Develop Reference

How to get request hostname from an HTTP-triggered, Python Azure Function request header?

I would greatly benefit from a list of all the available headers that App Service can forward to my (keyword PYTHON) Function. Or if someone knows how to "list-all", that would be awesome.
Through asking questions on SO, I see that the request IP addressed can be gleaned using:
req.headers.get("X-FORWARDED-FOR").
I need the Hostname that a request is coming from.
Looks like this is possible using C# Functions. But I either did it wrong using req.headers.Host or its not available for Python.
Is it possible using Python?

For this requirement, you just need to use req.headers.get("host"). I test it in my side, it works fine on azure portal.

GAE unsubscribe from a user's presence

Is there a way to unsubscribe from a user's presence? I no longer want to receive updates on /_ah/xmpp/presence/... for a particular user. I can't seem to find a simple API call to do that.
After digging around the XMPP protocol I found this which seems to indicate that doing a send_presence with presence type of 'unsubscribe' should work. Unfortunately digging into the GAE's xmpp API it appears that it defines
_VALID_PRESENCE_TYPES = frozenset([PRESENCE_TYPE_AVAILABLE,
PRESENCE_TYPE_UNAVAILABLE,
PRESENCE_TYPE_PROBE])
Which means I can't even do a send_presence(user_to_remove, status="", presence_type="unsubscribe") (PRESENCE_TYPE_AVAILABLE and others are just strings like "available" as per the xmpp specificiation)
Has anyone come across this issue or know how to achieve this ?

It seems that you can't. The docs (and the docstring) confirm that presence_type accepts a subset of the types defined in RFC 3921.
You can submit this as a feature request to the issue tracker.

As an experiment, you could re-implement your own "send_presence" that does the same thing as the existing function, without the check for valid presence types. Not officially sanctioned but worth a try.
One thing to note is that this won't block clients from re-subscribing from your bot or from badly-behaved clients ignoring it.
And as Drew mentioned, please do submit an issue on the issue tracker.

How should support for alternate credential types in twisted.pb be implemented?

My project has been trying to implement a credential checker using scrypt. We've tried implementing our own credentials and checker objects, but we've had a lot of trouble getting pb to use them.
Pb seems hard-coded to use MD5 hashes over the wire, which absolutely won't work in our implementation; we don't have a way to get the correct password in plaintext on the server side, since we're using scrypt, so we need a way to transmit the password to be verified in plaintext instead. We've tried using twisted.cred.credentials.UsernamePassword with our credential checker, but it doesn't seem to make it to the server. (we still get _PortalAuthChallenger instead)
The ticket at http://twistedmatrix.com/trac/ticket/4398 seems to indicate that a PBServerFactory subclass is needed in order to support custom credential checkers in pb, but so far I have been completely unable to figure out what to override in order to make it use a different ICredentials implementation. Are there any examples (or even just documentation) of how to get pb to use a different credentials class?

PB isn't exactly hard-coded to use MD5 hashes over the wire; that's just the authentication protocol as it's currently implemented. You can do pretty much whatever you want by implementing your own authentication protocol - which, in PB, just means an object that you get to call some authentication methods on.
Make your own object that implements IPBRoot, and pass it to PBServerFactory. This just means you need to implement a method called rootObject which returns the root object for a particular connection (and then declare that implementation with Zope Interface, of course).
Your IPBRoot implementation should wrap a Portal, similar to similar to _PortalRoot in Twisted's implementation.
Then, make a remote method on the object returned from rootObject suitable to your application; maybe something like remote_loginPlaintext. In this method you can authenticate users however you want, then call login on your particular Portal with whatever credentials are derived from that interaction and make sense for your requirements (and whatever interface, although for hopefully obvious reasons, IPerspective is what I'd recommend).
The fact that the somewhat inflexible _PortalRoot (which only supports 2 credential types; IAnonymous and IUsernamePassword) is registered as the adapter for Portal, making it seem a bit more official than it really is. Don't think of it is as the "official" PB/Cred integration mechanism, just the "default" one.
It would be great if you could contribute a more flexible authentication mechanism (perhaps a full SASL implementation?) for PB so that we could support other authentication types. I hope that you'll consider doing that when your application's particular needs are met.

Here's a link to the preliminary fix we came up with: http://paste.skewedaspect.com/show/20/
Note that this requires the custom credential to be Copyable, and allows the default MD5 key exchange behavior to be controlled by the keyword arg useMD5Challenge.
Note: in our implementation we're leaving the checking entirely to the Checker and having our Credential object contain nothing but username and password, so no actual code is being serialized.

