\part Preliminaries\endpart \chapter Introduction\label{INTRO}\endchapter This manual is intended for \tex\ wizards pondering the intricacies of various \lamstex\ constructions, as well as for \tex nicians designing style files for \lamstex\ who need more detailed information than that provided by the \lamstex~Style File Designers Manual. Although certain points about \tex\ receive detailed explanation, many sections presuppose considerable \tex pertise, which it would be impractical to try to provide within the scope of this already lengthy manual. Despite the manual's lengthiness, the division into chapters and sections allows specialized constructions used for one part of \lamstex\ to be separated from those used in other parts. Of course, the various chapters are not completely independent, and Part~I should probably be perused by everyone. \section{\amstex\ Conventions\label{K}} We will not be analyzing the file *amstex.tex* itself, since a detailed description of \amstex\ is given in the file *amstex.doc*. Nevertheless, certain \amstex\ conventions must be mentioned here, because they are used throughout *lamstex.tex*. First of all, \amstex\ uses the ``scratch'' tokens *\next*, *\next@*, *\nextii@*, *\nextiii@*, \dots. In order to keep the number down, many definitions will, for example, define *\nextiv@* back in terms of *\next@*, *\nextii@*, etc. The *amstex.doc* file mentions the peculiar contortions that are used to avoid difficulties that might arise when a definition has a clause like ** \def\next@{ ... \next ... } ** since a previous *\futurelet\next* may have let *\next* be something that is *\outer*. In \lamstex@, on the other hand, it is simply quite out of the question to allow anything to be *\outer*. Something like *\claim* can't be outer, for example, because then things like *\newpre\claim* wouldn't work. But even something like *\bye* can't be *\outer* because it might easily occur right after a point where \lamstex\ has to subject the next token to some sort of test (see the small print notes on pages~\ref{BADOUTER} and ~\ref{BADOUTER2}). Consequently, although we will continue to reserve *\next* as the token of choice in all *\futurelet* constructions, we will finesse this whole problem by making sure that {\it nothing\/} in \lamstex\ is *\outer*. The only *\outer* things in *plain* \tex\ are the {\ninerm ASCII} form-feed~ *^^L*, the *\new...* constructions (*\newcount*, *\newdimen*, \dots\,), the *\+* from the *\settabs* construction, the *\beginsection* and *\proclaim* constructions, and *\bye*. \lamstex\ redefines *^^L*, the *\new...* constructions, *\+*, and *\bye* so that they are not *\outer*, and it makes *\beginsection* be undefined (\lamstex\ has its own system of ``heading levels''), while \amstex\ has already made *\proclaim* be undefined (until a style file is read in). It should be mentioned that although we no longer have to worry about *\next* being *\outer*, some precautions are still in order---see page~\ref{NEWPRECAUTION}. \medskip Another \amstex\ convention involves constructions like ** \if...\def\next@{\csa}\else\def\next@{\csb}\fi\next@ ** (It is assumed that the user of this manual understands why this is required instead of simply *\if...\csa\else\csb\fi* whenever *\csa* or *\csb* has an argument.) In TUG{\ninerm BOAT}, Volume~8, No.~2, Kabelschacht\pagelabel{KABEL} points out that this can be replaced by ** \if...\expandafter\csa\else\expandafter\csb\fi ** We will call this the ``K-method''; it is often used without explicit mention. (As pointed out in the *amstex.doc* file, however, this method is not always valid or practicable). \medskip Another frequently used convention of \amstex\ is ``compressed format''. We often have to make definitions of the form \setbox0\hbox{\} ** \def\cs{\futurelet\next\cs@} \def\cs@{\ifx\next"copy0% \def\next@{...\cs@@...}\else \def\next@{...\cs@@@...}\fi \next@} \def\cs@@{...} \def\cs@@@{...} ** But this uses up three new control sequence names, *\cs@*, *\cs@@*, and *\cs@@@*, just for this one construction. The ``compressed format'' uses the same names *\next@*, *\nextii@*, etc., over and over again, simply redefining them within each definition: ** \def\cs{% \def\next@{\ifx\next"copy0% \def\next@{...\nextii@...}\else \def\next@{...\nextiii@...}\fi \next@}% \def\nextii@{...}% \def\nextiii@{...}% \futurelet\next\next@} ** Notice that the ``first'' clause *\futurelet\next\next@* has to be made last (and, although it looks strange at first, it's perfectly legitimate to have *\next@* defined in terms of *\next@* in this situation). Compressed format makes things go a little slower, since *\next@*, etc., have to be redefined all the time, but seems worth it, especially since it is usually used for major formatting constructions that introduce a lot of space anyway. \medskip \amstex\ also uses the construction ** \Invalid@\controlseq ** to make *\controlseq* give an error message. As explained in *amstex.doc*, this is recommended for any control sequences (often discovered via a *\futurelet*) that function as ``syntax'' for other control sequences, and consequently shouldn't be encountered on their own. \section{Constructions from \amstex\label{AC}} Some more specific \amstex\ code should also be mentioned. \Litbox0=** \ifx\amstexloaded@\relax\catcode`\@=\active \endinput\else\let\amstexloaded@=\relax\fi ** First of all, the code $$ \vcenter{\box0} \tag "\style{\bf A}" \pagelabel{A} $$ appears near the beginning of the *amstex.tex* file. This prevents the file *amstex.tex* from being loaded twice by making *\amstexloaded@* be undefined if *amstex.tex* hasn't been loaded, but *\relax* if it has. This is a necessity because of the two lines ** \let\ic@=\/ \def\/{\unskip\ic@} "8itcorr"8 ** that occur later (compare {\it The \tex book}, pp.~382@--383). Testing *\amstexloaded@* also allows other macro packages to tell whether *amstex.tex* has already been loaded, which is important for \lamstex@, as we will see in the next chapter. \amstex\ also introduces two new counters, and a new token list, ** \newcount\count@@ \newcount\count@@@ \toksdef\toks@@=2 ** in addition to the counter *\count@* and token list *\toks@* provided by *plain.tex*; these are also used in \lamstex\ (see section~\Sref{LANDG} for the choice of~*2* in the *\toksdef*). Furthermore, \amstex\ introduces the abbreviations ** \def\FN@{\futurelet\next} "2 \def\DN@{\def\next@} \def\DNii@{\def\nextii@} "2 \def\RIfM@{\relax\ifmmode} "8RIfM"8 \def\RIfMIfI@{\relax\ifmmode\ifinner} "2 \def\setboxz@h{\setbox\z@\hbox} \def\wdz@{\wd\z@} \def\boxz@{\box\z@} ** These are used throughout \lamstex\ also. When we show *lamstex.tex* code, however, we will usually expand out these definitions, to make things easier to read. Similarly, certain control sequences from *plain*, like *\z@*, *\p@*, etc., will usually be expanded out for the sake of readability. Moreover, in constructions like ** \count@=... \let\next@=\relax ** and so forth, we will often add the optional *=* signs that are normally omitted in the code. We will frequently use the \amstex\ control sequence *\eat@* defined by ** \def\eat@#1{} ** \amstex\ introduces the token *\space@* that has been *\let* equal to a space. It is often used after ** \futurelet\next\foo ** constructions where *\foo* has to do something special if the next token is a space. In many cases, *\foo* must skip over that space, and then execute *\goo*. The standard \amstex\ way of doing this is with the code ** \ifx\next\space@\def\next@. {\goo}\else \def\next@.