The Atlas ALGOL Preprocessor for Non-Standard Dialects

In the summer of 1964 the Atlas Computer Laboratory accepted delivery of the S.R.C. Atlas and started to provide a service to the Universities, Colleges of Technology and some Government Departments. The service was intended to cater for programs which were either too large or, in some way, too awkward to be run on the customer's own computer. The bulk of the work originally programmed for other computers was written in the three languages Mercury Autocode, FORTRAN and ALGOL. Mercury Autocode was acceptable on Atlas with virtually no change as a compatible Atlas EMA (Extended Mercury Autocode) compiler had been provided by I.C.T. Ltd. FORTRAN programs tended to require slight modification for, although incorporating many dialects as subsets in the HARTRAN system, the compiler did not accept every dialect. However, almost all FORTRAN programs are punched on cards with standard punching conventions so that changes, when required, tended to be systematic and easily made by card replacement. The situation for ALGOL was the worst. The two machines for which ALGOL programs were most frequently written were the English Electric KDF9 and the Elliott 803. In each case both the input/output facilities provided and the punched form of the program were completely different from the Atlas ALGOL conventions. The Atlas compiler had a small set of input/output procedures reminiscent of Mercury Autocode. The KDF9 set was more comprehensive including extremely flexible formats for output and also magnetic tape procedures. The Elliott 803 dialect included two non-ALGOL statements read and print. These statements could have a variable argument list consisting of expressions, strings and procedure calls. Furthermore, whereas the Atlas compiler only accepted I.C.T. 7-hole tape the KDF9 programs were punched on 8-hole tape and the 803 programs on 5-hole tape. The preprocessor was therefore initially designed to translate these two dialects into a form acceptable to the Atlas compiler. Also, the possibility of readily incorporating additional dialects of ALGOL, if required, was borne in mind.