Re: [Bug-apl] Complex number display

[Juergen Sauermann]

Hi Fred,

thanks, fixed in SVN 280.

/// Jürgen


On 05/21/2014 03:28 AM, Frederick H. Pitts wrote:

Hello Juergen, Elias,

        I think there are still three flies in the "complex number display"
ointment.  Please see the attached tar file containing a APL test file
and a log file generated from it using GnuAPL svn 279.  Only one of the
4 complex numbers entered in polar form and that have a zero real part
is displayed with a 0 for the real part.  The remaining 3 numbers
display with a real part having a E¯16 scale factor.  For these 3
numbers, the real part magnitude is definitely 10 orders of magnitude
less than the imaginary part magnitude and should display with a 0 for
the real part.

Regards,

Fred

On Tue, 2014-05-20 at 18:09 +0200, Juergen Sauermann wrote:

Hi Fred, Elias,

fixed in SVN 279.

I dared to print the (non-zero) imaginary part even if the standard
tells otherwise.

/// Jürgen


On 05/20/2014 07:35 AM, Frederick H. Pitts wrote:

Elias,

        The part before the "Then:" states the obvious.  Numeric output
conversion converts numbers in internal format (implementation-defined))
to external literal decimal format.

        The part after the "Then:" is totally bogus.  What is the point of
performing complex number calculations and then summarily throwing away
the imaginary part?  The imaginary part could be the part the user is
interested in.

        Sincerely, I do not see anything here that legitimately contradicts the
descriptions cited in the IBM APL2 Language Reference.  The IBM document
says 1) display the real part as 0 if the real part magnitude is ⎕PP
orders of magnitude less than the imaginary part magnitude and 2)
suppress the display imaginary part if its magnitude is ⎕PP orders of
magnitude less than the real part.  The exception is when ⎕PP is at
maximum; then both parts are displayed at full precision.

Regards,

Fred

On Tue, 2014-05-20 at 12:47 +0800, Elias Mårtenson wrote:

ISO/IEC 13751:2000 seems to give an implementation lots of flexibility
here.


I'm honestly not sure how to interpret this. Section 15.2.2 says:


          Informal Description: Numeric output conversion converts
          numeric values represented as numbers—numeric quantities whose
          format is implementation-defined—into the same numeric
          quantities represented in decimal notation as lists of
          characters.


Then:


          Note: Any imaginary-part of the numeric input to
          Numeric-Output-Conversion is simply ignored; only the
          real-part of a complex-number is used.


Regards,
Elias


On 20 May 2014 12:24, Frederick H. Pitts <fred.pi...@comcast.net>
wrote:
          Juergen,

Under "Display of Complex Numbers:" on page 13 of the

          "APL2
          Programming: Language reference", the document says:

"In J notation, the real or imaginary part is not

          displayed if it is
          less than the other by more than ⎕PP orders of magnitude
          (unless ⎕PP is
          at its maximum)".

In SVN 268, neither of the presented examples work.

           The smaller
          magnitude part is not zeroed if its the real part or
          suppressed if its
          the imaginary part.

In addition, complex numbers entered in polar notation

          as described at
          the top page 12 do not display correctly per the description
          on page 13.
          1D90 and 1R1.5707963267948965 should display as 0J1 assuming
          ⎕PP is less
          than maximum.  Currently the real part is displayed as some
          number times
          10 to the -16 power, i.e. zero for all intents and purposes.

Regards, Fred

          Retired Chemical Engineer