GEMINI: A Generalised Connection Table Language and Interpreter

Abstract

The design of a generalised program for interpreting connection tables is presented, including a language for describing the external encoding of structural information. Such a program is useful for format conversion or as a front-end to chemical application software. This program has been implemented (GEMINI), and is discussed with examples.

There are only three fundamental ways in which connection table formats differ: what information is stored, how that information is represented, and how those representations are encoded. GEMINI achieves a high degree of generality by dividing its task into these three parts.

Internally, GEMINI uses a rich set of data types with well-defined relationships which can include redundancy or ambiguity. Such information includes atomic number, connectivity, bond order, chirality, charge, co-ordinates, etc. Fortunately, there are very few different representations that are used for each kind of information, and algorithms are readily available for their interconversion.

A string-processing language is used to specify a connection table format. This language allows succinct specification of both the content and encoding of most external formats that we have encountered, but is currently limited to character-oriented files.