A little update on transit routing in Maps

So, I thought it is high time for a little update about the transit routing project in gnome-maps (thought I should make some post while we're still in April).
I talked a bit with Mattias Bengtsson before, and since he had been contemplating using OpenTripPlanner (OTP) for his GSoC project a couple of year ago and found it didn't scale too well for general turn-based routing, he was quite excited about my idea of combing GraphHopper and OTP, using OTP with just transit data (loaded from GTFS feeds).

The basic idea here (when in transit mode) is to first run a query against OTP and as a further step do a ”refinement“ by re-running the parts of the routes where OTP selects walking between two transit locations.

An additional step has proven nessesary, since OTP, when running without OpenStreetMap ”street data” will approximate the walking ”legs“ of the trip as a straight line, which can be too optimistic in some instances. Therefore we do an extra safe-calculation to see if a particular itinerary seems reasonable with respect to time needed for walking (this is needed when there's walking in the middle of an itinerary, and there's an upcoming transit section that needs to be ”caught”).

Anyway, some screenshots:

Here we can see an itinerary where there's some walking in the start, it also tries to recalculate walking parts in the start and end by using the actual starting points as selected in the routing pane, instead of relying on the ones returned by OTP, which are based on the nearest transit stop.
The reason the transit part of trip (the solid line) is “jagged“ is that the data used here doesn't include shapes for the transit lines, so in this case OTP will interpolate with intermediate stops passed by in between.

A bit more beatiful routes can be obtained using data from the bus company serving the island of Honolulu:

Currenly it is hard-wired to always show the first returned trip. Also, as you can see it still doesn't render any intructions lists and the combobox to select departing/arrival time is not hooked up to anything yet.

As before the code can be found in the wip/mlundblad/transit-routing branch.

Also, there's some bug that sometimes gives a segmentation fault in Clutter (possibly there's some race condition in the code I rewrote to allow dynamically creating routing layers, to allow showing the dashed and solid lines interchangably). Unfortunatly debugging these things from JS is not all that fun…

Experimenting some more with map layers

One of the new nice features of Maps 3.20 is the ability of loading map layers in GeoJSON, KML, and GPX formats.
As I wrote about earlier, I have been experimenting with transit routing support. For that I have been using OpenTripPlanner, which has a notion of ”routers” (or “graphs“ as it could also be refered too) being discrete graphs consisting of nodes and connecting paths (such as transit routes) making up separate transit networks.
I thought it might be useful to get a visual picture of the sample data, so I got the idea of building a little tool that pulls the router information from a running OTP instance and generates a GeoJSON representation.

The code can be found here: https://github.com/mlundblad/otp-routers-to-geojson

This script, which by the way is my second case of using GJS as a scripting language, is rather simple and just takes as the one and only argument a URL to a running OTP instance (i.e. http://localhost:8080/otp when running a local server on your machine) and writes the output on stdout.

The script will skip any case where the graph's bounding polygon only consists of two points (not sure if that really would happen in reality).

Loading a generated layer into Maps can result in something like this:

The big diamond-shaped graph is probably the result of some malformed data, as that stems from the national transit feed from Estonia.

I hope you find this useful! :-)

View your tracks in Maps

Thought I should mention another nice feature that we have landed for 3.20, thanks to the nice work done by Hashem. Namely that you can now load GPX tracks recorded using i.e. a smart phone into Maps.

In the layers popover there is now a button to load geographical annotated data, it also supports loading geometries in GEO-Json format and KML (the format used in Google Earth).

This screenshot shows parts of a track recorded during a train trip (as you can see the track makes some jittery motions occasionally, but that is due to a bit shaky GPS reception).

Loaded layers shows up in the layers popover as shown below:

So now you can visualize trips and workout sessions directly in Maps. Also with shared-mime-info version 1.6 (which should be included in distros when shipping GNOME 3.20) double-clicking on a .gpx file should bring up Maps directly with the track loaded, as it registers itself as a mime handler for this mime-type.

The GPX track shown is awailable here https://cloud.gnome.org/index.php/s/G2YtwrveFo9RRjq
if you like something to play with.

A little teaser for future Maps

We've just entered the UI and string freeze for 3.20, we have a lot of new stuff coming up for Maps in 3.20 that I'm really excited about to see ”out the door”.

One thing that I've been wanting in Maps for a while is support for transit routing (for using public transportation options) in addition to our current ”turn-based” routing for car, bicycle, and walking powered by GraphHopper.

For a brief period in late summer took some time investigating the GTFS feed specification (for timetable data). The OpenTripPlanner project seemed rather fitting.

After a while of manually downloading some GTFS feed data and adding them to OTP's graph directory structure, I got a bit bored of that and wrote a little script to do the job https://github.com/mlundblad/otp-updater. It is quite crude and would probably need some better documentation (including a sample for the feed list configuration and such) and also currently requires quite a few command line parameters to point out the location for the OTP installation (used to run the graph rebuild command from there), these should probably be read from a global configuration by default…

Now, when things have calmed down a bit for 3.20 (and I wanted to try to keep main focus during my spare-time coding sessions getting the OpenStreetMap editing up to shape) I sat down and transformed the thoughts I've had in my head when it comes to integrating with an OTP server into a little proof-of-concept. The code is available in the https://git.gnome.org/browse/gnome-maps/log/?h=wip/mlundblad/transit-routing branch. The code is quite full of debug output code, is for the time being hard-wired to query an OTP instance running on localhost, and there's some duct tape here and there, but at least it does something…

OK, so enough of talking, now a little screenshot:

As you can here, there is now a new mode button to switch to transit mode. Currently, the only visible output is a raw debug dump of the JSON output from OTP. By the way, the blue route line shown in the screenshot is not actually from the transit routing, it's there since I first did the search using car mode and then switched to transit. And the reason it does a query also using the Estonian graph, seems to be some off coordinate in the data, I think.

