Yannis Smaragdakis

Yannis Smaragdakis (also Ioannis Smaragdakis, Γιάννης Σμαραγδάκης)

Professor
Mailing address (shortened but valid): Dept. of Informatics, University of Athens, Ilisia, 15784, Greece
Phone: +30 210 7275311
Email: yannis@smaragd.org (or smaragd@di.uoa.gr)

Events:

Current Service: ISSTA'22 (PC Chair).
YouTube channel for my lab

Teaching:

Greece: Compilers (S11,S12,S13,S14,S15,S16,F17,S19,S20,S21,S22); OO Languages and Systems/Advanced Programming Techniques (F10, F11, F12, F13, F14, F15, F17, F18, F19, F20, F21); Advanced Compilers and Static Analysis (S19, S20,S21,S22); Operating Systems (F17); Systems Programming (S14,S15,S16); SE (S11,S12,S13).
US: SE (S09); OO Languages and Systems (F09, S09, S08, S07, F05, F04, F03, F02, F01); Graduate Seminar (S07, F08); Distributed Systems (F06, F07); Intro OO (F05, S05); Advanced OS (S04, S03, S02, S01, S00); Memory Management (F00).

Research

Programming languages and software engineering:
- Program analysis (static analysis, test generation, invariant inference, symbolic execution)
- Language mechanisms for abstraction (declarative languages, program generation, DSLs, modules and components, generics, extensible languages, multi-paradigm programming)
- Languages and tools for systems (programming models for concurrency, language support for distributed computing, memory management and program locality)

Gratefully acknowledging funding by the ERC and the HFRI.
(Perpetually outdated) CV

