PLplot

Alan W. Irwin writes:
 > Could we have some discussion about what the configuration defaults should be?
 > 
 > I suspect most of the following are completely ignored at the moment with some
 > superseded by variables set up by autotools.  Shouldn't the ignored ones
 > be completely removed from configure.ac (including the configuration
 > printout at the end which prints them out even though they are
 > meaningless)?
 > 
 > with_debug=no
 > with_opt=yes
 > with_profile=no
 > with_shlib=yes
 > with_warn=no
 > with_dbmalloc=no
 > with_fseek=no
 > with_pkgdir=
 > 
 > Below are some additional options along with how I feel they should be
 > changed (or not).  I lean toward a default of with_double=yes, but if a
 > ...
 > with_double=no				yes?
 > with_gcc=no				?
 > with_freetype=no			yes?
 > enable_dyndrivers=no			yes
 > enable_java=no				no?
 > enable_octave=no			yes
 > enable_gnome=no				no
 > enable_ntk=no				yes?
 > 
 > Are there any other defaults that should be changed?

These are fine with me personally, yes even with_double.  Although in that
case you may trip up some people who are used to the old behavior.

with_debug
with_opt
with_profile
with_warn

Should all stay, as I will get to support them again one of these days.
Probably won't be until April.  

with_shlib should probably also stay if it is currently supported -- does
anyone know?

with_dbmalloc can go AFAIAC.  These days we have access to better tools
(e.g. valgrind) and it has been some years since I've used a debugging
malloc.

BTW I will soon (?) add a few more: with_readline and with_libtecla, both for
supporting command line editing in the pltcl shell.

-- 
Maurice LeBrun    mjl@...
Research Organization for Information Science and Technology of Japan (RIST)

On Thu, 27 Feb 2003, Maurice LeBrun wrote:

> These are fine with me personally, yes even with_double.  Although in that
> case you may trip up some people who are used to the old behavior.

OK.  I left with-double=no.

However, in my recent commit I changed with_freetype, enable_dyndrivers,
enable_octave, and enable_ntk to yes.

I also replaced the non-working has_x11 logic with the have_x logic that is
implemented by the AC_PATH_XTRA.  I tested that new logic by trying the
standard configure option --without-x which for the first time now does the
right thing (no tk, no xwin).  I assume from this good test, that if anybody
ever tries to configure on a system without X, that the right thing will be
done in that case as well, but I don't have access to such a system to test
this.

>
> with_debug
> with_opt
> with_profile
> with_warn
>
> Should all stay, as I will get to support them again one of these days.
> Probably won't be until April.

OK.

>
> with_dbmalloc can go AFAIAC.  These days we have access to better tools
> (e.g. valgrind) and it has been some years since I've used a debugging
> malloc.

OK. I removed dbmalloc.

I also removed with_fseek because I couldn't find a trace of any use of it.

I left with_pkgdir= alone because apparently it is a part of the tcl
configuration logic.  (I hope that logic gets cleaned up before 5.2.1, but
that is a subject for another day.)

That still leaves the original question about with_gcc.  I think it is
completely ignored at the moment so it should probably just be dropped
unless somebody wants to maintain it.

> [Out of order...]
> with_shlib should probably also stay if it is currently supported -- does
> anyone know?

I believe the standard --enable-shared completely replaces with_shlib. Thus,
I don't think we should maintain a parallel with_shlib. If everyone agrees,
I will make the appropriate changes in configure.ac and sysloc.in to replace
the with_shlib logic with the corresponding enable-shared logic.

Does anyone else have additional configure defaults they want to change?

Alan

__________________________
Alan W. Irwin
email: irwin@...
phone: 250-727-2902

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

On Friday 28 February 2003 19:51, Alan W. Irwin wrote:
| On Thu, 27 Feb 2003, Maurice LeBrun wrote:
| > These are fine with me personally, yes even with_double.  Although
| > in that case you may trip up some people who are used to the old
| > behavior.
|
| OK.  I left with-double=no.

I don't think that Maurice really meant to keep with-double=no. After 
all, when we change the API the consequences are much worse, and that 
don't inhibits us from doing it.
And we have the configure summary, that clearly states the float/double 
status.
Also, with the new nn/csa/qhull libraries, the better is to use double 
by default.
So, I vote for with-double=yes.

