eiga-com

name	eiga_com
description	Platform rules, per-theater listing strategy, and Taiwan filter for the 映画.com 台湾映画 scraper
applyTo	scraper/sources/eiga_com.py

映画.com 台湾映画 Scraper Skill

Platform Profile

Field	Value
Site URL	https://eiga.com
Rendering	Static HTML — no JS required
Auth	None
Rate limit	0.5 s per movie, 0.3 s per area/theater page
Source name	eiga_com
Source ID format	eiga_com_{movie_id}_{theater_id} (e.g. eiga_com_82162_3018)

Listing Strategy

1. Fetch https://eiga.com/search/%E5%8F%B0%E6%B9%BE/movie/ (and /2/, /3/ up to 5 pages)
2. All results from this search URL contain "台湾" in the title — inherently Taiwan-relevant
3. Filter by pub_date window: today - 90 days ≤ pub_date ≤ today + 180 days
4. Stop paginating early if a page yields no new in-window films

Per-Theater Event Strategy

One event is created per theater (not per movie):

/movie/{id}/theater/                    → area links
  └─ /movie-area/{id}/{pref}/{area}/    → div.movie-schedule blocks (per theater)
       └─ div.more-schedule a.icon.arrow → /movie-theater/{id}/{pref}/{area}/{theater_id}/
            └─ table.theater-table       → 住所 row

source_id: eiga_com_{movie_id}_{theater_id} — stable across weekly runs (upserts update end_date)

start_date: movie's 劇場公開日 (p.date-published strong)

end_date: last td[data-date] in the current week's schedule (updated each run via upsert)

Fallback: if /movie/{id}/theater/ returns no area links, emit one movie-level event with location_name=None and source_id=eiga_com_{movie_id}

HTML Structure

Listing Page

<ul class="row list-tile">
  <li class="col-s-3">
    <a href="/movie/82162/">
      <p class="title">台湾ハリウッド</p>
      <small class="time">2026年4月25日</small>
    </a>
  </li>
</ul>

Movie Detail Page

<h1 class="page-title">台湾ハリウッド</h1>
<p class="date-published"><strong>2026年4月25日</strong> 劇場公開日</p>
<p class="data">2013年製作／124分／G／台湾
  原題または英題：…
  配給：…
</p>
<!-- Synopsis: first <p> without class and len > 80 -->

Area Page (/movie-area/{id}/{pref}/{area}/)

<div class="movie-schedule" data-theater="K's cinema" data-title="台湾ハリウッド">
  <table class="weekly-schedule">
    <td data-date="20260427">…</td>   <!-- YYYYMMDD format -->
    <td data-date="20260428">…</td>
  </table>
</div>
<div class="more-schedule">
  <a class="icon copy"  href="/movie-theater/82162/13/130201/3018/mail/">コピー</a>
  <a class="icon print" href="/movie-theater/82162/13/130201/3018/print/">印刷</a>
  <a class="icon arrow" href="/movie-theater/82162/13/130201/3018/">すべてのスケジュールを見る</a>
</div>

Theater Detail Page (/movie-theater/{id}/{pref}/{area}/{theater_id}/)

<table class="theater-table">
  <tr><th scope="row">住所</th><td>東京都新宿区新宿3-35-13 3F<br/><a>映画館公式ページ</a></td></tr>
</table>

Taiwan Relevance Filter

Search URL %E5%8F%B0%E6%B9%BE = 台湾 — all results are inherently Taiwan films. Additional filter: pub_date within [today-90, today+180] to exclude very old releases.

Field Mappings

Event field	Source
name_ja	h1.page-title (or listing p.title as fallback)
raw_title	same as name_ja
source_id	eiga_com_{movie_id}_{theater_id}
source_url	/movie-area/{id}/{pref}/{area}/ URL
start_date	p.date-published strong → pub_date (movie release date)
end_date	last td[data-date] in weekly schedule
location_name	div.movie-schedule[data-theater]
location_address	table.theater-table th="住所" + td (a-tags stripped)
raw_description	開催日時: start〜end\n\np.data text\n\nsynopsis
category	["movie"]
is_paid	True

Date Extraction Notes

• Listing small.time format: "2026年4月25日" → _DATE_RE extracts Y/M/D
• Detail p.date-published strong format: same
• Schedule dates: td[data-date] = "YYYYMMDD" → _yyyymmdd_to_date()
• raw_description prefix: 開催日時: YYYY年MM月DD日〜YYYY年MM月DD日\n\n
• The schedule only shows 1 week (~5 days). Daily scraper runs keep end_date current via upsert.

Key Selectors

Purpose	Selector
Theater name	div.movie-schedule[data-theater]
Schedule dates	div.movie-schedule td[data-date]
All-schedule link	div.more-schedule a.icon.arrow[href*='/movie-theater/']
Theater ID	_THEATER_ID_RE on arrow link href
Address row	table.theater-table th:contains("住所") + td

Troubleshooting

Issue	Cause	Fix
0 theater events, 1 fallback event	theater() page has no area links (movie not yet released)	Fallback event is emitted — correct behavior
location_address includes "映画館公式ページ"	<a> not stripped from td before get_text()	Call td.find_all("a") → a_tag.decompose() before get_text()
location_address is JS code	Regex matched JS-embedded 東京都	Use table.theater-table approach, not page-wide regex
theater_id = "unknown"	No a.icon.arrow in more-schedule	Check if area page structure changed
Old Taiwan films returned	Pub date filter window too wide	Adjust _LOOKBACK_DAYS / _LOOKAHEAD_DAYS
403 errors	Missing User-Agent	Session headers include Chrome UA

Pending Rules

• Monitor whether eiga.com changes its search URL format (path vs query param).
• If listing structure changes, update ul.row.list-tile li.col-s-3 selectors.
• When a movie has multiple theaters (e.g. wide release), each theater gets a separate event with its own source_id.