help links subscribe changes refresh edit

	IssueTracker
#40 broken input file testing ←	↑	→ #402 binary for fedora core 6

#401 unparse generates Fortran function names

Edit detail for #401 unparse generates Fortran function names revision 5 of 9

	1 2 3 4 5 6 7 8 9
Editor: page Time: 2007/11/12 20:33:57 GMT-8
Note: test

added:
The function::

  parse(s)==interpret(ncParseFromString(s)$Lisp::INFORM)

parses abd evaluates a string 's', returning the result as ANY.
We can use this to test that 'unparse'

added:
parse(s)==interpret(ncParseFromString(s)$Lisp::INFORM)

changed:
-unparse(y::INFORM)
-unparse(sqrt(x2/x1)::INFORM)
ys := unparse(y::INFORM)
test(parse(ys)::Expression Integer = y)
x := sqrt(x2/x1)
test(parse(unparse(x::INFORM))::Expression Integer = x)
unparse(parse("exp(z+1)")::Expression Integer::INFORM)="exp(z+1)"
test %
unparse(parse("1/exp(z+1)")::Expression Integer::INFORM)="1/exp(z+1)"
test %

In an email with subject:

  Axiom: Printing multiple variables in axiom.

to axiom-math on Nov 11, 2007 3:10 PM Constantine Frangos reported that:

I am getting the result below for y.

However, the output I actually want is:

  "cos(x2) + 2*x1"

or even better:

  cos(x2) + 2*x1

Is this possible ?

axiom
y := cos(x2) + 2*x1

(1)

Type: Expression Integer

axiom
unparse(y::INFORM)

(2)

Type: String

I think the problem originats in the routine form2String1 which is called via form2String from unparse. form2String1 calls fortexp0 defined in newfort.boot. This routine calls expression2Fortran which transforms the InputForm? via:

  fortranCleanUp exp2Fort1 segment fortPre exp2FortOptimize outputTran e

fortPre is responsible for renaming functions like cos to DCOS for Fortran double or single float compatibility in a routine called mkFortFn. segment is intended to handle lines longer than allowed in Fortran, and outputTran renames some operations to be subsequently remapped by fortPre to the appropriate operator depending on the precision setting. None of this is necessary if the intention is to "unparse" an Axiom expression.

The patch to format.boot below avoids this problem.

bootform2String1 u ==
  ATOM u => 
    u=$EmptyMode or u=$quadSymbol => formWrapId specialChar 'quad
    IDENTP u =>
      constructor? u => app2StringWrap(formWrapId u, [u])
      u
    SUBRP u => formWrapId BPINAME u
    STRINGP u => formWrapId u
    WRITE_-TO_-STRING formWrapId u
  u1 := u
  op := CAR u
  argl := CDR u
  op='Join or op= 'mkCategory => formJoin1(op,argl)
  $InteractiveMode and (u:= constructor? op) =>
    null argl => app2StringWrap(formWrapId constructorName op, u1)
    op = "NTuple"  => [ form2String1 first argl, "*"]
    op = "Map"     => ["(",:formatSignature0 [argl.1,argl.0],")"]
    op = 'Record => record2String(argl)
    null (conSig := getConstructorSignature op) =>
      application2String(constructorName op,[form2String1(a) for a in argl], u1)
    ml := rest conSig
    if not freeOfSharpVars ml then
      ml:=SUBLIS([[pvar,:val] for pvar in $FormalMapVariableList
        for val in argl], ml)
    argl:= formArguments2String(argl,ml)
      -- extra null check to handle mutable domain hack.
    null argl => constructorName op
    application2String(constructorName op,argl, u1)
  op = "Mapping" => ["(",:formatSignature argl,")"]
  op = "Record" => record2String(argl)
  op = 'Union  =>
    application2String(op,[form2String1 x for x in argl], u1)
  op = ":" =>
      null argl => [ '":" ]
      null rest argl => [ '":", form2String1 first argl ]
      formDecl2String(argl.0,argl.1)
  op = "#" and PAIRP argl and LISTP CAR argl =>
    STRINGIMAGE SIZE CAR argl
  op = 'Join => formJoin2String argl
  op = "ATTRIBUTE" => form2String1 first argl
  op='Zero => 0
  op='One => 1
  op = 'AGGLST => tuple2String argl
  op = 'BRACKET =>
    argl' := form2String1 first argl
    ["[",:(atom argl' => [argl']; argl'),"]"]
  op = "SIGNATURE" =>
     [operation,sig] := argl
     concat(operation,": ",formatSignature sig)
  op = 'COLLECT => formCollect2String argl
  op = 'construct =>
    concat(lbrkSch(),
           tuple2String [form2String1 x for x in argl],rbrkSch())
  op = "SEGMENT" =>
    null argl => '".."
    lo := form2String1 first argl
    argl := rest argl
    (null argl) or null (first argl) => [lo, '".."]
    [lo, '"..", form2String1 first argl]
  --isBinaryInfix op => fortexp0 [op,:argl]
  isBinaryInfix op => fortranCleanUp exp2Fort1 exp2FortOptimize [op,:argl]
  -- COMPILED_-FUNCTION_-P(op) => form2String1 coerceMap2E(u1,NIL)
  application2String(op,[form2String1 x for x in argl], u1)

boot

Value = T
Value = 7136

The function:

  parse(s)==interpret(ncParseFromString(s)$Lisp::INFORM)

parses abd evaluates a string s, returning the result as ANY. We can use this to test that unparse

axiom
parse(s)==interpret(ncParseFromString(s)$Lisp::INFORM)

Type: Void

axiom
y := cos(x2) + 2*x1

(3)

Type: Expression Integer

axiom
ys := unparse(y::INFORM)

(4)

Type: String

axiom
test(parse(ys)::Expression Integer = y)

axiom

Compiling function parse with type String -> Any

(5)

Type: Boolean

axiom
x := sqrt(x2/x1)

(6)

Type: Expression Integer

axiom
test(parse(unparse(x::INFORM))::Expression Integer = x)

(7)

Type: Boolean

axiom
unparse(parse("exp(z+1)")::Expression Integer::INFORM)="exp(z+1)"

(8)

Type: Equation String

axiom
test %

(9)

Type: Boolean

axiom
unparse(parse("1/exp(z+1)")::Expression Integer::INFORM)="1/exp(z+1)"

(10)

Type: Equation String

axiom
test %

(11)

Type: Boolean

same problem occurs in graphics --page, Mon, 12 Nov 2007 18:25:31 -0800 reply

The display of function names in legends generated by Axiom graphics also suffers from Fortranish names like DSIN etc. The patch to format.boot above also fixes that.