Maurice, some time ago you argued that we should not drop float support. 
Could you please "argue" again? Sourceforge mailing lists don't have a 
search facility, what makes difficult to find something we know is 
there.

|
| However, in my recent commit I changed with_freetype,
| enable_dyndrivers, enable_octave, and enable_ntk to yes.
|
| I also replaced the non-working has_x11 logic with the have_x logic
| that is implemented by the AC_PATH_XTRA.  I tested that new logic by
| trying the standard configure option --without-x which for the first
| time now does the right thing (no tk, no xwin).  I assume from this
| good test, that if anybody ever tries to configure on a system
| without X, that the right thing will be done in that case as well,
| but I don't have access to such a system to test this.
|
| > with_debug
| > with_opt
| > with_profile
| > with_warn
| >
| > Should all stay, as I will get to support them again one of these
| > days. Probably won't be until April.
|
| OK.

And meanwhile should they appear when we do ./configure --help? It will 
only confuse users! I think that they should hibernate until Maurice 
revives them.
./configure --help says to set CC, CFLAGS, etc.

| > with_dbmalloc can go AFAIAC.  These days we have access to better
| > tools (e.g. valgrind) and it has been some years since I've used a
| > debugging malloc.
|
| OK. I removed dbmalloc.
|
| I also removed with_fseek because I couldn't find a trace of any use
| of it.
|
| I left with_pkgdir= alone because apparently it is a part of the tcl
| configuration logic.  (I hope that logic gets cleaned up before
| 5.2.1, but that is a subject for another day.)
|
| That still leaves the original question about with_gcc.  I think it
| is completely ignored at the moment so it should probably just be
| dropped unless somebody wants to maintain it.

Yes, I also think so. gcc must be the default. For other situations 
CC=xx ./configure should do the trick.

|
| > [Out of order...]
| > with_shlib should probably also stay if it is currently supported
| > -- does anyone know?
|
| I believe the standard --enable-shared completely replaces
| with_shlib. Thus, I don't think we should maintain a parallel
| with_shlib. If everyone agrees, I will make the appropriate changes
| in configure.ac and sysloc.in to replace the with_shlib logic with
| the corresponding enable-shared logic.

Isn't it the same thing?

| Does anyone else have additional configure defaults they want to
| change?

Not that I can remember,
Joao

|
| Alan
|
| __________________________
| Alan W. Irwin
| email: irwin@...
| phone: 250-727-2902
|
| Astronomical research affiliation with Department of Physics and
| Astronomy, University of Victoria (astrowww.phys.uvic.ca).
|
| Programming affiliations with the Canadian Centre for Climate
| Modelling and Analysis (www.cccma.bc.ec.gc.ca) and the PLplot
| scientific plotting software package (plplot.org).
|
| __________________________
|
| Linux-powered Science
| __________________________
|
|
|
| -------------------------------------------------------
| This sf.net email is sponsored by:ThinkGeek
| Welcome to geek heaven.
| http://thinkgeek.com/sf
| _______________________________________________
| Plplot-devel mailing list
| Plplot-devel@...
| https://lists.sourceforge.net/lists/listinfo/plplot-devel

On Fri, 28 Feb 2003, [iso-8859-1] Jo=E3o Cardoso wrote:

> On Friday 28 February 2003 19:51, Alan W. Irwin wrote:
> | On Thu, 27 Feb 2003, Maurice LeBrun wrote:
> | > These are fine with me personally, yes even with_double.  Although
> | > in that case you may trip up some people who are used to the old
> | > behavior.
> |
> | OK.  I left with-double=3Dno.
>
> I don't think that Maurice really meant to keep with-double=3Dno.

Yes, from his words above he was clearly willing to go along.  But the only
comment I heard at the point when I made the changes this morning was his
slightly negative one so I decided to be conservative with the change.

However, now that you think double is a better default I am willing to go
along with that.  If nobody else objects in the next few hours I will
just go ahead and make the change.

