Semantic Web using Magritte - smallwiki

Philippe Marschall

5 Dec 5 Dec

3:23 p.m.

Hi For audioscrobbler I did the opposite, from given XML to Magritte: http://source.lukas-renggli.ch/audioscrobbler.html The audioscrobbler xml is quite gay and I was surprised when I got it working. It is also used for feed parsing of the Seaside homepage. Cheers Philippe 2007/12/5, smalltalk(a)fixinsbar.com <smalltalk(a)fixinsbar.com>om>:

...

I was wondering if anyone tried to do any kind of XML mapping using Magritte. I have the requirement to generate RDF from my business objects, so I was thinking of ways to simplify this. My idea was to add something in the Magritte descriptions for each object that pointed to the appropriate URI and schema. This may be more trouble than its worth. I suppose anyone who has done anything with generating XML from Magritte could maybe provide some insight on any hurdles they faced, design ideas, etc. I also am looking at Magritte for simplifying XML serialization in general. My problem overall is that I have to generate all these formats (XML for web service, RSS, ATOM, RDF) and I do not want to have to write a complex parser to even do this 1/2 way dynamically. Are there any other libraries out there that do this already? For instance, I did notice that there is that library (name escapes me) that lets you port your objects between Smalltalk versions by dumping to XML files, so that is somewhat close to what I need to do. I am thinking since I already have a large number of Magritte descriptions, I might as well expand upon that. This will also let me do things like change a schema dynamically and provide easy access memento objects to do things like create a preview of whatever format I am trying to produce. Thoughts? _______________________________________________ SmallWiki, Magritte, Pier and Related Tools ... https://www.iam.unibe.ch/mailman/listinfo/smallwiki

Reply

Jason Johnson

7:39 p.m.

Are your descriptions going to still be relevant for PDF in the same way they are for a web page? Magritte descriptions usually provide a GUI for entering or editing them, no? If the descriptions do fit (or if they don't you could make new ones) then I think the thing to do would be to add a renderer. Right now there is one for Seaside and one for Morphic. So you would need one for each format you need to output to. Perhaps there is already a package that does something like Seaside's canvas for XML (maybe XANO or whatever it's called?). On Dec 5, 2007 3:33 PM, <smalltalk(a)fixinsbar.com> wrote:

...

Thanks for the input, I had definitely considered that. I was even looking at XSLT but I really like keeping everything in Smalltalk. I do indeed need to go many directions since there are many entry points and services with this application. I also have explored creating new canvases and doing some things with Readers/Writers like in Pier. I would do exactly what your describing, but I still am left with the problem of generating good output from my objects. So for example, sticking with music, say we have an Album object. I have all kinds of stuff already for Magritte in there for this object. Now I want to do something like in my seaside control provide an RDF button that when clicked, generates the RDF output. I would need to traverse the entire object/description graph and output the RDF with the album name, title, etc. and the track listing, and then artist info, etc. It definitely would suck without the metadata to identify which RDF URI corresponds with each object in the graph. Essentially, I am thinking of creating a simpler serializer/deserializer driven by Magritte descriptions similar to what I've done before in .NET and Java. I am just not sure if that is very Smalltalk like as I like to always think there is a better way than resorting to what I have done in those languages :) For instance, there is a JSON serializer in .NET where I can say Album.ToJSON(), or the XML serializer Album.ToXML(). It so happens that depending on the serializer/deserializer, it will read attributes I put on my class like [XmlIgnore]. That stuff is really primitive I think compared to what I can do in Magritte. I guess I am trying to prevent a future meltdown because as this application is developed, I am going to have to start going back and forth with lots of web services to communicate with Java and Flash in particular. I want to set a good example on the Smalltalk side by generating everything automagically. Quoting Philippe Marschall <philippe.marschall(a)gmail.com>om>:

Hi For audioscrobbler I did the opposite, from given XML to Magritte: http://source.lukas-renggli.ch/audioscrobbler.html The audioscrobbler xml is quite gay and I was surprised when I got it working. It is also used for feed parsing of the Seaside homepage. Cheers Philippe 2007/12/5, smalltalk(a)fixinsbar.com <smalltalk(a)fixinsbar.com>om>:

