Thinking on version 3.0

[cvvergara]

Hi all,
I have osm2pgRouting 2.3 on develop branch
I would like it to be tested.

• Changed a lot of code removing C arrays and converting to C++ containers, so no leaks.
• It still slow
• it still does not work for big files
• added a couple of flags to try forking when dealing with the database and only works for small files
• it still works only on .osm files

I will releasing soon. (after Jenkins tests pass)

I really don't want to continue to "make improvements" and "bug fixes" (aka, no 2.4) I would prefer to do a complete rewrite:

There has being the idea of using libosmium to develop osm2pgrouting with a new design and change the way how it works

• currently works on osm files
  • libosimum allows to read other file formats in a very transparent way
• currently its 1 pass on osm file, so all nodes need to be stored in memory
  • change to 2 passes, so only required nodes are processed
• currently the configuration is read from the file only
  • be able to read from the configuration from the database when adding more information
  • or complain if the configuration is different from the one on the database
• separate tasks:
  • addnodes, addways, add relations only do that task, and not split ways
• and of course, not to have the long list of issues that the current implementation has

All this is start from scratch, and libosimum requires C++11 for compilation.

Opinions and ideas are welcome

[woodbri]

+1
I think this is the correct approach. It does not make sense to add a lot of code that will get thrown out when it gets rewritten. I have built a couple of apps that use libosimum and it is very easy and intuitive to use.

[dkastl]

I also agree and think it's a good idea.
Running in seperate steps is good. It's also good to use external tools like libosmium, that we don't need to reinvent the wheel.

[rohithsankepally]

I too agree with the idea. I recently started using libosmium and found that it has a lot of functionality which can be used to improve osm2pgrouting.