GB1399875A - Sorting operations on digital data processing systems - Google Patents

Abstract

1399875 Data processing; sorting INTERNATIONAL BUSINESS MACHINES CORP 29 Nov 1972 [22 Dec 1971] 55080/72 Heading G4M A digital data processing system sorts, by replacement selection, a string of input data from input file 54 (e.g. magnetic tape or disc) into a plurality of blocks in core store 50 and builds an index of keys each relating to the first record in its associated block, thereafter selecting the blocks by key order to merge them into a sorted string on output file 55. In replacement selection a number of input records are placed in the record storage area, the highest (or lowest) key record is removed to start a string, another record from the input file replaces it, the next highest (or lowest) record is added to the string, it is replaced from the input file and so on. When no further record can be added to a string a fresh string is started, the various strings being subsequently merged after the initial sort phase has ended. Initial sort phase (Figs. 2A-2C, not shown).- In the present disclosure a replacement selection sort is combined with building an index 52 (e.g. in core store). Where a block is defined as one or more records ordered in numerical sequence in a string, the index entries contain the location in work file 51 (e.g. a drum or disc) and the lowest key of each string block written in work file 51 from the replacement selection sort. Final sort phase (Figs. 3A-3C, not shown).- Index file 52 is accessed to identify the location in work file 51 containing the block with the lowest key and that block is transferred via block 58 to store 53. Then the next lowest keyed block is transferred (and so on) until area 53 is filled. The record with the lowest key in area 53 is selected, the lowest indexed record remaining in file 51 is located and the key numbers are compared. If the former is equal to or less than the latter, it is compared to the records in block 58. If it is again lower or equal, it is removed to the output file or added to an intermediate string. If it is higher the record is returned to file 51. The completed string in file 55 can be considered to be an intermediate string for use in further merging. The work file 51 may be organized as virtual storage with paging capability.