I was wondering if anyone tried to do any kind of XML mapping using Magritte. I have the requirement to generate RDF from my business objects, so I was thinking of ways to simplify this. My idea was to add something in the Magritte descriptions for each object that pointed to the appropriate URI and schema. This may be more trouble than its worth. I suppose anyone who has done anything with generating XML from Magritte could maybe provide some insight on any hurdles they faced, design ideas, etc. I also am looking at Magritte for simplifying XML serialization in general. My problem overall is that I have to generate all these formats (XML for web service, RSS, ATOM, RDF) and I do not want to have to write a complex parser to even do this 1/2 way dynamically. Are there any other libraries out there that do this already? For instance, I did notice that there is that library (name escapes me) that lets you port your objects between Smalltalk versions by dumping to XML files, so that is somewhat close to what I need to do. I am thinking since I already have a large number of Magritte descriptions, I might as well expand upon that. This will also let me do things like change a schema dynamically and provide easy access memento objects to do things like create a preview of whatever format I am trying to produce. Thoughts? _______________________________________________ SmallWiki, Magritte, Pier and Related Tools ... https://www.iam.unibe.ch/mailman/listinfo/smallwiki

_______________________________________________ SmallWiki, Magritte, Pier and Related Tools ... https://www.iam.unibe.ch/mailman/listinfo/smallwiki

Reply

smalltalk＠fixinsbar.com

6 Dec 6 Dec

2:33 p.m.

Quoting Jason Johnson <jason.johnson.081(a)gmail.com>om>:

...

Are your descriptions going to still be relevant for PDF in the same way they are for a web page? Magritte descriptions usually provide a GUI for entering or editing them, no?

Not sure if we're talking about the same things totally. My descriptions have little if anything to do with the GUI. I am looking at adding automagic form generation in a few places, but mainly the forms in my application are hand-crafted right now. We do use Magritte in conjunction with our forms (ex: memento validation), just not to generate the actual form UI.

...

If the descriptions do fit (or if they don't you could make new ones) then I think the thing to do would be to add a renderer. Right now there is one for Seaside and one for Morphic. So you would need one for each format you need to output to. Perhaps there is already a package that does something like Seaside's canvas for XML (maybe XANO or whatever it's called?).

I have explored this strategy before for output. The thing is I really need to produce the RDF/XML/FOAF/RSS/ATOM or whatever else is required given a particular use case on demand relevant to the context. Many situations, Seaside won't play much of a part if any, however Magritte and my back-end (eventually Gemstone) will. There are many channels already to the application, so I want to define how to provide data in each format once regardless of the channel.

Reply

smalltalk＠fixinsbar.com

8:22 p.m.

Thanks Lukas, that is what I am doing now for our exports to text files as well as some custom XML to send to specific web services. Basically, we just wrote a writer that visits the object hierarchy in the way we want and calls the appropriate visit methods. This is similar to some things I have seen in Pier. I am less worried about that part as I am the design of my actual descriptions and how easy it will be to read/write into these formats without writing too much manual stuff. In other words, I don't want to write separate ways of handling different kinds of objects, I want descriptions that fit all the object types I have. As far as I have tested, this is relatively simple. It's just in my experience going down these roads, you write about 90% then comes an uh-oh moment if you are not careful. I am most interested in how others have done similar things as well as affirmation of my ideas. In any case, glad to hear that it is certainly possible/practical. Quoting Lukas Renggli <renggli(a)iam.unibe.ch>ch>:

...

Are your descriptions going to still be relevant for PDF in the same way they are for a web page? Magritte descriptions usually provide a GUI for entering or editing them, no?

Not sure if we're talking about the same things totally. My descriptions have little if anything to do with the GUI. I am looking at adding automagic form generation in a few places, but mainly the forms in my application are hand-crafted right now. We do use Magritte in conjunction with our forms (ex: memento validation), just not to generate the actual form UI.

It is easy to generate dumps of described objects to specific output formats. It is almost too simple to show a text exporter here: aModel description do: [ :desc | aStream nextPutAll: (desc label); nextPutAll: ’: ’; nextPutAll: (desc toString: (desc accessor readFrom: aModel)); cr ] For a description of this example see page 6 in http://www.iam.unibe.ch/~scg/Archive/Papers/Reng07aMagritte.pdf . Cheers, Lukas -- Lukas Renggli http://www.lukas-renggli.ch _______________________________________________ SmallWiki, Magritte, Pier and Related Tools ... https://www.iam.unibe.ch/mailman/listinfo/smallwiki

Reply

smalltalk＠fixinsbar.com

8:47 p.m.

Agreed, you are definitely correct. I think depending on implementation it would be specific to what you are doing. For instance if we're visiting different domain objects and their descriptions to create the PDF, we might have certain kinds of objects to represent the output we want. Additionally, you are mentioning one use for Magritte. Lukas mentioned some others in one of his posts. Magritte to me is simply a good implementation of meta-programming constructs I have used in other languages such as .NET, Java, Lisp, Ruby, C++, etc. For instance, .NET has reflection and attributes which allows for some similar constructs as Magritte, but definitely not as flexible. Previously, I used .NET attributes to create a simple Object to RDF converter. Obviously though I'd rather work with Smalltalk :) Output formats is just one goal for me, another one is searching my object graph, for example as done in Pier. Ramon's implementation of Magritte to generate Glorp descriptions is another. I'd be happy to share anything that I produce regarding conversions. I brought some of this up because after exploring what I could find in Squeak for RDF in particular, it was frustrating. I did not want have to do things like create anything cased out or with specialized extra objects. I wanted to just use my existing objects and extend them in a way that didn't interfere with what I was already doing. Since my Magritte descriptions are already there and exist on the meta level, it's a good choice I think. Quoting Jason Johnson <jason.johnson.081(a)gmail.com>om>:

...

