23 November 2007RCEAS 2007, Budapest1 Embedded Functional Programming Gergely Patai Budapest University of Technology and Economics Department of Control Engineering and Information Technology

23 November 2007RCEAS 2007, Budapest2 Outline  Functional programming in a nutshell  Motivation: desirable properties  Reasons of neglect  Real-life examples in embedded systems development  Further directions

23 November 2007RCEAS 2007, Budapest3 Functional programming  Declarative programming paradigm  Programs defined as series of equations  Typical characteristics:  Lack of side effects (a.k.a. ‘purity’)  Strong, static typing and type inference  Succinctness  Some languages: Haskell, ML family, Scheme, Clean, F#

23 November 2007RCEAS 2007, Budapest4 FP properties: purity  Referential transparency  Function return values depend only on the explicit arguments, not the context  Context-free expressions: easier testing and debugging, richer static analysis possible  No variables  Comparatively little state space  Safe concurrency  Loops expressed with recursion  I/O clearly separated from logic

23 November 2007RCEAS 2007, Budapest5 FP properties: type system  Strong typing  Protection against misinterpreting data  Type inference  Types don’t need to be specified explicitly  Every subexpression has a well-defined type  Inconsistencies detected at compile time

23 November 2007RCEAS 2007, Budapest6 FP properties: succinctness  Typically little boilerplate  Code/structure reuse possible at a small level of granularity  Loops are often abstracted away  In general shorter by a factor of two to ten

23 November 2007RCEAS 2007, Budapest7 Why is it not widely used?  FP originates around 1960 with LISP  Too resource intensive for that time (garbage collection, reflection…)  Imperative languages pervaded the industry  Programmers got used to thinking imperatively

23 November 2007RCEAS 2007, Budapest8 Is it really not used?  Modern languages keep adopting features of functional languages  Managed memory  Type inference  Lambda expressions (unnamed functions)  Why not adopt the whole paradigm?

23 November 2007RCEAS 2007, Budapest9 Example: Lava  Haskell library to aid hardware design  Generating VHDL code from functional description  Mary Sheeran: Hardware Design and Functional Programming: a Perfect Match

23 November 2007RCEAS 2007, Budapest10 Example: Erlang  Concurrent functional language  Asynchronous IPC (message passing)  Fault tolerance mechanism  Hotswap support 

23 November 2007RCEAS 2007, Budapest11 Example: Lustre  Synchronous dataflow language, a declarative relative of Esterel  Used in commercial safety critical products (mostly avionics) since 1993 

23 November 2007RCEAS 2007, Budapest12 Example: Hume  Experimental language, still developing  Programs: boxes connected with wires  Wires: persistent state, one input and one output each  Boxes: buffered combination networks  Compile-time resource limit guarantees  Execution: VM, native code (C), FPGA  

23 November 2007RCEAS 2007, Budapest13 Where to go from here?  Functional languages viable in a wide range of applications  An underrated (or just unknown) paradigm  Our ongoing experiments:  Hume on embedded platforms (Tmote Sky, Mindstorms NXT, mitmót)  Haskell on the desktop for remote controlling robots  What about you?

23 November 2007RCEAS 2007, Budapest14 Thank you for your attention!