%&latex
% File cmolddig.tex
% Copyright 1999 Rowland McDonnell
% rowland.mcdonnell@physics.org
%
% Part of the cmolddig package
%
% This file can redistributed and/or modified under the terms
% of the LaTeX Project Public License Distributed from CTAN
% archives in directory macros/latex/base/lppl.txt; either
% version 1 of the License, or (at your option) any later version.

\documentclass[a4paper,11pt]{article}

\usepackage{array,xspace}

\makeatletter
\newcommand{\ttlb}{{\ttfamily\def\@tempcmda{OT1}%
\ifx\f@encoding\@tempcmda\char"7B\else\{\fi}}

\newcommand{\ttrb}{{\ttfamily\def\@tempcmda{OT1}%
\ifx\f@encoding\@tempcmda\char"7D\else\}\fi}}

\newcommand{\ttbs}{{\ttfamily\def\@tempcmda{OT1}%
\ifx\f@encoding\@tempcmda\char`\\\else\textbackslash\fi}}
 
\newcommand*{\comname}[2][VF500F2E]{{\ttfamily\ttbs #2%
\def\@tempcmda{VF500F2E}\def\@tempcmdb{#1}%
\ifx\@tempcmda\@tempcmdb\else\comarg{#1}\fi}}
\makeatother

\newcommand*{\comarg}[1]{{\ttfamily\ttlb$\langle$%
{\rmfamily\itshape#1}\/$\rangle$\ttrb}}

\newcommand*{\comoptarg}[1]{{\ttfamily[$\langle$%
{\rmfamily\itshape#1}\/$\rangle$]}}

\newcommand*{\filename}[1]{\texttt{#1}}
\newcommand*{\packname}[1]{\texttt{#1}}
\newcommand*{\optname}[1]{\texttt{#1}}
\newcommand*{\newsgroup}[1]{\texttt{news://#1}}
\newcommand*{\dvidriver}{dvi driver}

\newcommand*{\cmolddig}{\packname{cmolddig}\xspace}

\author{Rowland McDonnell}
\title{The \cmolddig package}
\date{23rd August 1999}

\begin{document}

\maketitle
\tableofcontents

\section{Introduction}

The \packname{cmolddig} package sets things up so that numbers are 
printed using old style digits -- $\oldstylenums{1234567890}$ -- 
rather than the usual lining digits -- $1234567890$. It only works 
with the original Computer Modern founts. You can choose whether 
numbers are printed as old style or lining digits in text or in maths 
mode.

You can get old style digits without any extra help by using the 
command \comname{oldstylenums} in maths mode, but this package 
arranges things so that all numbers are printed using old style digits 
without you needing to do that.

To cut a very involved tale down to the barest essentials: old style 
digits look better and are easier to read in text and the like; while 
lining digits are more suitable for maths, in tables, and the like.  
Both these assertions are argued over by typographers.

If you have the T1 encoded European Modern founts installed, you might 
be interested in the \packname{eco} package, which is the T1 
equivalent of \cmolddig. It avoids many of the limits of this package. 
The \cmolddig package doesn't contain any new characters: you need to 
have the OT1 encoded Computer Modern Roman founts installed before 
this package is useful. As far as I know, all \TeX\ systems come with 
these founts installed by default. It doesn't matter if you're using 
Metafont, PostScript Type 1, or TrueType versions of the Computer 
Modern founts; but you do need a \TeX\ system able to use virtual 
founts. Almost all modern (1999) \TeX\ systems can use virtual founts; 
the only exception I know of is Y\&Y \TeX's version.

\packname{Cmolddig} works by switching to a set of virtual founts that 
have the normal lining digits replaced with old style digits from the 
corresponding maths italic fount. These digits are conventional 
upright digits despite living in a nominally italic fount.

The original Computer Modern founts have old style digits available 
for the Computer Modern Roman family only, and even then only for the 
upright shapes: Computer Modern Roman upright, bold extended, bold, 
and small caps. Because of this, this package can't give you old style 
digits for Computer Modern Sanserif or Computer Modern Typewriter; nor 
does it give you old style digits if you're using any italic or 
slanted fount.

Options exist to use old style digits in maths or not, independently 
of the style of numbers in normal text.  You can also choose to make 
use of the extra small caps sizes included with the AMS fount set if 
you have it installed.

\section{Installation}

You need to put three directories of files in the right places on 
your \TeX\ system's directory tree: with that done, installation is 
complete.

