LATEX-L Archives

Mailing list for the LaTeX3 project


Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Condense Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
text/plain; charset="iso-8859-1"
Mon, 9 May 2005 22:54:16 +0200
Mailing list for the LaTeX3 project <[log in to unmask]>
Lars Hellström <[log in to unmask]>
Mailing list for the LaTeX3 project <[log in to unmask]>
text/plain (110 lines)
At 12.44 +0200 2005-04-27, Achim Blumensath wrote:
>I forgot one feature:
>o Being able to load packages from FD files (e.g., keyval).

In the past I've sort of wanted something like that myself, but in view of
how intricate the contexts can be in which NFSS may decide to load an FD
file, I now rather think it is a good thing that one cannot start loading
packages at such times. Packages are loaded in the preamble, and may modify
the user environment. FD files should rather be considered a representation
of data tables, and the only reason these tables are dynamically loaded
when needed is to conserve memory.

This is however not the same thing as saying that FD files should only
contain data for the tables set by \DeclareFontFamily and
\DeclareFontShape, as is the current official position, since this, as we
have observed, is often too restrictive. However about a week ago it
occurred to me that there is a clean way of writing FD files so that they
(i) do contain extra information and (ii) can be loaded by LaTeX without
any extensions to the core. The trick is simply that the FD file must
contain dummy definitions (that end up gobbling their arguments) of any
nonstandard data declaration commands used. Users who want to take
advantage of the extra information would still need to accomplish this by
loading (in the preamble) a package that (re)defines commands and whatsnot,
but that package doesn't need to know all fonts it can support and the
fonts aren't usable only when that package has been loaded, so there is a
reasonable independence.

As a simple example (but this is typed directly in this mail, so don't 100%
expect it to work), consider the default definition of \bfseries:


Suppose now that there is a package XX which redefines this (and presumably
also its sibling commands) as

           \csname XX@\string\bfdefault/\f@encoding/\f@family\endcsname

and provides a command \XXDeclareFamilyDefault defined as


With those in place, the declaration


will have the effect of changing the series selected by e.g. \textbf from
bx to b, but only for the encoding/family combination T1/ptm, which means
it makes sense to place that declaration in t1ptm.fd. All one has to do is
to write it as


since the \providecommand does nothing if the XX package has been loaded,
and the net effect of them both is to do nothing if XX has not been loaded.
(The dummy definition of \XXDeclareFamilyDefault disappears when the group
ends inside which the FD file is loaded.) Either way, it works out for the

Obviously this is a very simple example, but the idea can be utilized also
in implementing additional font axes on top of the NFSS shape and series.
To begin with one needs a new and richer set of user level commands for
font selection, but that's straightforward. The tricky part is how to do
substitution when a requested font does not exist, and here additional
tables (such as that implemented above) of information about particular
font families can be very useful. It certainly needs a lot of thought in
designing a good model, but it should be possible to do in a completely
generic fashion.

>Internally all axes could be implemented as shape parameter. I suggest
>replacing the current shape value by a comma separated list with one
>entry for each axis:
>  \fontshape{it,sc,osf}\selectfont

A disadvantage with using comma as the separator is that all comma
separated lists in LaTeX (that I can think of right now: class/package
options, package lists, keyval options) are unordered, in the sense that
relative order of the items is almost irrelevant. This is not the case
here; instead each axis has a fixed position. Perhaps a period `.' would be
more appropriate as separator?


One final point to note is that even for a font family with some elaborate
scheme of additional axes in place, there should be a set of font
declarations with "standard" series and shapes, for use if the font is
accessed using the standard font selection scheme of m versus bx and n vs.
it vs. sc vs. sl. These declarations could (preferably?) be set up using
the ssub "size function".

Lars Hellström