                                                     fixtest.txt 4-97
            
            
            
                              FIX 5.XX PROOF TEST        Page 1 of 1


           PURPOSE
           -------
            The central algorithm in FIX uses a buffer to find and
            process strings in a continuous input stream of unlimited
            length.  The technically inclined user will recognize the
            potential for insidious undetected omission errors under
            these circumstances.  The input stream must be processed
            one block or record at a time.  But the target string may
            be located anywhere in the input stream, without respect
            to record boundaries.  What happens when the target string
            spans two records?  If you are processing a delimited
            string, what happens when the string falls at the begin-
            ning or end of the record so that a possibly delimiting
            character falls in an adjacent record?  The test data file
            provided here enables the interested user to verify FIX
            performance under these most challenging data conditions.
            
            
           TEST DESIGN
           -----------
            The data test file has been designed to maximize the
            probability that every positional relationship between
            findstring and the inter-record boundary will be tested.
            Test data consists of a continuously repeating test
            pattern that is 5 characters in length.  There are 500K
            identical character patterns in the file resulting in a
            file size of 2.5M bytes.  The FIX processing record size
            is determined dynamically at run time and, for short
            strings, varies from 18K to 26K depending on wether the
            search is case sensitive or not.  This means that least 96
            buffer cycles will be required to process the test file.
            The density of test patterns in the data together with the
            number of buffer cycles make it highly likely that any
            "bugs" in the find/replace algorithm will be revealed.
            
            Test Data File:      TESTDATA.TXT  2,500,001 bytes (1 eof)
            Test Pattern:        "ABcd "  (not including quote marks)
            Pattern Repetitions: 500,000
            
            
           TEST EXAMPLE
           ------------
            >fix  'ABcd '  'ABcdef '  testdata.txt  testdat2
             TESTDATA.TXT  TESTDAT2  Occurrences replaced: 500,000
            
            >fix  'ef'  ''  testdat2  testdat3
             TESTDAT2  TESTDAT3  Occurrences replaced: 500,000
            
            >comp  testdata.txt  testdat3                NovaLogic(TM)
            Comparing TESTDATA.TXT and TESTDAT3          Beaverton, OR
            Files compare OK                             (503)524-9184






::
