GB1378982A - Data processing apparatus - Google Patents

Abstract

1378982 Data sorting and merging system INTERNATIONAL BUSINESS MACHINES CORP 6 Oct 1972 [17 Nov 1971] 46151/72 Heading G4M A programmed data processor using memory discs is used to sort strings of data and merge the sorted strings into a final ordered string. The sorting is carried out in four Phases. In the first Phase, Phase 0, various parameters required for the sorting process are calculated. Blocking Factor B equal to the number of records that can be read into or out of store in a single operation, G, equal to the number of records that can be held in store at one time and equal to the storage area less the area required for buffers and stored programs and M2 and M3, the merge orders for Phase 2 and Phase 3 and equal to the number of sequences that can be formed in one pass to form a longer sequence. Phase 1 is the Internal Sort (Figs. 9A, 9B, not shown). Data from the input file is read into the store and the lowest numbered record is fed into a work file. The next record, if it is higher than the last record in the work file, is read into store and the procedure repeated. If a new record is lower than the last record in the work file it is read into store and tagged as unuseable. The procedure is repeated until either the output buffer is full or all the records in store are tagged and so produce a sequence break. In the latter case the tags are cleared and a new output string is produced. In the former the data is stored and the sorting proceeds. If the final number of strings produced is less that or equal to M3 then Phase 3 is performed. Otherwise Phase 2 is performed. Phase 2 is the External Sort or Intermediate Merge (Figs. 12A, 12B, 12C, not shown) in which the strings from Phase 1 are merged to form a number of strings equal to or less than M3. An optional sequence for merging is calculated so that the minimum number of record handlings occur to produce the M3 strings and this may involve using null strings, i.e. strings containing no data so that less than M2 actual strings are used in the merge (Figs. 16, 17, not shown). When the M3 strings are produced Phase 3 performs the final merge.