All the files in the \filename{texinputs} directory should go 
into a directory that \TeX\ searches for input files. All the files in 
the \filename{tfm} directory should go into a directory that 
\TeX\ searches for \filename{tfm} (fount) files. All the files in 
the \filename{vf} directory should go into a directory that 
\TeX\ searches for \filename{vf} (virtual fount) files. If you don't 
know where these places are, have a read of your \TeX\ system's 
documentation.

Moving these files to the right places is all you need to do.  Once 
you've done this, try running \LaTeX\ on \filename{odshrtst.tex}: you 
should see some old style digits in the output, showing that you've 
installed the package correctly.

You can ignore all the files in the \filename{source} directory 
unless you're interested in how I created the virtual founts at the 
heart of this package.  The \filename{source} directory can be thrown 
away if you like.

\section{Using the \cmolddig package}

Make sure you include the \cmolddig package by saying:
\begin{verbatim}
  \usepackage{cmolddig}
\end{verbatim}
in the preamble of your document. This will switch you to Computer 
Modern Roman with old style digits in normal text only, not in maths 
mode.

\subsection{Controlling the package}

You can switch between Computer Modern Roman with old style digits and 
with normal lining digits inside your document using the new commands:
\comname{cmoldstyledigits} and \comname{cmliningdigits}.
\begin{verbatim}
   \cmoldstyledigits  35,000 cabbages with old style digits,
   \cmliningdigits    or 15 turnips with lining digits.
\end{verbatim}

By default, the \cmolddig package doesn't affect numbers in maths mode.  
If you want old style digits in maths mode, you can give the package 
the \optname{cmoldstyledigits} option:
\begin{verbatim}
  \usepackage[cmoldstyledigits]{cmolddig}
\end{verbatim}
the \optname{cmliningmathdigit} option forces lining digits in 
maths mode; this is what \cmolddig usually does, so there's probably 
no need to know about this option.

You can switch between Computer Modern Roman with old style digits in 
maths and with normal lining digits in maths inside your document 
using the commands: \comname{cmoldstyledigits} and 
\comname{cmliningdigits}.
\begin{verbatim}
   \cmoldstylemathdigits $35 \pi = x$%       old style digits
   \cmliningmathdigits   $17x^{2}-32x+17=y$% lining digits
\end{verbatim}

You might like to note that it's \emph{very} unusual to use old style 
digits in maths. I have come across one typographer who asserts that, 
despite this, maths is more readable when typeset with old style 
digits. Whatever the truth of the matter, it's probably best to keep 
to lining digits in maths if you are producing work that needs to meet 
any standard of conventionality.

If you have the AMS founts which come with Computer Modern Roman caps 
and small caps in two extra sizes -- 8\,pt and 9\,pt -- you can tell 
\cmolddig about it with the \optname{amsfonts} option:
\begin{verbatim}
    \usepackage[amsfonts]{cmolddig}
\end{verbatim}
will tell \cmolddig to use these extra sizes if needed. The \cmolddig 
package won't use these two extra founts unless you tell it to. The 
\optname{extracsc} option is identical in effect to the 
\optname{amsfonts} option.

The package also includes the \optname{noamsfonts} and 
\optname{noextracsc}: a pair of options which tell \cmolddig to avoid 
using the extra AMS small caps founts.  I can't see any need for 
anyone to use them, because the \cmolddig package avoids using these 
extra founts by default.

\section{How it works and stuff}

There are three separate things that can be considered: the virtual 
fount mechanism in \TeX, which makes this package possible; the 
creation of the particular virtual founts used by \cmolddig; and what 
the package actually does when you use it.

\subsection{Virtual founts}

A warning: the subject of \TeX\ and founts is complicated and 
never-ending.  This short section is inevitably inadequate, but I hope 
it sheds a little light.

To begin with, all \TeX\ knows about a fount is the \filename{tfm} 
file. When you select a fount for typesetting, \TeX\ reads the 
appropriate \filename{tfm} file which tells it how much space each 
character takes up (height, depth, and width), how much space to add 
or subtract between certain pairs of letters (kerning information), 
which pairs of characters should be replaced by a different character 
(ligature information; replacing \texttt{ff} with an `ff' ligature, 
for example), and similar things.  There is no information about the 
shape of any letters in this file.