{\goo}\fi\next@. "8swallow"8 ** The *.* after the *\next@* makes the space `visible' to \tex. As we shall see in section~\Sref{FIX}, \lamstex\ introduces a somewhat more economical approach to this problem. By the way, a case like this, where something is part of the syntax for *\next@*, is one of the situations where the K-method would not work. \medskip There are a few more \amstex\ devices that are important in \lamstex@, but their discussion has been deferred until Chapter~\ref{CA}, since these devices are actually additions intended for later versions of \amstex. \section{Changes to \amstex; \CS{local} and \CS{global} assignments\label{LANDG}} Numerous lines of *amstex.tex* have been deleted in *amstexl.tex* because they are not used, or are modified, by *lamstex.tex*. Major changes of this sort are discussed at the appropriate points. There are also numerous small changes. For example, all *\relaxnext@*'s have been omitted, since \lamstex\ no longer needs that device for dealing with *\outer* constructions. One change was necessary to avoid a conflict: *\roman* is now used in \lamstex\ for a numbering control sequence, whereas it previously had a different (extremely unlikely) use in \amstex@, as a control sequence to be used in math to produce a roman letter. The latter has now been changed from *\roman* to *\rom*. One other change should be mentioned explicitly: near the beginning of *amstexl.tex* the definitions ** \def\height{height} "8HWD"8 \def\width{width} \def\depth{depth} ** have been inserted, so that `*height*', `*width*' and `*depth*' can be replaced by the corresponding single tokens in the specifications for various *\hrules* and *\vrules*; these replacements save even more memory space in *lamstex.tex*, where rules occur much more frequently. Finally, I have now conscientiously adhered to {\it The \tex book\/}'s recommendations (see pages~301 and~346) that assignments of variables either always be global or always be local. In most cases, the necessary changes have been minor (like changing some *\xdef*'s to *\edef*'s, or vice versa), but somewhat more extensive changes were required to ensure that *\setbox*\,{\it n} is always local for {\it n\/} even and global for {\it n\/} odd; these changes occur in the definitions of *\insplit@*, *\rendsplit@*, *\lendsplit@*, *\lmultline@@@*, *\rmultline@@@*, *\binrel@*, *\sideset@*, *\r@@t*, *\pmb@*, and perhaps one or two other places. We will exercise comparable care regarding assignments of variables\linebreak throughout \lamstex. \chapter Getting started with \lamstex\label{GETSTART}\endchapter The first thing in *lamstex.tex*, after the copyright notice, is \C** \catcode`\@=11 ** to make *@* a letter, in order to create ``private'' control sequences that the casual user cannot type, as well as to access such private control sequences from *plain.tex* and *amstexl.tex*. We will always adhere to the convention introduced here, using horizontal lines when we print actual *lamstex.tex* code, as opposed to examples, pieces of code, etc. We will often use different line breaks from the actual *lamstex.tex* code, so that it will fit better on these printed pages. (It should also be noted that when we give preliminary pieces of code we will often omit *%*~ signs at the ends of lines, although they are meticulously added when needed in the code itself.) Since \lamstex\ is not supposed to be loaded unless \amstex\ has already been loaded, the next code, \C** \ifx\amstexloaded@\relax\else \errmessage{AmS-TeX must be loaded before LamS-TeX}\fi ** produces an error message if it hasn't---see the discussion of the code ({\bf A}) on page~\ref{A}. We will adopt a different scheme for preventing *lamstex.tex* from being read in twice, one that doesn't create a new control sequence name, by using the fact that *lamstex.tex* will eventually define the control sequence *\laxread@* (see page~\ref{lax}), while \amstex\ makes certain that *\undefined* is always undefined (see *amstex.doc*): \C** \ifx\laxread@\undefined\else\catcode`\@=\active\endinput\fi ** [Other macro packages that need to know whether or not \lamstex\ has been loaded can use a similar test, or, if the status of *\undefined* isn't clear, they can use the test ** \expandafter\ifx\csname laxread@\endcsname\relax ** which is false when \lamstex\ has been loaded, but true when it hasn't been loaded, since the control sequence produced by *\csname...\endcsname* is given the value *\relax* if it hasn't already been defined.] Next we redefine *\err@* from \amstex\ to produce error messages saying `*LamS-TeX error:*' instead of `*AmS-TeX error*:' \C** \def\err@#1{\errmessage{LamS-TeX error: #1}} ** *\Err@*, which is *\err@* with \amstex's ``default help message'', will now also produce such error messages (see section~\Sref{STER}). As indicated in section~\Sref{K}, we redefine *^^L*, the *\new...* constructions, *\+*, and *\bye* from *plain* so that they are not *\outer*, \C** \def^^L{\par} \let\+=\tabalign "2 \def\newcount{\alloc@0\count\countdef\insc@unt} "2 \def\newdimen{\alloc@1\dimen\dimendef\insc@unt} "2 \def\newskip{\alloc@2\skip\skipdef\insc@unt} "2 \def\newmuskip{\alloc@3\muskip\muskipdef\@cclvi} "2 \def\newbox{\alloc@4\box\chardef\insc@unt} "2 \let\newtoks=\relax \def\newhelp#1#2{\newtoks#1#1\expandafter{\csname#2\endcsname}} \def\newtoks{\alloc@5\toks\toksdef\@cclvi} "2 \def\newread{\alloc@6\read\chardef\sixt@@n} "2 \def\newwrite{\alloc@7\write\chardef\sixt@@n} "2 \def\newfam{\alloc@8\fam\chardef\sixt@@n} "2 \def\newlanguage{\alloc@9\language\chardef\@cclvi} "2 \def\newinsert#1{\global\advance\insc@unt by\m@ne \ch@ck0\insc@unt\count \ch@ck1\insc@unt\dimen \ch@ck2\insc@unt\skip \ch@ck4\insc@unt\box "2 \allocationnumber=\insc@unt \global\chardef#1=\allocationnumber \wlog{\string#1=\string\insert\the\allocationnumber}} "2 \def\newif#1{\count@\escapechar \escapechar\m@ne \expandafter\expandafter\expandafter \edef\@if#1{true}{\let\noexpand#1=\noexpand\iftrue}% "2 \expandafter\expandafter\expandafter \edef\@if#1{false}{\let\noexpand#1=\noexpand\iffalse}% \@if#1{false}\escapechar\count@} % the condition starts out false "slip \def\bye{\par\vfill\supereject\end} ** and then we \C** \let\beginsection=\undefined ** to make *\beginsection* undefined (see *amstex.doc*). \small In *plain* \tex@, the *\let\newtoks=\relax* is inserted before the definition of *\newhelp* so that *plain.tex* can be read in twice. Even though we are not allowing *lamstex.tex* to be read in twice, this is still required, since we read it in after *plain.tex*! \endsmall \chapter Changes to \amstex\label{CA}\endchapter The next part of \lamstex\ contains various changes to \amstex@. Some of the changes should, and probably eventually will, be made in \amstex@, though they didn't get included in \amstex~version *2*, while other changes are relevant only for \lamstex. (Further changes to \amstex\ will be made later, at the relevant points.) \section{\CS{Err\@}\label{STER}} \amstex's definition of *\Err@*, {\litindent=0pt ** \def\Err@#1{\errhelp\defaulthelp@\errmessage{AmS-TeX error: #1}} ** has been deleted in *amstexl.tex*,} because it can obviously be shortened to \C** \def\Err@#1{\errhelp\defaulthelp@\err@{#1}} ** \section{\CS{atdef\@}\label{ATDEF}} \amstex's original mechanism for defining the active *@* character can be improved considerably. First of all, we want the active *@* to mean ** \futurelet\next\at@ ** the problem being that we need to make this definition while *@* is active ** {\catcode`\@=\active \def@{...