Hacker News Reader

Comment by lasryaric | Hacker News Reader

It was not clear in my question but I was mostly talking about how do people visualise there data while they are working on it? Lets say you are a developer working on the well known Shazam application. You want to visualize your signals and frequencies for debugging purpose, how do you quickly visualize them?

In my case, while working on signal processing using C and objective C, I had no easy solution to visualise my data in a iterative way (code, visualize, repeat).

I was looking for a simple objective-c library to include, in order to visualise my data in a web browser.

by the way, thanks for your input, this is very valuable.

View on HN

If you can work with Python/Julia/R then I would still recommend the IPython notebook, since you will not get anything more interactive and better suited to data analysis and visualization in one package. It basically allows you to run arbitrary snippets of code and interactively work with your data (code, visualize repeat). In general, I would not try to develop your algorithms directly within your app but instead generate some example data, export that data and then work with it in IPython (or another tool) to iteratively build the algorithm. As soon as you have a working solution you can then integrate/translate the code back into your application.

Seconding the IPython Notebook. I've used matplotlib but also like Plot.ly for its prettier, more sharable charts.

It's also not a huge amount of work to set up a JSON-Flask-D3 visualisation pipeline out of test data; that's exactly what I did for my current data project, using cubism.js to easily visualise a time series. Again, benefits are the ability to easily share.

Also, sometimes, Google Spreadsheets/Excel can be the easiest solution to throw up quickly - or using built-in data explorations in RStudio. I've also thrown graph data into Neo4j because it gives you a nice visual interface for free.

Good point!

Ipython notebook completely changed the way I do data science, specifically because its iterative/exploratory capabilities are so powerful. You can also hook it up to large scale service like hadoop/spark. Embedded Matplotlib charts have spoiled me. I cringe at the thought that I used to write scripts to output csv, run them on the command line, then copy the output to excel to graph the data. Never again!

I've often wanted to build a cmdline tool that would read from stdin and popup a live interactive chart of the values. This would make debugging/understanding complicated data-dependent algorithms possible by just adding some printf style logging.

Reply on news.ycombinator.com