Sophie

Sophie

distrib > Mandriva > 2010.2 > i586 > media > contrib-release > by-pkgid > 6009aaa8e5ab2df861ebfa6faf6af1ce > files > 16

python-parsing-1.5.2-2mdv2010.1.noarch.rpm

"""
Purpose:   Linear Algebra Parser
Based on:  SimpleCalc.py example (author Paul McGuire) in pyparsing-1.3.3
Author:    Mike Ellis
Copyright: Ellis & Grant, Inc. 2005
License:   You may freely use, modify, and distribute this software.
Warranty:  THIS SOFTWARE HAS NO WARRANTY WHATSOEVER. USE AT YOUR OWN RISK. 
Notes: Parses infix linear algebra (LA) notation for vectors, matrices, and scalars.
       Output is C code function calls.  The parser can be run as an interactive 
       interpreter or included as module to use for in-place substitution into C files 
       containing LA equations.  

       Supported operations are:
       OPERATION:              INPUT                    OUTPUT
       Scalar addition:        "a = b+c"                "a=(b+c)"
       Scalar subtraction:     "a = b-c"                "a=(b-c)"
       Scalar multiplication:  "a = b*c"                "a=b*c"
       Scalar division:        "a = b/c"                "a=b/c"
       Scalar exponentiation:  "a = b^c"                "a=pow(b,c)"
       Vector scaling:         "V3_a = V3_b * c"        "vCopy(a,vScale(b,c))"
       Vector addition:        "V3_a = V3_b + V3_c"     "vCopy(a,vAdd(b,c))"
       Vector subtraction:     "V3_a = V3_b - V3_c"     "vCopy(a,vSubtract(b,c))"
       Vector dot product:     "a = V3_b * V3_c"        "a=vDot(b,c)"
       Vector outer product:   "M3_a = V3_b @ V3_c"     "a=vOuterProduct(b,c)"
       Vector magn. squared:   "a = V3_b^Mag2"          "a=vMagnitude2(b)"
       Vector magnitude:       "a = V3_b^Mag"           "a=sqrt(vMagnitude2(b))"
       Matrix scaling:         "M3_a = M3_b * c"        "mCopy(a,mScale(b,c))"
       Matrix addition:        "M3_a = M3_b + M3_c"     "mCopy(a,mAdd(b,c))"
       Matrix subtraction:     "M3_a = M3_b - M3_c"     "mCopy(a,mSubtract(b,c))"
       Matrix multiplication:  "M3_a = M3_b * M3_c"     "mCopy(a,mMultiply(b,c))"
       Matrix by vector mult.: "V3_a = M3_b * V3_c"     "vCopy(a,mvMultiply(b,c))"
       Matrix inversion:       "M3_a = M3_b^-1"         "mCopy(a,mInverse(b))"
       Matrix transpose:       "M3_a = M3_b^T"          "mCopy(a,mTranspose(b))"
       Matrix determinant:     "a = M3_b^Det"           "a=mDeterminant(b)"

       The parser requires the expression to be an equation.  Each non-scalar variable
       must be prefixed with a type tag, 'M3_' for 3x3 matrices and 'V3_' for 3-vectors. 
       For proper compilation of the C code, the variables need to be declared without 
       the prefix as float[3] for vectors and float[3][3] for matrices. The operations do
       not modify any variables on the right-hand side of the equation.

       Equations may include nested expressions within parentheses. The allowed binary 
       operators are '+-*/^' for scalars, and '+-*^@' for vectors and matrices with the
       meanings defined in the table above.

       Specifying an improper combination of operands, e.g. adding a vector to a matrix,
       is detected by the parser and results in a Python TypeError Exception. The usual cause
       of this is omitting one or more tag prefixes. The parser knows nothing about a 
       a variable's C declaration and relies entirely on the type tags. Errors in C 
       declarations are not caught until compile time.

