Stringifying using different arrayFormats, then parsed, create different things

[wholien]

It seems stringifying my metadata object with arrayFormat: brackets, then parsing it, creates a different object to stringifying with arrayFormat: indices then parsing that string.

> const qs = require('qs');

> metadata = [{key: "a", value: "2"}, {key: "b", value: "3"}]
[ { key: 'a', value: '2' }, { key: 'b', value: '3' } ]

> brackets = qs.stringify({"metadata": metadata}, { arrayFormat: "brackets"})
'metadata%5B%5D%5Bkey%5D=a&metadata%5B%5D%5Bvalue%5D=2&metadata%5B%5D%5Bkey%5D=b&metadata%5B%5D%5Bvalue%5D=3'

> parsed_brackets = qs.parse(brackets)
{ metadata: [ { key: [Array], value: [Array] } ] }

> parsed_brackets["metadata"]
[ { key: [ 'a', 'b' ], value: [ '2', '3' ] } ]

> indices = qs.stringify({"metadata": metadata}, { arrayFormat: "indices"})
'metadata%5B0%5D%5Bkey%5D=a&metadata%5B0%5D%5Bvalue%5D=2&metadata%5B1%5D%5Bkey%5D=b&metadata%5B1%5D%5Bvalue%5D=3'

> parsed_indices = qs.parse(indices)
{ metadata: [ { key: 'a', value: '2' }, { key: 'b', value: '3' } ] }

Am I doing something wrong? Is this expected? Is this just an issue with stringifying/parsing nested objects?

[Darwin-Gowago]

Having the same issues here. Is there any update on this?

[ljharb]

It's entirely reasonable that the same string produces different results when you use different array formats.

In this case, it seems like "brackets" does have a bug, whether in stringify, parse, or both.

[wholien]

judging from other query string libraries, it might require specifying the arrayFormat in both stringify and parse

[ljharb]

ohhhh right, it absolutely does :-) i missed that. indices is the default arrayFormat.

@wholien what happens if you specify brackets in your first parse?

[wholien]

@ljharb doing it with brackets:

> metadata = [{key: "a", value: "2"}, {key: "b", value: "3"}]
[ { key: 'a', value: '2' }, { key: 'b', value: '3' } ]

> brackets = qs.stringify({"metadata": metadata}, { arrayFormat: "brackets"})
'metadata%5B%5D%5Bkey%5D=a&metadata%5B%5D%5Bvalue%5D=2&metadata%5B%5D%5Bkey%5D=b&metadata%5B%5D%5Bvalue%5D=3'

> parsed_brackets = qs.parse(brackets, { arrayFormat: "brackets"}) // <---- here
{ metadata: [ { key: [Array], value: [Array] } ] }

> parsed_brackets["metadata"]
[ { key: [ 'a', 'b' ], value: [ '2', '3' ] } ]

I was reading through the readme and some of the parse code - doesn't look like there is an option for setting arrayFormats, though I could be wrong.

[ljharb]

Seems like we do have an issue here then.

I agree with the general premise that for any format parse accepts, or stringify produces, the other should be able to round trip (ie, stringify(parse(q)) === q modulo normalizations, and parse(stringify(q)) =~ q where =~ is a reasonable deep equal algorithm modulo normalizations).

Happy to accept a PR here, but I'm not sure concretely what it should change.

Failing test cases are probably the best place to start!