On 2018-11-20 11:15-0700 Steve Schwartz wrote:
Hi Plplot
We have been pretty quiet, happily employing plplot in our QSAS software. We
recently decided to update our plplot version, which we don’t do regurarly
since it requires a few tweaks to work with/embed in our system.
We (meaning my colleague and software person Tony Allen – in cc) have
discovered a bug in the handling of exponential label placement. While we
haven’t prepared a minimal example, the exchange below describes the symptoms.
The attached plbox.c includes Tony’s correction.
Hi Steve and Tony:
@Steve:
Thanks for passing along Tony's fix for a bug in PLplot that he noticed.
@Tony:
Most of this is addressed to you. But before you respond please take
a few minutes to subscribe to plplot-devel so everybody subscribed to
that list (not just those you directly e-mail) and those looking at
the list archive later can see what you say. (See
<https://sourceforge.net/projects/plplot/lists/plplot-devel> for the
simple subscription directions and remember to always use your
subscription address as the return address for your posts to this list
or otherwise your message will be lost to everybody other than the
list manager [who happens to be me].)
I took your version of plbox.c (as forwarded by Steve), changed from
Windows to Unix line endings (git assumes native line endings, e.g.,
Windows on Windows and Unix on Linux), commented out some debugging
print statements, changed one comment reference from "qsas" to "QSAS"
to be consistent with the same comment in another routine, and styled
that result. So after those further changes here are the differences
with the PLplot HEAD version (from the "git diff --ignore-all-space
--unified=10" command with that last-parameter used to give lots of
context).
--------------------------------------
diff --git a/src/plbox.c b/src/plbox.c
index b2b25840f..01c5aee7b 100644
--- a/src/plbox.c
+++ b/src/plbox.c
@@ -1470,21 +1470,21 @@ label_box( PLCHAR_VECTOR xopt, PLFLT xtick1,
PLCHAR_VECTOR yopt, PLFLT ytick1 )
ldy = plP_stsearch( yopt, 'd' );
lfy = plP_stsearch( yopt, 'f' );
liy = plP_stsearch( yopt, 'i' );
lly = plP_stsearch( yopt, 'l' );
lmy = plP_stsearch( yopt, 'm' );
lny = plP_stsearch( yopt, 'n' );
lty = plP_stsearch( yopt, 't' );
lvy = plP_stsearch( yopt, 'v' );
loy = plP_stsearch( yopt, 'o' );
lxy = plP_stsearch( yopt, 'x' );
-
+ //printf( "=== yopt %d %d %d \n", lny, lty, lvy );
plP_xgvpw( &vpwxmin, &vpwxmax, &vpwymin, &vpwymax );
// vpwxmi always numerically less than vpwxma, and
// similarly for vpwymi
vpwxmi = ( vpwxmax > vpwxmin ) ? vpwxmin : vpwxmax;
vpwxma = ( vpwxmax > vpwxmin ) ? vpwxmax : vpwxmin;
vpwymi = ( vpwymax > vpwymin ) ? vpwymin : vpwymax;
vpwyma = ( vpwymax > vpwymin ) ? vpwymax : vpwymin;
// Write label(s) for horizontal axes.
if ( ( lmx || lnx ) && ( ltx || lxx ) )
@@ -1819,21 +1819,24 @@ label_box( PLCHAR_VECTOR xopt, PLFLT xtick1,
PLCHAR_VECTOR yopt, PLFLT ytick1 )
// Write separate exponential label if mode = 1.
if ( !lly && !ldy && !loy && ymode )
{
snprintf( string, STRING_LEN, "(x10%su%d%sd)", esc_string, (int)
yscale, esc_string );
if ( custom_exponent_placement )
{
height = ( (PLLabelDefaults *) plsc->label_data
)->exp_label_disp;
pos = ( (PLLabelDefaults *) plsc->label_data
)->exp_label_pos;
just = ( (PLLabelDefaults *) plsc->label_data
)->exp_label_just;
+ plmtex( "t", height, pos, just, string ); // added for QSAS
}
+ else
+ {
if ( lvy )
{
offset = 0.1; // more space to clear labels in "v" mode
}
else
{
offset = 0.02;
}
// Left axis exponent
if ( lny )
@@ -1907,20 +1910,21 @@ label_box( PLCHAR_VECTOR xopt, PLFLT xtick1,
PLCHAR_VECTOR yopt, PLFLT ytick1 )
}
else
{
plmtex( "r", height, pos, just, string );
}
}
}
}
}
}
+}
//--------------------------------------------------------------------------
// void label_box_custom()
//
// Writes numeric labels on side(s) of box in custom locations
//--------------------------------------------------------------------------
void
label_box_custom( PLCHAR_VECTOR xopt, PLINT n_xticks, PLFLT_VECTOR xticks,
PLCHAR_VECTOR yopt, PLINT n_yticks, PLFLT_VECTOR yticks )
{
@@ -1931,21 +1935,21 @@ label_box_custom( PLCHAR_VECTOR xopt, PLINT n_xticks,
PLFLT_VECTOR xticks, PLCHA
PLFLT vpwxmin, vpwxmax, vpwymin, vpwymax;
PLFLT tn, offset;
PLCHAR_VECTOR timefmt;
PLINT i;
PLINT xdigmax, xdigits, xdigmax_old, xdigits_old;
PLINT ydigmax, ydigits, ydigmax_old, ydigits_old;
PLINT lxmin, lxmax, lymin, lymax;
PLINT pxmin, pxmax, pymin, pymax;
PLFLT default_mm, char_height_mm, height_mm;
PLFLT string_length_mm = 0.0, pos_mm = 0.0;
-
+ //printf( "=== custom\n" );
// Assume label data is for placement of exponents if no custom
// label function is provided.
