I'm seeing a "URIError: Malformed encodeURI input" error at encodeURIComponent when emoji are passed in as the value to encode(value).

It looks like the error is coming from line 7, .replace(/[^ !'()~*]/gu, encodeURIComponent), when encode(value) is called within nest on line 34, f = encode(name) + '=' + encode(value);.

The error doesn't occur on all emoji.

Here are a few emoji that have triggered the error in React Native on iOS and Android: 😎, 😅, 😍, 👀

Now i can be wrong here, but i see a difference when Postman does encoding on a value P@55word which results into password=P@55word vs when this library encodes the value it results password=P%4055word which of this is right? and or if there any options flag that i am missing to get this value encoded as postman..
To verify open the code section on postman in HTML

I would suggest disabling the wiki. I was able to just create a page and am not a contributor.

Currently the source provided is ES6. The default config of webpack & babel-loader is not to transpile anything in node_modules. And when using ES6, this package breaks when compiling/uglifying.

Usually packages transpile their code using a postinstall command, "postinstall": "babel -d lib/ src/" and point the main field to lib. lib is in .gitignore too.

Open to this?

With the newest release v4.5.0, the following line now returns an undefined:
const formEncode = require('form-urlencoded').default

Reverting to 4.4.2 works fine again. It seems like the changes for esm module support has broken it for CommonJS. Removing .default works fine actually: const formEncode = require('form-urlencoded') when on the latest version.

It now depends if this is a deliberate change. If yes, it would probably make more sense to bump it to v5 as it is a breaking change for any packages and projects that use .default import

Bug

Repro Steps

Pass an object into the form-urlencoded function where at least one of the properties is a string and contains a space. That space will be replaced with a + when it should be %20 as when you decode a + with url encoding it is still a +.

Example:

const formurlencoded = require('form-urlencoded')

const str = formurlencoded({
  grant_type: 'grant_type:phone',
  phone_code: '585858',
  phone_number: '(318)  470-1137'
})

console.log(str)
// str will equal
// grant_type=urn%3Alyft%3Aoauth2%3Agrant_type%3Aphone&phone_code=718629&phone_number=%28318%29++470-1137
// and should be
// grant_type=urn%3Alyft%3Aoauth2%3Agrant_type%3Aphone&phone_code=718629&phone_number=%28318%29%20%20470-1137

When I have a multi-value attribute like

const formData = {
  name: ['apple', 'bananas']
}

and run this through formurlencoded(formData, { skipIndex: true })

then I expect this to be

name=apple&name=bananas
// in parsed form:
// name: apple
// name: bananas

but not

name%5B%5D=apple&name%5B%5D=bananas
// in parsed form:
// name[]: apple
// name[]: bananas

Hi @iambumblehead,

I am just wondering why space is replace by + sign instead of %20?

Thanks

Doddy

Hi!

I'm replacing some code using jQuery ajax method, and I find your package very useful for this endeavor — thanks a lot!

Would you consider adding an option to skip empty arrays? I can do it myself and make a PR if you want.

Best regards! :)

fwiw, the recent change to module.exports.default = ... breaks rollup.

i'm not entirely sure about other envs, but with rollup, you need module.exports = fn... and the default is implied. (i believe webpack is the same, though it may sniff for .default. donno.)

after this change, you need to do this to support latest release and past releases:

import { default as _formurlencoded } from 'form-urlencoded';
const formurlencoded = 'default' in _formurlencoded ? _formurlencoded.default : _formurlencoded;

The "module" field in the package.json file points to a src/form-urlencoded.js file, which doesn't exists in either the repository or the npm package. This is currently conflicting with some functionality in rollup.

This field should be pointing to the esm version of the library, which is now the form-urlencoded.js file in the root folder.

Has version 3.0.1 been removed or accidentally published? It still exists in NPM and seems to use the untranspiled version of form-urlencode which breaks in older browsers.
It seems to relate to https://github.com/iambumblehead/form-urlencoded/blob/master/package.json#L4

function arrnest(name, arr) {
    return arr.length ? filterjoin(arr.map(function (elem, index) {
        return nest(name + '[' + index + ']', elem);
    })) : encode(name + '[]');
}

Data structure:
FCartId:[1,2,3,4,5,6]

After urlencoded:
FCartId[0]:1
FCartId[1]:2
FCartId[2]:3
FCartId[3]:4
FCartId[4]:5
FCartId[5]:6

I see jQuery serialize function return:
FCartId[]:1
FCartId[]:2
FCartId[]:3
FCartId[]:4
FCartId[]:5
FCartId[]:6

Looking for you reply!

Really, is mandatory the name of attributes of type array always ending with []?

