14.5. Updating Sequential Files

Data that is formatted and written to a sequential file as shown in Section 14.3 cannot be modified without the risk of destroying other data in the file. For example, if the name “White” needs to be changed to “Worthington,” the old name cannot be overwritten without corrupting the file. The record for White was written to the file as

If this record were rewritten beginning at the same location in the file using the longer name, the record would be

The new record contains six more characters than the original one. Therefore, the characters beyond the second “o” in “Worthington” would overwrite the beginning of the next sequential record in the file. The problem is that, in the formatted input/output model using the stream insertion operator << and the stream extraction operator >>, fields—and hence records—can vary in size. For example, values 7, 14, –117, 2074, and 27383 are all ints, which store the same number of “raw data” bytes internally (typically four bytes on 32-bit machines and eight bytes on 64-bit machines). However, these integers become different-sized fields, depending on their actual values, when output as formatted text (character sequences). So, the formatted input/output model usually is not used to update records in place. Sections 14.6–14.10 show how to perform inplace updates with fixed-length records.

Such updating can be done awkwardly. For example, to make the preceding name change, the records before 300 White 0.00 in a sequential file could be copied to a new file, the updated record then written to the new file, and the records after 300 White 0.00 copied to the new file. Then the old file could be deleted and the new file renamed. This requires processing every record in the file to update one record. If many records are being updated in one pass of the file, though, this technique can be acceptable.