |
| 534 | d.aliasDepth++ |
| 535 | good = d.unmarshal(n.Alias, out) |
| 536 | d.aliasDepth-- |
| 537 | delete(d.aliases, n) |
| 538 | return good |
| 539 | } |
| 540 | |
| 541 | var zeroValue reflect.Value |
| 542 | |
| 543 | func resetMap(out reflect.Value) { |
| 544 | for _, k := range out.MapKeys() { |
| 545 | out.SetMapIndex(k, zeroValue) |
| 546 | } |
| 547 | } |
| 548 | |
| 549 | func (d *decoder) null(out reflect.Value) bool { |
| 550 | if out.CanAddr() { |
| 551 | switch out.Kind() { |
| 552 | case reflect.Interface, reflect.Ptr, reflect.Map, reflect.Slice: |
| 553 | out.Set(reflect.Zero(out.Type())) |
| 554 | return true |
| 555 | } |
| 556 | } |
| 557 | return false |
| 558 | } |
| 559 | |
| 560 | func (d *decoder) scalar(n *Node, out reflect.Value) bool { |
| 561 | var tag string |
| 562 | var resolved interface{} |
| 563 | if n.indicatedString() { |
| 564 | tag = strTag |
| 565 | resolved = n.Value |
| 566 | } else { |
| 567 | tag, resolved = resolve(n.Tag, n.Value) |
| 568 | if tag == binaryTag { |
| 569 | data, err := base64.StdEncoding.DecodeString(resolved.(string)) |
| 570 | if err != nil { |
| 571 | failf("!!binary value contains invalid base64 data") |
| 572 | } |
| 573 | resolved = string(data) |
| 574 | } |
| 575 | } |
| 576 | if resolved == nil { |
| 577 | return d.null(out) |
| 578 | } |
| 579 | if resolvedv := reflect.ValueOf(resolved); out.Type() == resolvedv.Type() { |
| 580 | // We've resolved to exactly the type we want, so use that. |
| 581 | out.Set(resolvedv) |
| 582 | return true |
| 583 | } |
| 584 | // Perhaps we can use the value as a TextUnmarshaler to |
| 585 | // set its value. |
| 586 | if out.CanAddr() { |
| 587 | u, ok := out.Addr().Interface().(encoding.TextUnmarshaler) |
| 588 | if ok { |
| 589 | var text []byte |
| 590 | if tag == binaryTag { |
| 591 | text = []byte(resolved.(string)) |
| 592 | } else { |
| 593 | // We let any value be unmarshaled into TextUnmarshaler. |
| 594 | // That might be more lax than we'd like, but the |
| 595 | // TextUnmarshaler itself should bowl out any dubious values. |
| 596 | text = []byte(n.Value) |
| 597 | } |
| 598 | err := u.UnmarshalText(text) |
| 599 | if err != nil { |
| 600 | fail(err) |
| 601 | } |
| 602 | return true |
| 603 | } |
| 604 | } |
| 605 | switch out.Kind() { |
| 606 | case reflect.String: |
| 607 | if tag == binaryTag { |
| 608 | out.SetString(resolved.(string)) |
| 609 | return true |
| 610 | } |
| 611 | out.SetString(n.Value) |
| 612 | return true |
| 613 | case reflect.Interface: |
| 614 | out.Set(reflect.ValueOf(resolved)) |
| 615 | return true |
| 616 | case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64: |
| 617 | // This used to work in v2, but it's very unfriendly. |
| 618 | isDuration := out.Type() == durationType |
| 619 | |
| 620 | switch resolved := resolved.(type) { |
| 621 | case int: |
| 622 | if !isDuration && !out.OverflowInt(int64(resolved)) { |
| 623 | out.SetInt(int64(resolved)) |
| 624 | return true |
| 625 | } |
| 626 | case int64: |
| 627 | if !isDuration && !out.OverflowInt(resolved) { |
| 628 | out.SetInt(resolved) |
| 629 | return true |
| 630 | } |
| 631 | case uint64: |
| 632 | if !isDuration && resolved <= math.MaxInt64 && !out.OverflowInt(int64(resolved)) { |
| 633 | out.SetInt(int64(resolved)) |
| 634 | return true |
| 635 | } |
| 636 | case float64: |