Usage: To process LA equations embedded in source files, import this module and 
       pass input and output file objects to the fprocess() function.  You can
       can also invoke the parser from the command line, e.g. 'python LAparser.py',
       to run a small test suite and enter an interactive loop where you can enter
       LA equations and see the resulting C code.
       
"""

import re,os,sys
from pyparsing import Word, alphas, ParseException, Literal, CaselessLiteral \
, Combine, Optional, nums, Or, Forward, OneOrMore, ZeroOrMore, \
  FollowedBy, StringStart, StringEnd, alphanums
import math

# Debugging flag can be set to either "debug_flag=True" or "debug_flag=False"
debug_flag=False

#----------------------------------------------------------------------------
# Variables that hold intermediate parsing results and a couple of
# helper functions.
exprStack = []      # Holds operators and operands parsed from input.
targetvar = None    # Holds variable name to left of '=' sign in LA equation.


def _pushFirst( str, loc, toks ):
    if debug_flag: print "pushing ", toks[0], "str is ", str
    exprStack.append( toks[0] )

def _assignVar( str, loc, toks ):
    global targetvar
    targetvar =  toks[0]

#-----------------------------------------------------------------------------
# The following statements define the grammar for the parser.

point = Literal('.')
e = CaselessLiteral('E')
plusorminus = Literal('+') | Literal('-')
number = Word(nums) 
integer = Combine( Optional(plusorminus) + number )
floatnumber = Combine( integer +
                       Optional( point + Optional(number) ) +
                       Optional( e + integer )
                     )

lbracket = Literal("[")
rbracket = Literal("]")
ident = Forward()
## The definition below treats array accesses as identifiers. This means your expressions
## can include references to array elements, rows and columns, e.g., a = b[i] + 5.
## Expressions within []'s are not presently supported, so a = b[i+1] will raise
## a ParseException.
ident = Combine(Word(alphas + '-',alphanums + '_') + \
                ZeroOrMore(lbracket + (Word(alphas + '-',alphanums + '_')|integer) + rbracket) \
                ) 

plus  = Literal( "+" )
minus = Literal( "-" )
mult  = Literal( "*" )
div   = Literal( "/" )
outer = Literal( "@" )
lpar  = Literal( "(" ).suppress()
rpar  = Literal( ")" ).suppress()
addop  = plus | minus
multop = mult | div | outer
expop = Literal( "^" )
assignop = Literal( "=" )

expr = Forward()
atom = ( ( e | floatnumber | integer | ident  ).setParseAction(_pushFirst) | 
         ( lpar + expr.suppress() + rpar )
       )
factor = Forward()
factor << atom + ZeroOrMore( ( expop + factor ).setParseAction( _pushFirst ) )
        
term = factor + ZeroOrMore( ( multop + factor ).setParseAction( _pushFirst ) )
expr << term + ZeroOrMore( ( addop + term ).setParseAction( _pushFirst ) ) 
equation = (ident + assignop).setParseAction(_assignVar) + expr + StringEnd()

# End of grammar definition
#-----------------------------------------------------------------------------
## The following are helper variables and functions used by the Binary Infix Operator
## Functions described below.

vprefix = 'V3_'
vplen = len(vprefix)
mprefix = 'M3_'
mplen = len(mprefix)

## We don't support unary negation for vectors and matrices
class UnaryUnsupportedError(Exception): pass

def _isvec(ident):
   if ident[0] == '-' and ident[1:vplen+1] == vprefix:
      raise UnaryUnsupportedError 
   else: return ident[0:vplen] == vprefix

def _ismat(ident): 
   if ident[0] == '-' and ident[1:mplen+1] == mprefix:
      raise UnaryUnsupportedError 
   else: return ident[0:mplen] == mprefix

def _isscalar(ident): return not (_isvec(ident) or _ismat(ident))

