Printing floating point numbers with exponents.

[wilbert]

I couldn't get your last asm version SciStrD() to run - Operand size error?

I made a few improvements (see above for updated code).
Maybe it solves the problem. I'm not able to replicate it.
I tried both x86 and x64 and it works fine here.

[skywalk]

Sorry, it ran once with debugger on, but not without.
My only compiler option is #PB_Compiler_unicode.
PB461b1 x86.

[wilbert]

Sorry, it ran once with debugger on, but not without.

Strange that there's a difference.
I made a small change and think it should be working now.

[skywalk]

Cool, runs now and is 2% faster overall.
But, weird bug only if 1st call to sciStrD(0,6,1) contains a 0?

Code: Select all

; Data.s                   StrDe(S,6)     StrDsci(S,6)    sciStrD(S,6)    
Data.s "0"                ;[          0]  [ 0.000000e+0]  [ 0.000000e310] ;<-- This 'e310' goes away if called later?
Data.s "0.0"              ;[          0]  [ 0.000000e+0]  [ 0.000000e0]   
Data.s "+0e0"             ;[          0]  [ 0.000000e+0]  [ 0.000000e0]   
Data.s "+0.0e0"           ;[          0]  [ 0.000000e+0]  [ 0.000000e0]   
Data.s "+0.0e+0"          ;[          0]  [ 0.000000e+0]  [ 0.000000e0]   
Data.s "1"                ;[ 1.00000e+0]  [ 10.000000e-1] [ 1.000000e+0]  
Data.s "1.1"              ;[ 1.10000e+0]  [ 1.100000e+0]  [ 1.100000e+0]  
Data.s "0.1"              ;[ 100.000e-3]  [ 10.000000e-2] [ 1.000000e-1]  
Data.s "11"               ;[ 11.0000e+0]  [ 1.100000e+1]  [ 1.100000e+1]  
Data.s "111"              ;[ 111.000e+0]  [ 1.110000e+2]  [ 1.110000e+2]  
Data.s "10"               ;[ 10.0000e+0]  [ 1.000000e+1]  [ 1.000000e+1]  
Data.s "100"              ;[ 100.000e+0]  [ 1.000000e+2]  [ 1.000000e+2]  
Data.s "1000"             ;[ 1.00000e+3]  [ 1.000000e+3]  [ 1.000000e+3]  
Data.s "1000000"          ;[ 1.00000e+6]  [ 1.000000e+6]  [ 1.000000e+6]  
Data.s "999"              ;[ 999.000e+0]  [ 9.990000e+2]  [ 9.990000e+2]  
Data.s "-999"             ;[-999.000e+0]  [-9.990000e+2]  [-9.990000e+2]  
Data.s "-3.5e-20"         ;[-35.0000e-21] [-3.500000e-20] [-3.500000e-20] 

[wilbert]

Cool, runs now and is 2% faster overall.
But, weird bug only if 1st call to sciStrD(0,6,1) contains a 0?

The bug should be fixed now.

For your engineering notation, maybe a real fixed length would be nice.
If support would be limited from 1e-99 to 1e+99, you could make the exponent always two digits.
Like 1.23456e-06

[Little John]

Cool, runs now and is 2% faster overall.

And a maximum of 2% of the forum members are able to understand that code.

Regards, LJ

[wilbert]

And a maximum of 2% of the forum members are able to understand that code.

To me it's like a puzzle. I learned some new things creating the code and that's the fun for me.
Most of the times ASM isn't required on modern computers since they are fast enough.

[skywalk]

For your engineering notation, maybe a real fixed length would be nice.
If support would be limited from 1e-99 to 1e+99, you could make the exponent always two digits.
Like 1.23456e-06

Yeah, I thought about forcing 2 exponent digits, but my data did not exceed "nano" or "giga" so e±9 is fine for me. Using "±xxx.yyyye±z" format, any number I print as a string can be retrieved with consistent numeric accuracy.

And a maximum of 2% of the forum members are able to understand that code.

[akj]

@wilbert:

Thank you for the fix which seems to have solved the problem.
(I don't have any better solution of my own.)

[Psychophanta]

This is the function i use since long time for this issue:

Code: Select all

Procedure$ StrE(fp.d,dec.b=14)
  ;Devuelve cadena alfanumérica en formato de notación científica
  Protected E.b,neg.b=0
  If dec<0:dec=0:ElseIf dec>20:dec=20:EndIf;<- Sanity check
  E=Round(Log10(Abs(fp)),#PB_Round_Down)
  If E
    fp*Pow(10,-E)
    ProcedureReturn RTrim(RTrim(StrD(fp,dec),"0"),".")+"E"+Str(E)
  EndIf
  ProcedureReturn RTrim(RTrim(StrD(fp,dec),"0"),".")
EndProcedure