A Benchmark for Evaluating C ParserAnalysers - PowerPoint PPT Presentation

1 / 9
About This Presentation
Title:

A Benchmark for Evaluating C ParserAnalysers

Description:

enum. union. struct. Variables. Functions. Templates. Operators ... C and Fortran. Embeded SQL. Generated Code. lex/yacc. Stateflow. Test Buckets (cont. ... – PowerPoint PPT presentation

Number of Views:30
Avg rating:3.0/5.0
Slides: 10
Provided by: susan404
Category:

less

Transcript and Presenter's Notes

Title: A Benchmark for Evaluating C ParserAnalysers


1
A Benchmark for Evaluating C Parser-Analysers
  • Susan Elliott Sim
  • University of Toronto
  • simsuz_at_cs.utoronto.ca
  • November 4, 2001

2
Background
  • xfig 3.2.1 structured demonstration at CASCON99
  • Children of xfig
  • SORTIE project
  • C Parser-Analyser Olympics
  • Workshop at CASCON2001
  • Tuesday, November 6, 245pm-530pm
  • http//www.cs.utoronto.ca/simsuz/cascon2001

3
Motivation
  • C is a complex language and writing
    parser-analysers for it is hard.
  • Implementations of (C) parser-analysers
    typically trade-off between robustness and
    accuracy
  • full parsing vs. lexical approaches
  • Create a benchmark to empirically characterise
    where a parser-analyser lies in this design space

4
Motivation (cont.)
5
Test Buckets
  • Accuracy (14)
  • Preprocessor (3)
  • C Syntax
  • Data Structures
  • Array
  • enum
  • union
  • struct
  • Variables
  • Functions
  • Templates
  • Operators
  • Exceptions
  • Inheritance
  • Namespaces
  • Robustness (10)
  • Incomplete Information
  • Missing source, header, library
  • Dialects
  • GNU g
  • MS Visual C
  • IBM VisualAge C
  • Heterogeneous Source
  • C and Fortran
  • Embeded SQL
  • Generated Code
  • lex/yacc
  • Stateflow

6
Test Buckets (cont.)
  • Many possible buckets not yet included
  • Language features
  • Analysis problems
  • Quantitative tests
  • Problem quantitative tests need an oracle

7
Participating Tool Teams
  • Ccia, ATT (U. of Waterloo)
  • Mike Godfrey and Andrew Trevors
  • cppx (U. of Waterloo)
  • Andrew Malton, Ian Bull, Ian Davis
  • Rigi (U. of Victoria)
  • Holger Kienle and Johannes Martin
  • Tk-SEE/SN (U. of Ottawa)
  • Sergei Marchenko and Tim Lethbridge

8
Objectives
  • For tool teams to find out who has the best
    parser
  • For me to find out how to make a benchmark to
    evaluate parsers

9
Preview of Results
  • I found bugs in their tools, they found bugs in
    my test cases.
  • Can your parser do X? I dont know, give me a
    test case.
  • Im learning a lot about how to build a
    benchmark. (I dont have one yet.)
  • There will be a winner
  • In the land of the blind, the one-eyed parser is
    king.
Write a Comment
User Comments (0)
About PowerShow.com