Mailman 3 November 2003 - Twisted

[Twisted-Python] Woven web question about clearing http request.args
by Hegedus, Matthew S 05 Nov '03

05 Nov '03

I've got multiple input buttons and their view is the same logic. The logic is delete some data on a button press, then rerender the page with the updated data. The data it will remove depends on the unique ID of the button pressed. When the web page is rerendered the view gets called for each button with the same http request, so the deletion logic will be evaluated for each button, even though it should be evaluated for just the button that was pressed. I don't like that. I want to clear request.args so the behavior will be as if there was only one http event, corresponding to a single button press. I tried "request.args.clear()", but that doesn't seem to have the expected effect. What's worse is when I click the browser's reload button, request.args still has the last http request that occurred. In other words, if I click my delete button the page will be rerendered with updated data, then if I click the browser's reload it will try to delete more data as if I had hit on of my delete buttons again. Can anyone help me with this problem? _______________________________________________ Twisted-Python mailing list Twisted-Python(a)twistedmatrix.com http://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python

2 1

[Twisted-Python] form and submit
by vicky lupien 03 Nov '03

03 Nov '03

I have a form with a submit button. I want to send my data from the form in a web page to an other web page. Also, I want to check if the data is correct. I'm new to twisted, so I need some help Vicky _________________________________________________________________ MSN Messenger : discutez en direct avec vos amis ! http://messenger.fr.msn.ca/

2 1

[Twisted-Python] Dynamic form with twisted
by vicky lupien 03 Nov '03

03 Nov '03

I really need help. I want to make a form dynamically. I don't want to use FormProcess and I do not understand very well how it works. If someone have documentation on forms (not the one on twisted matrix), it will be appreciated. Vicky _________________________________________________________________ MSN Messenger : discutez en direct avec vos amis ! http://messenger.fr.msn.ca/

3 2

[Twisted-Python] design issue wrt exceptional events
by Dennis Allison 03 Nov '03

03 Nov '03

I need a front-end to a complex of Zope systems. The front-end listens on an address/port and examines incoming http transactions, and, based upon http headers and other information, connects to one or another of the Zope systems. The underlying approach is straightforward -- a Listener waits for a http transaction, it spawns a receiver which gathers the message, decides what to do, spawns a sender to the destination. When the sender responds, the process is reversed. I currently use an adaptation of Antony Baxter's pydirector to do the job. I use his asynchat version, but have found that it leaks file descriptors in my application. I've not tried the twisted version. So I thought I'd try to craft a tailored, light-weight version using the twisted framework. I see how to manage things when everything works, but what if one or another of the endpoints fails? How do I get everything cleaned up?

2 3

[Twisted-Python] Adapting RemoteReferences
by Sam Jordan 03 Nov '03

03 Nov '03

Hi, What is the general opinion on having a standard way to adapt RemoteReferences to specific interfaces which basically just wraps callRemote for each respective method? The reason i'm bringing this up is I can imagine various use-cases for wanting to treat remote objects and local objects as semantically equivalent. For example, I may be employing the observer pattern for a simple system to separate UI components from data components. Currently if I wish to have both remote and local observers interoperate I would have to either store them in separate data structures and treat each differently when sending notifications, or, I could store them in a single data structure and use appropriate type-checking to dispatch the calls. IMHO both of these solutions are relatively ugly. Whether or not there are sufficient use-cases for a standard twisted way to do this is another story, so i'm wondering what the views of people are on this subject. Certainly one could just register their own adapter for RemoteReference in each case, but the code to do this is rather ugly and for those that _do_ have to deal with such cases I think it would be nicer to have a simple, standard way of doing it, e.g. remoteRef.adaptTo(Interface). I've written a function that outlines the functionality I think it should provide which is included below; while this implementation is rather ugly i'm more worried about the idea rather than how it would eventually be implemented if someone decides this is worth having. Of course, there are certain constraints, such as the method wrappers having no way of knowing if the remote object actually implements the desired interface or not, but I don't see this as a problem for something that already requires string based dispatch. I guess you could say this proposal is somewhat similar to java's 'stub' approach, only it doesn't force it :-). Cheers, Sam. ------------------------------------------ import types class Interface: def add(x,y): pass def mul(x,y): pass class Interface2(Interface): def div(x,y): pass def sub(x,y): pass class Foo: def callRemote(self, name, *args, **kw): print 'calling %s with %s %s' % (name, args, kw) # this would likely put put in RemoteReference # to provide context def adaptTo(self, interface): # obtain all interface methods methods = interface.__dict__ for base in interface.__bases__: for attrib,val in base.__dict__.iteritems(): if not methods.has_key(attrib) and isinstance(val, types.FunctionType): methods[attrib] = base.__dict__[attrib] class _Adapter: def __init__(self, original): self.original = original for m in methods: exec "def %s(self, *args, **kw): return self.original.callRemote('%s', *args, **kw)" % (m,m) del _Adapter.m del _Adapter.methods return _Adapter(self) def test(): a = adaptTo(Foo(), Interface2) a.add(1,2) a.mul(2,4) a.div(8,4) a.sub(3,5) test() ------------------------------------------