} ** even though we want to allow *@* as part of the control sequence name *\at@*. Now we can easily name *\at@* even when *@* is active, as ** \csname at\string@\endcsname ** Of course, we can't simply say ** \def@{\futurelet\next\csname at\string@\endcsname} ** since *@* would then simply *\let\next=a*---we need to have the combination *\csname...\endcsname* expanded out before the *\futurelet\next* takes effect. We could do this with ** \def@{\def\next{\futurelet\next}\expandafter\next \csname at\string@\endcsname} ** but that's somewhat unsatisfactory, since it requires the active~ *@* to make a subsidiary definition each time it is used. Another possibility is to use the triple *\expandafter* trick (see {\it The \tex book}, page~374), which we will be using later on. But for the present problem the simplest strategy is to use the code ** \edef\next{\gdef\noexpand@{\futurelet\noexpand\next "8ATD"8 \csname at\string@\endcsname} \next ** Here the *\edef* makes *\next* mean ** "9\gdef"9","9@"9{"9\futurelet"9","9\next"9\at@} ** where the boxed control sequences are not expanded out either because they are primitives or because they are preceded by *\noexpand*; the *\csname..."allowbreak"0\endcsname* {\it is\/} expanded out in the *\edef*, but the control sequence *\at@* that it expands to is made equal to *\relax*, since *\at@* hasn't been defined previously---so *\at@* isn't expanded further in the *\edef*. (Here we are using the fact that *lamstex.tex* won't be read in twice [Chapter~\ref{GETSTART}].) Consequently, when we then call *\next* we get this *\gdef*. Thus, to get the desired definition of the active *@* we just need \C** {\catcode`\@=\active \edef\next{\gdef\noexpand@{\futurelet\noexpand\next \csname at\string@\endcsname}} \next } ** The definition of *\at@* itself is now easy, with *@* back as a letter. We will call the very same routine, *\at@@*, when the next token is a letter, other character, or control sequence (or active character); for any other type of token we will call *\at@@@*, which will be an error message: \C** \def\at@{% \ifcat\noexpand\next a\let\next@=\at@@\else \ifcat\noexpand\next0\let\next@=\at@@\else \ifcat\noexpand\next\relax\let\next@=\at@@\else \let\next@=\at@@@\fi\fi\fi\next@} ** The error message *\at@@@* is simply \C** \def\at@@@{\errhelp\athelp@\err@{Invalid use of @}} ** using the help message *\athelp@* from \amstex@. On the other hand, *\at@@* \ will simply be the control sequence \setbox0\hbox{\} \setbox1\hbox{`}\setbox2\hbox{'} ** "box1"0\"box0@at"box2 ** if it has been defined, or an error message otherwise. Here we put quotes around *\*\*@at* to emphasize that it is a single control word, even when \ isn't a letter;\pagelabel{QTS1} in practice, of course, such control words have to be constructed using *\csname"allowbreak..."allowbreak\endcsname*: \C** \def\at@@#1{\expandafter \ifx\csname\string#1@at\endcsname\relax \let\next@=\at@@@ \else \def\next@{\csname\string#1@at\endcsname}% \fi \next@} ** Note that we use *\string#1* so that the token *#1* can be a control sequence or active character, as well as a letter or other character. Finally, *\atdef@*, the mechanism for defining the value of the active *@* on various tokens, is the same as in \amstex@, except that we add a *\string*: \C** \def\atdef@#1{\expandafter\def\csname\string#1@at\endcsname} ** There are two noteworthy things about this redefinition: \list \item The original definition of *\atdef@* remains in *amstexl.tex*, because it is used for ** \atdef@;{...} \atdef@,{...} \atdef@:{...} \atdef@!{...} \atdef@?{...} \atdef@.{...} \atdef@-{...} ** These *\atdef@*'s give the same results as the new *\atdef@* would give, since for these characters the *\string* is simply redundant---once *@* is active, *@;* and *@:* and so forth will work just as before. (The *\atdef@@\vert* is irrelevant: it is deleted in *amstexl.tex*, and not used in \lamstex@.) \item Later in \lamstex\ we are going to make *""* active. Nevertheless, \amstex's ** \atdef@"" ** will still make the combination *@""* work correctly,\pagelabel{SW} because *\string""* for *""* active gives the same result as \amstex's *""* when *""* is not active. Actually, we are going to give a new *\atdef@""* (section~\Sref{FIX}), but that will be done before *""* is made active, so the same principle still applies. \endlist \section{Tests\label{TESTS}} \amstex\ has the flag *\ifin@*, which is set only by the routine *\in@*, a test to determine whether a particular token appears in any sequence; this test, in turn, is used only by the routine *\tagin@* to check for the presence of *\tag* in a sequence. \lamstex\ has numerous tests that are always performed independently, so it is economical to have a single flag that will be used by all of them; this flag will also replace *\ifin@* from \amstex@. So *amstexl.tex* deletes the line ** \newif\ifin@ ** and in \lamstex\ we introduce the flag \C** \newif\iftest@ ** Moreover, in *amstexl.tex* we also delete ** \def\in@#1#2{ . . . } \def\tagin@#1{ . . . } ** while in \lamstex\ we redefine the *\tagin@* routine so that instead of using *\ifin@* it merely reproduces (an equivalent of) the definition: \C** \def\tagin@#1{\tagin@false \def\next@##1\tag##2##3\next@{\test@true \ifx\tagin@##2\test@false\fi}% \next@#1\tag\tagin@\next@ \tagin@false\iftest@\tagin@true\fi} ** \section{Spaces after control sequence names in error messages\label{SACN}} Numerous error messages in \amstex\ use constructions of the form ** . . . \string\controlseq\space . . . ** to get a space after the control sequence *\controlseq* in the error message. However, it saves one token to instead use ** . . . \noexpand\controlseq . . . ** ---the *\noexpand* prevents expansion of *\controlseq* in the error message, but we still get a space after *\controlseq*. This device is used throughout \lamstex@, and the requisite changes were also made directly in *amstexl.tex*, since they were so minor. \small The first change occurs in the definition of *\define@* where ** \err@{\string\define\space must be . . . } ** is replaced by ** \err@{\noexpand\define must be . . . } ** The next occurs in the definition ** \define@@#1 ** which takes a control sequence *#1* as its argument, where, for example \Litbox0=** \err@{\string#1 is already defined} ** $$\box0\tag"\style{\bf A}"$$ is changed to ** \err@{\noexpand#1is already defined} ** with no space before the `*is*', since it will appear when the error message is given. [*amstex.tex* actually has ** \err@{\string#1\space is already defined} ** which is unnecessary complicated, though it has the same number of tokens as ({\bf A}).] Similar replacements are made in the definitions of ** \vmodeerr@#1 \mathmodeerr@#1 "2 \dmatherr@#1 "2 \nondmatherr@#1 "2 \onlydmatherr@#1 "2 \nonmatherr@#1 "2 \nonvmodeerr@#1 \textonlyfont@#1#2 ** Finally, in the definition of *\boldkey* (which is actually redefined in *lamstex.tex*---see page~\ref{NEWBK}) the ** \Err@{\string\boldkey\space can't ...