## Binary infix operator (BIO) functions.  These are called when the stack evaluator
## pops a binary operator like '+' or '*".  The stack evaluator pops the two operand, a and b,
## and calls the function that is mapped to the operator with a and b as arguments.  Thus,
## 'x + y' yields a call to addfunc(x,y). Each of the BIO functions checks the prefixes of its
## arguments to determine whether the operand is scalar, vector, or matrix.  This information
## is used to generate appropriate C code.  For scalars, this is essentially the input string, e.g.
## 'a + b*5' as input yields 'a + b*5' as output.  For vectors and matrices, the input is translated to
## nested function calls, e.g. "V3_a + V3_b*5"  yields "V3_vAdd(a,vScale(b,5)".  Note that prefixes are
## stripped from operands and function names within the argument list to the outer function and
## the appropriate prefix is placed on the outer function for removal later as the stack evaluation
## recurses toward the final assignment statement.  

def _addfunc(a,b): 
   if _isscalar(a) and _isscalar(b): return "(%s+%s)"%(a,b)
   if _isvec(a) and _isvec(b): return "%svAdd(%s,%s)"%(vprefix,a[vplen:],b[vplen:]) 
   if _ismat(a) and _ismat(b): return "%smAdd(%s,%s)"%(mprefix,a[mplen:],b[mplen:]) 
   else: raise TypeError

def _subfunc(a,b):  
   if _isscalar(a) and _isscalar(b): return "(%s-%s)"%(a,b)
   if _isvec(a) and _isvec(b): return "%svSubtract(%s,%s)"%(vprefix,a[vplen:],b[vplen:]) 
   if _ismat(a) and _ismat(b): return "%smSubtract(%s,%s)"%(mprefix,a[mplen:],b[mplen:]) 
   else: raise TypeError

def _mulfunc(a,b):  
   if _isscalar(a) and _isscalar(b): return "%s*%s"%(a,b)
   if _isvec(a) and _isvec(b):    return "vDot(%s,%s)"%(a[vplen:],b[vplen:]) 
   if _ismat(a) and _ismat(b):    return "%smMultiply(%s,%s)"%(mprefix,a[mplen:],b[mplen:]) 
   if _ismat(a) and _isvec(b):    return "%smvMultiply(%s,%s)"%(vprefix,a[mplen:],b[vplen:]) 
   if _ismat(a) and _isscalar(b): return "%smScale(%s,%s)"%(mprefix,a[mplen:],b) 
   if _isvec(a) and _isscalar(b): return "%svScale(%s,%s)"%(vprefix,a[mplen:],b) 
   else: raise TypeError

def _outermulfunc(a,b):
   ## The '@' operator is used for the vector outer product.
   if _isvec(a) and _isvec(b): 
     return "%svOuterProduct(%s,%s)"%(mprefix,a[vplen:],b[vplen:])
   else: raise TypeError

def _divfunc(a,b):
   ## The '/' operator is used only for scalar division  
   if _isscalar(a) and _isscalar(b): return "%s/%s"%(a,b)
   else: raise TypeError

def _expfunc(a,b):  
  ## The '^' operator is used for exponentiation on scalars and
  ## as a marker for unary operations on vectors and matrices.
  if _isscalar(a) and _isscalar(b): return "pow(%s,%s)"%(str(a),str(b))
  if _ismat(a) and b=='-1':         return "%smInverse(%s)"%(mprefix,a[mplen:])
  if _ismat(a) and b=='T':          return "%smTranspose(%s)"%(mprefix,a[mplen:])
  if _ismat(a) and b=='Det':        return "mDeterminant(%s)"%(a[mplen:])
  if _isvec(a) and b=='Mag':        return "sqrt(vMagnitude2(%s))"%(a[vplen:])
  if _isvec(a) and b=='Mag2':       return "vMagnitude2(%s)"%(a[vplen:])
  else: raise TypeError

def _assignfunc(a,b): 
   ## The '=' operator is used for assignment
   if _isscalar(a) and _isscalar(b): return "%s=%s"%(a,b)
   if _isvec(a) and _isvec(b): return "vCopy(%s,%s)"%(a[vplen:],b[vplen:]) 
   if _ismat(a) and _ismat(b): return "mCopy(%s,%s)"%(a[mplen:],b[mplen:]) 
   else: raise TypeError