Every \filename{tfm} file has up to 256 different characters listed in 
it.  Each character lives in a numbered slot: 0--255.  There are a 
number of standard \emph{encodings} defined for \LaTeX\ which say which 
character should live in which numbered slot.  In OT1 encoding (used 
for the original Computer Modern Roman founts), character number 48 is 
the number `0', character number 49 is the number `1', and so on.

In the normal run of things, \TeX\ places information about the fount 
in the \filename{dvi} file, and when you come to print or preview that 
\filename{dvi} file, the driver program finds a real fount (bitmap, PS 
Type 1, TrueType, or whatever) and displays it. This approach has a 
few problems, particularly when you come to work with founts that (in 
their natural state) don't have the same characters in the same slots 
as the standard \TeX\ encoding. For example, character number 1 in an 
OT1 encoded fount is an upper-case Delta: $\Delta$. This is unique to 
that encoding: character number 1 in a normal PS type 1 fount is 
undefined. If you tried to use a normal PS Type 1 fount with \LaTeX's 
OT1 encoding and took no precautions, things just wouldn't work at all 
right and you'd get the wrong characters printed for lots of things.

The virtual fount mechanism is one method for getting round this (and 
other) problems. The way it works (in brief) is this. You create a 
pair of files: a \filename{tfm} file and a \filename{vf} file. The 
\filename{tfm} file is a perfectly ordinary \filename{tfm} file. The 
\filename{vf} (virtual fount) file is something new: it contains 
information on how to print each character referred to in the 
\filename{tfm} file. The instructions might be to get all the 
characters except the numbers from (say) cmr10, and to get the numbers 
from cmmi10. This is exactly what \filename{cmrj10.vf} does. These 
\filename{vf} files can contain fragments of \filename{dvi} file so 
all sorts of trickery is possible. A more common use for virtual 
founts is re-mapping: re-arranging the characters so that they're in 
the expected slots for the encoding you're using. In other words, you 
might be using (say) a PS Type 1 version of Times using the 
\packname{times} package and OT1 encoding. If you were to use a dot 
accent, you'd be asking for character number 95. In the `natural' PS 
text encoding, character number 95 is an underscore. The magic of 
virtual founts means that character number 95 in the fount 
(\filename{tfm} file) that \TeX\ sees `really' is a dot accent; when 
the \dvidriver comes to display that fount, it sees that 
it's a virtual fount, looks at the appropriate \filename{vf} file, and 
displays character number 199 from the `real' Times fount every time 
the \filename{dvi} file asks for character number 95. (For the sake of 
a comprehensible explanation, I've ignored the widespread use of 8r 
encoding as an intermediate step, and ignored re-encoding done by the 
\dvidriver).

The \filename{vf} (virtual fount) file is a new type of file for 
\TeX\ systems, but it's ignored by \TeX\ itself: \filename{vf} files 
are used only by \filename{dvi} drivers.  What happens is this: when 
a modern \filename{dvi} driver looks at a \filename{dvi} file, it 
checks to see if any of the \filename{tfm} files in it have 
corresponding \filename{vf} files.  If so, the driver looks in the 
\filename{vf} files and follows the instructions in them on how to 
print the characters referred to.  A \filename{vf} file can refer to 
any real fount or any virtual fount, so this business of looking 
things up in virtual founts can get complicated.

You can find out more about virtual founts by running \filename{weave} 
and then \TeX\ on these files from CTAN:
\begin{center}
\begin{tabular}{@{}p{\textwidth}@{}}
\filename{systems/knuth/etc/vftovp.web}\\
\filename{systems/knuth/etc/vptovf.web}
\end{tabular}
\end{center}

\subsection{What the package does}

A brief explanation to begin with: every fount in \LaTeX\ belongs to a 
\emph{family}. Every family has a short internal name by which it is 
referred to inside \LaTeX. The internal name of the Computer Modern 
Roman family is \texttt{cmr}. The \cmolddig package defines a new 
family, which has the internal name \texttt{cmrj}. The `j' is a code 
that means `with old style digits', so \texttt{cmrj} stands for 
`Computer Modern Roman with old style digits'. If you're interested, 
the fontname documentation available from CTAN explains the full 
details of this code.

