Mailman 3 June 2003 - Twisted

[Twisted-Python] PyProtocols 0.9
by Phillip J. Eby June 7, 2003

June 7, 2003

Do you hate having to write lots of if-then logic to test what type something is? Wouldn't it be nice if you could just declare "I want this object to have this behavior" and magically convert whatever value you have, to the type you need? PyProtocols lets you do just that, cleanly, quickly, and robustly -- even with built-in types or other people's classes. PyProtocols extends the PEP 246 adapt() function with a new "declaration API" that lets you easily define your own interfaces and adapters, and declare what adapters should be used to adapt what types, objects, or interfaces. In addition to its own Interface type, PyProtocols can also use Twisted and Zope X3's Interface types. (Of course, since Twisted and Zope interfaces aren't as flexible, only a subset of the PyProtocols API works with them. Specific limitations are listed in the documentation.) If you're familiar with Interface objects in Zope, Twisted, or PEAK, the Interface objects in PyProtocols are very similar. But, they can also do many things that no other Python interface types can do. For example, PyProtocols supports "subsetting" of interfaces, where you can declare that one interface is a subset of another existing interface. This is like declaring that somebody else's existing interface is actually a subclass of the new interface. Twisted and Zope don't allow this, which makes them very hard to use if you're trying to define interfaces like "Read-only Mapping" as a subset of "Mapping Object". Unlike Zope and Twisted, PyProtocols also doesn't force you to use a particular interface coding style or even a specific interface type. You can use its built-in interface types, or define your own. If there's another Python package out there with interface types that you'd like to use (CORBA? COM?), you can even create your own adapters to make them work with the PyProtocols API. PyProtocols is also the only interface package that supports automatic "transitive adaptation". That is, if you define an adapter from interface A to interface B, and another from B to C, PyProtocols automatically creates and registers a new adapter from A to C for you. If you later declare an explicit adapter from A to C, it silently replaces the automatically created one. PyProtocols may be used, modified, and distributed under the same terms and conditions as Python or Zope. PyProtocols resources include: * Release notes, installation instructions, and browsable API docs: http://peak.telecommunity.com/protocol_api/ * Source and Binary Releases: http://peak.telecommunity.com/dist/ * Reference Manual (HTML): http://peak.telecommunity.com/protocol_ref/module-protocols.html * Reference Manual (PDF): http://peak.telecommunity.com/protocol_ref.pdf * Browsable CVS Repository: http://cvs.eby-sarna.com/PyProtocols/ Version 0.9 of PyProtocols is a complete, working release, so you can download it now and try it out. But, the exact spellings of the API and other interfaces may be subject to change in the future. We would like feedback from the community on the current API, as some API names may not be clear or obvious yet. Please contact the author with your input.

1 0

[Twisted-Python] Interface.adaptWith and Interface.__adapt__
by Phillip J. Eby June 7, 2003

June 7, 2003

Hi. I'm working on a new interface adaptation package for Python, called PyProtocols. It's an extension of PEP 246 to include a standardized API for declaring support for particular interfaces, and how to adapt instances, types, and other interfaces to an interface. It includes its own interface library, but also has adapters and monkeypatching for Zope interfaces to make them usable with the API. That is, since the API is itself defined with interfaces, it uses itself to adapt Zope interfaces to the interface for interfaces it uses. :) Anyway, I'd like to add support for Twisted interfaces as well. I started today on trying to write adapters to map Twisted's interface/adapter API onto the PyProtocols API, and I ran into a bit of a snag. I was about to write the code to add an __adapt__ method to Twisted interfaces, when I noticed there is already some code related to __adapt__. But that code isn't at all compatible with the PEP 246 spec for __adapt__, as I understand it. PEP 246 wants to be able to call an object's __adapt__ method with a single argument, while MetaInterface expects to yank out 'im_func' and pass two arguments into it. The only way I can see to get this to work with PEP 246, is to monkeypatch MetaInterface.__call__ to find '__adapt__' by checking the interface's dictionary, rather than using __getattr__. In this way, I can give MetaInterface an __adapt__ method that will be found whenever you refer to ISomething.__adapt__, but MetaInterface.__call__ will still find any __adapt__ that was put in by a 'def' in the interface body. Essentially, I would be monkeypatching to change these two lines: if hasattr(self, '__adapt__'): adapter = self.__adapt__.im_func(adaptable, default) To: if '__adapt__' in self.__dict__: adapter = self.__dict__['__adapt__'](adaptable, default) Because self.__adapt__ would no longer be the same value as self.__dict__['__adapt__'], due to my adding a MetaInterface.__adapt__ method. I assume that no inheritance lookup would be required, because interface inheritance for adaptation purposes is reversed. That is, a subclass of an interface would not want to inherit its base interface's __adapt__, because a subclass interface has more stringent requirements for its implementors. Therefore, a superclass __adapt__ isn't guaranteed to work for a subclass. Anyway... I noticed that the only code in Twisted currently (that I could find) that actually uses the current __adapt__ setup is the test suite. Is there any way this might be changeable to something more compatible with PEP 246? I mean, if I jump through enough hoops building adapters and monkeypatching, I can probably make it so that this will work transparently, and anybody who uses PyProtocols will then be able to mix Twisted, Zope, and PyProtocols interfaces in their applications. But it's pretty messy, and I'm wondering about the copyright status if I have to copy a lot of code from MetaInterface.__call__ in order to monkeypatch a replacement. (I'd like to release PyProtocols under the PSF license.) Anyway, if you'd like some background on PyProtocols, the reference manual is at: http://peak.telecommunity.com/protocol_ref/ I don't know if you guys have previously looked at PEP 246 or thought about any of the adapter precedence/implication/declaration issues that PyProtocols tries to address. It may be that some of what I've done may be useful to you. OTOH, you may have already looked at this sort of thing and have different requirements for your component architecture. My goal is merely compatibility among Python frameworks, so that people who use more than one framework's "Interface" types (e.g. Zope, Twisted, PEAK, etc.) can have some common ground rules, similar to the DBAPI or the way most Python numeric types can be used in an expression together.