> | That still leaves the original question about with_gcc.  I think it
> | is completely ignored at the moment so it should probably just be
> | dropped unless somebody wants to maintain it.
>
> Yes, I also think so. gcc must be the default. For other situations
> CC=3Dxx ./configure should do the trick.

OK.  Unless I hear other objections in the next few hours, I will remove
with_gcc.


>
> |
> | > [Out of order...]
> | > with_shlib should probably also stay if it is currently supported
> | > -- does anyone know?
> |
> | I believe the standard --enable-shared completely replaces
> | with_shlib. Thus, I don't think we should maintain a parallel
> | with_shlib. If everyone agrees, I will make the appropriate changes
> | in configure.ac and sysloc.in to replace the with_shlib logic with
> | the corresponding enable-shared logic.
>
> Isn't it the same thing?

No.  There is configure logic in sysloc.in that now depends on with_shlib,
but that option is set by the user independently of --enable-shared.  So
currently it is a mess.

I think the best way to deal with this is to completely remove with_shlib
and change over to enable_shared logic in sysloc.in.  If nobody objects
to this change I will do it in the next few hours.

Alan
__________________________
Alan W. Irwin
email: irwin@...
phone: 250-727-2902

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting softwar=
e
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Jo=E3o Cardoso writes:
 > On Friday 28 February 2003 19:51, Alan W. Irwin wrote:
 > | On Thu, 27 Feb 2003, Maurice LeBrun wrote:
 > | > These are fine with me personally, yes even with_double.  Althou=
gh
 > | > in that case you may trip up some people who are used to the old=

 > | > behavior.
 > |
 > | OK.  I left with-double=3Dno.
 >=20
 > I don't think that Maurice really meant to keep with-double=3Dno. Af=
ter=20
 > all, when we change the API the consequences are much worse, and tha=
t=20
 > don't inhibits us from doing it.
 > And we have the configure summary, that clearly states the float/dou=
ble=20
 > status.
 > Also, with the new nn/csa/qhull libraries, the better is to use doub=
le=20
 > by default.
 > So, I vote for with-double=3Dyes.

I didn't actually say I objected; what I wrote was:

> These are fine with me personally, yes even with_double.  Although in=
 that
> case you may trip up some people who are used to the old behavior.

So the latter is just a warning that this is yet another change that ma=
y cause
some transition problems for some people.  Change is always ok if there=
 is a
large enough benefit.. is that the case here?  It's easy enough to spec=
ify
--with-double[=3Dyes].  Are there other benefits to be gained, other th=
an
working with the nn/csa/qhull libraries?

Either way it won't affect me, as I always install both float & double
versions.

 > Maurice, some time ago you argued that we should not drop float supp=
ort.=20
 > Could you please "argue" again? Sourceforge mailing lists don't have=
 a=20
 > search facility, what makes difficult to find something we know is=20=

 > there.

If your code uses large floating point arrays and you're developing on =
a
personal machine the 50% memory reduction gained (and some cpu time gai=
ned) by
using (float *) rather than (double *) can be very significant.

--=20
Maurice LeBrun    mjl@...
Research Organization for Information Science and Technology of Japan (=
RIST)

On Friday 28 February 2003 21:12, Maurice LeBrun wrote:
> Jo=E3o Cardoso writes:
>  > On Friday 28 February 2003 19:51, Alan W. Irwin wrote:
>  > | On Thu, 27 Feb 2003, Maurice LeBrun wrote:
>  > | > These are fine with me personally, yes even with_double.  Althou=
gh
>  > | > in that case you may trip up some people who are used to the old
>  > | > behavior.
>  > |
>  > | OK.  I left with-double=3Dno.
>  >
>  > I don't think that Maurice really meant to keep with-double=3Dno. Af=
ter
>  > all, when we change the API the consequences are much worse, and tha=
t
>  > don't inhibits us from doing it.
>  > And we have the configure summary, that clearly states the float/dou=
ble
>  > status.
>  > Also, with the new nn/csa/qhull libraries, the better is to use doub=
le
>  > by default.
>  > So, I vote for with-double=3Dyes.
>
> I didn't actually say I objected; what I wrote was:
> > These are fine with me personally, yes even with_double.  Although in
> > that case you may trip up some people who are used to the old behavio=
r.
>
> So the latter is just a warning that this is yet another change that ma=
y
> cause some transition problems for some people.  Change is always ok if
> there is a large enough benefit.. is that the case here?  It's easy eno=
ugh
> to specify --with-double[=3Dyes].  Are there other benefits to be gaine=
d,
> other than working with the nn/csa/qhull libraries?

