--- gforth/float.fs 1994/05/31 07:26:23 1.3 +++ gforth/float.fs 1994/06/01 10:02:22 1.4 @@ -4,19 +4,18 @@ [ 1 floats 1- ] Literal + [ -1 floats ] Literal and ; : falign ( -- ) - here dup aligned swap + here dup faligned swap ?DO bl c, LOOP ; : f, ( f -- ) here 1 floats allot f! ; \ !! have create produce faligned pfas : fconstant ( r -- ) - Create falign f, - DOES> faligned f@ ; + falign here f, Create A, + DOES> @ f@ ; : fvariable - Create falign 0 f, - DOES> faligned ; + falign here 0. d>f f, AConstant ; : fdepth ( -- n ) f0 @ fp@ - [ 1 floats ] Literal / ; @@ -34,30 +33,31 @@ : -zeros ( addr u -- addr' u' ) BEGIN dup WHILE 1- 2dup + c@ '0 <> UNTIL 1+ THEN ; -: f. ( r -- ) scratch represent 0= - IF 2drop scratch 3 min type EXIT THEN - IF '- emit THEN dup >r 0<= - IF '0 emit +: f\$ ( f -- n ) scratch represent 0= + IF 2drop scratch 3 min type rdrop EXIT THEN + IF '- emit THEN ; + +: f. ( r -- ) f\$ dup >r 0< + IF '0 emit ELSE scratch r@ min type r@ precision - zeros THEN '. emit r@ negate zeros scratch r> 0 max /string 0 max -zeros type space ; \ I'm afraid this does not really implement ansi semantics wrt precision. \ Shouldn't precision indicate the number of places shown after the point? -: fe. ( r -- ) scratch represent 0= - IF 2drop scratch 3 min type EXIT THEN - IF '- emit THEN 1- s>d 3 fm/mod 3 * >r 1+ >r +: fe. ( r -- ) f\$ 1- s>d 3 fm/mod 3 * >r 1+ >r scratch r@ min type '. emit scratch r> /string type 'E emit r> . ; -: fs. ( r -- ) scratch represent 0= - IF 2drop scratch 3 min type EXIT THEN - IF '- emit THEN 1- >r - scratch 1 min type '. emit scratch 1 /string type - 'E emit r> . ; - -: fnumber ( string -- r / ) dup count >float 0= - IF defers notfound ELSE drop THEN ; +: fs. ( r -- ) f\$ 1- + scratch over c@ emit '. emit 1 /string type + 'E emit . ; + +: fnumber ( string -- r / ) + ?dup IF dup count >float 0= + IF defers notfound + ELSE drop state @ + IF postpone FLiteral THEN THEN THEN ; ' fnumber IS notfound