Review and merge with odf(?)

[mtennekes]

Hi Robin (cc @agila5 and @rogerbeecham),

Great to see your progress on the od package!

Finally, I have made some progress on 'my' od package, which I called odf. As you will see below, odf doen't have as much features as od. I have put the emphasis on the data structure, for which I used a graph structure. However, I have a different approach than I had in mind during my visit in Leeds: now I'm following a much more light-weighted approach.

Could you take a look and see if it's useful for your package? I've rendered a pdf from the vigette and also uploaded our draft paper. See also the latest version of the 'halfline' doughnut visualization, for which I've partly used odf.

I have the feeling that we can merge our approaches quite well. Let me know what you think.

[Robinlovelace]

Hi @mtennekes, many thanks for raising this and the offer to join forces. I will take a look and let you know my thoughts. First impression: as stated in the README the od package has no class system but is something I would like to at least consider. The fact that you have developed a class system is very attractive and, as you say, I think the code bases are not only compatible but complimentary. Suggest that a 'minihack' one afternoon or early evening could be the best way forward on this. Suggest inviting some others - will send an email now to help make it happen!

Thanks again, this will provide a strong basis for od work in R 🚀

[mtennekes]

Awesome!

[mtennekes]

For time-series and/or multi-categorical OD data, an interesting approach is to use vector cubes from the stars package. See https://keen-swartz-3146c4.netlify.app/raster.html#datacubes (section 4.3.2.).

[Robinlovelace]

That is a very interesting approach. I've still not fully understood the potential of that way of doing things but I imagine it opens doors, thanks for flagging it! Another approach, more on the side of visualisation, is this: https://github.com/rogerbeecham/od-flowvis-ggplot2

I recall @rogerbeecham was happy to 'donate' the relevant curve-producing code to a good cause but wonder if vis and class systems / processing should be kept separate?

[rogerbeecham]

Sorry haven’t been keeping up with this one. Yes happy to have the curve producing incorporated into this. Also I think the spatially arranged OD matrices is a nicer idea which sort of lends itself quite nicely to tidyverse style templates. Challenge when packaging up is incorporating an algorithm for automatically allocating spatial units into a grid. A ex-colleague of mine (https://www.win.tue.nl/~wmeulema/) has written these though and sure would be able to incorporate some of his into this… Roger From: Robin <notifications@github.com> Reply to: ITSLeeds/od <reply@reply.github.com> Date: Monday, 19 October 2020 at 07:03 To: ITSLeeds/od <od@noreply.github.com> Cc: Roger Beecham <R.J.Beecham@leeds.ac.uk>, Mention <mention@noreply.github.com> Subject: Re: [ITSLeeds/od] Review and merge with `odf`(?) (#20) That is a very interesting approach. I've still not fully understood the potential of that way of doing things but I imagine it opens doors, thanks for flagging it! Another approach, more on the side of visualisation, is this: https://github.com/rogerbeecham/od-flowvis-ggplot2 I recall @rogerbeecham<https://github.com/rogerbeecham> was happy to 'donate' the relevant curve-producing code to a good cause but wonder if vis and class systems / processing should be kept separate? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#20 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AELPQHWNOTZIZAGHOTSEYKLSLPJBLANCNFSM4SOMKPHA>.

[Robinlovelace]

Next step: include functions to convert to 'edge-node' format consisting of a list of 2.

[Robinlovelace]

Could also convert to sfnetwork objects, heads-up @luukvdmeer.