No.

But lets see the other point of view. If users don't specify double, they=
 will=20
loose the best 50% of plgriddata(). And I belive that the majority of use=
rs=20
use doubles, and so the default should be for them.

But this is just a default. We could warn users against using the default=
 !?!.

To sumarise the facts:

1-plgriddata() has 3 native methods that will use either float or doubles=
=2E
2-The 3 other csa/nn/qhul methods are clearly best in most circumstances =
but=20
require doubles.
3-csa is a single file that depends on nothing and that we can "sed" to s=
uit=20
our taste (this is an just an hypothesis).
4-nn/qhull are responsible for the best 2 methods.
   Qhull must exist in the user system for nn to work.
   Qhull is not generally available in most systems.
   Qhull can be compiled for floats, but it not recommended.
   It is not practical to change nn to accept floats.
5-Making plgriddata() convert its data from float to doubles is against w=
hat=20
the user specified at configure time.
6-Most people who uses floats are likely to be using fortran with legacy =
code.=20
plgriddata() is not for them.

Given all point 4 above limitations regarding Qhull, most users will not =
use=20
the plgriddata() related capabilities (but they will feel curious, I hope=
=2E=20
Lets watch the Qhull download statistics when we announce plplot)

Currently users are warned:
   plabort("plgriddata(): you must have Qhull to use GRID_NNI.");
I can also add
   plabort("plgriddata(): you must use doubles to use GRID_NNI.");

If the user wants to use those techniques he will ponder and rebuild plpl=
ot.
So, making float the default is not that bad. With a proper Announcement,=
=20
users will be warned from the very beginning and will configure according=
ly.=20
(But with a proper Announcement we could also say that doubles are now th=
e=20
default.)

Given all the above, I still think that doubles should be the default.

Pity we are not MS. We could make "configure" to send us the configure su=
mmary=20
so we would collect statistics ;-)

>
> Either way it won't affect me, as I always install both float & double
> versions.
>
>  > Maurice, some time ago you argued that we should not drop float supp=
ort.
>  > Could you please "argue" again? Sourceforge mailing lists don't have=
 a
>  > search facility, what makes difficult to find something we know is
>  > there.
>
> If your code uses large floating point arrays and you're developing on =
a
> personal machine the 50% memory reduction gained (and some cpu time gai=
ned)
> by using (float *) rather than (double *) can be very significant.

Yes, 50% is a lot, specially in my salary :)
Joao

On Fri, 28 Feb 2003, Alan W. Irwin wrote:

> However, now that you think double is a better default I am willing to go
> along with that.  If nobody else objects in the next few hours I will
> just go ahead and make the change.
>
> OK.  Unless I hear other objections in the next few hours, I will remove
> with_gcc.
>
> I think the best way to deal with this is to completely remove with_shlib
> and change over to enable_shared logic in sysloc.in.  If nobody objects
> to this change I will do it in the next few hours.

All three changes have now been done.  A complete build and
install from a clean start gave a good plplot-test.sh result.

Enjoy the new default configuration!

BTW, Maurice.  When I was tracking down all the with_gcc stuff it looks like
that variable is used for other things in autotools.  It is possible that
name clash was killing your attempts to use KCC.  Anyhow, now we no longer
attempt to use with_gcc, it is probably worth another try with KCC.  I will
be most interested in the results of that experiment.

Alan
__________________________
Alan W. Irwin
email: irwin@...
phone: 250-727-2902

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

On Sat, 1 Mar 2003, Joao Cardoso wrote:

> 5-Making plgriddata() convert its data from float to doubles is against what
> the user specified at configure time.
> 6-Most people who uses floats are likely to be using fortran with legacy code.
> plgriddata() is not for them.

