1. Forum
  2. Usenet
  3. COMP.GRAPHICS.APPS.GNUPLO

  • SET FORMAT default

    From =?UTF-8?Q?Hans-Bernhard_Br=c3=b6ker@21:1/5 to All on Thu Oct 10 20:56:07 2019
    Am 10.10.2019 um 20:29 schrieb layzarc@aol.com:
    The manual says the default for SET FORMAT is "% h". Not being a C programmer, can someone give me a brief explanation of what this format does?

    The documentation already does that for us. Right below where you found
    the above information, there's (a link to) the documentation section
    "format specifiers", which see.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From layzarc@aol.com@21:1/5 to All on Thu Oct 10 11:29:05 2019
    The manual says the default for SET FORMAT is "% h". Not being a C programmer, can someone give me a brief explanation of what this format does?

    Al

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From layzarc@aol.com@21:1/5 to All on Thu Oct 10 12:16:40 2019
    I saw that but what causes SET TABLE for example to print only 6 significant digits? I'm guessing because its printing a single precision "float". Is there anyway to get more precision in the TABLE from SPLOT contours?

    Al

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Karl Ratzsch@21:1/5 to All on Thu Oct 10 22:26:24 2019
    Am 10.10.2019 um 21:16 schrieb layzarc@aol.com:
    I saw that but what causes SET TABLE for example to print only 6 significant digits? I'm guessing because its printing a single precision "float". Is there anyway to get more precision in the TABLE from SPLOT contours?

    very simple, e.g.

    set format ".12f"
    set table
    plot <your function>

    "% h" has some automagic that switches between fp and exponential
    notation, which isnt really meant to be used to print out numerical
    data, but for labelling axis tics.

    hth, Karl

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From layzarc@aol.com@21:1/5 to All on Thu Oct 10 14:19:18 2019
    But does that really help if the value being printed comes from a single precision "float"? I guess what I'm asking does Gnuplot use "double" floating point thoughout the SPLOT contour calculation and its printing to the TABLE?

    Al

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Karl Ratzsch@21:1/5 to All on Thu Oct 10 23:57:34 2019
    Am 10.10.2019 um 23:19 schrieb layzarc@aol.com:
    But does that really help if the value being printed comes from a single precision "float"? I guess what I'm asking does Gnuplot use "double" floating point thoughout the SPLOT contour calculation and its printing to the TABLE?

    gnuplot has been using double precison fp math since some time in
    the last century I think.

    If the contour calculation doesnt, that would constitute a bug. Do
    you have evidence for that?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Ethan Merritt@21:1/5 to Karl Ratzsch on Sat Oct 12 00:06:25 2019
    On Thu, 10 Oct 2019 23:57:34 +0200, Karl Ratzsch wrote:

    Am 10.10.2019 um 23:19 schrieb layzarc@aol.com:
    But does that really help if the value being printed comes from a single precision "float"? I guess what I'm asking does Gnuplot use "double" floating point thoughout the SPLOT contour calculation and its printing to the TABLE?

    gnuplot has been using double precison fp math since some time in
    the last century I think.

    If the contour calculation doesnt, that would constitute a bug. Do
    you have evidence for that?

    There is one exception that I know of.
    If a binary matrix is read in, it is stored as (float) rather than (double). There's no good reason for that, but the binary input code is quite tangled
    so fixing it would require some disentangling. It would be a good project
    for someone (hint hint).

    Ethan

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)