3 6

[Twisted-Python] t.web.woven.widgets.Anchor and .Link
by screwtape＠froup.com June 7, 2003

June 7, 2003

Playing with Woven the other day, I tried to put a series of links into a page (think "navigation bar"). I spent quite some time coaxing and pleading with the Anchor widget before I discovered the Link widget which did more or less exactly what I wanted. After poking around with the Anchor widget a little more, I've decided it does absolutely nothing at all, since it lacks a setUp method. Is Anchor really as useless as it appears? Is it for backwards- compatibility? Shouldn't some note be made in its docstring? -- ___________ ____________________________ | Screwtape | Reply-To: munged on Usenet |________ ______ ____ __ _ _ _ | | "Earthworms are really good with dirt.. Is this too technical?" -- M*A*S*H |

1 0

[Twisted-Python] Removing broken/unused packages
by Itamar Shtull-Trauring June 6, 2003

June 6, 2003

I propose we remove coil and popsicle. zoot should probably go too, all it does is let you make useless gnutella ping/pong taps, and the actual gnutella protocol code, such as it is, is in twisted.protocols.gnutella. How about sibling? Anyone disagree? -- Itamar Shtull-Trauring http://itamarst.org/ http://www.zoteca.com -- Python & Twisted consulting

2 1

[Twisted-Python] Strange behavior with Woven, getDynamicChild, addSlash
by Alex Levy June 5, 2003

June 5, 2003

I have two Page classes I'm trying to make play nicely, and they won't. class Folder(page.Page): def initialize(self, *args, **kwargs): self.items = kwargs.get('items', {}) def getDynamicChild(self, name, request): return self.items.get(name) class Album(page.Page): # whatever... items = {'album': Album()} resource = Folder(items = items) Straightforward enough, or so I thought. Folder's template creates a list, with an anchor tag for each one of its "items". These links work fine, but when clicked, all the children ignore the fact that addSlash=1. So when I navigate to "/folder", it automatically redirects to "/folder/". However, when "album" is a child returned by my Folder, navigating to "/folder/album" does _not_ add the extra slash -- and I can't figure out why. Navigating manually to "/folder/album/" renders normally. Also, the slash is automatically added when Album is not a child of Folder, but instead its own "base resource" in my web server's docroot. That is, navigating to "/album" redirects to "/album/" normally. [The reason I'm doing this is to make a site where _every_ page shares the same layout; even directory listings need to be templated. Directory listings won't always be tied to real filesystem paths, so extending static.File seems like the wrong approach.] Any ideas? -- Alex Levy WWW: http://mesozoic.geecs.org "Never let your sense of morals prevent you from doing what is right." -- Salvor Hardin, Isaac Asimov's _Foundation_

1 0

[Twisted-Python] twisted newbie question about woven and model
by Philippe Lafoucrière June 5, 2003

June 5, 2003

hi all I'am trying to display some information via a .rpy script. My model is a complex object with several properties such as : object.title object.name object.creationDate ... The only way I found to display title & name is <span model="object" view="None"> <span model="title" view="Text" /> ... </span> Is it possible to access directly object.title ? model="object.title" doesn't work of course. Thanks for your help. -- Philippe Lafoucrière <lafou(a)wanadoo.fr> InFuzzion