Those two arguments do not convince me....;-)

Alan
__________________________
Alan W. Irwin
email: irwin@...
phone: 250-727-2902

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

* Alan W. Irwin <irwin@...> [2003-02-28 20:30]:

> > 5-Making plgriddata() convert its data from float to doubles is against what
> > the user specified at configure time.
> > 6-Most people who uses floats are likely to be using fortran with legacy code.
> > plgriddata() is not for them.
> 
> Those two arguments do not convince me....;-)

You put a smile at the end of your sentence above, so I hope you are joking.
If that is the case, I am afraid I did not get the joke.  At any rate, both
of Joao's arguments above are perfectly correct.

-- 
Rafael

* Joao Cardoso <jcard@...> [2003-03-01 03:30]:

> [snip]
>
> If the user wants to use those techniques he will ponder and rebuild
> plplot. So, making float the default is not that bad. With a proper
> Announcement, users will be warned from the very beginning and will
> configure accordingly.  (But with a proper Announcement we could also say
> that doubles are now the default.)
> 
> Given all the above, I still think that doubles should be the default.

I am pretty convinced by your arguments.  I thought that it would be
straightforward (and acceptable) to cast back and forth all the data
processed with nn/csa.

> Pity we are not MS. We could make "configure" to send us the configure
> summary so we would collect statistics ;-)

Well, you can always implement this and this would work for non-MS systems
too.  Of course, we should not play the Insidious PLplot Big Brother, but
letting users to choose if they wish to send the informations or not.  There
is a package in Debian, called popularity-contest, that sends regularly
reports about what is installed in the system to a specific email address.
We are using this to collect statistics.

Another possibility would be to make a pool at our website.

-- 
Rafael

Hi,

1) --with-double does not appears when we do=20
=20
    ./configure --help

but it is honoured if we configure --with-double=3Dno.

 Users that require floats might think that the option is no longer suppo=
rted.

2) Also, --disable-drivers seems to not sort  effect, as=20

    gd ljii ljiip ntk pbm ps pstex tk tkwin xfig xwin

are built.

3) ./configure --help | wc -l

reports 180 lines of help!!!, 7 screen pages!
The options count are:
   29  are --with,=20
   49 are --enable,
   6  are  --disable
   1  is --without

Is it possible to categorize this with sub-help options? I guess not.

Joao

On Saturday 01 March 2003 09:22, Rafael Laboissiere wrote:
> * Joao Cardoso <jcard@...> [2003-03-01 03:30]:
> > [snip]
> >
> > If the user wants to use those techniques he will ponder and rebuild
> > plplot. So, making float the default is not that bad. With a proper
> > Announcement, users will be warned from the very beginning and will
> > configure accordingly.  (But with a proper Announcement we could also=
 say
> > that doubles are now the default.)
> >
> > Given all the above, I still think that doubles should be the default=
=2E
>
> I am pretty convinced by your arguments.  I thought that it would be
> straightforward (and acceptable) to cast back and forth all the data
> processed with nn/csa.

To be accurate, it is straightforward, as data structures have to be fill=
ed=20
forth and back, but I don't think that's acceptable if the user specified=
=20
with-float at configure time.
Another example where memory is importante is in plimage(), that works=20
internally with shorts and chars!

> > Pity we are not MS. We could make "configure" to send us the configur=
e
> > summary so we would collect statistics ;-)
>
> Well, you can always implement this and this would work for non-MS syst=
ems
> too.  Of course, we should not play the Insidious PLplot Big Brother, b=
ut
> letting users to choose if they wish to send the informations or not.=20

That was what I meant.
And on what service would you rely? sendmail? With most desktop users rel=
ying=20
in they mail-reader ability to send and receive mail directly from their=20
organization mailer, I think that sendmail is not correctly setup on most=
=20
systems.

> There is a package in Debian, called popularity-contest, that sends
> regularly reports about what is installed in the system to a specific e=
mail
> address. We are using this to collect statistics.
>
> Another possibility would be to make a pool at our website.

That approach relies on a positive action, deliberately filling a form, a=
nd as=20
such it would undersample our users universe; on the contrary, asking at =
the=20
end of configure "send ... yes/no?" only requires a keypress.

* Joao Cardoso <jcard@...> [2003-03-02 20:01]:

> 1) --with-double does not appears when we do 
>  
>     ./configure --help
> 
> but it is honoured if we configure --with-double=no.

