Multi-line arrays in square brackets cause YAMLSemanticError

[klingtnet]

The library fails to parse multi-line arrays in square brackets. I stumbled upon this issue when speccy used this library for parsing the OpenAPI specification. Following is a minimal example to reproduce the issue:

package.json:

{
  "dependencies": {
    "fs": "0.0.1-security",
    "yaml": "^1.5.1"
  }
}

example.yaml:

---
foo:
  bar:
    enum: [
      "abc",
      "cde"
    ]

test.js:

const fs = require('fs');
const YAML = require('yaml');

const file = fs.readFileSync('./example.yaml', 'utf8');
try {
  YAML.parse(file)
} catch (error) {
  console.log(error)
}

$ node test.js
{ YAMLSemanticError: Insufficient indentation in flow collection
    at FlowCollection.parse (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/FlowCollection.js:77:60)
    at ParseContext._defineProperty (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/ParseContext.js:141:25)
    at CollectionItem.parse (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/CollectionIte



m.js:81:19)
    at ParseContext._defineProperty (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/ParseContext.js:141:25)
    at Collection.parse (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/Collection.js:199:20)
    at ParseContext._defineProperty (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/ParseContext.js:157:29)
    at CollectionItem.parse (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/CollectionItem.js:81:19)
    at ParseContext._defineProperty (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/ParseContext.js:141:25)
    at Collection.parse (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/Collection.js:199:20)
    at ParseContext._defineProperty (/home/anli/code/industrial-fulfillment-api/node_modules/yaml/dist/cst/ParseContext.js:157:29)
  name: 'YAMLSemanticError',
  message: 'Insufficient indentation in flow collection',
  source: 
   FlowCollection {
     context: 
      ParseContext {
        parseNode: [Function],
        atLineStart: false,
        inCollection: false,
        inFlow: false,
        indent: 4,
        lineStart: 16,
        parent: [CollectionItem],
        root: [Document],
        src: '---\nfoo:\n  bar:\n    enum: [\n      "abc",\n      "cde"\n    ]' },
     error: [Circular],
     range: Range { start: 26, end: 58 },
     valueRange: Range { start: 26, end: 58 },
     props: [],
     type: 'FLOW_SEQ',
     value: null,
     items: [ [Object], [QuoteDouble], [Object], [QuoteDouble], [Object] ],
     resolved: YAMLSeq { items: [Array], range: [Array], type: 'FLOW_SEQ' } } }

Ruby parses the YAML without problems:

$ cat example.yaml | ruby -e 'require "yaml"; puts YAML.load($stdin)'
{"foo"=>{"bar"=>{"enum"=>["abc", "cde"]}}}

The same goes for PyYAML:

import yaml

with open('example.yaml') as f:
       print(yaml.parse(f.read()))

[eemeli]

I went through the spec yet again, and it's not completely clear that something like this is in fact valid YAML:

foo: {
  bar: baz
}

The test suite does include a test that makes it clear that this would be wrong:

foo: {
bar: baz
}

... while this is completely fine:

{
bar: baz
}

Tracing through the relevant constructions would indicate that even the terminal ] or } char should be indented as much as the preceding parts, which effectively becomes relevant only when a flow collection is the value of a block sequence or mapping, and starts on the same line as its sequence indicator or its key value.

But that's a bit dumb, and as you figured most other libraries don't actually require that -- or maybe I just made a mistake when reading the spec. In any case, I've hopefully fixed this now, and will let it parse without error at least until someone complains.

Other relevant issues on this: #108, #114.

[klingtnet]

My issue was about arrays split over multiple-lines (see title) and not maps/dictionaries.

# ...
    enum: [
      "abc",
      "cde"
    ]

[perlpunk]

@eemeli Most libraries even don't care about flow collection indentation at all.
See the output of this example:

nested:
  a:
    b: [
flow style sequence
]

tested with the yaml-editor:https://gist.github.com/6686d1623cb6fc09757932bfbe3a61b5

I can understand why this:

  foo: [
  ]

looks valid, but not this:

  foo: [
]

The YAML spec (as I understand it) is very consistent there. The closing ] or } is part of the flow collection, and there is no exception for it to be not indented.

But it seems a lot of YAML documentes are out there that use this, so your decision to allow it is probably reasonable.

[eemeli]

@klingtnet Apologies for any confusion; [] flow sequences are handled by the same code as {} flow mappings, and should now work as intended. I also added a test for the example you gave:

yaml/tests/cst/corner-cases.js

Lines 236 to 259 in f8d35f4

	test('eemeli/yaml#113', () => {
	const src = source`
	---
	foo:
	bar:
	enum: [
	"abc",
	"cde"
	]
	`
	const doc = parse(src)
	const barValue = doc[0].contents[0].items[1].node.items[1].node
	expect(barValue.items[1].node).toMatchObject({
	error: null,
	items: [
	{ char: '[' },
	{ type: 'QUOTE_DOUBLE' },
	{ char: ',' },
	{ type: 'QUOTE_DOUBLE' },
	{ char: ']' }
	],
	type: 'FLOW_SEQ'
	})
	})