Add function to convert gtfs object to sp (similar to gtfs2sldf())

[dantonnoriega]

Suggestion: explore options for converting gtfs data classes into spatial objects such as SpatialLinesDataFrame.

Note: stplanr has a basic function for this: gtfs2sldf().

[mdsumner]

sf will be a better target now, sp has a subset of what sf can store. I'm happy to explore this, I've got my sf skillz pretty well sorted.

[dantonnoriega]

The help would be very much appreciated! My skills are very low. On Jan 5, 2017 7:18 PM, "Michael Sumner" <notifications@github.com> wrote: sf will be a better target now, sp has a subset of what sf can store. I'm happy to explore this, I've got my sf skillz pretty well sorted. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#24 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFaR2Y1OXtGKKr5ES1sIKPmV4TzEYg7zks5rPYhUgaJpZM4KKt16> .

[mdsumner]

One way is to use the re-compose function spbabel::sp, which is looking for the generic names "object_", "branch_", "order_", "x_", "y_" which correspond to your shape_id (one branch per object), shape_pt_sequence, shape_pt_lon and shape_pt_lat.

library(spbabel)
sp_line <- sp(transmute(NYC$shapes_df, object_ = shape_id, branch_ = shape_id, order_ = shape_pt_sequence, x_ = shape_pt_lon, y_ = shape_pt_lat), crs = "+proj=longlat +ellps=WGS84 +datum=WGS84 +no_defs")
sp_line$shape_id <- unique(NYC$shapes_df$shape_id)
sp_line$rownumber_ <- NULL

That can be converted to sf using sf::st_as_sf, and more directly the dev version of spbabel has stubs for the sf() function, that I haven't implemented :)

But, for that shapes_df it's more or less like this:

library(sf)
library(dplyr)
sf_line <- distinct(NYC$shapes_df, shape_id)
sf_line[["geometry"]] <- st_sfc(lapply(split(NYC$shapes_df, NYC$shapes_df$shape_id), function(x) st_linestring(as.matrix(x[order(x$shape_pt_sequence), c("shape_pt_lon", "shape_pt_lat")]))), crs = "+proj=longlat +ellps=WGS84 +datum=WGS84 +no_defs")
sf_line <- st_as_sf(sf_line)
## make shape_id a factor so plot works (but watch out for ggplot2::geom_sf coming soon)
sf_line$shape_id <- factor(sf_line$shape_id)

I hope that helps, I'm happy to go deeper if you need, but I kind of need to hear what work flows you'll want. I am interested, I've been watching the progress being made a Sydney Transit recently!

You can't really have a line with properties on the vertices or on the segments, though you can have Z and M but they have to be floating point (same as x, y) so you can't store arbitrary stuff there. In the past I've used the trip package as a really primitive way of doing that, and recently I've worked on more abstract forms in spbabel and rangl, but I can't quite pin it down yet.

You can easily break these lines into vertices and segments and store them right in a dataframe, and grouping and recombining them in sf world is pretty good. Feel free to ping me with your ideas about what would be really helpful and I'll try to do it. It's reasonably easy to merge in elevation and time and plot these things in 3- and 4-D and I'm very interested to start doing that with examples like this.

[tbuckl]

@mdsumner thanks for those snippets! @dantonnoriega thanks for this excellent package! as you can see above, i'm poking around with questions like this. let me know if you have any questions.

[dantonnoriega]

Added @tombuckley gtfs routes to sf function. @mdsumner, once the release is live, would you mind checking out the results? I'm not familiar enough with sf data to understand the benefit.

[mdsumner]

I can do that. One thing I learnt just now is to ensure lower level classes in sp don't have named lists which is a side effect of some approaches I have used. Fwiw running mapview on a sprint object is a good way to visualize it, check for basic sense.

[dantonnoriega]

Awesome! Thanks!

[mdsumner]

It definitely looks fine to me, @mpadge will have stronger insights about the best ways to manage the conversions, but I'd say that can be put off.

These functions (in convert_gtfs.R) definitely need examples and stronger title and description slots. They could be bundled together in one Rd.