3 3

[Twisted-Python] Socket Read Write
by Ted Patrick 03 Nov '03

03 Nov '03

I am writing a chat socket server and need to make sure that there are no conflicts when writing to other attached clients. I am concerned that when the server receives lots of messages from many socket connections, several sockets may encounter write contention. Is there a queue for the write method of a socket? Should one expect errors in this regard? Does the Twisted architecture eliminate this class of error? Using twisted.protocol.basic can 2 lineReceived events happen simultaneously? How is this addressed? Thanks, Theodore E. Patrick PowerSDK - Code for a distributed world http://www.powersdk.com

2 1

[Twisted-Python] SNMPProtocol, request for review...
by Mike C. Fletcher 02 Nov '03

02 Nov '03

I've built an SNMP Protocol for Twisted (based on a pattern from Patrick K. O'Brien), but as I'm a rather new Twisted developer, I'm wondering if someone more "in the know" could take a look at the package and see if I got the "ideas" right. That is, whether I'm going about the entire thing incorrectly/correctly, or whether I've missed anything that would be obvious to an experienced twisted developer. The protocol provides for sending get, getnext and getbulk requests, but doesn't yet support trap or answering any query (i.e. in SNMP-eese it's a manager-only protocol at the moment, not an agent-side protocol). I'm particularly interested in: * whether the timeouts are handled properly (we do a crude callLater and check for deferred.called for them), is there a more elegant Twisted mechanism for handling timed-out operations? * whether the message IDs are handled properly (we just store all "live" messages in a dictionary and delete them when they are timed out or responded to), is there some more elegant Twisted solution? * whether the entire approach is okay. I put most of the code and API on the protocol object, with a helper object for managing tabular downloads. Is there a more "Twisted" way of doing this? Should the get method on the protocol be implemented as a separate object driving the protocol as well? What should be in the "protocol" level, (as opposed to higher-level "application" code)? Package is here: http://members.rogers.com/mcfletch/programming/TwistedSNMP-0.1.tar.gz It requires pysnmp-3.3.5, available here: http://download.sourceforge.net/pysnmp/ You'll need an SNMP agent against which to test (as well as some OIDs for that agent), as the ips in the scripts won't work without the proper community IDs (or from unrecognised manager IPs, for that matter). Feedback appreciated, Mike _______________________________________ Mike C. Fletcher Designer, VR Plumber, Coder http://members.rogers.com/mcfletch/

2 1

[Twisted-Python] connectTCP/listenTCP socks/ssh
by Kev 01 Nov '03

01 Nov '03

I'm writing a socks proxy that uses ssh tunnels for connections, I can get it working by making my connectClass listenTCP with a forwarding.SSHListenForwardingFactory and then ClientCreator a SOCKSv4Outgoing to connectTCP, but obviously that's a bad method. How can I cut the listen/connect? Is there a way I can do it at that level or do I need to implement my own .connectSSH method? It seems I need to get both sides placed in the reactor somehow, but SSHChannel.conn doesn't have a fileno() so addReader doesn't accept it. (Is addReader the right function? I found it in the curses example and am assuming it's just a generic file descriptor handler) Thanks much

2 2