Project Description

Contributors: Giorgos Daskalopoulos, Panagiotis Mavridis.

This project uses the BF library. BF stands for 'Block File'. BF library provides a way to store data into fixed-size blocks. Usually libraries like BF, are used from Databases and not for general Input/Output.

Aim of the project is building a static library for indexing, using B+ Tree schemes.

AM library interface (Implemented by Panagiotis Mavridis):

• AM_Init() : Initialization of global variables.
• AM_CreateIndex(...) : Creation of a new index (Block-File).
• AM_DestroyIndex(...) : Deletion of an existing index (Block-File).
• AM_OpenIndex(...) : Opens a block file, reads header info, setting global variables properly in order to access data.
• AM_CloseIndex(...) : Deletes any dynamically allocated data are associated with that file and close Block-File.
• AM_InsertEntry(...) : Inserts entry at index.
• AM_OpenIndexScan(...) : Setting up a new Scan into index.
• AM_FindNextEntry(...) : Calling that function for the first time of Scan,returns the first tuple that validates the scanning 'query'. Every time that AM_FindNextEntry function is called,returns the next tuple. If there are no more tuples returns null.
• AM_CloseIndexScan(...) : Deletes any dynamically allocated data are associated with that scan.
• AM_PrintError(...) : Prints the latest error that occurred.
  

All above functions' arguments are explained in AM.h header

AM library inner functions (Implemented by Giorgos Daskalopoulos):

• keycmp(key1*,key2*) : Compares two key values based on their type (string,int,float).
• find_pointer(tree_block*,key*) : Iterates tree block to find id of next tree or data block that may contains key value.
• new_entry(block*,entry*) : Inserts entry into data or tree block.Returns proper error code if block is full.
• split(block1*,block2*) : Given a full block1 and an uninitialized block2, transfers 1/2 of records to second block.
• insert_rec(block_id,entry*,empty_temp_entry*) : Traverse recursivly tree, store data at proper data block, while returning to previous recursive calls, adds new entries at tree blocks if it is necessary.
• search_rec(block_id,key*) : Traverse recursivly tree , returns block_id number that may contain entries with attribute1=key.
• PrintFile() : Prints all tuples that there are stored into that file. Implemented mostly for debugging.

Current implementation limitations:

• B+ Tree doesn't support overflow buckets. Inserting many identical records may lead to unexpected behavior.
• Index is not thread safe.
• Number of max concurrent scans is set at header, before compilation.
• Maximum amount of indexes that library can handle is set at header, before compilation.
• Entries have only two attributes.
• String attributes are stored as fixed-size char arrays.