} ** is replaced by ** \Err@{\noexpand\boldkey can't ...} ** with a similar change for *\boldsymbol*. Note, by the way, that these substitutions cannot be made in ** \newhelp\athelp \newhelp\defahelp ** which end up putting things inside *\csname...\endcsname*. \endsmall \section{Line breaking\label{LBK}} The original \amstex\ definition of *\nolinebreak* had an extra element *\refskip@*, which was initially *\relax*, but which was changed for the bibliography. In version~*2*, that aspect of the bibliography macros (sections~\Sref{BASICD} and ~\Sref{BIBMACS}), as well as the indexing macros (section~\Sref{IPAGE}), has been improved. Consequently, the four control sequences ** \nolinebreak \allowlinebreak \linebreak \newline ** will all have something added; it will suffice to add the same thing, which we will call *\lkerns@*, to the first three, and something that we will call *\nkerns@* to the fourth; like the old *\refskip@*, these are both initially *\relax*. The definitions of these four line-breaking macros are deleted in *amstex.tex*, and we now add the new definitions. They differ from the original definitions (see *amstex.doc*) in the inclusion of *\lkerns@* and *\nkerns@*; however, these do not occur in quite the place that *\refskip@* occurred in the original definition of *\nolinebreak*---that was, in fact, incorrect. We have also simplified the definition of *\newline*---as indicated in *amstex.doc*, the case of *\newline\par* really isn't worth worrying about. \C** \let\lkerns@=\relax "slip \def\nolinebreak{\relax \ifmathmode@ \mathmodeerr@\nolinebreak\else "2 \ifhmode \saveskip@=\lastskip \unskip \nobreak \ifdim\saveskip@ > 0pt \hskip\saveskip@\fi \lkerns@ \else\vmodeerr@\nolinebreak\fi\fi} "2 \def\allowlinebreak{\relax \ifmathmode@ \mathmodeerr@\allowlinebreak\else "2 \ifhmode \saveskip@=\lastskip \unskip \allowbreak \ifdim\saveskip@ > 0pt \hskip\saveskip@\fi \lkerns@ \else\vmodeerr@\allowlinebreak\fi\fi} \def\linebreak{\relax \ifmathmode \mathmodeerr@\linebreak\else "2 \ifhmode \unskip\unkern\break\lkerns@ \else\vmodeerr@\linebreak\fi\fi} "slip \let\nkerns@=\relax "slip \def\newline{\relax \ifmathmode \mathmodeerr@\newline\else "2 \ifhmode \unskip\unkern\null\hfill\break\nkerns@ \else\vmodeerr@\newline\fi\fi}% ** \section{\CS{alloc\@\@}, \CS{newcount\@}, and \CS{newbox\@}\label{NEWNEW}} Near the beginning of \amstex@, *\alloc@* is redefined so that it doesn't write anything to the *.log* file, while the original definition is reinstated at the end. This means that the *\new...* constructions used to create new counters, \ registers, etc., within the file do not write anything to the *.log* file. After redefining *\alloc@*, \amstex\ also uses ** \let\alloc@@=\alloc@ ** to make *\alloc@@* {\it that\/} version of *\alloc@*, even if called after *\alloc@* has been redefined at the end. In the definition of *\loadmsam*, for example, the code ** . . . \alloc@@8\fam\chardef\sixt@@n\msafam . . . ** functions as a replacement for *\newfam\msafam*; this not only gets around the problem that *\newfam* is still *\outer* in \amstex@, it also ensures that nothing gets written to the *.log* file even if *\loadmsam* is used after *\alloc@* is restored to its old definition. In the definition of *\accentedsymbol*, however, a non-outer *\newbox* is needed, because it appears in a construction like ** \expandafter\newbox\csname ... \endcsname ** where we can't simply use the code for *\newbox*. So *amstex.tex* used *\newbox@* as a non-outer version of *\newbox*. However, the \amstex\ definition, ** \def\newbox@{\alloc@4\box\chardef\insc@unt} ** because it used *\alloc@* rather than *\alloc@@*, wasn't really the right choice anyway. In \lamstex\ we rectify this situation. First of all, the definitions of *\newbox@* and *\accentedsymbol* are deleted in *amstexl.tex*. Then in *lamstex.tex*, we \C** \def\newbox@{\alloc@@4\box\chardef\insc@unt} ** and also, for later use, \C** \def\newcount@{\alloc@@0\count\countdef\insc@unt} ** Then we redefine *\accentedsymbol* using *\newbox@*. In addition, the combination ** \expandafter\eat@\string ** in the definition is replaced by *\exstring@*, since we will introduce this as an abbreviation for that combination later (section~\Sref{CNT}) (we also take the opportunity to eliminate two unnecessary *\expandafter*'s from the original code): \C** \def\accentedsymbol#1#2{\expandafter \newbox@\csname\exstring@#1@box\endcsname \setbox\csname\exstring@#1@box\endcsname =\hbox{$\m@th#2$}% \define#1{\copy\csname\exstring@#1@box\endcsname{}}} ** \section{Lists\label{LISTS}} \amstex\ has two lists, *\alloclist@*, and *\fontlist@*, of the type introduced in {\it The \tex book}, page~378, and it defines *\rightappend@*, which is like *\rightappenditem* from that page. *\alloclist@* is maintained for the *\showallocations* command, which is really only for the use of \tex nicians, and \lamstex\ dispenses with this feature, in order to save space; consequently, in *amstexl.tex* all material related to *\alloclist@* is deleted. *\fontlist@* is used for the *\syntax* command, which \lamstex\ retains, but for this list of control sequence names it is more efficient to use a list of the type described on page~379 of {\it The \tex book}: \setbox0\hbox{name$_{\text1}$}% \setbox1\hbox{name$_{\text2}$}% ** \\\"box0\\\"box1... ** Several other lists of this sort will be used in \lamstex@; in some cases, we will even have a list of the form *\\_",_",_\\_",_",_ ...