#DB-Impl

This repository contains the solutions for the homeworks of the course Database Systems on Modern CPU Architectures.

##Building

There a two types of builds: a debug build and a release build. The release build is optimized for speed and does not contain debugging informations. For the debug build all optimizations are turned off (-O0), debugging informations (-g3) are created and special debugging code is enabled (using -DDEBUG).

You can create a release build using make. For the debug build enter make BUILD_TYPE=debug.

The binaries will be created in the bin/ folder. Executables built with the debugging flags will have the _debug suffix.

##Testing

There are two types of test cases:

• GoogleTest tests: use these tests for unit and integration tests. All files with a name which matches *Test.cpp in the directory tests/ and its subdirectories are automatically compiled and added to the test suite. The compiled testsuite will be named bin/runTests respectively bin/runTests_debug.
• scripted tests: these tests are intended for testing the commands provided to bash scripts. Test scripts must be placed in the tests/ directory or one of its subdirectories and their name must match *Test.sh.

make test executes both types of tests. All test cases are automatically executed by the continous integration service TravisCI as soon as new commits are pushed. Nevertheless, always test your commits before pushing them!

##Unused code

There is some code which I wrote but which is not used currently. You can find it in the unused directory.

##External sorting

For the first assignment an external sort algorithm had to be implemented. We chose to implement a merge sort algorithm. The binary is called bin/externalSort.

###Useful shell commands

od can be used in order to inspect the contents of a file.

od -A n -t x2 -v <fileName> #create a hex dump
od -A n -t u8 -v <fileName> #print the file as uint_64t
od -A n -t u8 -w8 -v <fileName> #print the file as uint_64t, one uint64_t per line

You can combine od with sort in order to sort the values in a file:

od -A n -t u8 -w8 -v <fileName> | sort -n

##Buffer manager

A thread-safe buffer manager can be found in the buffer directory. It uses the 2Q strategy for page replacement.

##Schema

Schema definitions can be stored in the database. There are three different tools to deal with the schemas:

• bin/parseSchema <inputFile>: parses the schema from <inputFile> and prints it to the console.
• bin/loadSchema <inputFile>: parses the schema from <inputFile> and stores it within the database. All previously loaded schema information will be lost.
• bin/showSchema: shows the schema currently stored in the database.

Only a subset of the valid SQL schema definitions are accepted by parseSchema and loadSchema. One accepted schema definition can be found in exampleSchema.sql.

##Slotted pages

An implementation of slotted pages can be found in slottedPages.

##B+-Tree

A template implementation of a B+-Tree can be found in bTree. It uses the existing implemtation of the buffer manager to store its nodes. Concurrent access is provided by lock coupling.

A simple CLI is available as bin/btreeVisualizer. The file btreeVisualizer.input.txt contains some example commands.

#Code generation

A code generator for arithmetic expressions can be found in the codegen folder.

It provides classes for forming an operator tree, generating LLVM code out of this tree and executing it.

It also contains a MetaExpression class which provides a more convenient interface for building the operator trees by using C++ operator overloading.