In this section of code is forcing always to set at the end of the name of arrays attribute the []

arrnest = (name, arr) => arr.length
            ? filterjoin(arr.map((elem, index) => skipIndex
                ? nest(name + '[]', elem)
                : nest(name + '[' + index + ']', elem)))
            : encode(name + '[]'),

When I sending attributes with Jersey - JAVA - I'm doing in this way:

for (String user : users) {
    formParam.add("usersId", user);
}

The next image show data collected from server:

After reading some of the closed issues #8 , specially the end of the #7, I changed the code with:

formurlencoded(formData, { skipIndex: true }),

That worked nearly the expected, but the name includes always the [] at the end.

And here my question, something might be wrong, or my code in Jersey or [] added by your package should not be mandatory ( might be there's some RFC what forces the [] at the end if is an array? ). Or maybe both scenarios are acceptable?

I make a small improvement in the server side to working with both name format, but I will appreciate your opinion about it.

Thanks for your time.

Hi, Thanks for a very useful libs. However, when I tried to use the library on my project. I come into a requirement like the following

sample json:

{
   "a" : [
        {
             "aa" : 1,
             "ab" : 2
        },
        {
            "ac" : 3
        }
    ]
}

The current code will format the field name as "a[0][aa]" before the encoding.
What I need was in the format of "a[0].name" before the encoding.

I added the another option for this in my fork repo. Let me know if this is something that can be merged.

Thanks

Hi, i'm having troubles with version 2.0.7 it seems to be exporting an object with { default }
I had to roll back to 2.0.4

Have this happened to you?
(im using node 8.9.4)

How do I output obj.prop2.prop3=elm as obj[prop2][prop3]=elm.

So UglifyJS has this tendency to replace any \uXXXX substring with the UTF-8 equivalent. However, code points such as \uDC00, \uDFFF, \uD800, \uDBFF are surrogate characters and therefore don't have a UTF-8 equivalent. UglifyJS therefore replaces these substrings with �, which turns the regex line into:

s = new RegExp(['(?:[\0-�"-&+-}�-퟿-�]|', "[�-�][�-�]|[�-�](?![�-�])|", "(?:[^�-�]|^)[�-�])"].join(""), "g"),

This regex is not the same as the regex in the source code. The �-� part which used to mean \uDC00-\uDFFF now essentially means \uFFFD-\uFFFD...

Now of course you could argue that this is something UglifyJS should fix on their end, but it has been mentioned several times and they don't seem to ever fix it. They just suggest to use the --ascii-only parameter to suppress this transformation.

I think a library such as this one will be used in a lot of projects where the source code will be uglified. People won't notice that this transformation is invalid and therefore won't use the --ascii-only parameter either.

You can fix this in the library itself by escaping the backslashes. Basically this:

encodechar = new RegExp(['(?:[\\0-\\x1F"-&\\+-\}\\x7F-\\uD7FF\\uE000-\\uFFFF]|',
    '[\\uD800-\\uDBFF][\\uDC00-\\uDFFF]|[\\uD800-\\uDBFF](?![\\uDC00-\\uDFFF])|',
    '(?:[^\\uD800-\\uDBFF]|^)[\uDC00-\\uDFFF])'].join(''), 'g')

Alternatively you can replace it with regex notation. UglifyJS won't transform those:

encodechar = /(?:[\0-\x1F"-&\+-\}\x7F-\uD7FF\uE000-\uFFFF]|[\uD800-\uDBFF][\uDC00-\uDFFF]|[\uD800-\uDBFF](?![\uDC00-\uDFFF])|(?:[^\uD800-\uDBFF]|^)[\uDC00-\uDFFF])/g

So since this technically is not a bug on your end, you can just ignore this if you want. However, I am hoping you will apply this fix to prevent more headaches in the future.

Hi
i try to encode the object that include this word นาย กิตติทัต อุปพงศ์
the right encoded must be %B9%D2%C2+%A1%D4%B5%B5%D4%B7%D1%B5+%CD%D8%BB%BE%A7%C8%EC
but i got %E0%B8%99%E0%B8%B2%E0%B8%A2+%E0%B8%81%E0%B8%B4%E0%B8%95%E0%B8%95%E0%B8%B4%E0%B8%97%E0%B8%B1%E0%B8%95+%E0%B8%AD%E0%B8%B8%E0%B8%9B%E0%B8%9E%E0%B8%87%E0%B8%A8%E0%B9%8C

can you help me with this?

When I use this module and provide an option skipIndex. I see issue with urlencoded data.

The use case that I have is for input payload like x=y&x=y, traditional body parsers convert this into a JSON as {x: [y,y]}. If I convert this into urlencoded data again using this module with option skipIndex set to true. I expect the response to be like x=y&x=y, but instead it is like x[]=y&x[]=y. Is the array indexing is still needed here? Referring the implementation where the indexing added here https://github.com/iambumblehead/form-urlencoded/blob/master/form-urlencoded.js#L27