rht
commented
7 years ago On which format/language to use: R, python, julia (and maybe js for d3) have been used the most for data science, viz, scientific publications, and so either would make the most sense for a standardized choice. The "grammar of graphics" (for proper visualization) was initially implemented in R, but this has long been ported to both python and julia as well (while gnuplot doesn't have such framework, hence, excluded).
In this case, here are the reasons for python, rather than the rest:
- It has tons more libraries for science in addition to statistics (ggplot/bokeh, pandas for large-scale ops, networkx for graph, ... and who knows, tensorflow for benchmark on bitswap decision engine)
- py-ipfs-api already exists if needed, while it wouldn't make sense to write an api for R (and julia's is still pending)
(note: personally, I wish I could go with julia, but it is the constraint imposed by the ecosystem (libraries) that led me to opt for scipy)
Plots are very important, there are the simplest way to present data like that.
Deciding on one format/language to generate those plots from is also important as then in a lot of cases the plot recipes can be reused, re-purposed and this makes much easier for others to include new plots. I had great success with plotting multi-million data points plots with R. The syntax is simple, it includes a lot, if not all, necessary statistics functions. Examples of plots:
https://ipfs.io/ipfs/QmaiKNgDt1eUkH6R3C2rvd4NeeWdRb7opiWBAD5aSMMfLT/512way.blue-all-shards.red-two-item-shards-vs-N.png https://ipfs.io/ipfs/QmZwrT8aDZJ9LdT3xXKBxwxLvKkGb41fFjDumjpV2Q5CXT/
Most importantly in many cases the recipes can be reused with minor tweaks.