“Time is an illusion.”
― Albert Einstein
Timeshape is a Java library that can be used to determine to which time zone a given geo coordinate belongs. It's based on data published at https://github.com/evansiroky/timezone-boundary-builder/releases, which itself is inherited from the OpenStreetMap data.
Timeshape is published on Maven Central. The coordinates are the following:
<dependency>
<groupId>net.iakovlev</groupId>
<artifactId>timeshape</artifactId>
<version>2018d.5</version>
</dependency>
The user API of library is in net.iakovlev.TimeZoneEngine
class. To use it, follow these steps:
Initialize the class with the data for the whole world:
import net.iakovlev.TimeZoneEngine;
TimeZoneEngine engine = TimeZoneEngine.initialize();
Or, alternatively, initialize it with some bounding box only, to reduce memory usage:
TimeZoneEngine engine = TimeZoneEngine.initialize(47.0599, 4.8237, 55.3300, 15.2486);
It is important to mention that for the time zone to be loaded by the second method, it must be covered by the bounding box completely, not just intersect with it.
During initialization, the data is read from resource and the index is built. Initialization takes a significant amount of time (several seconds), so do it only once in program lifetime.
Once initialization is completed, you can query the ZoneId
based on latitude and longitude:
import java.util.Optional;
import java.time.ZoneId;
Optional<ZoneId> maybeZoneId = engine.query(52.52, 13.40);
This section describes how the Timeshape is designed, which technologies it uses, and how to build it yourself. This might be helpful if you want to contribute to it, or just to understand how it works.
Timeshape is built around the idea of taking the source data with time zones in GeoJSON format, convert it into something more efficient to store and read than GeoJSON, and package converted data together with the code that can read and query it into a single artifact (JAR file). Efficiency is the key word here, because the source data is quite big, and using it as is would impose too high memory and artifact size requirements on the users of Timeshape.
Timeshape currently uses compressed protocol buffers
(a.k.a. protobuf) as the target data format. The protobuf data is compressed using 7Zip method, which allows to reach
relatively small artifact size: 19 MB total JAR size vs 57 MB for the source data only (GeoJSON compressed with zip).
The biggest win in terms of size is, however, not due to efficiency of protobuf vs GeoJSON, but due to the fact that float
is used instead of double
to store geo coordinates in protobuf.
This means, only 4+4=8 bytes are required for each point (latitude + longitude), instead of 8+8=16 bytes for double
.
Precision of float
is good enough for the source data.
At runtime, the code reads the packaged data and build a spatial index for querying. It uses quad tree for indexing, provided by the Esri geometry API Java library.
Timeshape uses sbt as build system. The sbt build definition has 4 main projects:
- geojson-protobuf
- core
- builder
- testApp
Below you'll find some information about those individual projects.
This project contains the protobuf definitions corresponding to GeoJSON format.
Those definitions are in file geojson-proto/src/main/protobuf/geojson.proto
. Java code to read and write such protobuf
messages is generated during compile time by sbt-protobuf sbt plugin.
Other projects (core
and builder
), which must read or write the protobuf, use those generated Java classes, and therefore
depend on geojson-protobuf
in classpath sense.
This project contains the logic to read the data into a quad tree and provide API for querying it. It's the main project
with which the library users interact, and provides the main published artifact. It uses sbt feature called
resource generator
to create the protobuf file containing the time zone data. The code that actually generates the
protobuf data file is in the builder project. The resource generator is run by sbt automatically when necessary.
This project is responsible for downloading the source data from Github and converting it from GeoJSON to protobuf format.
It's usually called from the resource generator of core
project, but can be run independently
(it's a standard Java application, after all).
It's a playground, more or less. It's used to experiment with the main Timeshape artifact produced
by the core
project, particularly for the purpose of estimating its memory usage (see Memory usage),
and maybe something else in the future.
If you want to build and run the Timeshape locally, follow these steps:
- Install sbt and JDK. It's proven to work on JDK 8, but newer versions might work too. Use latest versions, because changes of time zones happen regularly in real world, and only the latest JDK build might reflect them.
- Go to the directory where the source code is checked out and run
sbt
command there. - When sbt finishes to load the build definition and you see sbt console, you have several options:
- run
core/test
to execute the tests, they should pass. - run
testApp/run
to run the test app. It will query the data for one time zone and print the memory usage. - run
core/publishLocal
if you've made local modifications and want to use the modified version in your program. Version must be set to snapshot for this to work best.
- run
The testApp
project provides memory usage estimate by using JOL.
The current version's estimated footprint is roughly 128 MB of memory when the data for the whole world is loaded.
It is possible to further limit the memory usage by reducing the amount of time zones loaded. This is implemented by a call to
TimeZoneEngine.initialize(double minlat, double minlon, double maxlat, double maxlon)
.
Version of Timeshape consist of data version and software version, divided by a '.' symbol. Data version is as specified at https://github.com/evansiroky/timezone-boundary-builder/releases. Software version is an integer, starting from 1 and incrementing for each published artifact.
The code of the library is licensed under the MIT License.
The time zone data contained in library is licensed under the Open Data Commons Open Database License (ODbL).