Content-Length is missing #223
Comments
|
Should be good now |
import requests
URL = "http://github.com"
g_url = "http://clients6.google.com/rpc?key=AIzaSyCKSbrvQasunBoV16zDH9R33D88CeLr9gQ"
params = {
'method': 'pos.plusones.get',
'id': 'p',
'params': {
'nolog': 'true',
'id': URL,
'source': 'widget',
'userId': '@viewer',
'groupId':'@self'
},
'jsonrpc': '2.0',
'key': 'p',
'apiVersion': 'v1'
}
headers = {
'Content-type': 'application/json'
}
r = requests.post(g_url, params=params, headers=headers)
print r.request.headerscontent length still isn't set in this code with version 0.8.3 |
|
You're not uploading any body. Why would there be a content-length? |
|
kennethreitz, iirc - the Content-length header should always be set (even if it's zero) when a POST is involved. But by looking at the code pasted by zoranzaric, he is indeed sending a POST with a request body - so the content-length should surely be included? Am I missing something here?? Cal |
|
foxx, I thought setting params was correct for a POST. What am I doing wrong? kennethreitz, the Google API that I'm talking to responds with 411 (Length Required) |
|
@zoranzaric: |
|
@kennethreitz ok yeah snap... with |
|
@zoranzaric, no worries! A lot of other libraries are quite inconsistent, so I completely understand the confusion. That's the whole reason I started Requests :) |
|
Ah - sorry I missed the 'params' / 'data' difference. My bad! |
|
This is still broken with regard to sending a blank POST request: r = requests.post(full_uri) While a Content-Length header is sent, the value is blank (it should be 0 per the RFC). The request as the server see is below: |
|
Are you working on the invalid content-length for POST requests without data payload issue? I was just bit by this bug while trying to POST and PUT to a WCF service. I could luckily work around it by specifying a content-length of 0 via the 'headers' keyword argument to the 'post' and 'put' functions. |
|
@aknuds1 what version are you using? This is what I see in trunk: |
|
@slingamn Ah, maybe it's fixed in trunk then. I'm just using the version installed via pip, i.e. 0.11.2. |
|
Oh, weird, I get Can you try the above test case and post the result? |
|
I tried your case and get the following output: Is there a bug at httpbin.org then, since there is clearly a difference in the request when I define the content-length header myself? |
|
Sorry, I'm confused. Isn't that the expected output? |
|
@slingamn It's the desired output, yes, but why is there a difference towards WCF when I define the content-length header myself? I mean, httpbin.org reports that the request has defined content-length as "0", even though the same request directed at WCF fails with error 411. |
|
I'm inclined to suspect a bug with the WCF service, since I get the same result using http://requestb.in. and Requestbin saw the following headers: |
|
But there must be a difference wrt. content-length, since defining that header myself in the client library (requests) results in a request accepted by WCF. I'll see if I can get hold of the POST request received by WCF. |
|
One possible explanation: this results in the headers appearing in a different order, and the WCF service (incorrectly) takes note of this. To get the POST request, you could (as a last resort) try something like Wireshark. |
|
According to Fiddler, the POST request without user-defined content-length header looks like so: If I define the content-length header, via the 'headers' argument to requests.post, however, the request looks like so: |
|
Interesting! The line of code that produces the bad behavior is just |
|
I'm on Python 2.7.2. Here's what I see in my httplib.py: def _set_content_length(self, body):
# Set the content-length based on the body.
thelen = None
try:
thelen = str(len(body))
except TypeError, te:
# If this is a file-like object, try to
# fstat its file descriptor
try:
thelen = str(os.fstat(body.fileno()).st_size)
except (AttributeError, OSError):
# Don't send a length if this failed
if self.debuglevel > 0: print "Cannot stat!!"
if thelen is not None:
self.putheader('Content-Length', thelen)If you can't provide a reproducible test case for security/privacy reasons (totally understandable), can you try and monkeypatch your httplib and see if that |
|
My code is as simple as this (the HTTP proxy is Fiddler): If you run this script against Fiddler on your machine, you would probably see the same request as I do. I can't see that httplib._set_content_length is called at all. I injected some code in there to see if it was called by requests, but that code wasn't hit. |
|
And Fiddler shows a |
|
|
We've clarified this in HTTPbis (although it's there in 2616 too). Content-Length is not required on requests; if it's missing, the default is a length of 0. See: |
|
I need to set the "Content-Length" for a GET request. Anybody knows how to fix this? |
|
@nettok something along the lines of: from requests import get
r = get('http://httpbin.org', headers={'Content-Length': '0'})But of course if you're sending data, it wouldn't be zero. If you're sending a body, e.g., r = get('http://httpbin.org/get', data='spam')
r.json['headers']['Content-Length'] == '4' # True |
|
Also, just realized this had bitten me a while back here. If @kennethreitz still wants a PR attaching Content-Length to everything, I'd be happy to do so. |
|
Yes. |
|
@kennethreitz where should I branch from, develop or adapters? |
|
develop, adapters will be a while :) |
|
Roger that. |
|
I am getting a 411 while using netflix API too. I am getting this while trying to get the request token using the requests-oauth module. Things were working fine, but the API end point have changed recently which may indicate additional changes in server side as well. Code: https://github.com/amalakar/pyflix2/blob/master/pyflix2/pyflix2.py#L76 Surprisingly even though I am passing post data as Headers sent are: I am on requests-0.14.0, not sure if the issue is in requests or the oauth module. I would be happy to provide additional information. |
|
@kennethreitz Looks like the content-length is being set in case of GET request as well. Where the data being passed is going as query params in the URL. This could be a 400 Bad Request case (which is what is happening for me). Following is an example of headers sent: And this was for a GET request. |
|
Can you provide the call @amalakar? This must be a mistake on my part and I'd like to fix it. |
|
To clarify @amalakar, testing against HTTPBIN, with which is a bit bizarre but probably a bug on HTTPBIN's end. (The fact that Content-Length is |
|
Notice I am using data argument here, not params. The server side script is a simple php script that prints all the http headers it received in the http request. You can see that it got the Content-Length as 7. |
|
Well this seems to be more of a problem with the server if I'm reading RFC https://tools.ietf.org/html/rfc2616#page-119 The Content-Length entity-header field indicates the size of the An example is Applications SHOULD use this field to indicate the transfer-length of (And section 4.4 doesn't mention anything about sending data on a GET Any Content-Length greater than or equal to zero is a valid value. Note that the meaning of this field is significantly different from |
|
Yup, the server is doing the wrong thing here. (Though, in its defense, your request is semantically meaningless, as @piotr-dobrogost has pointed out. =D ) |
Well, on the contrary. The semantics of such a request is clearly defined :) |
|
Well from the documentation I get the following:
Given this can I conclude that if I provide the dictionary as Well I have another problem it seems. So far I have been using the https://github.com/maraujop/requests-oauth and the Notice that here the data dictionary has become query params as seen in the URL. This seems inconsistent wrt ``requests.get` in which case data dictionary is not sent as query params but as the body. |
|
As far as I know, data has always been for the body and parameters for the query string. requests-coauthor must have messed that up unless he did it purposefully. |
Please see the following example
In this example, the 'Content-Length' is missing.
Haven't got any spare time to try and patch the bug - as this was only a quick test to explore new libs.
Cal
[revised post 3 - multiple edits due to mistake on bug report]
The text was updated successfully, but these errors were encountered: