Growing Languages with Metamorphic Syntax Macros

1
Growing Languages with Metamorphic Syntax Macros

• Claus Brabrand Michael I. Schwartzbach
• BRICS, Dept. of Computer Science
• University of Aarhus,Denmark

Presented by Guoliang Xing
Center for Distributed Object Computing Department
of Computer Science Washington University
2
Outline

• Related Work Survey
• Designing a Macro Language
• Growing Language Concepts
• Metamorphisms
• Growing New Languages
• Implementation

3
Related Work Survey

• General properties
• Syntax Properties
• Type Properties
• Definition Properties
• Invocation Properties
• Implementation properties

4
Related Work Survey contd.

• CPP - C's preprocessor
• M4 - the Unix macro preprocessor
• TEXs built-in macro mechanism
• The macro mechanism of Dylan
• The C templates
• Hygienic macros of Scheme
• Macro mechanism of the Jakarta Tool Suite, JTS
• Meta Syntactic Macro System, MS2

5
General properties

• Lexical vs. Syntactical
• Language dependence
• Programmable
• Explicit programming on parse trees vs. Pattern
  matching and substitution

6
Syntax Type Properties

• How liberal an invocation syntax could be?
• CPP requires parenthesized and comma separated
  arguments
• Dylan almost arbitrary invocation syntax
• Result types and argument types
• Non-terminals of host language
• Type checking in syntactical macro systems

7
Definition properties

• Multiple definitions
• Definition scope
• One-pass vs. Two-pass
• Recursion
• Direct vs. Indirect
• Argument structure
• Fixed number vs. Variable number

8
Invocation Properties

• Body expansion
• Eager vs. Lazy
• Order of expansion
• Inner vs. Outer vs. Pre-scan
• Parsing ambiguities
• Hygienic expansion
• Alpha Conversion

9
Implementation Properties

• Transparence
• Error trailing
• Pretty printing
• Package

10
(No Transcript)
11
Designing a Macro Language

• Host language ltbigwiggt
• http//www.brics.dk/bigwig/
• Syntax
• macro syntax ltnontermgt id ltparamgt body
• param token
• ltnonterm idgt

12
Simple Examples

• syntax ltfloatconstgt pi
• 3.1415927
• syntax ltstmgt maybe ltstm Sgt
• if (random(2)1) ltSgt
• syntax ltstmgt repeat ltstm Sgt until (ltexp Egt)
  
• bool first true
• while (first !ltEgt)
• ltSgt
• first false

13
Parsing Definitions

• Two phases
• Collecting macro headers
• Parsing macro bodies in topological order

14
Parsing Invocations

• Tasks
• Interpreting the macro parameter list
• Matching required tokens
• Collecting actual argument parse trees
• Multiple definitions
• Interpreting a set of parameter lists
• Challenging each parameter list with the input
  tokens
• Choosing the most specific one among survived
  lists

15
Growing Language Concepts

• Growing new concepts and constructs in host
  language ltbigwiggt
• Concurrency control between session threads

• A stack of macros

16
Metamorphisms

• Two aspects of Macro definitions
• Syntax definition
• Syntax transformation aspect
• Metamorphisms
• User defined non-terminal along with a rule
  specifying how the new macro syntax is morphed
  into host language syntax

17
Metamorphismscontd.

• syntax ltdeclsgt enum ltid Xgt
• const int ltXgt 0
• syntax ltdeclsgt enum ltid Xgt , ltid Ygt
• const int ltXgt 0
• const int ltYgt 1
• syntax ltdeclsgt enum ltid Xgt , ltid Ygt , ltid Zgt
  
• const int ltXgt 0
• const int ltYgt 1
• const int ltZgt 2

18
Metamorphismscontd.

• decls enum id
• enum id , id
• enum id , id , id
• Extending the grammar by notation N
• decls enum id
• Extending the grammar by introducing a
• user-defined non-terminal enums
• decls enum id enums
• enums , id enums
• ?

19
Revised Macro Syntax

• macro syntax ltnontermgt id ltparamgt body
• metamorph ltnontermgt id --gt
  ltparamgt body
• param token
• ltnonterm idgt
• ltid nonterm idgt

20
Example Revised

• syntax ltdeclsgt enum ltid Igt ltenums decls Dsgt
  
• int e 0
• const int ltIgt e
• ltDsgt
• metamorph ltdeclsgt enums --gt , ltid Igt ltenums
  decls Dsgt
• const int ltIgt e
• ltDsgt
• metamorph ltdeclsgt enums --gt

21
Growing New Language

• Designing new language with metamorphism
• Host language ltbigwiggt
• Very domain-specific language (VDSL)
• Customizable with low cost
• No host language grammars are needed in new
  language

22
Example of VDSL--Backgroud

• Bachelor's Contract at computer science
  department of University of Aarhus
• CS students must complete a BS in one of several
  fields Physics, Mathematics etc.
• The requirements are complicated
• Web-based service needed to guide student
• Must be easy to maintain

23
Example of VDSL Impl.

• require "bach.wigmac"
• studies
• course Math101
• title "Mathematics 101"
• 2 points fall term
• ...
• course Lab304
• title "Lab Work 304"
• 1 point fall term
• exclusions
• Math101 ltgt MathA
• Math102 ltgt MathB
• prerequisites
• Math101,Math102 lt Math201,Math202,Math203,Math204
• CS101,CS102 lt CS201,CS203
• CS202,CS203 lt CS301,CS302,CS303,CS304
• Phys101,Phys102 lt Phys201,Phys202,Phys203,Phys301
• Lab101,Lab102 lt Lab201,Lab202
• field "CS-Mathematics"

24
Implementation

• Transparent Representation
• Parse tree of macro could be transparent to
  following compiling phases
• syntax ltidsgt xIDy ( ltids Isgt ) X,ltIsgt,Y

25
Implementationcontd.

• Generic Pretty Printing
• Four indent directives control the pretty
  printing of macros
• param ltwhitespacegt \n \ \-
• Example
• syntax ltstmgt switch (ltexp Egt) \\nltswbody
  stm Sgt\-\n ...
• Error Reporting
• Error messages can be viewed with or without
  macro expansion

26
Conclusion and Future Work

• Contribution A macro language that can grow
  domain-specific extensions of host languages or
  even entire new languages
• Future Work
• Invocation constraints that restrict the possible
  uses of macros.
• Implementations for other host languages (Java
  etc).
• parser generator