Fixed in my last cvs commit.  It is impressive that we have never included a
MY_ARG_WITH for the --with-double option.

> 3) ./configure --help | wc -l
> 
> reports 180 lines of help!!!, 7 screen pages!
> The options count are:
>    29  are --with, 
>    49 are --enable,
>    6  are  --disable
>    1  is --without

I do not see a problem with this.  I usually do "./configure --help | less".
What we have absolutely to do is to update and improve the documentation in
doc/docbook/src/os_notes.xml.

> Is it possible to categorize this with sub-help options? I guess not.

I do not know exactly what you mean.  If you are suggesting that configure
should accept new sub-help options like --help-foo and --help-bar, I think
we are out of hope.  If you are talking about a better presentation when
./configure --help is run, we could do something.

For instance, if I modify the MY_ARG_WITH(double,...) that I just committed
to read:

MY_ARG_WITH(double,[
Options that will affect the behavior of the library:
]AC_HELP_STRING([--with-double],
    [Compile the library with double precision floating point
     ($with_double)]))
 
then I have:

./configure --help
[...]
  --with-mkoctfile=file   Specify mkoctfile
  --with-octavex=file      Specify Octave

Options that will affect the behavior of the library:
  --with-double           Compile the library with double precision floating
                          point ($with_double)
  --with-www-user=NAME    User name at WWW host
[...]

Of course, this will only be meaninful if we do a reorganisation of
configure.ac and sysloc.in to regroup the calls to MY_ARG_WITH and
MY_ARG_ENABLE in a logical order.

-- 
Rafael

* Joao Cardoso <jcard@...> [2003-03-02 20:01]:

> To be accurate, it is straightforward, as data structures have to be filled 
> forth and back, but I don't think that's acceptable if the user specified 
> with-float at configure time.

I understand your argument, and I would also defend it.  However, Alan
pointed me out in a private discussion that there may be cases where users
do not care about efficiency or memory occupation, but would like to
configure --without-double, because they need to use some legacy code.

> > Another possibility would be to make a pool at our website.
> 
> That approach relies on a positive action, deliberately filling a form, and
> as such it would undersample our users universe; on the contrary, asking at
> the end of configure "send ... yes/no?" only requires a keypress.

To be frank, I would be pissed off if some project I configure & make asked
me such questions.  People are not used to that and I think that PLplot
should not make an exception.  Besides that, the configure script must be
non-interactive, at any case.

What about including a command in the distribution (like a shell script,
with text AC_SUBSTituted by configure and which email the informations to
us) and we ask our users in the mailing lists to run it?  Okay, it relies
also on positive action, but is less hassle than filling forms in a web
site.

-- 
Rafael

On Sunday 02 March 2003 22:10, Rafael Laboissiere wrote:
> * Joao Cardoso <jcard@...> [2003-03-02 20:01]:
> > To be accurate, it is straightforward, as data structures have to be
> > filled forth and back, but I don't think that's acceptable if the use=
r
> > specified with-float at configure time.
>
> I understand your argument, and I would also defend it.  However, Alan
> pointed me out in a private discussion that there may be cases where us=
ers
> do not care about efficiency or memory occupation, but would like to
> configure --without-double, because they need to use some legacy code.

I understand that situation. How could it be solved?

1-Adding another configure option --short-in-memory or perhaps=20
--with-huge-data? In that case, if --with-double=3Dno --with-huge-data=3D=
yes were =20
specified, then nn/csa would not be built; if only --with-double=3Dno, th=
an=20
floats would be converted to doubles and vice versa.

2-Even if --with-double=3Dno, we always convert floats<->doubles, but rel=
y that=20
users know what they are doing and would not call the nn/csa based griddi=
ng=20
algorithms? If they would they would do it just once -- learn by error.

3-Or making --with-qhull=3Dno by default? I vote against this last one :(=
)

The second option seems reasonable to me.

