Best of British :: Cars :: About

This site is the result of a programming exercise. It provides a central location to look for (British) classic and high performance cars currently for sale.

Each ad is analyzed individually against a database containing more than the makes and models found on normal car sites. Try looking for a Super Seven (using the category) and you'll be surprised how many different makes there are. The make and model detection is supported in the quick search box as well – How about searching for a Lotus Esprit Turbo?

Enjoy

Konrad Holl

Advantages

This site offers (at least) these advantages:

• Car categories to identify car genres, e.g. Super Seven or Cobra.
• Detailed car makes and models.
• Cars which are offered on several platforms are aggregated into a single entry providing links to each individual ad.
• Car prices are automatically converted to €, mileages to Kilometers.
• The prices and even the price refiners are calculated based on today's currency exchange rates.
• The search result page also offers links to manufacturer home pages, car enthusiast clubs, and Wikipedia pages.
• Search index contents are based on a number of used car platforms ^*
• The index is updated on a daily basis.

* Only platforms which permit crawling (based on robots.txt).

Currently Crawled Car Sites

Are you missing a site? Drop me a note on bob-cars@gmx.de.

System Architecture

• Ubuntu 16.04 LTS Operating System
• Apache Tomcat 7.x Application Server
• Apache httpd 2.x Web Server
• Elasticsearch 2.x Search Engine, replacing
  my Lucene extension - LDK (Lucene Development Kit):
  • Dynamic Facets, used for refining search results based on price using correct currency exchange rates
  • Topic Maps, used to define car categories like Super Seven or Cobra
  • Result Clusters, used to aggregate ads featuring the same car
  • Advanced Text Analytics, used to detect make and model in car ads
• MySQL Database Server
• GeoIP Geolocation Database
• FamFamFam Icons and Flags
• jQuery mobile
• jQuery UI, enhanced by Chosen
• <htmlcleaner/> and Jericho HTML parsers

Current Hardware

• Intel Core i5 760 CPU, 8 GB RAM
• 120 GB & 160 GB SATA hard disks, RAID 1
• 2.304 kB/s Internet Uplink

Releases

v2.1

• Improved Elasticsearch queries
• Enabled HTTPS protocol

v2.0

• Migration from LDK to Elasticsearch 2.x

v1.4

• Migration from Lucene 4.4 to Lucene 4.10.4
• Performance improvements (search time decreased by 50%)

v1.3

• Introduced LDK document IDs.
• Introduced LDK document ID based facets, reducing latency while index is updated.

v1.2

• Introduced trie based autocomplete function for quick search.
• Introduced advanced color detection instead of regular expressions.

v1.1

• Introduced fuzzy replica detection, used to be Boolean only.
• New automated crawler mode.

v1.0

• First production version after 2 years beta state
• Implemented classifieds

beta 13

• Added dropdown boxes to slider facets (mileage, price, and year)

beta 12

• Migration from Lucene 3.6.1 to Lucene 4.4

beta 11

• Various bug fixes
• JavaScript/jQuery optimizations

beta 10

• Added language selection
• Optimized page loading time using Google PageSpeed Tools

beta 9

• Added option to change result list sort order
• Improved result clustering

beta 8

• Added search/type-ahead to category, make, and model selection
• New option to change the currency (preselected by user location)
• Finalized color refinement

beta 7

• Improved user interface (CW's change request)
• Upgraded hardware
• Improved make/model detection
• Changed category, make, and model selection
• Changed category implementation from facets to topic sets
• Experimental mobile version

beta 6

New make/model detection with scoring based on text position and Levenshtein distance.

beta 5

Updated layout using jQuery UI for value sliders.

beta 4

Make/model suggestions using AJAX and jQuery UI.

beta 3

First public beta.