## End of BIO func definitions
##----------------------------------------------------------------------------

# Map  operator symbols to corresponding BIO funcs 
opn = { "+" : ( _addfunc ),
        "-" : ( _subfunc ),
        "*" : ( _mulfunc ),
        "@" : ( _outermulfunc ),
        "/" : ( _divfunc),
        "^" : ( _expfunc ), }


##----------------------------------------------------------------------------
# Recursive function that evaluates the expression stack
def _evaluateStack( s ):
  op = s.pop()
  if op in "+-*/@^":
    op2 = _evaluateStack( s )
    op1 = _evaluateStack( s )
    result = opn[op]( op1, op2 )
    if debug_flag: print result
    return result
  else: 
    return op

##----------------------------------------------------------------------------
# The parse function that invokes all of the above.
def parse(input_string):
    """
    Accepts an input string containing an LA equation, e.g., 
    "M3_mymatrix = M3_anothermatrix^-1" returns C code function
    calls that implement the expression.
    """
    
    global  exprStack
    global targetvar

    # Start with a blank exprStack and a blank targetvar
    exprStack = []
    targetvar=None
  
    if input_string != '':
      # try parsing the input string
      try:
        L=equation.parseString( input_string )
      except ParseException,err:
        print >>sys.stderr, 'Parse Failure'
        print >>sys.stderr, err.line
        print >>sys.stderr, " "*(err.column-1) + "^"
        print >>sys.stderr, err
        raise
      
      # show result of parsing the input string
      if debug_flag: 
        print input_string, "->", L
        print "exprStack=", exprStack
  
      # Evaluate the stack of parsed operands, emitting C code.
      try: 
        result=_evaluateStack(exprStack)
      except TypeError:
        print >>sys.stderr,"Unsupported operation on right side of '%s'.\nCheck for missing or incorrect tags on non-scalar operands."%input_string
        raise
      except UnaryUnsupportedError:
        print >>sys.stderr,"Unary negation is not supported for vectors and matrices: '%s'"%input_string
        raise
  
      # Create final assignment and print it. 
      if debug_flag: print "var=",targetvar
      if targetvar != None:
          try:
            result = _assignfunc(targetvar,result)
          except TypeError:
            print >>sys.stderr,"Left side tag does not match right side of '%s'"%input_string
            raise
          except UnaryUnsupportedError:
            print >>sys.stderr,"Unary negation is not supported for vectors and matrices: '%s'"%input_string
            raise

          return  result
      else:
        print >>sys.stderr, "Empty left side in '%s'"%input_string
        raise TypeError

##-----------------------------------------------------------------------------------
def fprocess(infilep,outfilep):
   """
   Scans an input file for LA equations between double square brackets,
   e.g. [[ M3_mymatrix = M3_anothermatrix^-1 ]], and replaces the expression
   with a comment containing the equation followed by nested function calls
   that implement the equation as C code. A trailing semi-colon is appended.
   The equation within [[ ]] should NOT end with a semicolon as that will raise
   a ParseException. However, it is ok to have a semicolon after the right brackets.

   Other text in the file is unaltered.

   The arguments are file objects (NOT file names) opened for reading and 
   writing, respectively.
   """
   pattern = r'\[\[\s*(.*?)\s*\]\]'
   eqn = re.compile(pattern,re.DOTALL)
   s = infilep.read()
   def parser(mo): 
      ccode = parse(mo.group(1))
      return "/* %s */\n%s;\nLAParserBufferReset();\n"%(mo.group(1),ccode)

   content = eqn.sub(parser,s)
   outfilep.write(content)
 