Python - Open default mail client using mailto, with multiple recipients

I'm attempting to write a Python function to send an email to a list of users, using the default installed mail client. I want to open the email client, and give the user the opportunity to edit the list of users or the email body.
I did some searching, and according to here:
http://www.sightspecific.com/~mosh/WWW_FAQ/multrec.html
It's apparently against the RFC spec to put multiple comma-delimited recipients in a mailto link. However, that's the way everybody else seems to be doing it. What exactly is the modern stance on this?
Anyhow, I found the following two sites:
http://2ality.blogspot.com/2009/02/generate-emails-with-mailto-urls-and.html
http://www.megasolutions.net/python/invoke-users-standard-mail-client-64348.aspx
which seem to suggest solutions using urllib.parse (url.parse.quote for me), and webbrowser.open.
I tried the sample code from the first link (2ality.blogspot.com), and that worked fine, and opened my default mail client. However, when I try to use the code in my own module, it seems to open up my default browser, for some weird reason. No funny text in the address bar, it just opens up the browser.
The email_incorrect_phone_numbers() function is in the Employees class, which contains a dictionary (employee_dict) of Employee objects, which themselves have a number of employee attributes (sn, givenName, mail etc.). Full code is actually here (Python - Converting CSV to Objects - Code Design)
from urllib.parse import quote
import webbrowser
....
def email_incorrect_phone_numbers(self):
email_list = []
for employee in self.employee_dict.values():
if not PhoneNumberFormats.standard_format.search(employee.telephoneNumber):
print(employee.telephoneNumber, employee.sn, employee.givenName, employee.mail)
email_list.append(employee.mail)
recipients = ', '.join(email_list)
webbrowser.open("mailto:%s?subject=%s&body=%s" %
(recipients, quote("testing"), quote('testing'))
)
Any suggestions?
Cheers,
Victor

Well, since you asked for suggestions: forget about the mailto: scheme and webbrowser, and write a small SMTP client using Python's smtplib module. It's standard, fully supported on all systems, and there's an example included in the documentation which you can practically just copy-and-paste pieces out of.
Of course, if you're using smtplib you will have to ask the user for the details of an SMTP server to use (hostname and port, and probably a login/password). That is admittedly inconvenient, so I can see why you'd want to delegate to existing programs on the system to handle the email. Problem is, there's no system-independent way to do that. Even the webbrowser module doesn't work everywhere; some people use systems on which the module isn't able to detect the default (or any) browser, and even when it can, what happens when you provide a mailto: link is entirely up to the browser.
If you don't want to or can't use SMTP, your best bet might be to write a custom module that is able to detect and open the default email client on as many different systems as possible - basically what the webbrowser module does, except for email clients instead of browsers. In that case it's up to you to identify what kinds of mail clients your users have installed and make sure you support them. If you're thorough enough, you could probably publish your module on PyPI (Python package index) and perhaps even get it included in a future version of the Python standard library - I'm sure there are plenty of people who would appreciate something like that.

As is often the case in Python, somebody's already done most of the hard work. Check out this recipe.

In the following line, there shouldn’t be a space after the comma.
recipients = ', '.join(email_list)
Furthermore, Outlook needs semicolons, not commas. Apart from that, mailto never gave me grief.
The general tip is to test mailto URLs manually in the browser first and to debug URLs by printing them out and entering them manually.

Python: Sending a large dictionary to a server

I have an application that should communicate status information to a server. This information is effectively a large dictionary with string keys.
The server will run a web application based on Turbogears, so the server-side method called accepts an arbitrary number of keyword arguments.
In addition to the actual data, some data related to authentication (id, password..) should be transmitted. One approach would be to simply urlencode a large dictionary containing all this and send it in a request to the server.
urllib2.urlencode(dataPlusId)
But actually, the method doing the authentication and accepting the data set does not have to know much about the data. The data could be transmitted and accepted transparently and handed over to another method working with the data.
So my question is: What is the best way to transmit a large dictionary of data to a server in general? And, in this specific case, what is the best way to deal with authentication here?

I agree with all the answers about avoiding pickle, if safety is a concern (it might not be if the sender gets authenticated before the data's unpickled -- but, when security's at issue, two levels of defense may be better than one); JSON is often of help in such cases (or, XML, if nothing else will do...!-).
Authentication should ideally be left to the webserver, as SpliFF recommends, and SSL (i.e. HTTPS) is generally good for that. If that's unfeasible, but it's feasible to let client and server share a "secret", then sending the serialized string in encrypted form may be best.

I think the best way is to encode your data in an appropriate transfer format (you should not use pickle, as it's not save, but it can be binary) and transfer it as a multipart post request
What I do not know if you can make it work with repoze.who. If it does not support sign in and function call in one step, you'll perhaps have to verify the credentials yourself.
If you can wrap your data in xml you could also use XML-RPC.

Why don't you serialize the dictionary to a file, and upload the file? This way, the server can read the object back into a dictionary .

Do a POST of your python data (use binary as suggested in other answers) and handle security using your webserver. Apache and Microsoft servers can both do authentication using a wide variety of methods (SSL client certs, Password, System accounts, etc...)
Serialising/Deserialising to text or XML is probably overkill if you're just going to turn it back to dictionary again).

I'd personally use SimpleJSON at both ends and just post the "file" (it would really just be a stream) over as multipart data.
But that's me. There are other options.

Have you tried using pickle on the data ?