Jageocoder - A Python Japanese geocoder

日本語版は README_ja.md をお読みください。

This is a Python port of the Japanese-address geocoder DAMS used in CSIS at the University of Tokyo's "Address Matching Service" and GSI Maps.

Getting Started

This package provides address-geocoding and reverse-geocoding functionality for Python programs. The basic usage is to specify a dictionary with init() then call search() to get geocoding results.

>>> import jageocoder
>>> jageocoder.init(url='https://jageocoder.info-proto.com/jsonrpc')
>>> jageocoder.search('新宿区西新宿2-8-1')
{'matched': '新宿区西新宿2-8-', 'candidates': [{'id': 5961406, 'name': '8番', 'x': 139.691778, 'y': 35.689627, 'level': 7, 'note': None, 'fullname': ['東京都', '新宿区', '西新宿', '二丁目', '8番']}]}

How to install

Prerequisites

Requires Python 3.9.2 or later.

All other required packages will be installed automatically.

Install instructions

• Install the package with pip install jageocoder

To use Jageocoder, you need to install the "Dictionary Database" on the same machine or connect to the RPC service provided by jageocoder-server .

Install Dictionary Database

When a dictionary database is installed, large amounts of data can be processed at high speed. A database covering addresses in Japan requires 20 GB or more of storage.

• Download an address database file compatible with that version from here

jageocoder download-dictionary https://www.info-proto.com/static/jageocoder/20250423/v2/jukyoall20250423_v22.zip

• Install the dictionary with install-dictionary command

jageocoder install-dictionary jukyoall20250423_v22.zip

If you need to know the location of the dictionary directory, perform get-db-dir command as follows. (Or call jageocoder.getdbdir() in your script)

jageocoder get-db-dir

If you prefer to create the database in another location, set the environment variable JAGEOCODERDB2DIR before executing install_dictionary to specify the directory.

export JAGEOCODERDB2DIR='/usr/local/share/jageocoder/db2'
install-dictionary <db-file>

Connect to the Jageocoder server

Since dictionary databases are large in size, installing them on multiple machines consumes storage and requires time and effort to update them. Instead of installing a dictionary database on each machine, you can connect to a Jageocoder server to perform the search process.

If you want to use a server, specify the server endpoint in the environment variable JAGEOCODERSERVERURL. For a public demonstration server, use the following

export JAGEOCODERSERVERURL=https://jageocoder.info-proto.com/jsonrpc

However, the server for public demos cannot handle heavy traffic, so we have set a limit on the number of requests per second. If you want to process a large number of requests, please refer to here to set up your own Jageocoder server. The endpoint is '/jsonrpc' on the server.

Uninstall instructions

Remove the directory containing the database, or perform uninstall-dictionary command as follows.

jageocoder uninstall-dictionary

Then, uninstall the package with pip command.

pip uninstall jageocoder

How to use

Use from the command line

Jageocoder is intended to be embedded in applications as a library and used by calling the API, but a simple command line interface is also provided.

For example, to geocode an address, execute the following command.

jageocoder search 新宿区西新宿２－８－１

You can check the list of available commands with --help.

jageocoder --help

Using API

First, import jageocoder and initialize it with init().

>>> import jageocoder
>>> jageocoder.init()

The parameter db_dir of init() can be used to specify the directory where the address database is installed. Alternatively, you can specify the endpoint URL of the Jageocoder server with url. If it is omitted, the value of the environment variable is used.

Search for latitude and longitude by address

Use search() to search for the address you want to check the longitude and latitude of.

The search() function returns a dict with matched as the matched string and candidates as the list of search results. (The results are formatted for better viewing)

Each element of candidates contains the information of an address node (AddressNode).

