I just discovered impure.com and was very impressed by what can be achieved in terms of visualizing data using their workspace. Although the platform is free to use, but not open source yet. This causes a few problems because the API section works with .com sites for eg. for ebay, so I cannot visualize ebay listings for regional sites such as ebay.in because the link to ebay.com is hardcoded.
I searched around, but didn’t find other projects that offer a similar way to work with site APIs and other data sources with the kind of user interface and detail that impure.com does in a realtime way within a browser window.
This brings me to the questions:
What technologies would be involved in creating a similar kind of project
What are the open source tools that can help develop a fullscreen UI to render the workspace. Are there any alternatives to flash for this, and how do they compare
The goal would be to use standard python data structures, python scripts to do some processing on these data structures, scripts to gather data from csv, json and API sources such as google yahoo, wikipedia and flickr, and scripts to render graphs, tag clouds, network visualizations etc. Then bring them all together into a visual interface that supports drag drop and simple type checking
How would the python backend integrate with the UI
sorry for the rambling question, but i wanted to present as much detail as possible. I’d like to do this for the fun of it
So, restating, Impure is like OpenDX but running in a browser and you are interested in suggestions for an architecture to build your own web version of it?
I suggest using Cappuccino as your front end Web GUI. This is a toolkit for building RIAs in the browser that are completely based on HTML and Javascript (no Flash). It is modeled after Apple’s Cocoa so you can build some very sophisticated apps with it.
On the web server side, I would use the Python bindings to OpenDX to build a JSON API that is exposed to the web browser. Note, this Python module uses DXLink, which appears to be mainly a batch/command interface to OpenDX – you might still need to do some work to get visualizations out of it and back to the web browser. But, it’s all open source so it should be feasible. Here’s a discussion where others are attempting the same thing.
I don’t know for sure that OpenDX supports pulling data via web service calls but you could certainly handle that on the Python side and feed it into OpenDX.
Note, I don’t mention a Python web framework because, since your app will be mostly a JSON API almost any of them would work just as well. I suggest the thinnest and simplest one that appeals to you.