plotutils-graph-patches-lloda/doc/h-glyphs.txt

The GNU libplot library and the plotting utilities based on it, such as
GNU `graph', can display text strings in any of 21 `Hershey fonts', such as
HersheySerif and HersheySerif-Bold.  Hershey fonts are built from `Hershey
glyphs'.  There are two sorts of glyph: standard (`occidental') and
Japanese (`oriental').  The glyphs were digitized by Dr. Allen V. Hershey
at what is now the U.S. Naval Surface Weapons Center in Dahlgren, VA
(formerly the U.S. Naval Weapons Laboratory), mostly during the late
1960's.

There are many glyphs that are not part of any Hershey font.  Such glyphs
must be referred to by number.  For example, the string "\#H0744\#H0745"
uses escape sequences to refer to Hershey glyphs #744 and #745, which are a
shamrock and a fleur-de-lys respectively.  All occidental Hershey glyphs
have numbers in the 0..3999 range, but there are many gaps.  In this file
we explain how the array of occidental glyphs now in GNU libplot was
constructed, and give more information on the numbering scheme.

The array was assembled in large part from earlier (pre-GNU) public
distributions of the occidental glyphs.  These included the following.

#0. Distributions by Dr. Hershey himself, in the late 1960's and 1970's.
These were distributions of his typographic software, of which the glyph
distribution was a part.  Initial distributions were on punched cards and
later ones were on magnetic tape.  At least 120 copies of the typographic
software were distributed.  It is not known how many distinct releases
there were, either of the software or of the glyphs.

#1. The U.S. NBS [U.S. National Bureau of Standards, since renamed the
National Institute of Standards and Technology] publication "A Contribution
to Computer Typesetting Techniques", dated 1976, and the accompanying
magnetic tape.  This presumably included all glyphs in use at that time.
In Figure 15 of that publication, 43 additional glyphs (digitized by Norman
Wolcott at NBS) were shown, but the data for them were not given.  (Glyphs
2250, 2260 were probably included in Figure 15 erroneously; they were also
listed among the standard glyphs).

#2. The Usenet distribution of Pete Holzmann <pete@xc.org> and Jim Hurt
(to the newsgroup mod.sources, now defunct).  Circa 1985, but based on the
above 1976 tape.  It introduced the scheme for encoding vector glyphs as
strings which is now used in libplot.  Almost identical to the above (none
of the 43 new Wolcott glyphs), but included three additional glyphs: 997,
998, 999.  These were merely horizontal strokes.

#3. The glyph database incorporated in the PGPLOT subroutine library of Tim
Pearson <tjp@astro.caltech.edu>, still available from
ftp://ftp.astro.caltech.edu and from the astro.caltech.edu web server.
This was probably based, c. 1980, on a distribution obtained either from
Nelson Beebe (see #4 below) or directly from Dr. Hershey.  It included data
for the 43 new Wolcott glyphs, and 10 additional glyphs numbered 236 (an
cartographic-sized `@'), 590 (an underscore), 2078 (Aring), 3330, 3331,
3332, 3430, 3431, 3432 (umlaut-accented German letters), and 4000 (a boxed
inverted questionmark).  The last of these (#4000) apparently did not
originate with Dr. Hershey, but the others may have.  The horizontal stroke
glyphs 997, 998, 999 from distribution #2 were not present.  Ten glyphs
were revised: 1225, 1226, 1407, 1408, 2225, 2226, 2407, 2408, 3010, and
3159.  The first eight of these were braces.  There were 1642 separately
indexed glyphs in all.

#4.  The glyph database incorporated in the PLOT79 utility of Nelson Beebe
<beebe@math.utah.edu>.  Also dates back to c. 1980; based on a tape from
Dr. Hershey himself, with local additions.  All the 1642 glyphs from #3 are
present, except that 590 (an underscore) and 2078 (Aring) are missing.
(The Hershey tape may have been a slightly earlier one than the tape used
for #3?)  Also, 911..922 (twelve outline and filled arrows) and 923..926
(four large pointing hands) were added by Dr. Beebe.  As well, 256 small
fixed-width glyphs (marker symbols taken from a plotter in the 1970's, and
low-resolution alphabets based on the plotter alphabet) in the 1500..1627
and 1700..1827 ranges were added.  There were 1912 separately indexed
glyphs in all.