*, where *_",_",_* are not control sequence names. In \lamstex\ we will still be using *\rightappend@* on occasion, but we will also define the routine *\rightadd@#1\to#2* to add *#1* to one of these simpler lists *#2*: \C** \def\rightadd@#1\to#2{\toks@={\\#1}\toks@@=\expandafter{#2}% "8atl"8 \xdef#2{\the\toks@@\the\toks@}% \toks@={}\toks@@={}} ** In *amstexl.tex*, the definition of *\fontlist@* is deleted, and in \lamstex\ we instead define *\fontlist@* to be a list of this simpler sort: \C** \def\fontlist@{\\\tenrm\\\sevenrm\\\fiverm\\\teni\\\seveni \\\fivei\\\tensy\\\sevensy\\\fivesy\\\tenex\\\tenbf \\\sevenbf\\\fivebf\\\tensl\\\tenit} ** Similarly, the definition of *\font@* is deleted in *amstexl.tex* and replaced in \lamstex\ by \C** \def\font@#1=#2 {\rightadd@#1\to\fontlist@\font#1=#2 } "8FL"8 ** (Although *\font@* appears in *amstexl.tex*, it occurs only within other definitions, so the definition can be deferred to *lamstex.tex*.) Although, as mentioned in section~\Sref{TESTS}, the test *\ifin@* has been deleted in \linebreak *amstexl.tex*, for a list *#1* of control sequences, \setbox0\hbox{name$_{\text1}$}% \setbox1\hbox{name$_{\text2}$}% ** \\\"box0\\\"box1... ** we will need another (simpler) test to determine whether a control sequence *#2* is in the list. Basically we want to use ** "hskip-30pt\def\ismember@#1#2{\test@false\let\next@=#2% "hskip-30pt {\def\\##1{\let\nextii@=##1\ifx\nextii@\next@\global\test@true\fi}#1}% ** But since we normally set the flag *\iftest@* only locally, we don't want to use a *\global\test@true* in this one situation (compare section~\Sref{LANDG}). So instead we will use a new scratch token, *\Next@*, for which we will always use *\global* assignments. In addition, there are two further details that we will explore in a moment: \pagelabel{GNXT} \C** \def\ismember@#1#2{"9\global\let\Next@=F"9\let\next@"9=" "9#2% "8ismember"8 {\def\\##1{\let\nextii@=##1\ifx\nextii@\next@ "9\global\let\Next@=T"9\fi}#1}% \test@false\ifx\Next@ T\test@true\fi"9\let\next@=\relax"9} ** This test may compared with the test on page~379 of {\it The \tex book}. Using *\let\next@* instead of *\def\next@* allows the test ** \ismember@#1\next ** to be used after *\next* has been *\let* equal to some control sequence by a *\futurelet*\pagelabel{ISFL}, which will be important in section~\Sref{ZZZ} (on some occasions we will also be using *\ismember@#1#2* when *#2* is an explicit argument). But two precautions are then in order: \bullist \item\pagelabel{NEEDEQ} Although actual *lamstex.tex* code usually omits *=* signs after *\let*'s, in the above code we need both the boxed *=* sign {\it and the space after it\/}! Reason: Our *\futurelet* may have *\let\next* be a space token, which is thus a \ in the notation of {\it The \tex book}, page~269. According to the syntax rule on page~277, one such space (but only one) will be ignored after the *=* sign; if we simply had ** \let\next@=#2{\def\\ ... ** then the \ *#2* would be ignored, and *\next@* would end up being the *{*, which would then disappear, causing infinite confusion later on. \item Another important precaution is the *\let\next@=\relax* at the end. That is needed because the *\futurelet\next* may have *\let\next* equal something equivalent to *\iftrue* or *\iffalse*, so that *\next@* would then also be equal to *\iftrue* or *\iffalse*. If that situation were allowed to continue, havoc might ensue the next time we used a macro containing *\next@* within it. \endlist As an example of this latter phenomenon,\pagelabel{NEWPRECAUTION} note that the original *plain* \tex\ definitions ** "hskip-30pt{\catcode`\'=\active \gdef'{^\bgroup\prim@s}} "hskip-30pt\def\prim@s{\prime\futurelet\next\prim@m@s} "hskip-30pt\def\pr@m@s{\ifx'\next\let\next\pr@@@s \else\ifx^\next\let\next\pr@@@t "hskip-30pt \else\let\next\egroup\fi\fi\next} "hskip-30pt . . . ** later had to be modified by changing *\pr@m@s* to ** "hskip-30pt\def\pr@m@s{\ifx'\next\let\nxt\pr@@@s \else\ifx^\next\let\nxt\pr@@@t "hskip-30pt \else\let\nxt\egroup\fi\fi\nxt} ** For example, we might have ** $a'\iffirstset x\else y \fi$ ** where *\iffirstset* is some user-defined construction, and then the\linebreak *\futurelet\next* in *\prim@s* would *\let\next=\iffirstset*. Note that the appearance of *\next* after an *\ifx* test causes no problem, but its appearance within an *\if...* clause, even following a *\let* or *\def*, would make things go haywire. Similarly, in the definition of *\ismember@*, the *\next@* appears in safe places. To avoid such problems in general, after any *\futurelet* we will use only the token *\next*, and otherwise *\next* will not appear in any macros except after *\ifx* tests (or *\ifcat* tests). One definition in *amstex.tex* requires modification to adhere to this rule: the definition ** "hskip-30pt\gdef\comment@@@#1\comment@@@{\ifx\next\comment@@@\let\next\comment@ "hskip-30pt \else\def\next{\oldcodes@\endlinechar=`\^^M\relax}% "hskip-30pt \fi\next} ** has been changed in *amstexl.tex* to ** "hskip-30pt\gdef\comment@@@#1\comment@@@{\ifx\next\comment@@@\let\next@\comment@ "hskip-30pt \else\def\next@{\oldcodes@\endlinechar=`\^^M\relax}% "hskip-30pt \fi\next@} ** \section{Skipping spaces in \CS{futurelet}'s\label{FIX}} On page~\ref{swallow} we mentioned \amstex's device for skipping over space tokens in *\futurelet* constructions. Instead of using this device directly, which requires somewhat long definitions each time, \lamstex\ uses a special ``futurelet-next-skipping-spaces'' construction ** \FNSS@\foo ** which is like *\futurelet\next\foo*, except that any space tokens after *\foo* will be discarded, and *\foo* will be applied after *\next* has been *\let* equal to the first non-space token after *\foo*. *\FNSS@#1* begins by storing *#1* in *\FNSS@@*, and then applies a *\futurelet\next* construction, calling *\FNSS@@@*, which then does the checking for a space: \C** \def\FNSS@#1{\let\FNSS@@=#1\futurelet\next\FNSS@@@} "slip \def\FNSS@@@{\ifx\next\space@ \def\FNSS@@@@. {\futurelet\next\FNSS@@@}\else \def\FNSS@@@@.{\FNSS@@}\fi \FNSS@@@@.} ** Thus, when *\next* happens to be a space, we swallow the space and call the routine *\futurelet\next\FNSS@@@* again, to get the next non-space token; when we do get a non-space token, we simply apply *\FNSS@@*, the argument of *\FNSS@*. \medskip The \amstex\ *\atdef@""* (see *amstex.doc*) is deleted from *amstexl.tex*, because it can be shortened if we use this *\FNSS@* to get the first non-space token after *@""*: \C** \atdef@""{\unskip "2 \def\next@{\ifx\next`\def\next@`{\futurelet\next\nextii@}% \else\ifx\next\lq\def\next@\lq{\futurelet\next\nextii@}% \else\def\next@####1{\futurelet\next\nextiii@}\fi\fi \next@}% "2 \def\nextii@{\ifx\next`\def\next@`{\sldl@``}% \else\ifx\next\lq\def\next@\lq{\sldl@``}% \else\def\next@{\dlsl@`}\fi\fi\next@}% "2 \def\nextiii@{\ifx\next'\def\next@'{\srdr@''}% \else\ifx\next\rq\def\next@\rq{\srdr@''}% \else\def\next@{\drsr@'}\fi\fi\next@}% \FNSS@\next@} ** [In our definition of *\FNSS@@@* we used a new control sequence *\FNSS@@@@* instead of using a scratch token like *\next@* to allow the use of *\FNSS@* in such ``compressed format'' definitions.] There is only one other definition in \amstex\ that is (deleted from\linebreak *amstexl.tex* and) shortened using *\FNSS@*: \C** \def\root{% \def\next@{\ifx\next\uproot\let\next@=\nextii@\else "2 \ifx\next\leftroot\let\next@=\nextiii@\else "2 \let\next@=\plainroot@\fi\fi\next@}% "2 \def\nextii@\uproot##1{\uproot@##1\relax\FNSS@\nextiv@}% "2 \def\nextiv@{\ifx\next\leftroot\let\next@=\nextv@\else \let\next@=\plainroot@\fi\next@}% "2 \def\nextv@\leftroot##1{\leftroot@##1\relax\plainroot@}% "2 \def\nextiii@\leftroot##1{\leftroot@##1\relax \FNSS@\nextvi@}% "2 \def\nextvi@{\ifx\next\uproot\let\next@=\nextvii@\else \let\next@=\plainroot@\fi\next@}% "2 \def\nextvii@\uproot##1{\uproot@##1\relax\plainroot@}% \bgroup\uproot@\z@\leftroot@\z@ \FNSS@\next@} ** However, there are numerous places in \lamstex\ where the use of *\FNSS@* will similarly save space, definitely make the extra tokens used for *\FNSS@* worth while. Note, moreover, that the above definition of *\root* not only saves space, but also avoids introducing the scratch tokens *\nextviii@* and *\nextix@* that occur in the original definition, but which occur nowhere else in \amstex. \section{\CS{loop}\label{LOOP}} The same article that introduced the ``K-method'' (page~\ref{KABEL}) also introduced a new definition of *plain* \tex's *\loop...