> > > Another possibility would be to make a pool at our website.
> >
> > That approach relies on a positive action, deliberately filling a for=
m,
> > and as such it would undersample our users universe; on the contrary,
> > asking at the end of configure "send ... yes/no?" only requires a
> > keypress.
>
> To be frank, I would be pissed off if some project I configure & make a=
sked
> me such questions.  People are not used to that and I think that PLplot
> should not make an exception.  Besides that, the configure script must =
be
> non-interactive, at any case.
>
> What about including a command in the distribution (like a shell script=
,
> with text AC_SUBSTituted by configure and which email the informations =
to
> us) and we ask our users in the mailing lists to run it?

It could be run by configure, then users would just sent the result to us=
=2E
Stupid woman!(1) don't you see that we already have this? It's called=20
config.summary!

Joao

(1) from an old British comics series, "Hallo, Hallo", if I recollect wel=
l :)

On Sunday 02 March 2003 21:58, Rafael Laboissiere wrote:
> * Joao Cardoso <jcard@...> [2003-03-02 20:01]:
=2E..
>
> > Is it possible to categorize this with sub-help options? I guess not.
>
> I do not know exactly what you mean.  If you are suggesting that config=
ure
> should accept new sub-help options like --help-foo and --help-bar, I th=
ink
> we are out of hope.

That was my idea, but forget it, we have more important things to fix.
Joao

* Joao Cardoso <jcard@...> [2003-03-02 23:27]:

> I understand that situation. How could it be solved?
> 
> 1-Adding another configure option --short-in-memory or perhaps 
> --with-huge-data? In that case, if --with-double=no --with-huge-data=yes were  
> specified, then nn/csa would not be built; if only --with-double=no, than 
> floats would be converted to doubles and vice versa.
> 
> 2-Even if --with-double=no, we always convert floats<->doubles, but rely that 
> users know what they are doing and would not call the nn/csa based gridding 
> algorithms? If they would they would do it just once -- learn by error.
> 
> 3-Or making --with-qhull=no by default? I vote against this last one :()
> 
> The second option seems reasonable to me.

For me two, provided that you the data passed back and forth are casted with
PLFLT.  Single precision people should know that they are going to loose
something (in this case memory space) if they want to use the great
plgriddata.  However, they should no be prevented to use it because the API
only accepts double.

> It could be run by configure, then users would just sent the result to us.
> Stupid woman!(1) don't you see that we already have this? It's called 
> config.summary!

I know, I was thinking about something more complete, including the
architecture and other things.  And yes, config.summary is a good starting
point.

-- 
Rafael

* Joao Cardoso <jcard@...> [2003-03-02 23:35]:

> That was my idea, but forget it, we have more important things to fix.

Okay, let us keep this project dormant.  I still think that regrouping the
options in .c/onfigure --help and organising them by sections may be a good
idea.

-- 
Rafael

* Rafael Laboissiere <laboissiere@...> [2003-03-03 01:05]:

> For me two, provided that you the data passed back and forth are casted
> with PLFLT.  Single precision people should know that they are going to
> loose something (in this case memory space) if they want to use the great
> plgriddata.  However, they should no be prevented to use it because the API
> only accepts double.

Sorry, I have looked carefully into plgridd.c.  It seems that plgriddata
already uses PLFLT in a consistent way, so my comment above is not relevant.
However, why don't use PLINT instead of int?
 
-- 
Rafael

On Monday 03 March 2003 00:49, Rafael Laboissiere wrote:
> * Rafael Laboissiere <laboissiere@...> [2003-03-03 01:05]:
> > For me two, provided that you the data passed back and forth are cast=
ed
> > with PLFLT.  Single precision people should know that they are going =
to
> > loose something (in this case memory space) if they want to use the g=
reat
> > plgriddata.  However, they should no be prevented to use it because t=
he
> > API only accepts double.

I have already made the changes.
I made the cast explicit, so that casual readers will understand that the=
=20
types might be different.

I think that most compilers will just ignore a cast from a variable that =
is=20
already in the final type, like in:

=09x =3D (double) y;

The cast should represent no overhead if both x and y are doubles. Do you=
=20
agree?

> Sorry, I have looked carefully into plgridd.c.  It seems that plgriddat=
a
> already uses PLFLT in a consistent way, so my comment above is not
> relevant. However, why don't use PLINT instead of int?

I never worried about it, except in structure definitions, for padding=20
purposes, and in drivers, where it can matter.

Joao