This package's default behaviour is to define the \comname{rmdefault} 
family to be \texttt{cmrj}. This command defines the fount family used 
for typesetting normal text. What this change means is that when 
\LaTeX\ comes across a request to typeset normal text, it sees that it 
needs to use the \texttt{cmrj} family, it loads the file 
\texttt{ot1cmrj.fd} (this is assuming that you're using OT1 encoding 
-- if you're not, you'll get an error message instead). This file 
tells \LaTeX\ which \filename{tfm} file to load when you ask for any 
given fount. In this case, \LaTeX\ is told to use the new virtual 
founts where they exist. You can learn more about \filename{fd} files 
by reading \filename{fntguide.tex}, which is part of the standard 
\LaTeX\ distribution.

The \cmolddig package makes two new commands: 
\comname{cm\-old\-style\-digits}, which changes \comname{rmdefault} to 
\texttt{cmrj} and switches to \comname{rmfamily}; and 
\comname{cm\-lining\-digits}, which changes \comname{rmdefault} to 
\texttt{cmr} and also switches to \comname{rmfamily}

Maths is a bit more tricky.  By default, the \cmolddig package 
doesn't affect maths mode behaviour.  The package defines two 
commands for switching between the two sorts of digits in maths mode: 
\comname{cmoldstylemathdigits} and \comname{cmliningmathdigits}.  
These commands are used by the package when you use the 
\optname{cmoldstylemathdigits} and \optname{cmliningmathdigit} 
options.  What do they do?  They change the mathcode of the numbers 
0--9 to the appropriate values to select either normal lining digits 
from the usual place, or old style digits from the appropriate maths 
italic fount (which is where the normal upright roman old style digits 
are kept, for reasons best known to Donald Knuth).  If that makes 
almost no sense to you and you'd like to know more, the best solution 
is to read Knuth's \emph{The \TeX book}, the \LaTeX\ source code, and the 
\cmolddig package file.  In case you're wondering, I don't understand 
maths fount selection at all well, and I found out how to perform this 
particular switching by asking on \newsgroup{comp.text.tex}.

The maths digit switching doesn't require the fount files that come 
with this package: it uses the normal `real' Computer Modern founts 
without needing any virtual fount trickery.

\subsection{Creating the \cmolddig virtual founts}

You don't need to do this job: all the useful files that this process 
creates are part of the \cmolddig distribution.  I've included this 
section and the corresponding source files to satisfy the curiosity 
of people who want to know how I did the deed.

The contents of the virtual founts used by \cmolddig are pretty 
straightforward: rounding errors aside, they are identical to the 
Computer Modern originals with normal lining digits in the text 
founts replaced with old style digits from the appropriate maths italic 
fount. These old style digits are upright despite living in an italic 
fount; ask Dr Knuth if you want to know why. The job of creating them 
was non-trivial only because of certain eccentricities of 
\packname{fontinst} and the equally eccentric encodings used by Donald 
Knuth in his original Computer Modern family of founts.

I used \packname{fontinst} to help create the virtual founts used by 
this package; \packname{fontinst} is a set of \TeX\ macros available 
from CTAN: it'll run on any computer with a functioning \TeX\ on it.

To create the virtual founts the way I did, you need to install 
\packname{fontinst}, and use the files in \cmolddig's 
\filename{source} directory by following the instructions below. You 
also need \packname{tftopl} and \packname{vptovf} which should be part 
of your \TeX\ system.

Some of the files used to create the virtual founts at the heart of 
this package have names that break the `8+3' filename limit imposed by 
some archaic operating systems. Apologies for this; it's just the way 
they turned out, and I didn't notice until it was far too late.% 
\footnote{The computer I use now has a file system that dates back to 
1984; it has a 31 character filename limit} Since there's no 
\emph{need} to use the source files, I thought it'd be okay not to 
fix this.

You'll need to use \filename{tftopl} to create the following 
\filename{pl} files from the corresponding Computer Modern 
\filename{tfm} files:
\begin{center}
\begin{tabular}{@{}>{\raggedright\small}p{\textwidth}@{}}
\filename{cmb10.pl}\\
\filename{cmbx10.pl},
\filename{cmbx12.pl},
\filename{cmbx5.pl},
\filename{cmbx6.pl},
\filename{cmbx7.pl},
\filename{cmbx8.pl},
\filename{cmbx9.pl}\\
\filename{cmcsc10.pl}\\
\filename{cmmi10.pl},
\filename{cmmi12.pl},
\filename{cmmi5.pl},
\filename{cmmi6.pl},
\filename{cmmi7.pl},
\filename{cmmi8.pl},
\filename{cmmi9.pl}\\
\filename{cmmib10.pl},
\filename{cmmib5.pl},
\filename{cmmib6.pl},
\filename{cmmib7.pl},
\filename{cmmib8.pl},
\filename{cmmib9.pl}\\
\filename{cmr10.pl},
\filename{cmr12.pl},
\filename{cmr17.pl},
\filename{cmr5.pl},
\filename{cmr6.pl},
\filename{cmr7.pl},
\filename{cmr8.pl},
\filename{cmr9.pl}\\
\end{tabular}
\end{center}