\repeat* mechanism, which we will use in \lamstex: \C** \def\loop#1\repeat{% \def\iterate{#1\relax\expandafter\iterate\fi}% \iterate\let\iterate=\relax} ** This has the property that it allows constructions like ** \loop _",_",_ \if... _",_",_ \else _",_",_ \repeat ** where we repeat when the *\if...* test is {\it false\/} rather than true, which will turn out to be useful at several points in \lamstex@. \footnote{The alternative is to use *\if... \test@false \else \test@true \fi \iftest@ _",_",_ \repeat*.} Corresponding to this redefinition of *\loop*, we add a new definition of *\gloop@* (which is used by the *\cfrac* construction): \C** \def\gloop@#1\repeat{% \gdef\iterate@{#1\relax\expandafter\iterate@\fi}% \iterate@\global\let\iterate@=\relax} ** The main purpose of redefining *\gloop@* is so that it will use *\iterate@* rather than *\iterate*, so that *\iterate* will only be given a local definition, not both a local and a global one (compare section~\Sref{LANDG}). In *amstexl.tex* we delete the line ** \newif\ifbadans@ ** and the definition of *\printoptions*, because we can define *\printoptions* in terms of *\iftest@* (section~\Sref{TESTS}). We've deferred this redefinition until now because it is also a little bit more convenient to use this alternative *\loop* test: \C** \def\printoptions{\W@{Do you want S(yntax check), G(alleys) or P(ages)?^^JType S, G or P, follow by : }% "2 \loop \read -1 to\ans@ \edef\next@{\def\noexpand\Ans@{\ans@}}% \uppercase\expandafter{\next@}% \ifx\Ans@\S@\global\test@true\syntax\else \ifx\Ans@\G@\global\test@true\galleys\else \ifx\Ans@\P@\global\test@true\else \global\test@false\fi\fi\fi \iftest@ \else \W@{Type S, G or P, follow by : }% \repeat} ** \section{A l\'a fran\c cais\label{ALF}}Finally, certain changes are made to *amstex.tex* to accommodate French styles that make some or all of *;* and *:* and *!* and *?* into active characters. In this case, various \amstex\ macros that involve tests like ** \ifx\next! ** need to be changed. \footnote{In addition, in *lamstex.tex* we have the problem that certain control sequences, notably those for commutative diagrams and tables, use certain punctuation as part of their {\it syntax}. For example, the *\ds* option for arrows in a commutative diagram is typed in the form *\ds(h;v)* (see page~155 of the \lamstex~Manual). If we make a definition like *\def\ds(#1;#2){...}*, then \tex\ incorporates a type~*12*~ *;* as part of the syntax for *\ds*. So in a document where *;* is active, the *;* that the user types will not be recognized as the proper syntax element. The devices for handling this problem will be discussed at the appropriate time (in Volume~2).} Our goal is to allow all necessary changes to be indicated in a reasonably short file, say *french.tex*, which can be read in {\it after\/} *lamstex.tex* (so that it can be loaded on top of a \lamstex\ format file). The most reasonable approach is to have certain control sequences that have been *\let* equal to the active punctuation symbols, so that we can use these in an *\ifx\next...* test. Of course, this can only be done after the active punctuation symbols have been defined, not in *amstexl.tex*. To get around this problem, we will insist that the definition of the active punctuation symbols in *french.tex* are not made with an ordinary *\def*, but with a special *\APdef*, which will manage things properly for us. For the control sequences that will be *\let* equal to the active punctuation symbols we will use the following control words (which have to be created using *\csname...\endcsname*):~ `*\A@;*' and `*\A@:*' and `*\A@?*' and ~`*\A@!'*. We begin by assigning the non-active characters as default values: \C** \expandafter\let\csname A@;\endcsname=; \expandafter\let\csname A@:\endcsname=: \expandafter\let\csname A@?\endcsname=? \expandafter\let\csname A@!\endcsname=! ** When ** \APdef:{ ... } ** appears in *french.tex*, we want to (1)~\hbox{*\def:{ ... }*} and (2)~*\let*`*\A@:*'*=:* (here,\pagelabel{QTS2} as on page~\ref{QTS1}, we put quotes around *\A@:* to emphasize that it is a single control word; in actual code something like ** \expandafter\let\csname A@\string:\endcsname=: ** will be needed). To achieve this, we use \C** \def\APdef#1{\def\next@{\expandafter \let\csname A@\string#1\endcsname=#1}% \afterassignment\next@\def#1} ** Thus, for example, *\APdef:* defines *\next@* to mean \setbox0\hbox{`*\A@:*'} ** \let"box0=: ** and after the assignment *\def:*, which swallows the following \hbox{*{ ... }*}, we perform *\next@*, so that `*\A@:*' has now been *\let* equal to the active~*:*\,. So, we can produce control sequences that have the value of each active punctuation symbol that may occur in a file, assuming that *\APdef* has always been used in *french.tex*. Now consider the original \amstex\ definition ** \def\tdots@{\unskip \def\next@{$\m@th\mathinner{\ldotp\ldotp\ldotp}\, "2 \ifx\next,\,$\else\ifx\next.\,$\else "2 \ifx\next;\,$\else "2 \ifx\next:\,$\else "2 \ifx\next?\,$\else "2 \ifx\next!\,$\else $ \fi\fi\fi\fi\fi\fi}% \ \futurelet\next\next@} ** We want to supplement this with *\ifx\next* tests that check whether *\next* is an active punctuation symbol, in case *\tdots@* gets used in a file where that is the case. We can do this with tests like ** \expandafter\ifx\csname A@\string;\endcsname\next ** For greater flexibility, when we encounter an active punctuation symbol we will not necessarily insert the extra *\,* that the non-active symbol gets; we will instead insert *\fextra@*, which by default is \C** \let\fextra@=\, ** (but which *french.tex* can redefine, if desired). The original definition of *\tdots@* is deleted from *amstexl.tex*, and a new definition is given in *lamstex.tex*: \C** \def\tdots@{\unskip \def\next@{$\m@th\mathinner{\ldotp\ldotp\ldotp}\, \ifx\next,\,$\else\ifx\next.\,$\else "2 \ifx\next;\,$\else "9\expandafter\ifx\csname A@\string;\endcsname\next"9 "9\fextra@$\else"9 "2 \ifx\next:\,$\else "9\expandafter\ifx\csname A@\string:\endcsname\next"9 "9\fextra@$\else"9 "2 \ifx\next?\,$\else "9\expandafter\ifx\csname A@\string?\endcsname\next"9 "9\fextra@$\else"9 "2 \ifx\next!\,$\else "9\expandafter\ifx\csname A@\string!\endcsname\next"9 "9\fextra@$\else"9 $ \fi\fi\fi\fi\fi\fi"9\fi\fi\fi\fi"9}% \ \futurelet\next\next@} ** Similarly, the definition of *\extrap@* is deleted from *amstexl.tex*, and in *lamstex.tex* we add \C** \def\extrap@#1{% \ifx\next,\def\next@{#1\,}\else "2 \ifx\next;\def\next@{#1\,}\else "9\expandafter\ifx\csname A@\string;\endcsname\next"9 "9\def\next@{#1\fextra@}\else"9 "2 \ifx\next.\def\next@{#1\,}\else\extra@ \ifextra@\def\next@{#1\,}\else \let\next@"9="9#1\fi\fi\fi\fi"9\fi"9\next@} ** The boxed *=* sign in this code is another case (compare page~\ref{NEEDEQ}) where the *=*~ sign cannot be omitted, though now the reason is different: we might have *#1* being an *=* sign! There are similar required *=*'s in the (original) definitions of *\boldsymbol@*, *\boldkeydots@*, and *\boldsymboldots@*, and in the new definition of *\boldkey*, to follow. These *=* signs and the one in the definition of *\ismember@* are the only ones that actually occur after a *\let* in *lamstex.tex*. Similarly, the definitions of *\dotsc* and *\keybin@* are deleted from \linebreak *amstexl.tex*, and in *lamstex.tex* we add \C** \def\dotsc{\def\next@{\ifx\next;\plainldots@\,\else "9\expandafter\ifx\csname A@\string;\endcsname\next"9 "9\plainldots@\fextra@\else"9 \ifx\next.\plainldots@\,\else\extra@\plainldots@ \ifextra@\,\fi\fi\fi"9\fi"9}% \futurelet\next\next@} "slip \def\keybin@{\keybin@true \ifx\next+\else\ifx\next=\else\ifx\next<\else \ifx\next>\else\ifx\next-\else\ifx\next"1\else \ifx\next:\else "9\expandafter\ifx\csname A@\string;\endcsname\next\else"9 \keybin@false\fi\fi\fi\fi\fi\fi\fi"9\fi"9} ** The definition of *\boldkey#1* is also deleted from *amstexl.tex*.\pagelabel{NEWBK} In this case, where we have a control sequence with an argument, rather than one that has picked up the next token with a *\futurelet\next*, and where our tests are of the form ** \ifx#1! ** we have to start with ** \let\next=#1 ** so that we can then use things like ** \expandafter\ifx\csname A@\string!\endcsname\next ** to check for an active~*!*. The new code is: \C** \def\boldkey#1{\ifcat\noexpand#1A% \ifcmmibloaded@{\fam\cmmibfam#1}\else \Err@{First bold symbol font not loaded}\fi "2 \else \let\next"9="9#1% "2 \ifx#1!\mathchar""5\bffam@21 \else "9\expandafter\ifx\csname A@\string!\endcsname\next"9 "9\mathchar""5\bffam@21 \else"9 "2 \ifx#1(\mathchar""4\bffam@28 \else "2 \ifx#1)\mathchar""5\bffam@29 \else "2 \ifx#1+\mathchar""2\bffam@2B \else "2 \ifx#1:\mathchar""3\bffam@3A \else "9\expandafter\ifx\csname A@\string:\endcsname\next"9 "9\mathchar""3\bffam@3A \else"9 "2 \ifx#1;\mathchar""6\bffam@3B \else "9\expandafter\ifx\csname A@\string;\endcsname\next"9 "9\mathchar""6\bffam@3B \else"9 "2 \ifx#1=\mathchar""3\bffam@3D \else "2 \ifx#1?\mathchar""5\bffam@3F \else "9\expandafter\ifx\csname A@\string?\endcsname\next"9 "9\mathchar""5\bffam@3F \else"9 "2 \ifx#1[\mathchar""4\bffam@5B \else "2 \ifx#1]\mathchar""5\bffam@5D \else "2 \ifx#1,\mathchari@63B \else "2 \ifx#1-\mathcharii@200 \else "2 \ifx#1.\mathchari@03A \else "2 \ifx#1/\mathchari@03D \else "2 \ifx#1<\mathchari@33C \else "2 \ifx#1>\mathchari@33E \else "2 \ifx#1"1\mathcharii@203 \else "2 \ifx#1|\mathcharii@06A \else "2 \ifx#10\bold0\else\ifx#11\bold1\else\ifx#12\bold2\else "2 \ifx#13\bold3\else\ifx#14\bold4\else\ifx#15\bold5\else "2 \ifx#16\bold6\else\ifx#17\bold7\else\ifx#18\bold8\else "2 \ifx#19\bold9\else \Err@{\noexpand\boldkey can't be used with #1}% \fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi \fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi"9\fi\fi\fi\fi"9} ** \chapter Numbering styles\endchapter Next we define the standard \lamstex\ numbering styles: *\arabic*, *\alph*, *\Alph*, *\roman*, *\Roman*, and *\fnsymbol*. These numbering styles are meant to be applied to a {\it number}, not to a counter (in the \lamstex\ macros they are usually applied to *\number*\ for some \). *\arabic* is consequently quite trivial: \C** \def\arabic#1{#1} ** The definition of *\alph#1* uses the fact that the lower-case letters a--z have positions *97*--*122* (if some one were to design a perverse font for which this wasn't true, then *\alph* would have to be defined differently). *\alph* begins by setting the scratch counter *\count@* to the value of *#1*: ** \def\alph#1{\count@=#1\relax ** We always add the *\relax* as a precaution in such situations, even if it may not be strictly necessary ({\it The \tex book}, page~208, recommends a space, but *\relax* seems best, to avoid any anomalous situations where an unwanted space might somehow intrude itself into the token stream). Then we add *96* to the value of *\count@* and we print the character \linebreak *\char\count@*, but we give an error message if this would give us a character past *122* (this could easily happen if some one caused the counter to be augmented more than 25~times): \C** \def\alph#1{\count@=#1\relax\advance\count@ by 96 \ifnum\count@>122 \Err@{\noexpand\alph invalid for numbers > 26}\else\char\count@\fi} ** (See section~\Sref{SACN} for the use of *\noexpand\alph* rather than *\string\alph*.) We don't bother giving an error message if *\count@* ends up having a value less than *97*, since reasonable macros will normally start the counter for a particular construction at *1*. (So there is the possibility that some one will perversely *\Reset* some construction to *0*, say, and then have *\alph* produce a left quotation mark *`*! If this really seems bothersome, another check can easily be added to the code.) The definition \C** \def\Alph#1{\count@=#1\relax\advance\count@ by 64 \ifnum\count@>90 \Err@{\noexpand\Alph invalid for numbers > 26}\else\char\count@\fi} ** is exactly analogous, using the fact that the letters A--Z should have the positions *65*--*90*. For *\roman* we just have to \C** \def\roman#1{\romannumeral#1\relax} ** since \tex\ provides the *\romannumeral* primitive. (The *\relax* is essential here; otherwise something like *\roman{10}3* would be expanded into *\romannumeral103*, giving the result `ciii'.) But *\Roman* is more complicated, \C** \def\Roman#1{\uppercase\expandafter{\romannumeral#1}} ** Here the *\expandafter* causes expansion after the left brace *{*. Without the *\expandafter*, something like *\Roman3* would become ** \uppercase{\romannumeral3} ** which is just *\romannumeral3*, and thus `iii'. With the *\expandafter*, we get \setbox0\hbox{\} ** \uppercase{"box0} ** i.e., *\uppercase{iii}*, or `III'. Note that we don't need to add *\relax* after the *#1* in his definition, because *\romannumeral#1* is expanded within the *\uppercase{...}*. Finally, we have the *\fnsymbol* numbering style. As the code below shows, *\fnsymbol#1* successively sets \Reset\tag1 $$\align\hbox{\tt\bs count\@\@}&=\hbox{\tt\#1}\tag\\ \hbox{\tt\bs count\@}&=\left[\frac{\hbox{\tt\#1}-\hbox{1}}{\hbox{7}} \right]\tag\\ \hbox{\tt\bs count\@\@\@}&=\left[\frac{\hbox{\tt\#1}-\hbox{1}}{\hbox{7}} \right]+\hbox{1}\tag\\ \hbox{\tt\bs count\@\@}&=\hbox{\tt\#1}-\hbox{7} \cdot\left[\frac{\hbox{\tt\#1}-\hbox{1}}{\hbox{7}} \right].