Re: Learning to use the hash effectively

Two things I will recommend:

File locking. You have a very big race condition in your code: you read in the contents of the file, alter them, then write them back out. Between the time that you've read the file and you write the file, someone else (ie. another process) could have changed that file. Then you would overwrite that chance when you write the file w/ your file contents.
One way to fix this is to open the file in read/append mode, flock it, seek to the beginning of the file, read from it, alter the file contents in memory, seek back to the beginning, truncate it, then rewrite the contents from memory. The flock will prevent the race condition (at least w/ another version of your program that uses flock).
Another way to fix the problem is to use a semaphore file, like in tilly's Simple Locking. Here, you flock a semaphore file when you want to enter the "critical section" of your program, and then other processes of your program cannot enter that critical section until you have released the lock.
I would recommend the second approach.
My second suggestion is, you could just use a DBM file for this, particularly since you already have the notion of keys mapping to values. In particular, you could use MLDBM to serialize the data structure into the DBM format of your choice.