##-----------------------------------------------------------------------------------
def test():
   """ 
   Tests the parsing of various supported expressions. Raises
   an AssertError if the output is not what is expected. Prints the
   input, expected output, and actual output for all tests.
   """
   print "Testing LAParser"
   testcases = [
     ("Scalar addition","a = b+c","a=(b+c)"),
     ("Vector addition","V3_a = V3_b + V3_c","vCopy(a,vAdd(b,c))"),
     ("Vector addition","V3_a=V3_b+V3_c","vCopy(a,vAdd(b,c))"),
     ("Matrix addition","M3_a = M3_b + M3_c","mCopy(a,mAdd(b,c))"),
     ("Matrix addition","M3_a=M3_b+M3_c","mCopy(a,mAdd(b,c))"),
     ("Scalar subtraction","a = b-c","a=(b-c)"),
     ("Vector subtraction","V3_a = V3_b - V3_c","vCopy(a,vSubtract(b,c))"),
     ("Matrix subtraction","M3_a = M3_b - M3_c","mCopy(a,mSubtract(b,c))"),
     ("Scalar multiplication","a = b*c","a=b*c"),
     ("Scalar division","a = b/c","a=b/c"),
     ("Vector multiplication (dot product)","a = V3_b * V3_c","a=vDot(b,c)"),
     ("Vector multiplication (outer product)","M3_a = V3_b @ V3_c","mCopy(a,vOuterProduct(b,c))"),
     ("Matrix multiplication","M3_a = M3_b * M3_c","mCopy(a,mMultiply(b,c))"),
     ("Vector scaling","V3_a = V3_b * c","vCopy(a,vScale(b,c))"),
     ("Matrix scaling","M3_a = M3_b * c","mCopy(a,mScale(b,c))"),
     ("Matrix by vector multiplication","V3_a = M3_b * V3_c","vCopy(a,mvMultiply(b,c))"),
     ("Scalar exponentiation","a = b^c","a=pow(b,c)"),
     ("Matrix inversion","M3_a = M3_b^-1","mCopy(a,mInverse(b))"),
     ("Matrix transpose","M3_a = M3_b^T","mCopy(a,mTranspose(b))"),
     ("Matrix determinant","a = M3_b^Det","a=mDeterminant(b)"),
     ("Vector magnitude squared","a = V3_b^Mag2","a=vMagnitude2(b)"),
     ("Vector magnitude","a = V3_b^Mag","a=sqrt(vMagnitude2(b))"),
     ("Complicated expression", "myscalar = (M3_amatrix * V3_bvector)^Mag + 5*(-xyz[i] + 2.03^2)","myscalar=(sqrt(vMagnitude2(mvMultiply(amatrix,bvector)))+5*(-xyz[i]+pow(2.03,2)))"),
     ("Complicated Multiline", "myscalar = \n(M3_amatrix * V3_bvector)^Mag +\n 5*(xyz + 2.03^2)","myscalar=(sqrt(vMagnitude2(mvMultiply(amatrix,bvector)))+5*(xyz+pow(2.03,2)))")
      
     ]
   
   for t in testcases:
      name,input,expected = t
      print name
      print "   %s input"%input
      print "   %s expected"%expected
      result = parse(input)
      print "   %s received"%result
      print ""
      assert expected == result
    
   ##TODO: Write testcases with invalid expressions and test that the expected
   ## exceptions are raised.

   print "Tests completed!"
##----------------------------------------------------------------------------
## The following is executed only when this module is executed as 
## command line script.  It runs a small test suite (see above)
## and then enters an interactive loop where you
## can enter expressions and see the resulting C code as output.

if __name__ == '__main__':
  # run testcases
  test()

  # input_string
  input_string=''
  
  # Display instructions on how to use the program interactively
  interactiveusage = """
  Entering interactive mode:
  Type in an equation to be parsed or 'quit' to exit the program.
  Type 'debug on' to print parsing details as each string is processed.
  Type 'debug off' to stop printing parsing details
  """
  print interactiveusage
  input_string = raw_input("> ")
  
  while input_string != 'quit':
    if input_string == "debug on":
       debug_flag = True
    elif input_string == "debug off":
       debug_flag = False
    else:
      try:
        print parse(input_string)
      except:
        pass 

    # obtain new input string
    input_string = raw_input("> ")
  
  # if user types 'quit' then say goodbye
  print "Good bye!"