[Fhsst-authors] Hsalgebra Figures . . . and Stuff

[Peter Hutnick]

I checked some changes in this morning.  I intended to do it yesterday, 
but savannah ate my pubkey!

Before I go any further I'd like some feedback on the stuff I checked in.

I created figure 1.12 (in hsalgebra.tex).

I made every effort not to goof the TeX file up.  Used GNU nano (with 
-w) to edit.  Still builds fine.  Whew!

I also found an error in that example; the signs for all of the values 
of f(x) in the table were wrong.  It's fixed in the attached version.

I have a few questions/issues on the whole figures situation.  First, I 
am a bit confused about the naming convention.  I have been using wf for 
What is a Function . . . the name of the chapter.  I have been numbering 
them ex[figure number].  This seems to deviate somewhat from the 
existing figures, but I'm not sure what the right way is.

** Warning: Here comes some advice that makes up for its lack of quality 
with sheer quantity and audaciousness. **

I think that the naming scheme for these figures has some shortcomings  
(other than merely being opaque to me ).

First, a directory filled with files that are /all/ prefixed by the same 
string (hsa) indicates to me that there should be another subdir.  Like:

fhhst
  |-> maths
          |-> eps
                |-> hsa
                       |-> functions.hypb.1.gnuplot


("eps" should be figs IMO, since other dir names indicate content, not 
format . . .)

Using a more plain description will help avoid renumbering issues.  I 
presume this is why all the qi qf stuff was used in the first place.

In the same vein, the .gnuplot files are full of careful formatting 
commands, which makes the plots look very nice.  OTOH, it makes them 
fragile.  Ideally, changing a coefficient would cause the plot to 
automatically re-scale, shift labels around, etc.  GNUPlot's ability to 
do this seems quite limited.  In the absence of near AI abilities in 
GNUPlot to make everything pretty, I propose that prettyfication be left 
to the last.

Specifically I am talking about using coords with x and ylabel, and 
manually setting the x and yticks.  I understand that this is to avoid 
the crowded looking zeros at the origin.  Like I said, best left to last.

Incidentally, the PostScrpt terminal type has an eps option.  I think 
this will save a step.  And why aren't the EPS files generated by the 
make system?

I would be willing to make any of these changes that are agreeable to 
the group.  (Assuming I can track down that key . . .)

I am acutely aware that I am the new kid, and I didn't show up just to 
start telling the old hands what's what.  OTOH (how many hands is that 
so far?), I can't go ahead with practices that I think are going to 
cause problems down the road in good conscience without saying something.

** End unsolicited advice **

I also have some questions about GNUPlot.  I can do my own research on 
this stuff, but if someone here is an expert it would save me a bunch of 
groveling through docs and begging on mailing lists that I don't happen 
to already be subscribed to

The xlabel and ylabel settings work precisely opposite of the way I'd 
expect.  Namely, they put the xlabel at the end of the /negative/ /y/ 
axis, as opposed to the positive x axis.  What give with that?

Am I correct in my belief that the aspect ratio of these charts is fixed 
at 7:5?  Having to fiddle around with a proportion to get a "square" 
plot for things like circles seems sub-optimal.

Speaking of circles, my circle (hsa-wf-ex1.eps) is missing the parts 
that should be close to the x axis.  Is there a known work-around for 
this sort of thing?

Wow, that was a long email.  I hope I haven't ruffled any feathers or 
been otherwise annoying.  By all means, flame me if I have

Thanks for letting me monkey around with your project.

-Peter