What is the difference between those two requests?

Question

For a task automatization I needed to write some python code, which makes post request to a web service, then retrieves information etc..

First I tried to use requests.

import requests

def make_invoice_body(terminal, service, code, number):
    return {
        "terminalId": terminal,
        "serviceId" : service,
        "invoiceNumber": number,
        "invoiceCode": code,
        "paymentType": 1,
        "requestNumber": random.randint(1000000, 9999999)
    }

headers = {'Content-Type': 'application/json'}
body = make_invoice_body(51, "1001000", "FOO", "123456")

requests.post(url, headers=headers, data=body)   

Above code piece was recieved error as a response.

Then I tried with urllib2,

import urllib2
import json

req = urllib2.Request(url)
req.add_header('Content-Type', 'application/json')
resp = urllib2.urlopen(req, body) #body from above
print resp.read()

and it worked.

Now I am curious about differences of above snippets. Shouldn't they do the same task and recieve equivalent answers?

Answer 1

There is no "native dict" format in HTTP world.

The real difference is that requests is too smart.

When using data= in requests, you are trying to post multipart/form-data. In this case, a dict will be unzipped into key-value pairs in form-data format.

When using json= in requests, requests will automatically json.dumps your dict into a raw byte string and set Content-Type for you.

But in urllib, it does nothing more than you give. You need to manually json.dumps your dict. So I believe the example you give is wrong. it should be like the following:

import urllib2
import json

req = urllib2.Request(url)
req.add_header('Content-Type', 'application/json')
resp = urllib2.urlopen(req, json.dumps(body)) #body from above
print resp.read()

Thanks for @t.m.adam's reminding.