param=1&param=2 is the same as param[]=1&param[]=2

[Somebi]

This shouldn't be this way, the last param should be taken instead of uniting them into an array.

[tj]

there are no rules really, these are behaviours that other libraries support but I would agree

[Marak]

@Somebi - Disagree.

If you do a plain ole form post with multiple input fields off the same name, the browser will concat these values into a list, it won't just take the last value.

[Somebi]

I just accustomed to that behavior that param[]=value generates an array and param=value&param=value takes the last one. I have created a small helper on client-side, which removes duplicated names.

[Marak]

@Somebi - I'm curious, which other query string parsers act that way? I'm interested in making sure this query string parser behaves nicely.

[Somebi]

Well that was in php + apache.

[ghost]

@Marak This is also the behavior in Ruby/Rails... in fact I was trying to replicate in Express the good old trick of preceding a checkbox input with a hidden field of the same name so that when the checkbox isn't checked I received "0" and when it is I get "1"... but I was surprised to realize that instead I get ["0", "1"] when it's checked.

[fizker]

Just for information, the same question was asked here: http://stackoverflow.com/questions/1746507/authoritative-position-of-duplicate-http-get-query-keys

And they linked to a great document that lists different ways that it is handled: https://www.owasp.org/images/b/ba/AppsecEU09_CarettoniDiPaola_v0.8.pdf.
Check page 9 in that document.

The most used is keep either the first or the last value, but there really are no consensus.

Perhaps add an option for choosing which of the 4 different methods to use?

The 4 being:

1. Keep first (e.g. "1")
2. Keep last (e.g. "2")
3. Concatenate via comma (e.g. "1,2")
4. Concatenate as array (e.g. ["1", "2"])

[Orion98MC]

I Agree with fizker and Somebi.
This is expected behavior in rails and php/apache.

Please make it a useful convention to get rid of the extra controller code required so far.

[blah238]

Looks like the same issue as: #9

Any forks with the "keep last" behavior?