Make ✨ 🍰 ✨ requests act like jQuery when passing nested objects to data/payload

Once upon a time

I, once was stuck when using python-requests in order to poke a PHP backend.

You may currently face this issue as Requests does not support dict with depth greater than 1 on parameter data for a x-www-form-urlencoded based request.

PHP/jQuery $.params support this syntax out of the box. As requests is not about supporting the way PHP/jQuery $.params work internally, this PR maybe doomed.

To reproduce

from requests import post from pprint import pprint if __name__ == '__main__': response = post( 'https://api.ipify.org?format=json', { 'auth_user': 'test', 'auth_pwd': 'test', 'json_data': { 'operation': 'core/get', 'class': 'ServiceChange', 'key': 'SELECT ServiceChange WHERE id=429', 'output_fields': 'request_state', 'h': { 't': 1 } } } ) pprint( response.request.body ) 

What happen now

print(response.request.body) # auth_user=test&auth_pwd=test&json_data=operation&json_data=class&json_data=key&json_data=output_fields&json_data=h 

Something should happen when passing nested dict. This behavior is wrong. We loss data between passing and sending. At least raise a warning or whatnot.

✨ What should have happened

print(response.request.body) # auth_user=test&auth_pwd=test&json_data%5Boperation%5D=core%2Fget&json_data%5Bclass%5D=ServiceChange&json_data%5Bkey%5D=SELECT+ServiceChange+WHERE+id%3D429&json_data%5Boutput_fields%5D=request_state&json_data%5Bh%5D%5Bt%5D=1 

Why ?

• [x] requests is on a higher level than urllib, this does not make sense in urllib, but can in requests
• [x] PHP and Ruby on Rails support this syntax out of the box, and is used by more than 30 % of the remote backends.
• [x] This PR would make it more beginner friendly, advanced users that are looking for test=foo&test=baz are more likely to encode their payload manually. Not the other way around.
• [x] This change does not break or change actual behavior expect for the multivalued param case which is a very rare case in my opinion.

How to use this patch ?

Simply use my fork for the time being, you could support my PR too.

Patch : tree/patch-1
PR : https://github.com/psf/requests/pull/5253