TzWorld

Resolve timezones from a location using data from the timezone-boundary-builder project.

Upgrading from 1.x {: .warning}

The on-disk data format changed in 2.x: priv/timezones-geodata.tzw1 replaces priv/timezones-geodata.etf.zip. After upgrading you must run mix tz_world.update once to reinstall the data in the new format. Until you do, every lookup returns {:error, :time_zone_not_found}. The old .etf.zip and .dets files in priv/ are no longer read and can be deleted to reclaim disk space (≈ 900 MB).

Installation

Add tz_world to your list of dependencies in mix.exs:

def deps do
  [
    {:tz_world, "~> 2.0"}
  ]
end

After adding TzWorld as a dependency, run mix deps.get to install it. Then run mix tz_world.update to install the timezone data.

NOTE No data is installed with the package and until the data is installed with mix tz_world.update all calls to TzWorld.timezone_at/1 will return {:error, :time_zone_not_found}.

Configuration

There is no mandatory configuration required however four options may be configured in config.exs:

config :tz_world,
  # Configure a custom TzWorld backend. It will be used
  # as the default backend in calls to `TzWorld.timezone_at/1`
  default_backend: MyTzWorldBackend,
  # The default is the `priv` directory of `:tz_world`
  data_dir: "geodata/directory",
  # The default is either the trust store included in the
  # libraries `CAStore` or `certifi` or the platform
  # trust store.
  cacertfile: "path/to/ca_trust_store",
  # The default is no options, however one can set any `httpc` client options.
  httpc_opts: [
    proxy: {{String.to_charlist(proxy_host), proxy_port}, []}
  ]

Backend selection

TzWorld provides alternative strategies for managing access to the backend data. Each backend is implemented as a GenServer that needs to be either manually started with BackendModule.start_link/1 or preferably added to your application's supervision tree.

The recommended backend is TzWorld.Backend.SpatialIndex. It is also the default — applications that do not pin :default_backend will pick it up automatically.

For example:

defmodule MyApp.Application do
  @moduledoc false

  use Application

  def start(_type, _args) do
    children = [
      ...
      TzWorld.Backend.SpatialIndex
    ]

    opts = [strategy: :one_for_one, name: MyApp.Supervisor]
    Supervisor.start_link(children, opts)
  end
end

The following backends are available:

• TzWorld.Backend.SpatialIndex (recommended, default) resolves a point by querying an R-tree spatial index built once at startup and held in :persistent_term. Lookups read directly from :persistent_term and bypass the GenServer mailbox entirely. This is the fastest backend on every benchmarked workload — and dramatically faster (≈18×) than any other backend on no-match queries (e.g. ocean points), where the previous bounding-box-scan algorithms had to walk every shape.

• TzWorld.Backend.EtsWithIndexCache keeps the timezone shapes in a compressed :ets table, with an in-memory cache of every bounding box for candidate filtering. Useful when you want shapes shared across processes via :ets rather than via :persistent_term.

• TzWorld.Backend.DetsWithIndexCache keeps the shapes on disk in a :dets file, with the same in-memory bounding-box cache. Useful when memory is constrained — only the index is kept in memory and shapes are loaded from disk on demand.

Other backends can be implemented as long as they follow the TzWorld.Backend behaviour. Custom backends should be configured in config.exs or runtime.exs under the :default_backend key so that they will be considered as the default for calls to TzWorld.timezone_at/1. For example:

config :tz_world,
  default_backend: MyTzWorldBackend

Installing the Timezones Geo JSON data

Installing tz_world from source or from hex does not include the timezones Geo JSON data. The data is required and to install or update it run:

mix tz_world.update

This task will download, transform, zip and store the timezones Geo data. Depending on internet and computer speed this may take a few minutes.

By default mix tz_world.update will download geojson data that does not include time zone information for the oceans. There are two optional parameters that are accepted by mix tz_world.update that can be used to configure the desired behaviour:

• --include-oceans will download the geojson data, including data for the oceans. This give almost complete global coverage of time zone data. The default is --no-include-oceans which does not include data that covers the oceans. The geojson data including the oceans is about 10% larger than the data that does not include the oceans.

• --force will force an update to the geojson data even if the installed data is the latest release. This option can be useful if you choose to switch from the data without ocean coverage to the data with ocean coverage (and the reverse). The default is --no-force.

Updating the Timezone data

From time-to-time the timezones Geo JSON data is updated in the upstream project. The mix task mix tz_world.update will update the data if it is available.

A running application can also be instructed to reload the data by executing TzWorld.reload_timezone_data.

Usage

The primary API is TzWorld.timezone_at. It takes either a Geo.Point struct or a longitude and latitude in degrees. Note the parameter order: longitude, latitude. It also takes and optional second parameter, backend, which must be one of the configured and running backend modules. By default timezone_at/2 will detect a running backend and will raise an exception if no running backend is found.

iex> TzWorld.timezone_at(%Geo.Point{coordinates: {3.2, 45.32}})
{:ok, "Europe/Paris"}

iex> TzWorld.timezone_at({3.2, 45.32})
{:ok, "Europe/Paris"}

iex> TzWorld.timezone_at(%Geo.PointZ{coordinates: {-74.006, 40.7128, 0.0}})
{:ok, "America/New_York"}

# Assumes that the downloaded data does not include
# data for the oceans (the default)
iex> TzWorld.timezone_at(%Geo.Point{coordinates: {1.3, 65.62}})
{:error, :time_zone_not_found}

Performance

Version 2.0 lookups are 1.4×–18× faster than 1.x (R-tree spatial index), and mix tz_world.update peak memory is ≈ 13× lower (end-to-end streaming pipeline). See the Performance guide for measurements and methodology.