Rewriting Nearly Departed (Part 2: Tech Background)
This post will give a little of the technical background to Nearly Departed and explain some of the reasons for the rewrite.
This is part 2 of a multi-part series:
Headline tech features
- Written in Objective C
- Kind-of MVVM
- Very little unit-test coverage
- Made with a set of discrete service classes, injected using Objection
Why no WatchOS support any more?
The first WatchOS support was a WatchOS 1 glance. Communication between Watch and iPhone used shared
NSUserDefaults and Darwin Notifications were used by the two processes to notify each other that something had changed. (Remember that for WatchOS1 the Watch app process ran on the iPhone, effectively using the Watch as a separate screen. So Darwin Notifications could be used for interprocess comms because both processes were running on the same device)
WatchOS 2 introduced native third-party apps, moving the Watch app process onto the Watch itself, so the Darwin Notification code was broken.
- National Rail’s system for tracking the movements of passenger services
- A SOAP API for accessing real-time information from Darwin
The API gives a tonne of great real-time data, including:
- Departure/arrival boards (which return the same data you would see on displays at any of our 2500 stations
- Progress information for all services which call at these stations. This includes:
- Scheduled arrival/departure time at each calling point
- Actual or estimated arrival/departure time at each calling point
Hats off to them, the OpenLDBWS team have done a great job. I have a couple of minor complaints (such as times being sent like
08:18 without timezone or date information) - but they’ve managed to nicely model a proper complex set of data. Shame it’s SOAP/XML though!
While the OpenLDBWS return some station data in the responses, I still need to maintain my own set of station data which is bundled with the app. My requirements are:
- For each station, I need the following:
- Station name (such as “Stevenage”)
- The Computer Reservation System (CRS) code, which uniquely identifies this station in the API (example: “SVG”)
- location (latitude/longitude)
- The app needs this data in list format (so it can search for stations by name or CRS code)
- But also in a format which can be searched by nearest station to a given location
The data itself comes from two sources:
station_codes.csv is simply a list of station names and their CRS codes, whereas
RailReferences.csv also has location data in easting/northing format. Unfortunately the data doesn’t completely match between those two files, so I have a Ruby script which does a few things:
- matches data from the
- corrects inconsistencies between the two files (NaPTAN often has incorrect CSR codes)
- adds some data which is missing from one file or the other (usually because a new station has opened but the source data hasn’t yet been updated)
- filters out some non-mainline stations (such as heritage rail or ferry terminals)
- converts easting/northing locations to latitude/longitude
- creates a two-dimensional binary tree based on station locations, and writes it to a 670K
.plistfile. The tree is 12 levels deep, each level being split alternatively by longitude/latitude.
The app reads this plist into memory in its tree structure for location searches, and flattens it to an array for name/CRS searches.
- I frequently get asked to provide support for showing ServiceDetails data from OpenLWBWS - ie, data for a particular service - but the app was only designed to show DepartureBoard data
- The app wasn’t written to use other data sources - so if I ever wanted to add London Underground data, it would be very difficult to do so
- It’s just not very well written! ;-)
In the next post, I’ll discuss my high-level plans for the rewrite.