PLBOOL custom_exponent_placement = !plsc->label_func && plsc->label_data;
// pos, height, and just are unnecessarily set to quiet
// -O3 -Wuninitialized warnings that are obvious false alarms from
// the clarity of the code associated with the true or false
// result for custom_exponent_placement.
PLFLT pos = 0.0, height = 0.0, just = 0.0;
PLCHAR_VECTOR esc_string = plgesc_string();
@@ -2384,20 +2388,21 @@ label_box_custom( PLCHAR_VECTOR xopt, PLINT n_xticks,
PLFLT_VECTOR xticks, PLCHA
// Write separate exponential label if mode = 1.
if ( !lly && !ldy && !loy && ymode )
{
snprintf( string, STRING_LEN, "(x10%su%d%sd)", esc_string, (int)
yscale, esc_string );
if ( custom_exponent_placement )
{
height = ( (PLLabelDefaults *) plsc->label_data
)->exp_label_disp;
pos = ( (PLLabelDefaults *) plsc->label_data
)->exp_label_pos;
just = ( (PLLabelDefaults *) plsc->label_data
)->exp_label_just;
+ plmtex( "t", height, pos, just, string ); //added for QSAS
}
if ( lvy )
{
offset = 0.1; // more space to clear labels in "v" mode
}
else
{
offset = 0.02;
}
// Left axis exponent.
--------------------------------------
For the record, label_box_custom is a private function that is only
called by the static function draw_box in src/pllegend.c, which in
turn is only called by c_plcolorbar within that source file. So
label_box_custom is a misnomer, and I plan (after the current release
is done) to address that issue by changing that name to
label_box_colorbar. But from now on in this e-mail I will refer to
that function with its present name of label_box_custom.
(@ Everybody: Note that Tony's fix is limited to changes when
custom_exponent_placement
is true, i.e., when there is no custom label function but (custom?) label data
are still provided to change the placement of the exponential label.)
@Tony:
We really do need a simple test case for both the revised label_box and
label_box_custom
functions. Could you provide that? I suggest that simple example doesn't need
any
data to plot, but it does need to contain a labelled box and a simple colorbar
(see example 16 for an example of a simple colorbar) to exercise both label_box
and
label_box_custom. And, of course, both the box and colorbar should have
exponential
labels and appropriate label data should be provided in each case to exercise
the cases for both functions where custom_exponent_placement is true.
It has been a long time since I looked at plbox.c, but my impression
now is label_box_custom is quite similar to label_box code so I am
pretty sure label_box_custom badly needs documentation about exactly
how and why it is changed from label_box for the special needs of
plcolorbar. So providing such documentation is on my post-release
agenda, but in the absence of such documentation I am frankly
currently ignorant of the designed differences between label_box and
label_box_custom. However, with that caveat I don't understand why
your fix is so different in the two separate functions.
For example, prior to your fix label_box and label_box_custom had very
similar custom_exponent_placement logic, but now they are quite different
(e.g., the above added "else" in label_box that skips a huge amount
of code is not present in label_box_custom). Did you really mean
to have such quite different logic in those two cases? Even if
the present fix works, it would be best for the logic in those two
functions to be implemented as similarly as possible.
Assuming your reply to my questions includes another revised version
of plbox.c, I don't care about the line endings and whitespace styling
changes I made since those are easy to do again. However, please do
incorporate my further changes (commenting out of the print functions
and the qsas ==> QSAS change) in your revised code.
N.B. I am happy to take a quick break (as now) from getting out the
5.14.0 release to discuss these changes with you, but we are in deep
freeze now (i.e., it is obviously much too late in the present release
cycle to get these or virtually any other code change into that
release). But once you provide a good simple test example to make
sure all is well with these changes, it would be ideal to get these
changes accepted for PLplot early in the release cycle for 5.15.0.
Best wishes,
Alan
__________________________
Alan W. Irwin
Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________
Linux-powered Science
__________________________
_______________________________________________
Plplot-devel mailing list
Plplot-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/plplot-devel
