0

My requirements are to write binary records inside a file. The binary records can be thought of as raw bytes in memory. I need a way to delimit each record, so that i can do something similar to binary search on the file. For example start in middle of file, find the next record delimited and start the search.

My question is that can ASCII such "START-RECORD" be used to delimit the binary record ? START-RECORD, data-length, .......binary data...........START-RECORD, data-length, .......binary data...........

When starting from an arbitrary position within a file, i can simply search for ASCII String "START-DATA". Is this approach feasible?

Jimm
  • 8,165
  • 16
  • 69
  • 118
  • Maybe you could do something like this, but I don't see how it helps much. A binary chop seaarch only works if the records are sorted. The only fast way to do this is to maintain a separate index so that data can be easily added without a huge file move/copy. If it's indexed, you don't need to search for a start-field. – Martin James Apr 26 '12 at 20:17
  • I agree with your comments on binary search. But other way to search would be to divide the file into arbitrary chunks, and give those chunks to various threads to do sequential scan. The chunks do not have to contain same number of records. So one way would be to determine offsets in file which would contain full records and assign those offsets to threads for search – Jimm Apr 26 '12 at 20:21

1 Answers1

1

Not in a single pass, since you're reading in binary mode or not. If you insert some strings or another pattern as "delimiter", you'd need to search for the binary representation of it while reading the file.

ChristopheD
  • 112,638
  • 29
  • 165
  • 179