Mailman 3 February 2004 - Twisted-web

fcgi.py
by robin bryce 14 Feb '04

14 Feb '04

I needed to create a bridge from apache to a server implemented using the Twisted framework. I took a copy of fcgi.py from http://alldunn.com/python/fcgi.py and made come crude modifications to get it to work with the Twisted factory / protocol pattern. I got in touch with the orignal author, Robin Dunn, and he is happy for me to contribute the modification. Is there any interest here in this or is there a better way to get apache talking to a twisted server ? Best regards, Robin Bryce

3 3

How to upload files using HTML forms?
by Frantisek Fuka 13 Feb '04

13 Feb '04

How do I detect the files that were uploaded from WEB using Forms? When I look at request.args, I can see the filenames of the files, but where are the uploaded files themselves? I am also curious where in the documentation could I find answers to questions like these? I looked into server.py source code and didn't find any documentation or examples... -- Frantisek Fuka (yes, that IS my real name) (and it's pronounced "Fran-tjee-shek Foo-kah") ---------------------------------------------------- My E-mail: fuka(a)fuxoft.cz My Homepage: http://www.fuxoft.cz My ICQ: 2745855

1 0

allowNone for formless
by James Y Knight 13 Feb '04

13 Feb '04

Attached patch adds allowNone argument for Typed objects in formless. Tests updated accordingly. Also, boolean values are handled more sensibly (formless's boolean type coerce takes 'True' or 'False' strings instead of making freeform translate). Also added a python enumerate implementation for compatibility with python2.2 since enumerate is used in freeform.py. James

2 1

Cosmetic patch to nevow/freeform.py
by Alex Levy 12 Feb '04

12 Feb '04

This is insignificant, but helpful. Removes the description <div> element when there is no description. Makes things look a little more concise. Index: nevow/freeform.py =================================================================== RCS file: /cvs/Quotient/nevow/freeform.py,v retrieving revision 1.108 diff -u -r1.108 freeform.py --- nevow/freeform.py 11 Feb 2004 01:33:09 -0000 1.108 +++ nevow/freeform.py 12 Feb 2004 17:31:20 -0000 @@ -221,12 +221,14 @@ if data.typedValue.getAttribute('compact'): return theslotfilled - return [ + ayb = [ label(_class="freeform-label")[ data.label ], theslotfilled, # freeform-input span(_class="freeform-error")[ errormessage ], - div(_class="freeform-description")[ data.description ], ] + if data.description: + ayb[div(_class="freeform-description")[ data.description ]] + return ayb def input(self, context, slot, data, name, value): raise NotImplementedError, "Implement in subclass" -- 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 1

Slot/pattern/key changes
by James Y Knight 11 Feb '04

11 Feb '04

This is a set of *BACKWARDS-INCOMPATIBLE* changes that simplifies and clarifies the set of attributes nevow has for finding elements. Tests pass, and my app works (both after being updated slightly for the API changes). There used to be 3 attributes you could put on elements to locate them: slot, pattern, and key. They all did similar but slightly different things. - Filling slots with data is a useful concept but 'slot' shouldn' t be an attribute. A slot is something that gets replaced with the given data, and is thus better as a standalone element. It has been replaced with the following: slot('slotname') in a stan tree, or <nevow:slot name="slotname" /> in an (X)HTML template. context.slotted() (which could locate a slot) has been eliminated and replaced with context.fillSlots(slotname, stantree). Note that you cannot locate slots, you can just specify data for them to fill themselves with. - Assigning unique keys is also a useful concept, but it doesn't require the "context.keyed()" method. That method has been eliminated. HTMLRenderer used it to find the key="content" in HTML files used as sub-renderers, now it uses pattern="content". - Patterns remain as the only way to locate a sub-element of your renderer. There are now 3 methods to locate them, depending on what you want to do. patternGenerator() replaces patterns(), onePattern() semi-replaces keyed(), and allPatterns() replaces the direct use of specialMatches that the sequence renderer previous did. Please see the docs in context.py for these three methods for precise behavior. So, in summary, to convert your old code to new code: 1) If you had "div(style="border:1px", slot="myslot")" replace it with div(style="border: 1px")[slot("myslot")]. 2) If you had a template with "<div style="border:1px" nevow:slot="myslot" />" replace it with "<div style="border:1px"><nevow:slot name="myslot" /></div> 3) If you had "context.slotted("myslot").clear()[a(href="foo")["bar"]] replace it with context.fillSlot("myslot", a(href="foo")["bar"]] 4) If you used keyed(), convert key= to pattern= and keyed() to onePattern(). 5) If you used patterns(), just rename it to patternGenerator() Also included in this patch are doc cleanups, removing excess xml() wrappings, replacing yield by return when possible, and the ability to use the "<nevow:invisible>" element in a (x)html template file, with the same behavior as the invisible stan tag. James

2 3

RE: Writing a web application - how to make twisted/nevow "look good"
by Joel Boehland 11 Feb '04

11 Feb '04

Hi, I think developing some sort of widget library would be useful to the nevow community. I have done a little tinkering with nevow UI widgets as well. Mostly I'm deconstructing some of the ui elements in plone2.0 and trying to convert them to nevow. I've gotten their action-tabs, tables and portal boxes basically working. Anyhow I'd be interesting in some sort of project like this. The question is how should this be done? Any sort of reasonably complex web ui is going to have elements of nevow + xhtml+ js + css. The js and css may also be dynamic. How would be create a catalog/library to hold these ui widgets? Would it be a web demo sort of like pastebin where you can see a widget, and then cut-n-paste the code for the nevow/css/js/xhtml? Or would it be a more formalized library that you would download. I like the idea of a web-based catalog, since you could see how each of the widgets looks. Anyhow, I like the idea of some sort of widget library or catalog. --Joel