| 637 | if !isDuration && resolved <= math.MaxInt64 && !out.OverflowInt(int64(resolved)) { |
| 638 | out.SetInt(int64(resolved)) |
| 639 | return true |
| 640 | } |
| 641 | case string: |
| 642 | if out.Type() == durationType { |
| 643 | d, err := time.ParseDuration(resolved) |
| 644 | if err == nil { |
| 645 | out.SetInt(int64(d)) |
| 646 | return true |
| 647 | } |
| 648 | } |
| 649 | } |
| 650 | case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr: |
| 651 | switch resolved := resolved.(type) { |
| 652 | case int: |
| 653 | if resolved >= 0 && !out.OverflowUint(uint64(resolved)) { |
| 654 | out.SetUint(uint64(resolved)) |
| 655 | return true |
| 656 | } |
| 657 | case int64: |
| 658 | if resolved >= 0 && !out.OverflowUint(uint64(resolved)) { |
| 659 | out.SetUint(uint64(resolved)) |
| 660 | return true |
| 661 | } |
| 662 | case uint64: |
| 663 | if !out.OverflowUint(uint64(resolved)) { |
| 664 | out.SetUint(uint64(resolved)) |
| 665 | return true |
| 666 | } |
| 667 | case float64: |
| 668 | if resolved <= math.MaxUint64 && !out.OverflowUint(uint64(resolved)) { |
| 669 | out.SetUint(uint64(resolved)) |
| 670 | return true |
| 671 | } |
| 672 | } |
| 673 | case reflect.Bool: |
| 674 | switch resolved := resolved.(type) { |
| 675 | case bool: |
| 676 | out.SetBool(resolved) |
| 677 | return true |
| 678 | case string: |
| 679 | // This offers some compatibility with the 1.1 spec (https://yaml.org/type/bool.html). |
| 680 | // It only works if explicitly attempting to unmarshal into a typed bool value. |
| 681 | switch resolved { |
| 682 | case "y", "Y", "yes", "Yes", "YES", "on", "On", "ON": |
| 683 | out.SetBool(true) |
| 684 | return true |
| 685 | case "n", "N", "no", "No", "NO", "off", "Off", "OFF": |
| 686 | out.SetBool(false) |
| 687 | return true |
| 688 | } |
| 689 | } |
| 690 | case reflect.Float32, reflect.Float64: |
| 691 | switch resolved := resolved.(type) { |
| 692 | case int: |
| 693 | out.SetFloat(float64(resolved)) |
| 694 | return true |
| 695 | case int64: |
| 696 | out.SetFloat(float64(resolved)) |
| 697 | return true |
| 698 | case uint64: |
| 699 | out.SetFloat(float64(resolved)) |
| 700 | return true |
| 701 | case float64: |
| 702 | out.SetFloat(resolved) |
| 703 | return true |
| 704 | } |
| 705 | case reflect.Struct: |
| 706 | if resolvedv := reflect.ValueOf(resolved); out.Type() == resolvedv.Type() { |
| 707 | out.Set(resolvedv) |
| 708 | return true |
| 709 | } |
| 710 | case reflect.Ptr: |
| 711 | panic("yaml internal error: please report the issue") |
| 712 | } |
| 713 | d.terror(n, tag, out) |
| 714 | return false |
| 715 | } |
| 716 | |
| 717 | func settableValueOf(i interface{}) reflect.Value { |
| 718 | v := reflect.ValueOf(i) |
| 719 | sv := reflect.New(v.Type()).Elem() |
| 720 | sv.Set(v) |
| 721 | return sv |
| 722 | } |
| 723 | |
| 724 | func (d *decoder) sequence(n *Node, out reflect.Value) (good bool) { |
| 725 | l := len(n.Content) |
| 726 | |
| 727 | var iface reflect.Value |
| 728 | switch out.Kind() { |
| 729 | case reflect.Slice: |
| 730 | out.Set(reflect.MakeSlice(out.Type(), l, l)) |
| 731 | case reflect.Array: |
| 732 | if l != out.Len() { |
| 733 | failf("invalid array: want %d elements but got %d", out.Len(), l) |
| 734 | } |
| 735 | case reflect.Interface: |
| 736 | // No type hints. Will have to use a generic sequence. |
| 737 | iface = out |
