Fang.com (房天下) Real Estate Scraper

Scrapes fang.com — one of China's largest real estate portals — for second-hand (sale) listings, rental listings, and new-home projects.

Data Source

Public HTML pages of esf.fang.com (second-hand sales), zu.fang.com (rentals), and newhouse.fang.com (new-home projects). No login or API key required.

Modes

Mode	Inputs	Source
byCity (default)	listingType (sale/rent/newHome), city, rooms, sortOrder	city-subdomain list pages, e.g. esf.fang.com/house/i1/, zu.fang.com/bj/house/i1/, newhouse.fang.com/house/s/b2/
search	searchQuery, city	keyword listing search esf.fang.com/house/kw{kw}/
byListingUrl	listingUrls	sale esf.fang.com/chushou/{id}.htm, rent zu.fang.com/{city}/chuzu/{type}_{id}_1.htm, project newhouse.fang.com/loupan/{id}.htm

Output fields

sale / rent listings: title, price (万 CNY for sale, CNY/month for rent), pricePerSqm, rooms, area, location, community, orientation, floor, buildYear, publishDate, images (verified hotlink-accessible soufunimg.com URLs — real data-original lazy-load URLs, never loading placeholders), url, plus recordType, sourceUrl and scrapedAt on every record. Rent records also carry rentType and tags (e.g. 周边配套齐 / 采光好).

newHome projects: title, status (在售/待售), district, layout, areaRange, pricePerSqm, totalPrice, location, phone, images, url.

Detail records (mode=byListingUrl) add: houseId / projectId, community, district (GBK-decoded), agentName, purpose (esf), comarea + developer (newHome).

Filters

• minPriceWan / maxPriceWan — price range (client-side, applied to all records)
• minAreaSqm / maxAreaSqm — area range (client-side)
• rooms — 1–5+ bedrooms (server-side for sale; client-side for rent/new-home)
• sortOrder — server-side sort for sale/rent lists: default / latest / priceAsc / areaAsc

Reliability

• Exponential-backoff retries on 429/5xx and on the intermittent timeouts common to Chinese CDNs.
• On 403/429 the actor lazily engages the free Apify AUTO datacenter proxy group and retries.
• fang.com intermittently serves a JS slider-captcha check wall (check.3g.fang.com/check.html) with HTTP 200. The actor detects the wall, engages the AUTO proxy, and retries with backoff; if the wall persists it emits 0 records with a status message instead of garbage.
• Filtered-to-zero pagination guard: when client-side filters (rooms, minPriceWan, maxPriceWan, minAreaSqm, maxAreaSqm) exclude every listing on 3 consecutive full pages, pagination stops early instead of walking the entire city list, and the actor soft-fails with a status message.
• Invalid listing URLs produce typed recordType: "error" records (no silent drops).

Limitations

• city is applied as a subdomain for sale and new-home (jn.esf.fang.com, sh.newhouse.fang.com); rent uses the city in the path (zu.fang.com/sh/house/). city=national browses all cities for sale and new-home; rent has no national listing, so it falls back to Beijing (zu.fang.com/bj/house/).
• The three listing sites use different city codes. The actor auto-maps the schema's codes to each site's canonical code (verified Aug 2026), e.g. shenyang → sy.esf.fang.com / sy.newhouse.fang.com / zu.fang.com/sy/, wh → wuhan.esf.fang.com / wuhan.newhouse.fang.com / zu.fang.com/wuhan/ (wh.esf.fang.com is a dead host — fang.com's server refuses connections on it, so wh must map to wuhan on every site), su → suzhou, changsha → cs, haikou → hn. Requesting an unaliased wrong code used to yield a DNS failure (0 records) or — worst case for rent — fang.com's silent Beijing fallback page (wrong-city data); both are now prevented.
• fang.com has no coverage for Hohhot and Urumqi (no esf/newhouse subdomain, no zu path): these cities are excluded from the city dropdown, and rent browse for them emits 0 records with a status message instead of the site's Beijing fallback.
• New-home pagination uses b{page+1} URL segments observed on house/s/.
• List-page images are the site's 224x168c viewimage thumbnails; detail-page galleries are upsized to 512x384c viewimage variants.