3 2

Writing a web application - how to make twisted/nevow "look good"
by Paul Moore 09 Feb '04

09 Feb '04

I've been playing with twisted, and lately nevow, for a while now. My idea is that I'd like to be able to develop a replacement for an in-house database application, which is written in Oracle's HTML DB. The application code side of things is easy - it's a pretty simple application, with a few screens of tabular data (grab data from a SQL table, and display it in a HTML table), and corresponding update and "create new entry" forms. Plus delete buttons. I've not written the code in all its glory yet, but it's not going to be taxing. The problem is getting a good look and feel for the whole thing. My design skills are pretty ropey at best, and worse still when I'm coding HTML. But Oracle HTML DB comes with built-in flashy-looking designs - little "task list" boxes with rounded corners, tabbed webpages, etc. If I can't produce a nice look and feel, my Twisted code is going to look clumsy and amateurish in comparison - even though from a technical point of view, it may be more responsive, more maintainable, etc etc. But HTML DB applications *look* good, even when developed by someone like me, with the design skills of a prune. I know all the nice look comes from clever use of images, tables, CSS etc. I've read the page source for some of the HTML DB stuff, and it's an unreadable mess (probably because it's generated). So stealing it isn't really a possibility :-) I was wondering - is there anywhere an example of an open source Twisted/Nevow application which focuses strongly on look & feel, and which can usefully be used as a source of sample code? Alternatively, a nice idea might be a "widget library" in Nevow - little classes which can be used as higher level building blocks in stan. An example might be something like: document = ..., headed_list("The title")["line1", "line2"], ... The "headed list" could render as a table - something like <div> <table> <tr style="background-color: yellow"><td>The title</td></tr> <tr><td>line1</td></tr> <tr><td>line2</td></tr> </table> </div> (I told you my design skills weren't much good :-)) The idea being that you can build your pages at a higher level than raw HTML layout. And at the other extreme, I guess it should be possible to define a "page template", which defines a site's overall look and feel - tabbed format, left hand navigation bar, whatever. OK, I'm rambling now. I haven't the skill to build anything like this, but I'd *love* to be able to use such a thing. Any suggestions would be gratefully accepted! Paul. -- This signature intentionally left blank

1 0

Revamped ContextSerializer/TagSerializer and other contextish stuff
by James Y Knight 06 Feb '04

06 Feb '04

The HTTP output encoding is set to UTF-8 in renderer, and unicode strings are automatically encoded in utf-8 for output. If any browsers can't deal with UTF-8, screw them. ;) Revamping of TagSerializer and ContextSerializer to be more sensible. 1) Context chaining works properly now. - Precompiling a stan tree, html[body[ul(renderer=foo)[li[a(data="Bar")]]]] will result in the following (paraphrased): "<html><body>", WovenContext(parents=body,html, tag=ul(renderer=foo)[ "<li>",WovenContext(IData="Bar", parents=li, tag=a), "</li>"], "</body></html>" The thing to notice is that the parent context chain is cut off whenever a WovenContext is inserted into the actual tree, as it is redundant, and possibly wrong (since e.g. a renderer function could reparent its subtrees) In ContextSerializer, the rendering context is chained onto the end of the remembered context, thus creating the full context chain. 2) Because of the above change, stripContexts isn't needed anymore. 3) Contexts have an "isAttrib" field now, indicating whether the context is within an attribute value. This is used by StringSerializer to properly escape quotes. 4) Contexts don't erase their parent tag field when cloned. 5) ContextSerializer tells TagSerializer that the context is its own, and not to make a new context. 6) Removed a bunch of rendundant cloning. Renderer.precompile, WovenContext.patterns ContextSerializer uses shallow clone. 7) Tag.clone(deep=False) copies its children list. James This fixed the following bugs that noone noticed yet: - Attribute context doesn't contain tag. - Attribute values don't always get " quoted. - Extra contexts got put in the chain by TagSerializer sometimes. Ideas for new tests: 1) Render multiple times from one precompiled document, with mutation of context.tag by renderer function. 2) Attribute context is the context of their own tag. 3) Renderer can check its context chain to make sure that it looks proper (e.g. has the right parents from the stan tree). With and without precompilation and dynamic elements.

1 1

WG: Woven: Problem with Controler
by Uwe Schmitt 04 Feb '04

04 Feb '04

Hi, trying to implement a website I want to implement a controler, but setting the model inside the controler does not work. The (reduced) code is as follows: -------- START ------------------------------------------------- from twisted.web.woven import page, controller class MyController(controller.Controller): def handle(self,request): self.model.setSubmodel(request,'name', "uwe") self.model.notify() class MyPage(page.Page): appRoot = True template=""" <html controller="machine" model="/state" > <body> <p model="name" view="Text" /> </body> </html>""" def wcfactory_machine(self, request, node, model): return MyController(model) def wmfactory_state(self, req): return self resource = MyPage() -------- END ------------------------------------------------- When I request the resource, woven says: "Node had 'model=name' attribute, but the submodel was not found" Any hints ? Greetings, Uwe

3 2

Guard and flatsax patches
by James Y Knight 03 Feb '04

03 Feb '04

Guard patch: - Reinstates the request.setupSession attribute from the old guard, which I use in order to force a new session if the user changes their authentication data. Flatsax patch: - Require pyxml >= 0.8 - Handle entities correctly. James

1 1