| 738 | out = settableValueOf(make([]interface{}, l)) |
| 739 | default: |
| 740 | d.terror(n, seqTag, out) |
| 741 | return false |
| 742 | } |
| 743 | et := out.Type().Elem() |
| 744 | |
| 745 | j := 0 |
| 746 | for i := 0; i < l; i++ { |
| 747 | e := reflect.New(et).Elem() |
| 748 | if ok := d.unmarshal(n.Content[i], e); ok { |
| 749 | out.Index(j).Set(e) |
| 750 | j++ |
| 751 | } |
| 752 | } |
| 753 | if out.Kind() != reflect.Array { |
| 754 | out.Set(out.Slice(0, j)) |
| 755 | } |
| 756 | if iface.IsValid() { |
| 757 | iface.Set(out) |
| 758 | } |
| 759 | return true |
| 760 | } |
| 761 | |
| 762 | func (d *decoder) mapping(n *Node, out reflect.Value) (good bool) { |
| 763 | l := len(n.Content) |
| 764 | if d.uniqueKeys { |
| 765 | nerrs := len(d.terrors) |
| 766 | for i := 0; i < l; i += 2 { |
| 767 | ni := n.Content[i] |
| 768 | for j := i + 2; j < l; j += 2 { |
| 769 | nj := n.Content[j] |
| 770 | if ni.Kind == nj.Kind && ni.Value == nj.Value { |
| 771 | d.terrors = append(d.terrors, fmt.Sprintf("line %d: mapping key %#v already defined at line %d", nj.Line, nj.Value, ni.Line)) |
| 772 | } |
| 773 | } |
| 774 | } |
| 775 | if len(d.terrors) > nerrs { |
| 776 | return false |
| 777 | } |
| 778 | } |
| 779 | switch out.Kind() { |
| 780 | case reflect.Struct: |
| 781 | return d.mappingStruct(n, out) |
| 782 | case reflect.Map: |
| 783 | // okay |
| 784 | case reflect.Interface: |
| 785 | iface := out |
| 786 | if isStringMap(n) { |
| 787 | out = reflect.MakeMap(d.stringMapType) |
| 788 | } else { |
| 789 | out = reflect.MakeMap(d.generalMapType) |
| 790 | } |
| 791 | iface.Set(out) |
| 792 | default: |
| 793 | d.terror(n, mapTag, out) |
| 794 | return false |
| 795 | } |
| 796 | |
| 797 | outt := out.Type() |
| 798 | kt := outt.Key() |
| 799 | et := outt.Elem() |
| 800 | |
| 801 | stringMapType := d.stringMapType |
| 802 | generalMapType := d.generalMapType |
| 803 | if outt.Elem() == ifaceType { |
| 804 | if outt.Key().Kind() == reflect.String { |
| 805 | d.stringMapType = outt |
| 806 | } else if outt.Key() == ifaceType { |
| 807 | d.generalMapType = outt |
| 808 | } |
| 809 | } |
| 810 | |
| 811 | mapIsNew := false |
| 812 | if out.IsNil() { |
| 813 | out.Set(reflect.MakeMap(outt)) |
| 814 | mapIsNew = true |
| 815 | } |
| 816 | for i := 0; i < l; i += 2 { |
| 817 | if isMerge(n.Content[i]) { |
| 818 | d.merge(n.Content[i+1], out) |
| 819 | continue |
| 820 | } |
| 821 | k := reflect.New(kt).Elem() |
| 822 | if d.unmarshal(n.Content[i], k) { |
| 823 | kkind := k.Kind() |
| 824 | if kkind == reflect.Interface { |
| 825 | kkind = k.Elem().Kind() |
| 826 | } |
| 827 | if kkind == reflect.Map || kkind == reflect.Slice { |
| 828 | failf("invalid map key: %#v", k.Interface()) |
| 829 | } |
| 830 | e := reflect.New(et).Elem() |
| 831 | if d.unmarshal(n.Content[i+1], e) || n.Content[i+1].ShortTag() == nullTag && (mapIsNew || !out.MapIndex(k).IsValid()) { |
| 832 | out.SetMapIndex(k, e) |
| 833 | } |
| 834 | } |
| 835 | } |
| 836 | d.stringMapType = stringMapType |
| 837 | d.generalMapType = generalMapType |
| 838 | return true |
| 839 | } |
| 840 | |
| 841 | func isStringMap(n *Node) bool { |
| 842 | if n.Kind != MappingNode { |
| 843 | return false |
| 844 | } |
| 845 | l := len(n.Content) |
| 846 | for i := 0; i < l; i += 2 { |
| 847 | if n.Content[i].ShortTag() != strTag { |
| 848 | return false |
