使用RandomAccessFile删除记录

Question

what i currently able to do for deleting a record from my .txt file is to put all the characters as empty but unfortunately it just make make those unit as empty and it will not remove those units. do you have any idea how can i completely delete a row of record from my txt file without rewriting the whole file again ?

protected boolean DeleteObject(int mID)
{
    try {
        file = new RandomAccessFile(new File("UserDB.txt"), "rw");
        long FileSize = file.length();
        file.seek(0);
        long NUMRecords = FileSize / RECORD;
        for (int j=0; j < NUMRecords; j++) {

            sID = file.readUTF();
            for (int i = 0; i < 10 - sID.length(); i++) {
                file.readByte();
            }
            ID = Integer.parseInt(sID);

            fName = file.readUTF();
            for (int i = 0; i < 20 - fName.length(); i++) {
                file.readByte();
            }

            surname = file.readUTF();
            for (int i = 0; i < 20 - surname.length(); i++) {
                file.readByte();
            }

            sMajor = file.readUTF();
            for(int i=0; i<3-sMajor.length(); i++){
                file.readByte();
            }
            major = Integer.parseInt(sMajor);

            if(ID == mID)
            {
                file.seek(j*RECORD);
                for(int k =0; k<RECORD; k++)
                {
                    file.writeUTF("");
                }
                return true;
            }
        }
    }
    catch (IOException e) {
        System.out.println("ERROR IN DELETING FILE "+e);
        e.getStackTrace();
    }

    System.out.println("Nothing Found");
    return false;
}

Answer 1

You could design your file to contain a logical delete flag. When deleting the record you would then update the logical delete flag, marking the record as deleted rather than physically removing it from the file.

Answer 2

I once had a file intended as a queue. The concept is that deleting always occurred linearly since it is a queue. So the deleting process would just rewrite the file offset of the next non-deleted row.

I reserved the first 8 bytes of the file as a file pointer (whose value was 8 or above). This makes the file binary though, not txt anymore. You could write that offset in a sibling file though.

However, this being a queue, I had to create new files from time to time and garbecollect the entirely deleted ones.

Otherwise, no you cannot cut away a file portion. Each row must have it's own delete flag. And vacuuming/packing/managing free space (or rewriting a clean version) is a sad reality for every database out there.