My Ph.D. students: Anastasis Antoniadis.. Sifis Lagouvardos. Konstantinos Triantafyllou.
Ph.D. alumni: Brian McNamara, Eli Tilevich, Christoph Csallner, Shan Shan Huang, John Altidor, Kaituo Li, Aggelos Biboudis, George Balatsouras, George Kastrinis.
Research Associates: George Fourtounis,
Sample older B.Sc. and M.Sc. theses
Latest work:
Project areas, each with representative publications and links:
- Static Program Analysis: we express a rich set of static analyses completely declaratively (in the Datalog language).
- Program generators, generics in PL, extensibility: MorphJ is a language for safe "morphing", i.e., creating classes by iterating over members of other classes. Recent work also includes a generalization of the concept of "variance" in generics/templates: a way to marry parametric and subtype polymorphism. In the 90s, I also worked (as an intern and a funded Ph.D. student) on the Intentional Programming (IP) project at Microsoft.
- - "Stream Fusion, to Completeness" (POPL'17)
  - "Structured Program Generation Techniques" (GTTSE--Grand Timely Topics in SE)
  - "Automating Ad hoc Data Representation Transformations" (OOPSLA'15)
  - "Foo: A Minimal Modern OO Calculus" (FTfJP'15)
  - "jUCM: Universal Class Morphing" (MASPEGHI'15)
  - "Streams a la carte: Extensible Pipelines with Object Algebras" (ECOOP'15). Closely related to earlier performance analysis study: "Clash of the Lambdas" (ICOOOLPS'14)
  - "Refactoring Java Generics by Inferring Wildcards, In Practice" (OOPSLA'14)
  - "Forsaking Inheritance: Supercharged Delegation in DelphJ" (OOPSLA'13)
  - "Reified Type Parameters Using Java Annotations" (GPCE'13)
  - "Java Wildcards Meet Definition-Site Variance" (ECOOP'12)
  - "Morphing: Structurally Shaping a Class by Reflecting on Others": a comprehensive article (ACM TOPLAS, sans appendix)
    - "Expressive and Safe Static Reflection with MorphJ" (PLDI'08)
    - "Morphing: Safely Shaping a Class in the Image of Others" (ECOOP'07)
    - "Morphing Software for Easier Evolution" (RAM-SE'07)
  - "Statically Safe Program Generation with SafeGen" (SCP) shows a language that can certify a generator under all possible inputs to assert that it will only produce well-formed programs.
    - conference version at GPCE'05
  - "Taming the Wildcards: Combining Definition- and Use-Site Variance" (PLDI'11)
  - Meta-AspectJ (MAJ) is a state-of-the-art meta-programming tool for Java and AspectJ
    - "Domain-Specific Languages and Program Generation with Meta-AspectJ" (ACM TOSEM). Expands "Generating AspectJ Programs with Meta-AspectJ" (GPCE'04) and "Easy Language Extension with Meta-AspectJ" (ICSE'06 Emerging Results)
  - "cJ: Enhancing Java with Safe Type Conditions" (AOSD'07)
  - "Program Generators and the Tools to Make Them" (PEPM'04)
  - "A Personal Outlook on Generator Research" (Domain-Specific Program Generation)
  - "Application Generators": survey article in the Encyclopedia of Electrical and Electronics Engineering, (John Wiley and Sons)
  - "JTS: Tools for Implementing Domain-Specific Languages" (ICSR'98)
  - "Scoping Constructs for Program Generators" (GCSE'99)
  - "DiSTiL: a Transformation Library for Data Structures" in the Domain-Specific Languages conference (DSL)'97.
- Declarative languages: I am a late convert to declarative programming. An interesting direction is bounded, special-purpose declarative languages (e.g., for parallelism, or for GC-time computation).
- Program analysis and test generation for bug detection: recent work includes Residual Investigation: an approach that enhances static bug warnings with predictive dynamic analyses. My original work includes JCrasher and its followups, CnC and DSD-Crasher. JCrasher is a system for random testing of Java programs. CnC combines static checking (the ESC/Java system) with automatic test case generation. DSD-Crasher adds a dynamic invariant detection step to CnC. Finally, the ORM- modeling formalism allows efficient satisfiability checking and test data generation.
- Languages and language-inspired tools for distributed and concurrent computing: Some recent work includes the causally-precedes relation for sound, predictive race detection, and adaptive locks: a model for switching between mutex locks and transactions dynamically (either for performance or for correctness). See also the TIC model for transactional memory.
  Past work includes J-Orchestra: an automatic partitioning system for Java programs. It manages to enable distributed execution over a network, as if the program was running on a single VM, using only rewrite techniques over standard runtimes. See also NRMI: Java middleware offering a programming model for distributed computing that resembles single-machine programming.
- - "Sound Predictive Race Detection in Polynomial Time" (POPL'12)
  - "Adaptive Locks: Combining Transactions and Locks for Efficient Concurrency" (JPDC)
    - conference version at PACT'09, earlier version at TRANSACT'09, position paper at MSPC'08.
  - "Transactions with Isolation and Cooperation" (OOPSLA'07)
  - "NRMI: Natural and Efficient Middleware" (IEEE TPDS)
    - conference version at ICDCS'03
  - "J-Orchestra: Enhancing Java Programs with Distribution Capabilities": a comprehensive article (ACM TOSEM)
    - "J-Orchestra: Automatic Java Application Partitioning": the original J-Orchestra paper from ECOOP'02. Extended version in CoC Tech Report 02-17.
    - "Appletizing: Running Legacy Java Code Remotely From a Web Browser" (ICSM'05) uses J-Orchestra to convert a regular Java application into an applet-server combination.
    - "Portable and Efficient Distributed Threads for Java" (ACM Middleware'04) explains the distributed synchronization in J-Orchestra.
    - "Automatic Partitioning: A Promising Approach to Prototyping Ubiquitous Computing Applications" (IEEE Pervasive Computing)
    - "Automatic Application Partitioning: The J-Orchestra approach" (position paper from the 8th ECOOP workshop on Mobile Object systems)
  - "Transparent Program Transformations in the Presence of Opaque Code" (GPCE'06)
  - "Binary Refactoring: Improving Code Behind the Scenes" (ICSE'05): a refactoring-for-optimization technique that does not pollute the source code
  - "Aspectizing Server-Side Distribution": a description of GOTECH, (ASE'03)
- Memory management: I have worked on adaptive memory management algorithms, i.e., ways to combine two policies and get the best of both, with theoretical guarantees. In the past, I proposed the EELRU replacement algorithm for virtual memory systems, which consistently outperforms LRU. Additionally, I have worked on techniques for compressed virtual memory and for the reduction (i.e., simulation-specific compression) of program reference traces.
- - "Adaptive Caching: Effective Shaping of Cache Behavior to Workloads" (MICRO'06)
  - "General Adaptive Replacement Policies" (ISMM'04)
  - "The EELRU adaptive replacement algorithm" (Performance Evaluation--Elsevier)
  - - "EELRU: Simple and Effective Adaptive Page Replacement" (SIGMETRICS'99): the original EELRU paper
  - "Flexible Reference Trace Reduction for VM Simulations" (ACM TOMACS)
  - - "Trace Reduction for Virtual Memory Simulations" (SIGMETRICS'99): the original trace reduction paper
  - "The Case for Compressed Caching in Virtual Memory Systems", at the 1999 USENIX Annual Technical Conference
  - "Performing Replacement in Modem Pools" (USENIX'00): an algorithm for managing modem pools efficiently. Directly related to replacement algorithms in virtual memory!
  - - Some pointers to supplementary material.
- Multiparadigm Programming: FC++ is a really cool library for programming functionally in C++. It is the first library to fully combine higher-order and polymorphic functions, by neatly exploiting C++ type inference. LC++ is its small sibling.
- - "Functional Programming with the FC++ library" (Journal of Functional Programming)
  - - "Functional Programming in C++" (ICFP'00): the original FC++ paper
  - "FC++: Functional Tools for OO Tasks" (Software Practice & Experience) shows how common OO programming tasks benefit from FC++
  - The FC++ Web page contains pointers to more FC++ publications and information
  - Also related: "Static Interfaces in C++" (2000 C++ Template Programming Workshop): How can we enforce interface requirements in C++?
- Object-Oriented components: My Ph.D. proposed the concept of "mixin layers". A mixin layer is a static component encapsulating fragments of several different classes (mixins) so that all fragments are composed consistently. Mixin layers are an elegant implementation technique for layered designs.
- - "Implementing Layered Designs with Mixin Layers": the original mixin layers paper from ECOOP'98
  - "Implementing Large-Scale Object-Oriented Components" (Ph.D. Dissertation)
  - "Mixin Layers: An Object-Oriented Implementation Technique for Refinements and Collaboration-Based Designs" (ACM TOSEM)
  - "Mixin-Based Programming in C++" (GCSE'00)
  - "Layered Development with (Unix) Dynamic Libraries" (ICSR'02): Unix dynamic libraries (shared objects) offer great support for layered programming
  - - source code

Non-Technical:

"Effectiveness of Anonymization in Double-Blind Review" (CACM, Jun'18)
Why do a Ph.D. and how to pick an area (update of older "Ph.D. Rants and Raves")
Yannis's law: Programmer Productivity Doubles Every 6 Years. (Yes, I'm only half serious, but half is enough.)
My (favorite) academic humor