#5.  The glyph database accompanying Dr. Hershey's distribution in the
1980's of typographic software in True Basic, for PC's.  For details, see
his 1981 technical report `Advanced Computer Typography' and his 1995
technical report `Cartography and Typography with True Basic'.  The 43
Wolcott glyphs (see above) were not present, and the 10 extra glyphs from
the PGPLOT distribution were not present either.  However, an entire new
alphabet (Triplex Greek) was included.  The small upper-case alphabet
(`Cartographic Roman') was extended to include lower-case characters, and a
small `Cartographic Italic' alphabet was added as well.  Many alphabets
were extended by crafting such characters as @, <, >, #, and % for them.
Also, several glyphs were revised.  13 cartographic symbols in the 700..899
range were dropped, probably inadvertently.


The glyph array built into GNU libplot, and accessible to the GNU plotting
utilities, is a merged version.  It includes all glyphs from distribution
#5, the 43 Wolcott glyphs, and 7 of the 10 additional glyphs from
distribution #3, i.e., 2078 (Aring), and 3330, 3331, 3332, 3430, 3431, 3432
(umlaut-accented German letters).  The 13 cartographic symbols in the
700..899 range that were inadvertently dropped from distribution #5 have
been restored.

Many non-Hershey glyphs were also included in the GNU libplot array of
occidental glyphs, to facilitate the construction of ISO-Latin-1 Hershey
fonts.  The 4000..4194 range contains a large number of such non-Hershey
glyphs.  Nelson Beebe's large pointing hands (923..926 in distribution #4)
were included as 4040..4043.  An important source of non-Hershey glyphs was
Robert Beach's UGS [Unified Graphics System], which was developed at the
SLAC [Stanford Linear Accelerator Center] Computation Research Group in the
1970's.  (See Computer Graphics, Fall 1974, pp. 22-23.  The UGS source code
may still be available at ftp://ftp.slac.stanford.edu/software/ugs77/ .)
A number of glyphs in the 4000..4025 range were taken from the UGS glyph
repertory.

The following is a breakdown of the original Hershey glyphs (defined to
include the 43 Wolcott glyphs and the 10 additional glyphs mentioned
above).  They occupy the 1..3926 range of the occidental array, and may be
accessed individually by escape sequences in the range "\#H0001".."\#H3926".  
They fall into the following groups:

0001..0284	alphanumeric symbols, cartographic [small] size
0501..0746	alphanumeric symbols, principal [large] size
0750..0909	centered symbols
0910		alphanumeric symbols, principal [large] size
1001..1295	alphanumeric symbols, indexical [medium] size
1401..2312	alphanumeric symbols, principal [large] size
2317..2382	centered symbols
2401..3926	alphanumeric symbols, principal [large] size

`Alphanumeric symbols' means not just alphabetic symbols and numerals, but
also punctuation marks, mathematical symbols, etc.; in general, any symbol
that can be viewed as having a well-defined `baseline', and which could
appear in a string of characters resting on that baseline.

`Centered symbols' are different.  In practice they would seldom be
displayed in such a way that they are resting on a baseline.  Instead, they
would be drawn so as to be centered on a specified point.  Also, they would
seldom appear in a string.  An example would be the `sand' symbol, Hershey
glyph #0764, which is clearly designed to be drawn on a map rather than to
appear in text.

In the original distributions of the Hershey glyphs, no distinction was
made between alphanumeric symbols and centered symbols.  That was possible
because in the coordinate system used by Dr. Hershey, all are centered on
the point (0,0).  The `baseline' concept was not used.  Each alphanumeric
symbols had a well-defined baseline, but the baselines differed for the
three sizes.

When strings are drawn with the alabel() function of GNU libplot, e.g.,
when GNU `graph' labels axes, a baseline must be chosen.  For alphanumeric
symbols of all kinds and also centered symbols, we have chosen it to be the
baseline appropriate for the principal [large] size alphanumeric symbols.
This determines a choice of vertical elevation of each symbol, when it is
rendered as an element of a string.

If the `x' option to alabel() is used, to place characters on a baseline,
the vertical elevation may not be appropriate for indexical [medium] and
cartographic [small] alphanumeric symbols.  But nowadays the indexical and
cartographic glyphs are seldom accessed.  The only reason for drawing
smaller characters is to draw superscripts and subscripts, and our
rendering algorithm uses miniaturized principal-size characters for that.

If you are using the alabel() function to display strings, you may specify
`c' as the third argument to obtain vertical centering, instead of the
usual `x', which places characters on a baseline.  You would use `c' to
place a centered symbol at a specified point.  For example, the function
call

	alabel ('c', 'c', "\#H0764");

would place the `sand' symbol, glyph #764, at the current position.  The
first argument `c' requests horizontal centering, so the symbol will be
centered both horizontally and vertically.