Diff for /gforth/stuff.fs between versions 1.24 and 1.30

version 1.24, 2003/01/22 17:12:49 version 1.30, 2003/02/03 08:59:29
Line 129  AUser CSP Line 129  AUser CSP
     repeat      repeat
     - + dup >r resize throw r> ;      - + dup >r resize throw r> ;
   
   \ ]] ... [[
   
   : compile-literal ( n -- )
       postpone literal ;
   
   : [[ ( -- )
   \G switch from postpone state to compile state
       \ this is only a marker; it is never really interpreted
       compile-only-error ; immediate
   
   : postponer ( c-addr u -- )
       2dup find-name dup if ( c-addr u nt )
           nip nip name>comp
           2dup [comp'] [[ d= if
               2drop ['] compiler is parser
           else
               postpone,
           endif
       else
           drop
           2dup snumber? dup if
               0> IF
                   swap postpone literal postpone compile-literal
               THEN
               postpone Literal postpone compile-literal
               2drop
           ELSE
               drop no.extensions
           THEN
       then ;
   
   : ]] ( -- )
       \ switch into postpone state
       ['] postponer is parser state on ; immediate restrict
   
   \ f.rdp
   
   : push-right ( c-addr u1 u2 cfill -- )
       \ move string at c-addr u1 right by u2 chars (without exceeding
       \ the original bound); fill the gap with cfill
       >r over min dup >r rot dup >r ( u1 u2 c-addr R: cfill u2 c-addr )
       dup 2swap /string cmove>
       r> r> r> fill ;
   
   : f>buf-rdp-try { f: rf c-addr ur nd up um1 -- um2 }
       \ um1 is the mantissa length to try, um2 is the actual mantissa length
       c-addr ur um1 /string '0 fill
       rf c-addr um1 represent if { nexp fsign }
           nd nexp + up >=
           ur nd - 1- dup { beforep } fsign + nexp 0 max >= and if
               \ fixed-point notation
               c-addr ur beforep nexp - dup { befored } '0 push-right
               c-addr beforep 1- befored min dup { beforez } 0 max bl fill
               fsign if
                   '- c-addr beforez 1- 0 max + c!
               endif
               c-addr ur beforep /string 1 '. push-right
               nexp nd +
           else \ exponential notation
               c-addr ur 1 /string 1 '. push-right
               fsign if
                   c-addr ur 1 '- push-right
               endif
               nexp 1- s>d tuck dabs <<# #s rot sign 'E hold #> { explen }
               ur explen - 1- fsign + { mantlen }
               mantlen 0< if \ exponent too large
                   drop c-addr ur '* fill
               else
                   c-addr ur + 0 explen negate /string move
               endif
               #>> mantlen
           endif
       else \ inf or nan
           if \ negative
               c-addr ur 1 '- push-right
           endif
           drop ur
           \ !! align in some way?
       endif
       1 max ur min ;
   
   : f>buf-rdp ( rf c-addr +nr +nd +np -- ) \ gforth
   \G Convert @i{rf} into a string at @i{c-addr nr}.  The conversion
   \G rules and the meanings of @i{nr nd np} are the same as for
   \G @code{f.rdp}.
       \ first, get the mantissa length, then convert for real.  The
       \ mantissa length is wrong in a few cases because of different
       \ rounding; In most cases this does not matter, because the
       \ mantissa is shorter than expected and the final digits are 0;
       \ but in a few cases the mantissa gets longer.  Then it is
       \ conceivable that you will see a result that is rounded too much.
       \ However, I have not been able to construct an example where this
       \ leads to an unexpected result.
       swap 0 max swap 0 max
       fdup 2over 2over 2 pick f>buf-rdp-try f>buf-rdp-try drop ;
   
   : f>str-rdp ( rf +nr +nd +np -- c-addr nr ) \ gforth
   \G Convert @i{rf} into a string at @i{c-addr nr}.  The conversion
   \G rules and the meanings of @i{nr +nd np} are the same as for
   \G @code{f.rdp}.  The result in in the pictured numeric output buffer
   \G and will be destroyed by anything destroying that buffer.
       rot holdptr @ 1- 0 rot negate /string ( rf +nd np c-addr nr )
       over holdbuf u< -&17 and throw
       2tuck 2>r f>buf-rdp 2r> ;
   
   : f.rdp ( rf +nr +nd +np -- ) \ gforth
   \G Print float @i{rf} formatted.  The total width of the output is
   \G @i{nr}.  For fixed-point notation, the number of digits after the
   \G decimal point is @i{+nd} and the minimum number of significant
   \G digits is @i{np}.  @code{Set-precision} has no effect on
   \G @code{f.rdp}.  Fixed-point notation is used if the number of
   \G siginicant digits would be at least @i{np} and if the number of
   \G digits before the decimal point would fit.  If fixed-point notation
   \G is not used, exponential notation is used, and if that does not
   \G fit, asterisks are printed.  We recommend using @i{nr}>=7 to avoid
   \G the risk of numbers not fitting at all.  We recommend
   \G @i{nr}>=@i{np}+5 to avoid cases where @code{f.rdp} switches to
   \G exponential notation because fixed-point notation would have too
   \G few significant digits, yet exponential notation offers fewer
   \G significant digits.  We recommend @i{nr}>=@i{nd}+2, if you want to
   \G have fixed-point notation for some numbers.  We recommend
   \G @i{np}>@i{nr}, if you want to have exponential notation for all
   \G numbers.
       f>str-rdp type ;
   
   0 [if]
   : testx ( rf ur nd up -- )
       '| emit f.rdp ;
   
   : test ( -- )
       -0.123456789123456789e-20
       40 0 ?do
           cr
           fdup 7 3 1 testx
           fdup 7 3 4 testx
           fdup 7 3 0 testx
           fdup 7 7 1 testx
           fdup 7 5 1 testx
           fdup 7 0 2 testx
           fdup 5 2 1 testx
           fdup 4 2 1 testx
           fdup 18 8 5 testx
           '| emit
           10e f*
       loop ;
   [then]

Removed from v.1.24  
changed lines
  Added in v.1.30


FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>