ekidata-jp

The free CSV files of 駅データ.jp (https://ekidata.jp/), the railway station database of Value and Vision Co., Ltd. (株式会社バリューアンドビジョン): the railway companies of Japan, their lines, their stations with coordinates, the pairs of adjacent stations on each line, and a prefecture master. Each dated file the download page offers is its own subset, with a typed Parquet beside the CSV as served; stations also carry a point geometry.

Use, modification and redistribution are allowed by the site's terms, with one condition on redistribution: unmodified data only free of charge (see Licence).

What the data is

駅データ.jp offers its data in a free tier and a paid tier. This dataset holds the free tier only, as listed on the members' download page https://ekidata.jp/dl/?p=1. That page needs a login with a free member account; the files here were downloaded with the account of the person who built this dataset. No credential is part of this dataset or of the code that built it, which reads them from the environment.

table	file on the site	content	specification
company	companyYYYYMMDD.csv	railway companies (事業者)	https://ekidata.jp/doc/company.php
line	lineYYYYMMDDfree.csv	lines (路線)	https://ekidata.jp/doc/line.php
station	stationYYYYMMDDfree.csv	stations (駅)	https://ekidata.jp/doc/station.php
join	joinYYYYMMDD.csv	adjacent stations on a line (接続駅)	https://ekidata.jp/doc/join.php
pref	pref.csv	prefectures (都道府県マスタ)	https://ekidata.jp/doc/pref.php

The site's specification (https://ekidata.jp/doc/) says the free data lacks, compared with the paid data: part of line_color_c and line_color_t, all of line_type, all of station_name_k and station_name_r, and every Shinkansen station. Apart from these, the two are the same.

Editions

The download page offers several dated files of each table. Each is a subset named <table>.<date>, the date being the one in the file name. The latest edition of each table is also the subset named after the table alone (station is the default subset). pref.csv has no date and is the subset pref.

subset	rows	raw file
station.2026-07-31 (= station)	10,962	raw/2026-10-05/station20260731free.csv
station.2026-07-13	10,962	raw/2026-10-05/station20260713free.csv
station.2026-06-18	10,962	raw/2026-10-05/station20260618free.csv
station.2026-05-28	10,941	raw/2026-10-05/station20260528free.csv
station.2026-04-30	10,943	raw/2026-10-05/station20260430free.csv
station.2026-04-09	10,943	raw/2026-10-05/station20260409free.csv
line.2026-06-18 (= line)	624	raw/2026-10-05/line20260618free.csv
line.2026-04-09	623	raw/2026-10-05/line20260409free.csv
line.2026-03-23	622	raw/2026-10-05/line20260323free.csv
line.2025-06-04	622	raw/2026-10-05/line20250604free.csv
line.2025-05-29	622	raw/2026-10-05/line20250529free.csv
company.2026-04-09 (= company)	175	raw/2026-10-05/company20260409.csv
company.2025-10-15	174	raw/2026-10-05/company20251015.csv
company.2025-06-04	174	raw/2026-10-05/company20250604.csv
company.2025-05-29	176	raw/2026-10-05/company20250529.csv
company.2025-05-23	174	raw/2026-10-05/company20250523.csv
join.2026-09-14 (= join)	10,188	raw/2026-10-05/join20260914.csv
join.2026-06-18	10,189	raw/2026-10-05/join20260618.csv
join.2026-04-09	10,168	raw/2026-10-05/join20260409.csv
join.2026-02-26	10,168	raw/2026-10-05/join20260226.csv
join.2025-09-16	10,173	raw/2026-10-05/join20250916.csv
join.2025-09-03	10,172	raw/2026-10-05/join20250903.csv
pref	48	raw/2026-10-05/pref.csv

All files were fetched on 2026-10-05 (Japan time), which names the directory raw/2026-10-05/ and parquet/2026-10-05/. The site corrects its data and posts dated change notes (更新履歴) on its pages; the latest note on the download page that day was dated 2026-09-14. raw/2026-10-05/MANIFEST.json gives each file's URL, the category and date the page shows for it, size, sha256 and row count. The CSVs are the site's own files, byte for byte.

