Abstract The CLAIR data file currently in use at the University of Melbourne contains 21705 chemical analyses of igneous rocks of all types and ages from all over the world. The way in which these analyses have been collected and processed is described, together with the philosophy behind the file. Some “statistics” from the file indicate some disturbing trends, e.g. only 7% of the analyses contain some minor-element determinations with major-element analysis; 12.7% of analyses are published with incorrect totals; and 2.9% of the rock names on the file account for 59.7% of all analyses. Comments are also made on the completeness of the analyses, their geographical distribution and the references from which they came.