Question

I have an assignment where I have a tree where all the parent nodes contain yes or no questions and the leaf nodes contain an answer. There are two children per parent node, a yes node and a no node. This tree needs to be serialized and stored in a file via RandomAccessFile so that the whole tree isn’t written out to the file every time. If the leaf node doesn’t contain the answer the user is looking for, the user submits a new answer and a question to differentiate between the leaf’s answer and the answer he was thinking of. Then this leaf becomes a parent with its own question and two children a “yes” node and a “no” node (these two nodes are new leafs). This process is what’s giving me trouble because if a leaf is already written out to the file how can I overwrite the leaf without spilling into another node’s data (since the byte size of a leaf node changes when it becomes a parent). Note that the program can be killed at any time and its tree structure should still remain intact. Thanks

Answer 1

having a process write to a file, having in mind that the process can be killed at any time … -> journaling

the byte size of a node changes when it becomes a parent? let’s see …

a parent node has a question (reference to a string)
a parent node has a yes node (reference to a node)
a parent node has a no node (reference to a node)

a leaf node has an answer (reference to a string)
a leaf node may have 2 unused references to other nodes …

so a node has a string reference and 2 node references …

if both node refs are NULL, it’s a leaf node and the string ref is an answer … else it’s a parent node and the string ref is a question …

when you serialize that to a file:

you know a node has a fixed length: 3 pointers(references)
so your new string reference will be current position + 3 pointer length
write that address as string ref …
write 2 null references for now … (we don’t know yet where those nodes will be written)
write the string

traverse your tree and keep in mind that you have to update the skipped 2 node references when you know the position of the node in your file …

when you update a node (leaf becomes parent):

write the new question string to the file, and store it’s address
store the address of the current answer string …
replace the string reference in the old node with the new question string address
write a new node for the old answer (string ref is the stored address for the current answer) and update one of the old nodes node-references (yes or no, accordingly)
write a new node for the other answer (string reference can be stored 3 pointers later, so address is known)
write the new answer string
update the old nodes other node reference