Document Acties

Presentations at Zest

Opgeslagen onder:

This Friday we had various presentations and food at the Zest Software headquarters.

Every few months at Zest Software we have our so-called "eten en weten" meetings; literally: eating and knowing, but in Dutch it rhymes. Basically it means: presentations by colleagues at the end of the afternoon, then food, then more presentations, with some discussions sprinkled around. The later the evening, the more technical the presentations get, so the less technical colleagues can flee. :-) It is also a nice way to catch up with former colleagues who are invited; this time we were joined by Joris Slob; always a pleasure. Meet our strong team here.

Coaching

It started off with Esther and Jean-Paul introducing new coaching options for the team members. Everyone gets the chance to have a few talks per year with a coach, paid for by Zest. The coaches were also there to introduce themselves. A very nice option.

Fred: breathe in, breathe out

Then colleague Fred happily surprised us with a workshop about breathing techniques. In his spare time he sings in a choir, so he has experience here. Sorry, I don't think any pictures were taken. :-)

Food

It was dinner time so we enjoyed some food brought in by the Sun Sien restaurant.

Joris: the semantic web

Now the floor was for former colleague Joris Slob, currently working at the Leiden Institute of Advanced Computer Science. He talked about the semantic web and the technologies that float around in that area. Web 1.0 can be summarised as: smart people put things on the web; less smart people only look at it. With web 2.0 lots more people can dump content, whether it is smart or not. But there is nothing fundamentally new about this web: it is just a different use of the same old techniques. For that you can look at what some call web 3.0: the semantic web. Computers currently do not understand the semantics of web pages. They fetch an html page and see something like this:

<html>
  <head>
    <title>blah blah blah</title>
  </head>
  <body>
    <h1>blah blah blah</h1>
    <p>blah blah <em>important blah blah</em> blah</p>
  </body>
</html

Let Google search for images of jaguars and you will get both cars and animals. But try Google's Wonder wheel: it realises you could be looking for jaguar cars, jaguar parts, jaguar animals and more.

As human you have knowledge. What do you do with that knowledge? Where do you store it? Can you put it in a database? Our databases do not fit this reality very well.

In comes RDF, the Resource Description Framework. RDF offers user-defined relations, instead of a rigid database where all possible relations between items in tables have to be defined beforehand. RDF consists of triples: subject + predicate + object = fact. For example: Joris (subject) likes (predicate) Python (object). So the predicate is the relation between subject and object.

But how do you know who this "Joris" is? Lots of people are called Joris, at least in The Netherlands. So you use a URI, a Uniform Resource Identifier, like this: http://www.liacs.nl/~jslob/foaf.rdf#me

This URI uses the Friend Of A Friend (foaf) vocabularies. These describe relations between persons. In this example we get an xml representation of the RDF triples, but they can also be stored in the native RDF non-xml format.

Another part of the semantic web is OWL, the Web Ontology Language. (You may think this three letter acronym is in the wrong order, but the French would disagree with you; just ask our French colleague Vincent. Update: actually, see the section on the acronym in the encyclopedia for the real reason.) OWL gives greater expression possibilities. You also need reasoners. Reasoners can go through your knowledge base and find inconsistencies, or infer new relations. Fact 1: Joris is a human. Fact 2: humans have 2 legs. So we infer a third fact: Joris has 2 legs. Some inconsistencies can be resolved: "Obama is a good president" versus "Obama is a bad president" could become: 'John thinks "Obama is a good president"' and 'Lucy thinks "Obama is a bad president"'.

Then there are ontologies. Ontologies define possible relations in a knowledge domain. Examples are SIOC, Dublin Core, Foaf, Good Relations and Basic Geo Vocabulary. If you can reuse an existing one, that is very good, as you do not want to create your own if you can avoid it.

To get information out of a knowledge base, you can use SPARQL, which is a query language that looks a bit like SQL. To get everything, you would write:

SELECT ?s, ?p, ?o where {?s ?p ?o}

Another part of the semantic web is formed by microformats. These embed meaning in html. Examples are hCards, hCalendar, RDFa (pushed by Google, blocked by Adobe for HTML5). Also: Google Rich Snippets. This can for example be used to have information about ratings (4 out of 5 stars) show up in Google search results.

For the python programming language you want to look at the rdflib library. Most readers of this blog will likely be familiar with Zope, so I'll mention that you can also use the ZODB (Zope Object DataBase) as a back end.

Finally, if you want to learn more, you should read this book: Programming the Semantic Web, by Toby Segaran, Colin Evans and Jamie Taylor.

Vincent: jQuery, plus libraries, plus Proxy

At this time the non-technical people made good their escape, and the rest buckled up to learn about the wonderful world of jQuery, several jQuery libraries, and the soon to be released jQuery Proxy, made by Vincent. He made the presentation with live demos using Django; there was also a pony somewhere...

The jQuery principle is: write less, do more. You usually do not have to care about differences between browsers. It has got great documentation.

Some libraries are:

  • jQuery UI: the official UI for jQuery. It has themes, widgets, interactions (like draggable, droppable, resizable).
  • jQuery Tools: less effects than jQuery UI, more web oriented, with tabs, tooltip, overlay, expose, scrollable.

You can find lots more on http://plugins.jquery.com/ which is basically a pypi or CPAN for jQuery, including a demo for each plugin. Some that Vincent uses are cookie, color and lightbox.

