[pgrouting-dev] Re: Network Layering support

[Stephen Woodbridge]

On 3/4/2011 4:19 AM, Jay Mahadeokar wrote:
>
>
> On Fri, Mar 4, 2011 at 11:37 AM, Stephen Woodbridge
> <woodbri at swoodbridge.com <mailto:woodbri at swoodbridge.com>> wrote:
>
>
>     I think it takes a few things to do this.
>
>     1. declare this is our goal
>     2. start a discussion on a design and interfaces
>     3. break out some smaller incremental transformations that move in
>     that direction
>     4. iterate on the design and interfaces as as needed
>
>
> Looks like a nice idea. I am completely unaware of opengraphrouter and
> its background. Can we take this discussion to another new thread?
>
> @Stephen and Daniel, can you list out the people who will be willing to
> contribute / give desirable time to realise the above idea?

Jay,

I'm willing to provide my knowledge to task of design and review. I 
think my skills are more suitable to high-level concepts, use cases, and 
as a user of the tools. And on the other end of things, I can build a 
cgi-bin or fast-cgi module using the library (assuming a C-API exists). 
I can load data and test and debug.

Anton is probably the sole expert on pgRouting, although you have 
already implemented the APSP code which I have not tried out yet.

> Are planning to divide it into mini projects and float it out for GSoc
> 2011? This would surely encourage students to contribute and come up
> with required modules provided the mentors are available for guidance.

This year, I have decide not to take on the route as primary mentor. I 
will act as a backup mentor and provide as much help as I can via the 
list because I'm intereted in all aspects of this.

>     Mini-projects that would help with this:
>
>     1. can we repackage our solvers to run outside of postgres?
>     Why? This would help define what is common between the environments
>     and what is not. In turn this helps define the interface and common
>     facilities that need to be created to support the code in two
>     separate environments.
>
>     2. Have Jay get his project working in both a standalone and in
>     pgRouting. The code he has already works standalone, but it will
>     need to morth into code that will work in pgRouting. Again this is
>     similar to 1.
>
>
>
> Are you talking about the APSP code, or the CH code of Geisberger?  APSP
> would work with pgRouting readily I guess. It is working as of now, but
> still needs some validation checks etc.

I see the CH code as having some potential challenges with regards to 
pgRouting so it might make more sense to prep the data outside the 
database, especially if you are working with a large data set. I'm also 
not sure how well it will fit with the current methodolgy of build the 
graph on the fly solve it and through it away.

Licensing issues aside, I think it would be interesting to take the CH 
code potentially see what we can do with it. Some of the things I would 
look at are:

1. can the readers and writers for the various data streams easily be 
abstracted? or is it tangle through out the code? For example if I 
wanted to load data from shapefiles how easy/hard would that be? Or load 
data from a database table? Would it be easy to isolate this code so we 
could have virtual function table and set function pointers for the data 
readers?

2. does their code support turn restrictions? How are they defined for 
input?

3. can the graph be modified at runtime? If I want to change between 
shortest time or shortest distance? Can I add arbitrary avoidances, like 
an accident or a bridge out, etc?

4. what kind of analysis can be done with the CH graph? Dijkstra? A* 
search? Can I extract a Dijkstra solution tree for doing drive time 
analysis?

5. does the code you got support generating textual driving directions?

There are probably a 1000 more questions.

Best regards,
   -Steve

>         It's difficult to keep an open source project active, and I
>         would like
>         to know for example if Ashraf is still engaged in the
>         opengraphrouter
>         project. Is he?
>
>
>     I talked to him recently and he still has interest, but I have
>     dropped the ball on keeping him engaged because I have been insanely
>     busy. As I related to you, I will probably not be a primary mentor
>     this year.
>
>     -Steve
>
>
>
>
> --
> Regards,
> -Jay Mahadeokar
>
>
>
> _______________________________________________
> pgrouting-dev mailing list
> pgrouting-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/pgrouting-dev