If you have access to the AMS extra founts, you'll also need:
\begin{center}
\begin{tabular}{@{}>{\raggedright\small}p{\textwidth}@{}}
\filename{cmcsc8.pl},
\filename{cmcsc9.pl}
\end{tabular}
\end{center}

The \filename{source} directory contains these five files:
\begin{center}
\begin{tabular}{@{}p{\textwidth}@{}}
\filename{dostretch.mtx}\\
\filename{instcmrj.tex}\\
\filename{ot1noflig.etx}, \filename{ot1nofligcj.etx}, \filename{ot1nofligj.etx}
\end{tabular}
\end{center}

Put these five source files in a directory with the \filename{pl} 
files listed above. If you don't have the AMS extra founts, edit 
\filename{instcmrj.tex} and comment out the lines referring to cmcscj9 
and cmcscj8.

Then run Plain \TeX\ on \filename{instcmrj.tex}.  The result of this 
run will be a huge pile of files: run \packname{vptovf} on the files 
ending in \filename{vpl} (virtual property list; human-readable 
precursors to a machine-readable \filename{tfm}/\filename{vf} pair).  
The resulting \filename{tfm} and \filename{vf} files are the files 
that do the trick.  The \filename{fd} file created by 
\packname{fontinst} isn't particularly useful to begin with: compare 
it to the one supplied with \cmolddig to see what modifications I felt 
were needed.

You can throw away the debris now if you like: all the files 
(\filename{mtx}, \filename{pl}, and \filename{fd}) that have been 
worked with except the five files from \cmolddig's \filename{source} 
directory are machine-generated so you don't need to keep them.

\subsubsection{So how does it work?}

You really ought to understand \LaTeX\ founts and \packname{fontinst} 
to an extent before reading this section.

The \packname{fontinst} 
file that does the work contains lines like this:
{\small
\begin{verbatim}
\installfont{cmrj10}{cmr10,kernoff,cmmi10,kernon,dostretch}{OT1j}
            {OT1}{cmrj}{m}{n}{<10> <10.95>}
\end{verbatim}
}
The last 5 arguments (on the second line) are just the NFSS 
parameters used in the \filename{fd} file that's created by 
\packname{fontinst} for this installation.  The interesting bit is 
the first line: cmrj10 is the (virtual) fount about to be created.  
In this case, the glyphs are taken from cmr10 and cmmi10 
(\packname{fontinst} converts the \filename{pl} files to 
\filename{mtx} files before it does anything with them).  

The file \filename{kernoff.mtx} tells \packname{fontinst} to ignore 
kerning information from subsequent founts: this ensures that the 
newly created fount has kerning from cmr10 only. This is important, 
since maths fount kerning is entirely unsuitable for text. The file 
\filename{kernon.mtx} tells \packname{fontinst} to take notice of 
kerning again. It's redundant, since no kerning information is held in 
\filename{dostretch.mtx}, but kept there just for the sake of 
neatness. So: kerning information is taken from cmr10, not from 
cmmi10.

The file \filename{dostretch.mtx} tells \packname{fontinst} to take 
the new virtual fount's spacing parameters from those that have been 
set by the first fount read -- in other words, to take spacing from 
cmr10 in this case. The third argument is the encoding file used to 
build the fount: in this case, it's OT1j; that's OT1 encoding with old 
style digits. Rather than using the glyphs called `one', `two', 
`three', and so on, this \filename{mtx} file ensures that the glyphs 
called `oneoldstyle', `twooldstyle', `threeoldstyle' are used instead. 
The `oldstyle' digit glyphs are kept in the cmmi founts.

Some of these installation lines ask for the final encoding 
OT1NOFLIGj. This calls \filename{ot1nofligj.etx}, which in turn calls 
\filename{ot1noflig.etx}; this last file is a new encoding file I 
wrote for use with \packname{fontinst v1.8}. It defines the 
\textsc{tex text without f-ligatures} encoding, used in cmr5 and the 
caps and small caps founts.


\end{document}