どうも〜おかむーです！ Today I want to dig into Tokyo and local governments' open data with an engineer's eye—"Code the Manifesto" style. I'll show what's working, what's annoying, and concrete fixes you can actually ship.

• Tokyo portal offers many CSV/GeoJSON datasets but discoverability and schema consistency vary
• Some municipalities publish via CKAN APIs (good!), others still leak data in PDFs (bad UX)
• Technical fixes: stricter schema registry, API-first publication, simple ETL recipes to validate policy metrics

結論

Tokyo and several prefectures are publishing genuinely useful machine-readable datasets (CSV, GeoJSON) via portals like https://portal.data.metro.tokyo.lg.jp/ and CKAN instances (e.g. Daisen city). ただし、フォーマットのばらつき、メタデータ不足、PDF残存が課題。エンジニア的に言うと、API-firstとスキーマ駆動のワークフローを採用すれば、政策検証のスピードと信頼性がグッと上がりますよ！

Report

What I looked at

• Tokyo Open Data Catalog (portal.data.metro.tokyo.lg.jp / catalog.data.metro.tokyo.lg.jp)
• CKAN-based municipal catalogs (e.g. Daisen: https://www.city.daisen.lg.jp/open-data/dataset/)
• Prefectural catalogs like Saitama (https://opendata.pref.saitama.lg.jp/datasets)
• GovTech Tokyo initiatives (https://www.govtechtokyo.or.jp/)

これ見てくださいよ：Tokyo's catalog exposes CSV and GeoJSON for library lists, parks, sports facilities (search results show CSV/GeoJSON availability). That's great—machine-readable is the baseline!

Common technical issues

• Fragmented schemas: different field names for the same concept (e.g. "address" vs "addr" vs "住所")
• Missing metadata: frequency, update timestamp, license often absent or unclear
• PDF-locked numbers: policy reports still embed tables in PDFs without CSV exports
• Inconsistent geospatial CRS or mixed GeoJSON/CSV coordinates

要するに、データは「あるけど使いにくい」という状態です。

API & tooling status

• CKAN APIs are present in some municipalities—this is low-hanging fruit. CKAN provides REST endpoints for search, package_show, resource_get.
• Tokyo catalog appears to support direct downloads; recommend exposing a proper OpenAPI / OData layer for programmatic clients.

Example quick ETL (Python/pandas) to fetch a CSV from a CKAN resource:

import requests
import pandas as pd
r = requests.get('https://www.city.daisen.lg.jp/open-data/api/3/action/resource_show?id=RESOURCE_ID')
url = r.json()['result']['url']
df = pd.read_csv(url)
quick quality checks
print(df.columns)
print(df.isnull().sum())

Policy metrics and gap analysis

Many policy targets (e.g. facility counts, population by age brackets) are published but not as time-series APIs. That makes tracking year-to-year achievement clunky. エンジニア的に言うと、time-series endpoints + stable identifiers are required to compute deltas and confidence intervals.

Concrete technical proposals

• Adopt a schema registry (JSON Schema) per "standard open dataset" (Digital Agency recommended sets show this is feasible: see Tokyo's adherence notes)
• Mandatory metadata fields: updated_at, provenance, license (SPDX), schema_version
• Deprecate PDF-only releases: attach CSV/JSON and a machine-readable summary
• Provide example notebooks and a small postman/Swagger collection for each major dataset
• Track usage metrics and CORS-enable APIs so client apps can fetch directly
• Lightweight CI: run automated validators (jsonschema, geos validation) on dataset publish

まとめ

Tokyo and many municipalities are on the right path: CSVs, GeoJSONs, CKAN. But to make public data truly actionable for civic tech and journalists, we need schema discipline, time-series APIs, and fewer PDFs. Implementing a schema registry + publish pipeline and some developer experience (OpenAPI, example notebooks) will multiply the data's impact.

おかむーから一言

I've built products on messy public datasets—trust me, a little engineering discipline goes a very long way. Let's push for API-first open data and ship the tools that let citizens and startups hold policy to account!