Tilingsearch.org: a searchable repository of Islamic geometric patterns

Brian Wichmann

The paper in the proceeding is not being presented here, since it is too technical to be given in the short time available. Those interested can read the proceedings paper here: EVA2020.
Hence the presentation considers four issues which arose with the construction of the web site which I hope will be of interest. This material is available at: EVA2020 - 4 issues and will be presented via Zoom on 16th November 2020.

Introduction

Issue 1: Photos

Photographs are needed to validate the patterns produced. The photos are not shown on the web site due to copyright reasons. Hence the photos are retained in the documentation of each pattern.
A photo of a pattern is not necessarily available. A good example is the pattern recorded by Owen Jones in a visit to the Alhambra around 1830: Alhambra drawing. In this case, the tilingsearch web page contains a link to the V&A drawing - almost as good as a photo.
In many other cases, no photo or even a location may be available in the original published source. The original source of Bourgoin's work can be guessed from the sites he visited as recorded in: Possible Sources for Bourgoin (1879) "Le Trait des Entrelacs". (Tony Lee's work has been invaluable for the site).
Another problem is the the main photos available for the well-known sites are panoramas which do not give the detail required. Fortunately, David Wade's Patterns in Islamic Art does give a large collection with detail and so made the tilingsearch.org site possible.

Issue 2: Colour

Many patterns have no colour. Some may have been painted, but have lost that in time.
A specific pattern can be located in many places with varying colour or none. For instance, Bourgoin, Plate 171 is recorded as being at over 20 places. The original source is a drawing, while the version at the Agra Fort in India is carved pink stone, the version from Samarkand has individual patterned tiles. No dramatic colour here.
Dramatic colour can be found at: Bou Inania Madrasah, Fez, Morocco. The graphic has slight brighter colours. (Colour matching is tricky and varies with the lighting.) Complex patterns such as this are often only in one location. In this case, the graphic uses white interlacing which goes up/down at the straight cross-overs.
When a pattern has no colour in any location, then artificial colouring is used. The structure of the pattern can then be emphasised by the use of colour. For instance, consider: Bourgoin, Plate 177. As can be seen from David Wade's photo (EGY 1322), the structure is not obvious, while the graphic makes this clear.

Issue 3: Location

The National Art Library has been used to find patterns around the world. Books is seven languages have been used which results in some problems. One of these is that the same location has been found with very different names.
For some time now I have been attempting to add a location index to the system. However, without local knowledge this is difficult (or impossible!). With a local index, is is possible to link to the Archnet system which provides basic historic data which is useful for those using tilingsearch.org.
As a result of the French involvement in Egypt, the monuments in Cairo were numbered which provided a means of helping the construction of an index - very necessary in Cairo because of the large number of mosques. For the result of this work, see: Index to Egyptian sites. Further to this, for each site, a standardised name has been chosen from which a computerised set of indices has been generated. The current result for Egypt is: Index to Egypt. It is hoped the Geonames information will aid the production of location maps.
Work is ongoing in producing an index for each area. Substantial assistance has been forthcoming from Mustafa Bulut for Turkey and Mamoun Sakkal for Syria. In the case of the Alhambra, the individual rooms are listed in the index. Ensuring that the 131 patterns are located in the correct room has been made possible by Tony Lee's extensive knowledge of the site.

Issue 4: Character sets

An index to the sources was needed for the first version of the system. The table giving the data is in SQL using the UTF8 character set. This allows all the seven languages to be handled with ease by pasting in the text from major catalogues. At least one book referenced does not appear to be available in the UK.
For all the other software and data I type into the system I try to use Latin1. This can give rise to difficulties: the dotless i in Turkish is an example. It is not easy to ensure that no odd characters have not crept into the system since the result might be displayed after the data has been processed by several programs!
An earlier version of the system allowed users of the web site to undertake a text search. This proved to be unreliable for users who might type non-Latin1 characters or use a different spelling of a site name. It seems that little is lost by using only simple options for the user input.

v50