| 849 | } |
| 850 | } |
| 851 | return true |
| 852 | } |
| 853 | |
| 854 | func (d *decoder) mappingStruct(n *Node, out reflect.Value) (good bool) { |
| 855 | sinfo, err := getStructInfo(out.Type()) |
| 856 | if err != nil { |
| 857 | panic(err) |
| 858 | } |
| 859 | |
| 860 | var inlineMap reflect.Value |
| 861 | var elemType reflect.Type |
| 862 | if sinfo.InlineMap != -1 { |
| 863 | inlineMap = out.Field(sinfo.InlineMap) |
| 864 | inlineMap.Set(reflect.New(inlineMap.Type()).Elem()) |
| 865 | elemType = inlineMap.Type().Elem() |
| 866 | } |
| 867 | |
| 868 | for _, index := range sinfo.InlineUnmarshalers { |
| 869 | field := d.fieldByIndex(n, out, index) |
| 870 | d.prepare(n, field) |
| 871 | } |
| 872 | |
| 873 | var doneFields []bool |
| 874 | if d.uniqueKeys { |
| 875 | doneFields = make([]bool, len(sinfo.FieldsList)) |
| 876 | } |
| 877 | name := settableValueOf("") |
| 878 | l := len(n.Content) |
| 879 | for i := 0; i < l; i += 2 { |
| 880 | ni := n.Content[i] |
| 881 | if isMerge(ni) { |
| 882 | d.merge(n.Content[i+1], out) |
| 883 | continue |
| 884 | } |
| 885 | if !d.unmarshal(ni, name) { |
| 886 | continue |
| 887 | } |
| 888 | if info, ok := sinfo.FieldsMap[name.String()]; ok { |
| 889 | if d.uniqueKeys { |
| 890 | if doneFields[info.Id] { |
| 891 | d.terrors = append(d.terrors, fmt.Sprintf("line %d: field %s already set in type %s", ni.Line, name.String(), out.Type())) |
| 892 | continue |
| 893 | } |
| 894 | doneFields[info.Id] = true |
| 895 | } |
| 896 | var field reflect.Value |
| 897 | if info.Inline == nil { |
| 898 | field = out.Field(info.Num) |
| 899 | } else { |
| 900 | field = d.fieldByIndex(n, out, info.Inline) |
| 901 | } |
| 902 | d.unmarshal(n.Content[i+1], field) |
| 903 | } else if sinfo.InlineMap != -1 { |
| 904 | if inlineMap.IsNil() { |
| 905 | inlineMap.Set(reflect.MakeMap(inlineMap.Type())) |
| 906 | } |
| 907 | value := reflect.New(elemType).Elem() |
| 908 | d.unmarshal(n.Content[i+1], value) |
| 909 | inlineMap.SetMapIndex(name, value) |
| 910 | } else if d.knownFields { |
| 911 | d.terrors = append(d.terrors, fmt.Sprintf("line %d: field %s not found in type %s", ni.Line, name.String(), out.Type())) |
| 912 | } |
| 913 | } |
| 914 | return true |
| 915 | } |
| 916 | |
| 917 | func failWantMap() { |
| 918 | failf("map merge requires map or sequence of maps as the value") |
| 919 | } |
| 920 | |
| 921 | func (d *decoder) merge(n *Node, out reflect.Value) { |
| 922 | switch n.Kind { |
| 923 | case MappingNode: |
| 924 | d.unmarshal(n, out) |
| 925 | case AliasNode: |
| 926 | if n.Alias != nil && n.Alias.Kind != MappingNode { |
| 927 | failWantMap() |
| 928 | } |
| 929 | d.unmarshal(n, out) |
| 930 | case SequenceNode: |
| 931 | // Step backwards as earlier nodes take precedence. |
| 932 | for i := len(n.Content) - 1; i >= 0; i-- { |
| 933 | ni := n.Content[i] |
| 934 | if ni.Kind == AliasNode { |
| 935 | if ni.Alias != nil && ni.Alias.Kind != MappingNode { |
| 936 | failWantMap() |
| 937 | } |
| 938 | } else if ni.Kind != MappingNode { |
| 939 | failWantMap() |
| 940 | } |
| 941 | d.unmarshal(ni, out) |
| 942 | } |
| 943 | default: |
| 944 | failWantMap() |
| 945 | } |
| 946 | } |
| 947 | |
| 948 | func isMerge(n *Node) bool { |
| 949 | return n.Kind == ScalarNode && n.Value == "<<" && (n.Tag == "" || n.Tag == "!" || shortTag(n.Tag) == mergeTag) |
| 950 | } |