>>> jageocoder.search('新宿区西新宿２－８－１')
{
  'matched': '新宿区西新宿２－８－',
  'candidates': [{
    'id': 12299846, 'name': '8番',
    'x': 139.691778, 'y': 35.689627, 'level': 7, 'note': None,
    'fullname': ['東京都', '新宿区', '西新宿', '二丁目', '8番']
  }]
}

The meaning of the items is as follows

• id: ID in the database
• name: Address notation
• x: longitude
• y: latitude
• level: Address level (1:Prefecture, 2:County, 3:City and 23 district,

4:Ward, 5:Oaza, 6:Aza and Chome, 7:Block, 8:Building)

• note: Notes such as city codes
• fullname: List of address notations from the prefecture level to this node

Search for addresses by longitude and latitude

You can specify the latitude and longitude of a point and look up the address of that point (so-called reverse geocoding).

When you pass the longitude and latitude of the point you wish to look up to reverse(), you can retrieve up to three address nodes surrounding the specified point.

>>> import jageocoder
>>> jageocoder.init()
>>> triangle = jageocoder.reverse(139.6917, 35.6896, level=7)
>>> if len(triangle) > 0:
...     print(triangle[0]['candidate']['fullname'])
...
['東京都', '新宿区', '西新宿', '二丁目', '8番']

In the example above, the `level optional parameter is set to 7 to search down to the block (街区・地番) level.

> [!NOTE] > > Indexes for reverse geocoding are automatically created the first time you perform reverse geocoding. Note that this process can take a long time.

Explore the attribute information of an address

Use searchNode() to retrieve information about an address.

This function returns a list of type jageocoder.result.Result . You can access the address node from node element of the Result object.

<pre><code class="lang-python">&gt;&gt;&gt; results = jageocoder.searchNode(&#39;新宿区西新宿２－８－１&#39;) &gt;&gt;&gt; len(results) 1 &gt;&gt;&gt; results[0].matched &#39;新宿区西新宿２－８－&#39; &gt;&gt;&gt; type(results[0].node) &lt;class &#39;jageocoder.node.AddressNode&#39;&gt; &gt;&gt;&gt; node = results[0].node &gt;&gt;&gt; node.get_fullname() [&#39;東京都&#39;, &#39;新宿区&#39;, &#39;西新宿&#39;, &#39;二丁目&#39;, &#39;8番&#39;]</code></pre>

Get GeoJSON representation

You can use the as_geojson() method of the Result and AddressNode objects to obtain the GeoJSON representation.

<pre><code class="lang-python">&gt;&gt;&gt; results[0].as_geojson() {&#39;type&#39;: &#39;Feature&#39;, &#39;geometry&#39;: {&#39;type&#39;: &#39;Point&#39;, &#39;coordinates&#39;: [139.6917724609375, 35.68962860107422]}, &#39;properties&#39;: {&#39;id&#39;: 80223284, &#39;name&#39;: &#39;8番&#39;, &#39;level&#39;: 7, &#39;priority&#39;: 3, &#39;note&#39;: &#39;&#39;, &#39;parentid&#39;: 80223179, &#39;siblingid&#39;: 80223285, &#39;fullname&#39;: [&#39;東京都&#39;, &#39;新宿区&#39;, &#39;西新宿&#39;, &#39;二丁目&#39;, &#39;8番&#39;], &#39;matched&#39;: &#39;新宿区西 新宿２－８－&#39;}} &gt;&gt;&gt; results[0].node.as_geojson() {&#39;type&#39;: &#39;Feature&#39;, &#39;geometry&#39;: {&#39;type&#39;: &#39;Point&#39;, &#39;coordinates&#39;: [139.6917724609375, 35.68962860107422]}, &#39;properties&#39;: {&#39;id&#39;: 80223284, &#39;name&#39;: &#39;8番&#39;, &#39;level&#39;: 7, &#39;priority&#39;: 3, &#39;note&#39;: &#39;&#39;, &#39;parentid&#39;: 80223179, &#39;siblingid&#39;: 80223285, &#39;fullname&#39;: [&#39;東京都&#39;, &#39;新宿区&#39;, &#39;西新宿&#39;, &#39;二丁目&#39;, &#39;8番&#39;]}}</code></pre>

Get the local government codes

There are two types of local government codes: JISX0402 (5-digit) and Local Government Code (6-digit).

You can also obtain the prefecture code JISX0401 (2 digits).

