Option to dedupe multiple keys with the same name

[ljharb]

Option to dedupe multiple keys with the same name

[Marsup]

This is how parsing is supposed to be done AFAIK.

[RGBboy]

@Marsup as in it should get set to 'false'? Or the other way?

[Marsup]

If you have multiple parameters with the same name, it should become an array.

[RGBboy]

Yep that is what happens currently. However some parsers would deduplicate this and just set it to the last defined value. For example parsing JSON does this if you define a property twice.

Would you be open to a pull request to add this functionality behind an option?

[Marsup]

I don't think parsers for other formats are relevant here, look at the other query string parsers, all the other ones I know do that.

I'll let @nlf decide on that behavior change but I think it'll be a no.

[RGBboy]

Here is my use case if that helps justify the change:

I have an html form with a checkbox. Because browsers only pass the value of a checkbox if it is checked I also have a hidden input to set the unchecked value. This hidden input uses the same name as the checkbox input. When this data is then parsed by qs I get an array, not a single value which is what I would like.

[Marsup]

Bad design IMHO. If the value is missing, default to false.

[RGBboy]

Not if you have to cater for true, false and undefined.

[Marsup]

Then preprocess the query you're sending, but not with some tricks with hidden fields. It's not the server side that needs fixing.

[RGBboy]

Unfortunately I can't preprocess because we must cater for clients with no Javascript. Unfortunately the W3C spec forces developers down this road if they want to use checkboxes. If we could use Javascript I would just send JSON and not use query strings at all.

[Marsup]

Then why not radio buttons ? Tri-state checkbox in pure html don't exist, the only way to set it is through javascript, so no js, no tri-state. If your user has no possibility to set it himself, I don't see how this can make a difference between "false" and "never been touched false".

[RGBboy]

That is true for a single form, but not for multiple with the same endpoint. When a user submits a form with no checkbox we receive undefined (which we interpret as no change) where as if we receive true or false we update the value given.

[Marsup]

Then what about not using the same name and default to it with joi (considering you're using it), like :

var schema = Joi.object().keys({
  a: Joi.string().default(Joi.ref('adefault')),
  adefault: Joi.string()
});

[nlf]

I tend to agree that this shouldn't be qs's problem, it providing the server with all passed in parameters is the idea. I don't really like the idea of adding in options for edge cases, especially where a little extra processing after the parsing is done is not much work in your app.

[nlf]

Closing this unless there ends up being more demand for it