And now there is jQuery Proxy, written by Vincent. This gives easy integration of AJAX in python sites. At least it works in Django and Plone. You get jQuery syntax in python code and can send this back to the browser. It also allows using jQuery libraries/plugins. It could eventually replace KSS. The way you use it is for example this:

from jquery.pyproxy.django import JQueryProxy, jquery

# The @jquery decorator handles the transformation of your results
# into JSON so we can decode it on client side.

@jquery
def ajax_add_comment(request):
    # The JQuery proxy object helps us to manipulate the page the user sees.
    jq = JQueryProxy()
    # The data/form sent with Ajax just apear like a classical POST form.
    form = request.POST

    #we do some validation of the form.
    ...

    if errors:
        # We display an error message.
        jq('#my_error_message').show()
        return jq

    ...
    # We display a success message.
    jq('#my_success_message').show()
    return jq

The source is available on github and we will hopefully see a release next weekend. Very interesting!

Mark: DVCS

Next up was Mark, with a talk about Distributed Version Control Systems (DVCS), or as he liked to call it: git. :-) See his weblog for the slides. At Zest we are using subversion to store our code and share it among the developers. With subversion you have a central server that every commit goes to. But some would prefer doing local commits, so as not to disturb others with experiments or with a temporarily broken code state, or just because they are in a train and have no network connection. This is where distributed systems like git and mercurial or bazaar shine. You can still use a central repository if you want to though.

With git you have a complete local repository, which means it is much much faster, as you need less network traffic: basically the only times you need to connect to a different server are when you pull code (get new code from trunk or a branch) and when you push code (send your changes to a server). Checking in code is fast, because you do not need to do get fresh code first (svn up) to see if someone else changed some code, so you also do not get clashes when checking in, and the commit is local so that makes it extra fast.

Not everyone at Zest is sold on the idea of using git yet. For now we will keep using subversion. Those wanting to use git can use git-svn, and have been doing that already. If possible we'll try and keep our buildouts git-friendly; we have already created some client buildouts with the client packages directly in the buildout src directory instead of including it using infrae.subversion, mr.developer or svn:externals (though the last one is apparently less of a problem).

Maurits: collective.watcherlist

I was the last presenter of the evening. I talked about a client project, but there was a Non Disclosure Agreement involved, so I have probably already said too much. :-) So we will skip that and jump straight to my presentation about a new package I have created, called collective.watcherlist

collective.watcherlist is a package that enables you to keep a list of people who want to receive emails when an item gets updated. The main use case is something like Products.Poi, which is an issue tracker for Plone. That product lets you create an issue tracker. In this tracker people can add issues. The tracker has managers. Every time a new issue is posted, the managers should receive an email. When a manager responds to an issue, the original poster (and the other managers) should get an email. Anyone interested in following the issue, should be able to add themselves to the list of people who get an email.

The origins of collective.watcherlist also lie in the mentioned Products.Poi package, first created by Martin Aspeli and now maintained by me. A while ago I fixed some bugs in the email sending part of Poi, as sending international emails can be tricky. After this, I thought this part of the code was quite solid (the rest as well, actually :-)) and could be useful for other packages that needed to send out email. So I decided to factor this code out into a separate package. This also made some parts of Poi cleaner and simpler.

There is no release as of this writing, but the source code is in the collective. You may want to grab the buildout of a branch of Poi that uses it. That branch of Poi is meant for Plone 4. I should create a proper alpha release for that soon. The collective.watcherlist package itself works just fine on both Plone 3.3 and 4, though there are some minor test failures on Plone 3. The test coverage is at a solid 98 percent, so I am quite sure most bugs have been ironed out.

I showed what the package does by going through the code in the sample directory. For this blog entry I will resort to some copy-pasting from the readme file.

It is not a package for end users. Out of the box it does nothing. It is a package for integrators or developers. You need to write some python and zcml in your own package (like Poi now does) to hook collective.watcherlist up in your code.

collective.watcherlist might also be usable as a basis for a newsletter product. If you feel Singing and Dancing is overkill for you, or too hard to adapt to your specific needs, you could try writing some code around collective.watcherlist instead.

In its simplest form, the integration that is needed, is this:

  • Register an adapter from your content type to collective.watcherlist.interfaces.IWatcherList. In a lot of cases using the default implementation as factory for this adapter is fine: collective.watcherlist.watchers.WatcherList
  • Create an html form where people can add themselves to the watcher list.
  • Register a BrowserView for your content type, inheriting from collective.watcherlist.browser.BaseMail and override its properties subject, plain and/or html.
  • Create an event handler or some other code that gets the adapter for your content type and uses that to send an email with the subject and contents defined in the browser view you created.

That is it. There are currently no viewlets, portlets or other templates that you need to override, so it should be easy to fit into the theme of your website, provided you do not mind coming up with a UI yourself. The sample directory is a good spot to look for the basis of that though.

Future plans include:

  • an option to send to all members (though this should already be very easy to do yourself),
  • allow setting a different from-address,
  • personalised emails, even if just with a footer to point to an unsubscribe option,
  • optionally storing more info per subscriber, like a preference for plain text or html,
  • more newsletter-like functionality, for example double-opt-in (similar to what the PasswordResetTool does when creating a Plone member).