I don't think I expressed myself clearly. Magritte is (as I understand it) a way to provide extra meta "descriptions" of how your objects might be presented. Other components can take these components and turn them into some representation. For example, the Seaside Magritte renderer takes these descriptions and makes HTML. You could create e.g. a PDF renderer that reads these descriptions and generates a PDF item from them. I know that would be a lot of work, but once it's done one time we all have it. _______________________________________________ SmallWiki, Magritte, Pier and Related Tools ... https://www.iam.unibe.ch/mailman/listinfo/smallwiki

Reply

Lukas Renggli

10:46 p.m.

...

Just by analogy: did you have a look at the way PierBlog creates the RSS feed for the blog? Seems to be a similar problem...

It doesn't really use the descriptions to generate the blog items, it could (or better it should) though. Lukas

...

-- Matthias On Dec 5, 2007 3:33 PM, <smalltalk(a)fixinsbar.com> wrote:

Thanks for the input, I had definitely considered that. I was even looking at XSLT but I really like keeping everything in Smalltalk. I do indeed need to go many directions since there are many entry points and services with this application. I also have explored creating new canvases and doing some things with Readers/Writers like in Pier. I would do exactly what your describing, but I still am left with the problem of generating good output from my objects. So for example, sticking with music, say we have an Album object. I have all kinds of stuff already for Magritte in there for this object. Now I want to do something like in my seaside control provide an RDF button that when clicked, generates the RDF output. I would need to traverse the entire object/description graph and output the RDF with the album name, title, etc. and the track listing, and then artist info, etc. It definitely would suck without the metadata to identify which RDF URI corresponds with each object in the graph. Essentially, I am thinking of creating a simpler serializer/deserializer driven by Magritte descriptions similar to what I've done before in .NET and Java. I am just not sure if that is very Smalltalk like as I like to always think there is a better way than resorting to what I have done in those languages :) For instance, there is a JSON serializer in .NET where I can say Album.ToJSON(), or the XML serializer Album.ToXML(). It so happens that depending on the serializer/deserializer, it will read attributes I put on my class like [XmlIgnore]. That stuff is really primitive I think compared to what I can do in Magritte. I guess I am trying to prevent a future meltdown because as this application is developed, I am going to have to start going back and forth with lots of web services to communicate with Java and Flash in particular. I want to set a good example on the Smalltalk side by generating everything automagically. Quoting Philippe Marschall <philippe.marschall(a)gmail.com>om>:

Hi For audioscrobbler I did the opposite, from given XML to Magritte: http://source.lukas-renggli.ch/audioscrobbler.html The audioscrobbler xml is quite gay and I was surprised when I got it working. It is also used for feed parsing of the Seaside homepage. Cheers Philippe 2007/12/5, smalltalk(a)fixinsbar.com <smalltalk(a)fixinsbar.com>om>:

I was wondering if anyone tried to do any kind of XML mapping using Magritte. I have the requirement to generate RDF from my business objects, so I was thinking of ways to simplify this. My idea was to add something in the Magritte descriptions for each object that pointed to the appropriate URI and schema. This may be more trouble than its worth. I suppose anyone who has done anything with generating XML from Magritte could maybe provide some insight on any hurdles they faced, design ideas, etc. I also am looking at Magritte for simplifying XML serialization in general. My problem overall is that I have to generate all these formats (XML for web service, RSS, ATOM, RDF) and I do not want to have to write a complex parser to even do this 1/2 way dynamically. Are there any other libraries out there that do this already? For instance, I did notice that there is that library (name escapes me) that lets you port your objects between Smalltalk versions by dumping to XML files, so that is somewhat close to what I need to do. I am thinking since I already have a large number of Magritte descriptions, I might as well expand upon that. This will also let me do things like change a schema dynamically and provide easy access memento objects to do things like create a preview of whatever format I am trying to produce. Thoughts? _______________________________________________ SmallWiki, Magritte, Pier and Related Tools ... https://www.iam.unibe.ch/mailman/listinfo/smallwiki

_______________________________________________ SmallWiki, Magritte, Pier and Related Tools ... https://www.iam.unibe.ch/mailman/listinfo/smallwiki

-- Lukas Renggli http://www.lukas-renggli.ch

Reply