<pre><code class="lang-python">&gt;&gt;&gt; node.getcityjiscode() # 5-digit code &#39;13104&#39; &gt;&gt;&gt; node.getcitylocalauthoritycode() # 6-digit code &#39;131041&#39; &gt;&gt;&gt; node.getprefjiscode() # prefecture code &#39;13&#39;</code></pre>

Get link URLs to maps

Generate URLs to link to GSI and Google maps.

<pre><code class="lang-python">&gt;&gt;&gt; node.getgsimaplink() &#39;https://maps.gsi.go.jp/#16/35.689627/139.691778/&#39; &gt;&gt;&gt; node.getgooglemaplink() &#39;https://maps.google.com/maps?q=35.689627,139.691778&amp;z=16&#39;</code></pre>

Traverse the parent node

A "parent node" is a node that represents a level above the address. Get the node by attribute parent.

Now the node points to '8番', so the parent node will be '二丁目'.

<pre><code class="lang-python">&gt;&gt;&gt; parent = node.parent &gt;&gt;&gt; parent.get_fullname() [&#39;東京都&#39;, &#39;新宿区&#39;, &#39;西新宿&#39;, &#39;二丁目&#39;] &gt;&gt;&gt; parent.x, parent.y (139.691774, 35.68945)</code></pre>

Traverse the child nodes

A "child node" is a node that represents a level below the address. Get the node by attribute children.

There is always only one parent node, but there can be multiple child nodes. Therefore, children returns a list of address nodes.

Now the parent points to '二丁目', so the child node will be the block number (○番, △番地) contained therein.

<pre><code class="lang-python">&gt;&gt;&gt; type(parent.children) &lt;class &#39;list&#39;&gt; &gt;&gt;&gt; len(parent.children) 50 &gt;&gt;&gt; [child.name for child in parent.children] [&#39;1番&#39;, &#39;1番地&#39;, &#39;10番&#39;, &#39;10番地&#39;, &#39;11番&#39;, &#39;11番地&#39;, &#39;12番地&#39;, &#39;134番地&#39;, &#39;135番地&#39;, &#39;136番地&#39;, &#39;139番地&#39;, &#39;140番地&#39;, &#39;141番地&#39;, &#39;145番地&#39;, &#39;158番地&#39;, &#39;174番地&#39;, &#39;178番地&#39;, &#39;181番地&#39;, &#39;2番&#39;, &#39;2番地&#39;, &#39;3番&#39;, &#39;3番地&#39;, &#39;308番地&#39;, &#39;309番地&#39;, &#39;310番地&#39;, &#39;311番地&#39;, &#39;313番地&#39;, &#39;314番地&#39;, &#39;315番地&#39;, &#39;318番地&#39;, &#39;4番&#39;, &#39;4番地&#39;, &#39;5番&#39;, &#39;5番地&#39;, &#39;6番&#39;, &#39;6番地&#39;, &#39;673番地&#39;, &#39;674番地&#39;, &#39;7番&#39;, &#39;7番地&#39;, &#39;705番地&#39;, &#39;708番地&#39;, &#39;710番地&#39;, &#39;733番地&#39;, &#39;734番地&#39;, &#39;735番地&#39;, &#39;8番&#39;, &#39;8番地&#39;, &#39;9番&#39;, &#39;9番地&#39;]</code></pre>

For developers

Documentation

Tutorials and references are here.

Create your own dictionary

Consider using jageocoder-converter.

Tests

Run pytest for unit tests, pytest jageocoder/ --doctest-modules for testing sample codes in comments and pytest docs/source/ --doctest-glob=*.rst` for testing codes in the online manual document.

Contributing

Address notation varies. So suggestions for logic improvements are welcome. Please submit an issue with examples of address notations in use and how they should be parsed.

Authors

• Takeshi SAGARA - Info-proto Co.,Ltd.

License

This project is licensed under the MIT License.

This is not the scope of the dictionary data license. Please follow the license of the respective dictionary data.

Acknowledgements

We would like to thank CSIS for allowing us to provide address matching services on their institutional website for over 20 years.

We would also like to thank Professor Asanobu Kitamoto of NII for providing us with a large sample of areas using the older address system and for his many help in confirming the results of our analysis.