3 6

RE: [Twisted-Python] ReverseProxy does not forward GET URL parameters
by Lance Kurisaki June 5, 2003

June 5, 2003

> While you're at it, a set of tests would be great and help you find > remaining bugs ;) I will look into writing one. I needed an excuse to learn unittest... :-) Lance

2 2

[Twisted-Python] Patch to fix issues/web/models.py from CVS
by Patrick K. O'Brien June 5, 2003

June 5, 2003

Index: models.py =================================================================== RCS file: /cvs/Issues/issues/web/models.py,v retrieving revision 1.19 diff -c -u -r1.19 models.py --- models.py 9 May 2003 07:47:38 -0000 1.19 +++ models.py 5 Jun 2003 00:08:52 -0000 @@ -186,7 +186,7 @@ return form.FormDisplayModel(self.commentMethod) def getPerson(self, name): - return self.issue.issue.repository.buildPerson(name) + return self.issue.repository.buildPerson(name) def wchild_postComment(self, request): return form.FormProcessor(self.commentMethod, RedirectToParent) @@ -218,16 +218,16 @@ assert oldFixer is not None and newFixer is not None oldFixer = self.getPerson(oldFixer) newFixer = self.getPerson(newFixer) - self.issue.issue.setState(issue.InTransfer(oldFixer, newFixer)) + self.issue.setState(issue.InTransfer(oldFixer, newFixer)) self._comment(oldFixer, comment) def queue(self, email=None, comment=None, queue=None): self._comment(email, comment) assert queue is not None try: - q = self.issue.issue.repository.getQueue(queue) + q = self.issue.repository.getQueue(queue) except repo.NoSuchQueue: - q = self.issue.issue.repository.buildQueue(queue) + q = self.issue.repository.buildQueue(queue) self.issue.setState(issue.InQueue(q)) def converse(self, fixer=None, finder=None): Also attached as issues_models.diff. -- Patrick K. O'Brien Orbtech http://www.orbtech.com/web/pobrien ----------------------------------------------- "Your source for Python programming expertise." -----------------------------------------------

2 2

[Twisted-Python] Persistant Identities CRED
by Philippe Lafoucrière June 5, 2003

June 5, 2003

[from another post] I was looking for documentation on cred, since I want to persist identies created on a twisted server. The cred howto needs some rewrite with concrete exemples. I can't pickle identities with Shelf since Indenties have references to Auth, and Auth to Identities (infinite loop when pickeling). -- Philippe Lafoucrière <lafou(a)wanadoo.fr> InFuzzion

1 0

[Twisted-Python] mishandling windows connect error
by Andrew Dalke June 5, 2003

June 5, 2003

That error report I made earlier this evening is not reproducible under Linux. It's only under MS Windows. Here's some more information. Twisted isn't reporting the correct error message for when a TCP connection fails under MS Windows. Here's the reproducible E:\dalke>..\Python22\python.exe Python 2.2.3c1 (#41, May 22 2003, 19:16:28) [MSC 32 bit (Intel)] on win32 Type "help", "copyright", "credits" or "license" for more information. >>> from twisted.web.xmlrpc import Proxy >>> from twisted.internet import reactor >>> def printValue(value): ... print value ... reactor.stop() ... >>> proxy = Proxy("http://beatty.userland.com:34686") >>> proxy.callRemote('test.sumprod', 3, 5).addCallbacks(printValue, printValue) <Deferred at 0xc38f40> >>> reactor.run() [Failure instance: Traceback: twisted.internet.error.ConnectError, [160] ] >>> Under Linux, this reports "Connection refused" as expected. The ConnectError contains the value [160]. Looking at the stack trace, I found it was created in tcp.py in BaseClient.doConnect where it says # on windows failed connects are reported on exception # list, not write or read list. if platform.getType() == "win32": r, w, e = select.select([], [], [self.fileno()], 0.0) if e: self.failIfNotConnected(error.getConnectError(e)) return If there is an error then 'e' will contain a list of size one, which is the fileno. Now look at the getConnectError definition in internet/error.py def getConnectError(e): """Given a socket exception, return connection error.""" try: number, string = e except ValueError: return ConnectError(string=e) This will always give a ValueError (list of size 1 can't be assigned to two terms) so it returns a ConnectError with the 'string' set to the single element list. And this is the '[160]' reported in the ConnectError of the failure, which means it was on fileno 160. I do not claim to be a network person. It's all confusing to me. I tried commenting out that block, to see what the connect_ex does. It returns a EWOULDBLOCK. That's as far as I took it. It feels to me like there's a bug around there someplace. Andrew dalke(a)dalkescientific.com

3 4