Skip to content

Latest commit

 

History

History
57 lines (49 loc) · 2.85 KB

README.md

File metadata and controls

57 lines (49 loc) · 2.85 KB
Raw

Parses distrowatch into a graph

from this graph a nice svg image is created.

The example timeline

Note that you can use the search form from Distrowatch to create a more constraint variant (for example if you're only interested in Linux or the Debian family, that search form is very extensive).

Progress/Architecture

  • the fetchdists.py screenscrapes distrowatch into a JSON structure. This makes retrieving data a lot more convenient as distrowatch uses ancient/broken html. (but thats quite obvious I guess, it still functions so who cares)
  • start.sh glues the different python programs together. Its designed in this way so I just have to fetch the JSON once, and then can use a file, this gives a huge speedup in debugging.
  • You can specify criteria that the search.php form generates. See fetchdists.py --help for details.
  • the graph.py program eats the output of fetchdists.py. It then transforms that output json into a tree structure, however while doing it it modifies the dependson variables (there were some unstructured formating which is replaced with regexes, crude but effective). Currently this json graph could be used for other querring purposes in principle.
  • The svg.py program creates a csv output from the graph.py programs output.
  • gnu clad should eat the csv output to create the final svg.

I'm thinking about recreating the entire svg rendering in python and circumventing gnu clad, because gnuclad is written in C (or C++, I didn't check). (also the gnu clad project seems dead-ish and I want to add some features, such as small figures for each release. This is difficult to feed to gnu clad in csv format so I think its easier to just look at their source and reimplement it in python. We share the same license so it shouldn't be a problem), although looking at all the config options now I know that I won't be doing that any time soon.

Todo

  • Combine data from the found archive with the scraped data
  • Add the images from the original.
  • Add domain support (backgrounds for example for debian)
  • Add some sort of combine mechanism to supplement data from distrowatch (for example for android, since distrowatch ignores that for some reason)
  • Code sharing support? (the original had this but I didn't think it was that usefull honestly)
  • Create a custom renderer in html/css/js rather than svg, this will be much more flexible. I also have the idea to use a circular rendere in this case, where 1992 will be the center and the farther you go out the farther in time you go.
  • Add support for releases (giving another line type for rolling rleases)

How to help

Feel free to use this code according to the GNU license. You can use the github interface for creating pull requests, alternatively I also accept patches.

Please sponser either distrowatch or gnuclad.