I've also made a little list of feeds I've tried with here: https://wiki.gnome.org/Apps/Maps/Resources/GTFSFeeds

Running those feeds (plus a few more I had manually added before writing the updater script), the server takes around 7-8 GB RAM. So this would probably require a rather beefy machine to host the server.

Over and out for tonight! :-)

Maps shaping up for 3.20

So, we're soon approaching the UI freeze for GNOME 3.20. It's looking quite good when it comes to OpenStreetMap editing in Maps (among other things).

But first I thought I was going to show-case another improvement, namely the expanded place bubbles (show information about places you search for on the map).

This is what a place bubble will look like in 3.20 when you first open it, only the most basic information is visible, the name of the place, what it is (conveyed by the icon) and the street address. However if you click the ”more information button” (the dots), more stuff is revealed. Thanks Jonas for this!

Now we also show a link to the place's official website (if one is tagged in OSM, ofcourse) and likewise for the phone number (theoretically the phone number should turn into a clickable tel: URI link if you have an app installed that has registered a mime handler for that, but I didn't find have one, neither Empathy, Ekiga, or Skype seems to do it).

These fields are also possible to edit on OSM now, and if you copy a tel: URL from i.e. a web page, it would automagicall reformat it into the raw telephone number. Also note that the title of the editing dialog has been adjusted to better communicate the fact the changes are actually uploaded to the public OpenSteetMap database, and not stored locally or something like that. This was after user feedback we got after showing an earlier iteration.

If you click on the ”Add to OpenStreetMap” context menu item when not currently being zoomed in (right now to any of the two most zoomed-in levels), you get this notification with the option to zoom in to an appropriate level, centered on the position where you clicked.

Here we can see the POI type selection dialog, and this shows the types using the Swedish translation, so it probably looks like yibberish for a lot :-), but the reason I showed this is two-fold, first it shows how we get readily-translated names by parsing the data used by the iD web-based editor (thanks to Ole Laursen for that suggestion, by the way). Secondly, it shows a usage for another feature, namely how you can search for a value if happen to know the OSM tag used for that.

As you can see in the screenshot above, the ”hostel” tag value gets expanded to the title using the translation. Also, beneeth the search entry you can see the list of recently used types, selecting one of these will immediatly select that type and go back to the editing mode.

Here's how it looks when adding a new place, also this dialog title was adjusted to better reflect that stuff is added to the public OpenStreetMap database.

Before I forget, I should also mention the awesome work done by our Outreach participants.
Amisha Singla has been working on support for printing routes (with instructions) and Hashem Nasarat working on support for loading cust geojson map layers.

So, there's quite a lot of new features here. To those who regularily claim that ”GNOME just removes features in every new release” I would mention that Maps has grown by probably around 5000 LoC just for the OSM editing functionallity.

Borrowing POI types and translations from the iD OpenStreetMap editor

So, in a previos blog post I got comments from Ole Laursen and Zeeshan (zeenix) about why not fetching translations for location (POI) type names from the web-based OSM editor.
At first I was a bit sceptical, since our type concept will be a bit simplified (hiding the various tags that we consider as forming a ”POI type” under a single ”Type” selection).
But I got a little ”itch” for trying it out and managed to write a little script that mines this data from a git clone of iD and spits out a ”condensed“ JSON file with the type mappings with the translated names of tags.

And it works like this:

Editing an object with no type set currently, and it shows “None“ (as it would also do when creating a new location).

Clicking on the type button takes you to the type selection dialog. Start typing, and it will show matches (it should also fall back to matching the raw non-translated title if a translation for your current language is not added for a specific type, or there is no translation at all, as usual). As a little courtesy to nerdy OSM mappers, you can also type the name of a raw OSM tag value (such as ”bicyle_parking”) and it will show matches for this as well (this only works for tags that are included in the preset data from iD that are collected by the script though).

 

 

And selecting the type takes you back to the editor as before, with the type set to the one selected before.

Oh, and to be on the safe side, the type selector is only shown for existing objects if the type is already set to one of the defined types (and not for cases where there is a combination of tags, such as a place being defined as both an amenity and a shop, like say, a supermarket with a pharmacy department). This is done as a precausion to avoid any cases where tags would be overwritten in a non-obvious or hidden way. For objects where none of the type tags are defined, the selection will be permitted, and shows the initial ”None“ state.

A little helping hand when adding OpenStreetMap POIs

Since the last blog post I spent some time curing the ”amnesia“ of the POI type selection view. So, now it will show a list of (up to ten) most recently used types. And it will also save this list between runs.

Clicking one of the types in the list will select it and return to the editing view.
The list of available types is still very short and mostly interesting for testing purposes still. I might try to import some data from the iD editor. There is also some lookup code doing linear searching that might not scale so well, and possibly I might have to find a better solution for that, but I wanted to start out simple and not get lost in premature optimizations.

Another thing that will probably need some love is the search algorithm, currently it will only match the beginning of the type titles. While this probably works quite well for most cases in English, it might be awkward for languages with another word ordering, where for example different types of shops could start with the same string…