Files and columns

parquet/2026-10-05/<subset>.parquet holds one subset each. Column names are those of the CSV and keep its order. The type of a column comes from its name, never from inference:

columns	type
codes: company_cd, rr_cd, line_cd, station_cd, station_g_cd, pref_cd, station_cd1, station_cd2	INTEGER
company_type, line_type, e_status, e_sort, zoom	INTEGER
lon, lat	DOUBLE
everything else, including post, line_color_c, open_ymd, close_ymd	the text of the CSV
geometry (stations only)	added here: a point from lon and lat

The codes are identifiers. They are kept as integers because every value in every file is an integer without leading zeros, so no code loses anything; a value with a leading zero, or any value that is not an integer, stops the build instead of being changed. line_color_c stays text because it is a hex colour (000000 among them). open_ymd and close_ymd stay text because they hold 0000-00-00. An empty field is null.

table	columns
company	company_cd, rr_cd, company_name, company_name_k, company_name_h, company_name_r, company_url, company_type, e_status, e_sort
line	line_cd, company_cd, line_name, line_name_k, line_name_h, line_color_c, line_color_t, line_type, lon, lat, zoom, e_status, e_sort
station	station_cd, station_g_cd, station_name, station_name_k, station_name_r, line_cd, pref_cd, post, address, lon, lat, open_ymd, close_ymd, e_status, e_sort, geometry
join	line_cd, station_cd1, station_cd2
pref	pref_cd, pref_name

The meaning of every column is in the specification. In short: e_status is 0 in operation, 1 not yet open, 2 closed; company_type is 0 other, 1 JR, 2 major private, 3 semi-major private; station_g_cd groups stations of different lines that count as one station; rr_cd is a two-digit railway code; in line, lon, lat and zoom are the centre and Google Maps zoom of a map that shows the whole line, not a location of the line.

The geometry is a GeoParquet 1.0.0 Point column without a crs key, which the GeoParquet specification defines as OGC:CRS84 (WGS 84, longitude then latitude). The station specification says of lon and lat only 世界測地系 (the world geodetic system) and refers to the FAQ for the datum; the FAQ (https://ekidata.jp/faq.php, read 2026-10-05) has no entry on it. The coordinates are therefore read as WGS 84; JGD2000 and JGD2011, Japan's realisations of the world geodetic system, differ from it by much less than the precision of a station's position. Rows keep the order of the CSV.

-- DuckDB, straight off the Hub: stations in operation per prefecture, latest edition
SELECT p.pref_name, count(*) AS stations
FROM 'hf://datasets/yuiseki/ekidata-jp/parquet/2026-10-05/station.2026-07-31.parquet' s
JOIN 'hf://datasets/yuiseki/ekidata-jp/parquet/2026-10-05/pref.parquet' p USING (pref_cd)
WHERE s.e_status = 0
GROUP BY ALL ORDER BY stations DESC;

Things to know

• No Shinkansen stations. The free station files leave them out, as the specification says, but the free line and join files keep them: the 12 Shinkansen lines (four-digit line_cd, 1001 to 1012) are in line with no station, and 154 rows of join.2026-09-14 are on those lines, between six-digit station codes that are in no free station file. Two more lines of line.2026-06-18 have no station: 11405 JR北陸本線(富山～直江津) and 99207 三陸鉄道北リアス線, both closed (e_status 2).
• station_name_k, station_name_r and line_type are empty in every row (paid fields). line_color_c and line_color_t are filled for 48 of the 624 lines of line.2026-06-18.
• open_ymd is 0000-00-00 for 10,584 of the 10,962 stations of station.2026-07-31, and close_ymd is 0000-00-00 for 295 of its 497 closed stations; 0000-00-00 means no date is given.
• 加木屋中ノ池 (station_cd 3000922) has lat 35 in every station edition, a whole degree where every other station has at least three decimals (most have six), so its point is unlikely to be where the station is; it is kept as served.
• The specification names the address column add; the CSVs call it address. Addresses mix full-width and half-width digits.
• station_g_cd is not always a station_cd: 23 group codes of station.2026-07-31 are not the code of any station, which the specification allows. e_sort repeats (143 values of station.2026-07-31 occur more than once); the specification orders by e_sort and then the code.
• Editions differ in a few rows each. For example station.2026-07-31 differs from station.2026-07-13 in one row, the postcode of 島高松, as the change note of 2026-07-31 says; join.2026-09-14 drops two rows of 京阪本線 and adds one of あいの風とやま鉄道線, as the notes of 2026-09-14 say.
• The download page also carries links to editions that are commented out in its HTML and so are not shown to users. Among them are 2026-09-14 editions of the line and station files (f.php?t=3&d=20260914, f.php?t=5&d=20260914), and older editions. None of these was taken. The change notes of 2026-09-14 on the site's pages describe changes to station and line data that the shown station and line files therefore may not have yet.
• The paid tier (有料データ) is not here.

