Georg Berky

Hi, my name is Georg Berky. I have worked as a student assistant and TA for the Chair of Databases and Information Systems from winter 2003/04 until summer 2010.
Back then my first job had been  implementing a test lab for P-News to evaluate the effect of ontological query relaxation and to improve the editor for Hasse diagrams.


After finishing my pre-diploma and until spring 2010 I have worked as a TA for Database Systems 1, Search Engines and Oracle Database Programming.

I finished my diploma thesis bearing the title "Cost-based optimization of partial Preference SQL query evaluation on SQL database systems" in December 2009.

contact me:

• real world: room 2054, new computer science faculty
  
  
• email:
  
  
• jabber:
  
  
• twitter:  @dbgezwitscher (in German, contact me if you want tweets in English)

for myself:

 SQL syntax descriptions in BNF

My active tickets
RecentChanges

Areas of interest:

• PSQL Parser
• Implementation of new preference constructors
• the PSQL query tree
• cost-based PSQL query optimization
• SQL92 implementation details
• database system internals
• mobile database systems and development for embedded devices

pages I contributed to

ExecutePSQL
GetQuery
How to get used to PSQL code
How to create a test database with lots of tuples
CostBasedOptimizer
DataConnector
ColumnSet
PSQL Parser

---

weekly Agenda

Week 19 - 23

• fix bug #156 (ORDER BY refactoring and projection issue)
• fix GROUP BY bug: #167
• implement HAVING keyword: #159
• finish DBS dialect abstraction: #157
  • do final checks where other DBS type checks occur in the code
    • add CREATE TABLE methods to DBCs and modify CreateTableNode

• finish PostgreSQL DBC: #162
  • figure out how to access statistics tables
  • construct statistics data queries and implement into DBC

Week 15 - 18

• continue work on GROUP BY and aggregate functions: #151
  • parser
    • clean up parser code and implement parser logger to speed up future work: #152
    • add CUSTOM [ 'class name' ](column) option
  • tree builder
    • add GROUPBY node above projection node in tree builder's callback method
  • java classes
    • implement GROUPBY query tree node class
    • implement aggregator/groupby cursor class
    • implement other aggregation functions
  • handle COUNT(*) properly
• Work on DROP TABLE bug: #155 that's #157 now

Week 14

• fix TEMPORARY TABLE bug #150
• work on GROUP BY and aggregate functions #151
  • aggregation cursors exist in XXL already
  • grouping cursors exist in XXL already
    • complicated and hard to use

Week 13

• work on SELECT/INSERT issue #148
  • wrongly assumed this was a CREATE TABLE issue before
    • column order is correct, verified by SQLDeveloper
  • check parser
  • check tree builder
  • check query tree node
  • problem was caused by iterator() method of ColumnSet. See ticket #17 for details.

Week 12

• SQL LIKE
  • parser
  • tree builder
  • new RelExpression subclass
  • test

Week 11

• implement ORDER BY
  • Parser
  • TreeBuilder
    • implement OrderElement class
    • implement order element stack
    • make callback methods work
    • implement ORDER BY in Projection node, similar to distinc, only apply extra cursor if in root node
  • Code

Week 10

• fix column-relation association caching issue

Week 9

• fix bug in INSERT INTO
  • find what's causing the ArrayIndexOutOfBounds? exception
  • find where the negative index is coming from
  • find where the wrong column names in the RelExpColumn field is coming from
  • trace back where the misplaced RelExpColumn on the stack is coming from

Week 8

• implement CREATE TABLE with subqueries
  • modify parser to accept subqueries
  • modify tree builder to create node and attach child node
  • test getQuery()

• implement INSERT INTO with subqueries
  • parser
  • tree builder
    • check whether subtree gets attached properly
  • node class
    • test node class

Week 6 & 7

• fix urgent Bug in LESS for Markus Endres
  • check open() calls
  • test using other evaluation algorithms (BNL, Hexagon)
  • build test lab
  • find logical bugs in implementation

Week 5 & 6

• implement INSERT INTO with subqueries
  • parser
  • tree builder
  • node class

Week 3 & 4

• assistant
  • UNION/UNION ALL
    • implement UNION cursor methods
    • implement UNION ALL cursor methods
    • read SQL standard and solve alias vs. real column name confusion
    • let Relation class help ColumnSet with retrieving the datatypes
    • implement UNION ALL in parser
    • implement CORRESPONDING BY deferred

Week 2

• assistant
  • SQL UNION
    • better method to ensure compatibility of attributes
    • find efficient ways to retrieve SQL datatypes in one query where possible
    • implement CORRESPONDING BY if supported

Week 1

• assistant
  • merge code with HEAD version
  • test EXPLICIT with JGraphT and Object instead of String
  • preference selectivity for RANKED
  • preference selectivity for SEMIPARETO
  • cost function for SortMergeJoinWrapper
  • selectivity for RelExpMinMax

Weeks 47 - 53

• thesis:
  • write program to create appropriate databases for testing
  • do evaluation of query execution times
  • make fancy gnuplot diagrams
  • write down evaluation results and add to thesis paper
  • port selection of best query tree to PSQL's query optimizer

Week 46

• thesis:
  • test PSROp node
  • test with multiple relations
  • test with additional preferences in the tree

Week 45

• thesis:
  • fix "array index out of bounds" problem
  • edit optimizer rule match() and executeOperation()
  • implement recursive query construction method as discussed with Markus Endres
    • implement recursive canExecuteOnDBS() method
    • check Oracle SQL syntax diagrams, check how binary operators work
    • start to implement recursive query construction

• assistant:
  • SQL LIKE for PSQL

Week 44

• thesis:
  • work on getQuery() method in new node class: fix "array index out of bounds" problem
    • read BoundStatementCursor code
    • read WrappingResultSetMetaData code: understand role of ColumnSet

Week 43

• thesis:
  • implement executeOperation() method for optimizer rule
    • first test successful, test more :-)
      
      
  • add execution cost field to Cost class and update getCost() method
    • cancelled because of comparison problems (PSQL costs vs. Oracle costs)
      
      
  • continue writing the thesis paper if sufficient time
    
    

Week 42

I hope the fancy number will make this week yield new answers :-).

• thesis:
  • implement new wrapper class for Projection, Selection, Relation and Join
    • analyzed how getQuery works and wrote a Wiki page for it: GetQuery
    • implement getQuery() method: tested, works
      
      
  • implement optimizer rule which transforms existing query tree nodes into the new node type
    • implement match() method
    • implement executeOperation() method: still buggy, but general approach seems to be good
      
      
  • computation costs:
    • refactor getQuery() in Relation: create new method yielding the query string
    • enable further measuring of execution costs via EXPLAIN PLAN on Oracle if sufficient time
      
      
  • continue writing the thesis paper if sufficient time

• assistant work:
  • EXPLICIT: implement free of cycles test
  • LAYERED: implement disjoint sets test

Week 41

• thesis:
  • Gazed upon Trac in awe :-)
  • Chapter 2: turned bullets into prose
  • integrated Markus' feedback
  • thought about new node class for the query tree

• assistant work:
  • DUAL
  • LINEARSUM
  • CONTAINS (regex),
  • documented ExecutePSQL