hankyu-umeda

name	hankyu_umeda
description	Platform rules, HTML structure, and Taiwan relevance filter for the 阪急うめだ本店 scraper
applyTo	scraper/sources/hankyu.py

hankyu_umeda Scraper Skill

Platform Profile

Item	Value
Site URL	https://www.hankyu-dept.co.jp/honten/event/
API/Rendering	static-html — requests + BeautifulSoup (no Playwright needed)
Auth	None
Rate limit	Polite: _DELAY = 0.5s before each request
Source name	hankyu_umeda
Source ID format	hankyu_umeda_{slug} where slug = last path segment of detail URL; fallback hankyu_umeda_{sha1(title+date_str)[:10]}

HTML Structure (confirmed 2026-04-28)

article div.o-event[data-place="{code}"]
  └── a[href]                          # detail URL (may be external subdomain)
       └── p.o-event__title            # event title
       └── p.o-event__desc             # short description
  └── div.o-event__detail
       └── p  "◎4月22日（水）～27日（月）\n◎9階 催場"  # date + venue
Week header: p#week01, p#week02, … p#week05

Field Mappings

Event field	Source
raw_title / name_ja	p.o-event__title
raw_description	開催日時: …\n\n + p.o-event__desc + detail text
source_url	div.o-event > a[href] (absolute URL)
start_date	div.o-event__detail p:first — ◎M月D日 regex
end_date	same — range after ～
location_name	"阪急うめだ本店 " + 2nd ◎ line (e.g. "9階 催場")
location_address	hardcoded: "大阪府大阪市北区角田町8-7 阪急うめだ本店"

Taiwan Relevance Filter

_TAIWAN_RE = re.compile(r"台湾|台灣|Taiwan|taiwan|🇹🇼", re.IGNORECASE)

Applied to BOTH p.o-event__title AND p.o-event__desc. Returns None if neither matches.

Past Taiwan event names (for testing):

• 台湾ライフ (秋 年1回)
• URL slug: taiwan_life

Date Extraction Rules

Date format: ◎4月22日（水）～27日（月） (same-month) or ◎4月29日（水）～5月11日（月） (cross-month)

Priority:

1. Cross-month regex _DATE_DIFF_MONTH: ◎(\d+)月(\d+)日…～(\d+)月(\d+)日
2. Same-month regex _DATE_SAME_MONTH: ◎(\d+)月(\d+)日…～(\d+)日
3. Single-day regex _DATE_SINGLE
4. Year inferred by _infer_year(): same year unless month < current month AND not in Dec→Jan wrap

Key Rules

• Page shows 5 weeks only. Taiwan展（台湾ライフ等）is typically in autumn (September–November). During spring/summer the scraper returns 0 events — this is correct behavior.
• source_id uses the URL slug (taiwan_life) — stable across runs even if the page is re-scraped daily.
• No Playwright, no JS execution required.
• location_address is always hardcoded (Osaka, Kita-ku) — not extracted from page.
• category defaults to ["lifestyle_food"]; annotator will refine.

Troubleshooting

Symptom	Likely Cause	Fix
0 events in spring/summer	No Taiwan fair scheduled	Normal — check again in September
Date parse returns None	New date format variant	Add regex to _parse_date_range()
source_id changes between runs	URL not absolute (starts with /)	Check _build_source_id — ensure absolute URL passed
65 items but 0 Taiwan	No current Taiwan event	Expected; re-run in autumn

Pending Rules