Checked before upload

scripts/02_verify.py in the source repository checks:

• every raw file has the size and sha256 in MANIFEST.json, and its row count by Python's csv module equals the manifest's
• there is one Parquet per CSV and no other
• each Parquet has as many rows as the CSV
• every value of every row equals the value read from the CSV with the csv module and typed by the same rule (empty is null), row by row in CSV order
• every column has the type of the rule, in CSV order
• stations: GeoParquet 1.0.0 metadata naming a Point and no crs; every row has a point whose x and y are its lon and lat
• a DuckDB without extensions opens every file; no row group has a bloom filter

Licence

The data is under the terms of use of 駅データ.jp, not under an open licence. The terms (利用規約, https://ekidata.jp/agreement.php, last revised 2025-06-04 (改定日：2025年6月4日), read on 2026-10-05) say, in Articles 4 to 7:

第4条 (利用) 商用、非商用にかかわらず、第7条（禁止事項）に該当する内容を除き、本件データを弊社に連絡する必要なく利用することができる。また、本件データを利用する際に、弊社のデータであることを明記する必要はない。 但し、本件データを利用することにより生じたいかなる損害についても弊社は責任を負わない。 会員登録した法人は、駅データおよび弊社のサイトまたは案内資料（紙またはPDFデータ）で利用者実績や事例として、社名や利用目的が公開される場合がある。 また、公開時には利用者の法人ロゴの使用を弊社に認めるものとする。 【お願い】 弊社のデータであることを明記する必要はありませんが、弊社のデータであることの明記や、弊社サイトへのリンクを歓迎します。

第5条 (加工) 商用、非商用にかかわらず、本件データを加工して利用することができる。 但し、本件データを加工することにより生じたいかなる損害についても弊社は責任を負わない。

第6条 (第三者提供) 加工、非加工にかかわらず、本件データを第三者へ提供することができる。 但し、非加工で提供する場合は無償でなくてはならない。 ※ また、本件データを第三者提供することにより生じたいかなる損害についても弊社は責任を負わない。

第7条 (禁止事項) 本件データを利用するにあたり以下に該当することをしてはならない。 (1) 犯罪を勧誘又は助長するおそれのある利用 (2) わいせつ、裸体、死体、その他公序良俗に反する情報に関係する利用 (3) 消費者の判断に錯誤を与える恐れのある利用 (4) 第三者の財産権(知的所有権を含む)の侵害、プライバシーの侵害、誹謗中傷その他の不利益を与える利用 (5) 弊社または本件サービスへの誹謗中傷、弊社業務の運営・維持を妨げる利用 (6) ウイルスその他第三者に害悪を加えるおそれのある利用

and, for the mark in Article 6:

※「無償」とは、データのみで第三者へ提供する場合です。 非加工のデータを有償のシステムに組み込み、有償で販売しても問題ありません。但し、データは無償としてください。

In English, in short: the data may be used for any purpose, commercial or not, without contacting the company, except the uses Article 7 prohibits; the company is not liable for any damage. Attribution is not required, but credit or a link is welcome. The data may be modified. Modified or unmodified, it may be provided to third parties, but unmodified data only free of charge (free of charge refers to the data on its own; it may be built into a paid system). Article 7 prohibits uses that encourage crime; obscene or otherwise offensive uses; uses that may mislead consumers; uses that infringe third parties' property, privacy or reputation; defaming the company or its service or hindering its business; and uses that may spread viruses or otherwise harm third parties. The terms may be revised on the site, and use after a revision means agreement to it.

The raw CSVs here are unmodified data, provided free of charge on the Hub, which Article 6 allows. The Parquet files are modified data (Article 5): LICENSE beside this file says what was done. Whoever passes the unmodified CSVs on must not charge for them, and every use is subject to Article 7.

Citation

駅データ.jp, 株式会社バリューアンドビジョン (Value and Vision Co., Ltd.), https://ekidata.jp/, free data, editions as listed above, downloaded 2026-10-05.