\tag\endalign$$ Hence $$ \hbox{\tt\bs count\@\@}=\cases \hbox{1}&\hbox{for \tt\#1}=\hbox{1}, \hbox{8}, \hbox{15},\dotsc{}\\ \hbox{2}&\hbox{for \tt\#1}=\hbox{2}, \hbox{9}, \hbox{16}, \dotsc{}\\ \hdots\\ \hbox{7}&\hbox{for \tt\#1}=\hbox{7}, \hbox{14}, \hbox{21}, \dotsc{}\\ \endcases$$ so *\count@@* tells which of the symbols \ast, \dag, \ddag, \P, \S, $\|$, \# should be printed, namely, \ast~ for $\hbox{\tt\bs count\@\@}= \hbox{1}$, \dag~ for $\hbox{\tt\bs count\@\@}=\hbox{2}$, etc. On the other hand, $$ \hbox{\tt\bs count\@\@\@}=\cases \hbox{1}&\hbox{for \tt\#1}=\hbox{1},\dots,\hbox{7}\\ \hbox{2}&\hbox{for \tt\#1}=\hbox{8},\dots,\hbox{14}\\ \hdots\endcases$$ so *\count@@@* is the number of times that this symbol has to be printed. To print the symbol this many times, we use a loop, initially setting the counter *\count@* to *\count@@@*: \C** \def\fnsymbol#1{\count@=#1\relax \count@@=\count@ "2 \advance\count@ by -1 \divide\count@ by 7 "2 \count@@@=\count@ \advance\count@@@ by 1 "2 \multiply\count@ by 7 \advance\count@@ by -\count@ "2 \count@=\count@@@ {\loop \ifcase\count@@\or"1\or\dag\or\ddag\or\P\or\S\or \text{$\|$}\or\#\fi \advance\count@ by -1 \ifnum\count@>0 \repeat}} ** \amstex\ has already defined *\dag*, *\ddag*, *\P*, and *\S*, so that they can be used either in text or in math mode, where they will change size properly; *\text{$\|$}* simply does the same for the $\|$ symbol (*\#* needs no special treatment). Modifications of *\dag*, \dots\ may be necessary for fonts with different layouts. In the above definition we put the *\loop* inside a group just in case *\fnsymbol* happens to be used inside some *\loop* itself, a somewhat finicky precaution, since a nested *\loop* involving *\fnsymbol* will never be produced by any \lamstex\ construction. The same precaution will be used for any \lamstex\ *\loop* that might conceivably occur within another *\loop*.\pagelabel{POSSLOOP} \chapter Printing cardinal and ordinal numbers\endchapter This chapter is in some sense a companion to the previous one. First we define *\cardnine@#1* for printing the final part of the name of a cardinal number; it will be applied to a counter with a value from *1*, \dots, *9*: \C** \def\cardnine@#1{\ifcase#1\or one\or two\or three\or four\or five\or six\or seven\or eight\or nine\fi} ** The number *10* will be used so often in the macros of this chapter that we want to use *\newcount* to introduce a counter having that value. As we mentioned in section~\Sref{NEWNEW}, *amstex.tex* (and *amstexl.tex*) redefine *\alloc@* so that it doesn't write anything to the *.log* file; and ** \let\alloc@@=\alloc@ ** is used so that *\alloc@@* is {\it that\/} version of *\alloc@*, even if called after *\alloc@* has been restored to the old definition from *plain* \tex\ at the end.\pagelabel{NEWALLOC2} We now want to reinstate the new definition until the end of *lamstex.tex*. Instead of redefining it directly, we just have to \C** \let\alloc@=\alloc@@ "8NEWALLOC"8 ** since *\alloc@@* was permanently given the new definition. Since *\newcount* is defined in terms of *\alloc@*, which we have just made equivalent to *\alloc@@*, the next code \C** \newcount\ten@ \ten@=10 ** does not produce anything in the *.log* file of a file that has *\input lamstex* (nor will any of the other *\new...* constructions to follow). Although *\cardinal{...}* will normally be applied when *...* is a number, to be on the safe side, we first safely store the value in a counter, ** \def\cardinal#1{\count@=#1\relax ** and then proceed by cases. \C** \def\cardinal#1{\count@=#1\relax \ifnum\count@>99 \number\count@ \else \ifnum\count@=0 zero% \else \ifnum\count@<\ten@ \cardnine@\count@ \else \ifnum\count@<20 \advance\count@ by -\ten@ \ifcase\count@ ten\or eleven\or twelve\or thirteen\or fourteen\or fifteen\or sixteen\or seventeen\or eighteen\or nineteen\fi "2 \else \count@@=\count@ \count@@@=\count@@ "2 \divide\count@ by \ten@ \multiply\count@ by \ten@ \advance\count@@@ by -\count@ \divide\count@ by \ten@ "2 \ifcase\count@\or\or twenty\or thirty\or forty \or fifty\or sixty\or seventy\or eighty\or ninety\fi "2 \ifnum\count@@@=0 \else-\cardnine@\count@@@\fi \fi \fi \fi \fi} ** Thus, if $\hbox{*#1*}>\hbox{99}$, we simply typeset this number. If $\hbox{*#1*}=\hbox{0}$, we simply typeset `zero'. If $\hbox{0}<\hbox{*#1*}<\hbox{10}$, we just use *\cardnine@* to print the number. For $\hbox{10}<\hbox{*#1*}<\hbox{20}$, we must explicitly specify `ten', \dots, `nineteen'. \Reset\tag1 Finally, for $\hbox{20}<\hbox{*#1*}<\hbox{99}$, we have to do a little calculation. The next code successively sets $$\spreadlines\jot\align \hbox{\tt\bs count\@\@\@}=\hbox{\tt\bs count\@\@}&=\hbox{\tt\#1}\tag\\ \allowdisplaybreak \hbox{\tt\bs count\@}&=\hbox{10}\cdot\left[\frac{\hbox{\tt\#1}}{\hbox{10}} \right]\tag\\ \allowdisplaybreak \split\hbox{\tt\bs count\@\@\@}&=\hbox{\tt\#1}-\hbox{10}\cdot \left[\frac{\hbox{\tt\#1}}{\hbox{10}}\right]\\ &=\hbox{\tt\#1}\pmod{\hbox{10}}\endsplit\tag\\ \allowdisplaybreak \hbox{\tt\bs count\@}&=\left[\frac{\hbox{\tt\#1}}{\hbox{10}} \right]\tag\endalign$$ The first part of the word for *#1* is `twenty' if $\hbox{\tt\#1}= \hbox{20},\dots,\hbox{29}$, and hence *\count@=2*; it is `thirty' if $\hbox{\tt\#1}=\hbox{30},\dots,\hbox{39}$, and hence *\count@=3*; etc. If $\hbox{*\count@@@*}=0$ the name is complete; otherwise we must add `-one', `-two', \dots, depending on the value of *\count@@@*. \medskip *\ordnine@*, for ordinal numbers, is exactly analogous to *\cardnine@*: \C** \def\ordnine@#1{\ifcase#1\or first\or second\or third\or fourth\or fifth\or sixth\or seventh\or eighth\or ninth\fi} ** The ordinal numbers $<\hbox{100}$ can be treated in a manner exactly like the cardinal numbers. But a problem arises for ordinal numbers like `100th', `101st', `102nd', `103rd', \dots, `109th', `110th', `111th', `112th', \dots---now the proper suffix depends not only on the last digit of the number, but also on whether the next-to-last digit is a~*1*. The routine *\ordsuffix@* selects the right suffix in these cases, assuming that the number in question has been stored in *\count@*. The first part of the routine below divides *\count@* by~10, and then calculates *\count@@@* to be $\hbox{*\count@*}\pmod{\hbox{10}}$ [compare the previous calculations for *\cardinal*]. If this is~1, so that the next-to-last digit of the number in *\count@* was~1, `th' is selected. Because the original value of *\count@* is needed for the second part of the routine, we need to store it in yet another counter, *\count@@@@*, which we have to declare. This second part simply computes *\count@@@* as $\hbox{*\count@*}\pmod{\hbox{10}}$ [for the original value of *\count@*]; if this second part of the routine ends up being invoked, the correct choice of the suffix depends only on this value of *\count@@@*: \C** \newcount\count@@@@ "slip \def\ordsuffix@{\count@@@@=\count@ \divide\count@ by \ten@ \count@@@=\count@ \count@@=\count@ "2 \divide\count@@ by \ten@ \multiply\count@@ by \ten@ \advance\count@@@ by -\count@@ "2 \ifnum\count@@@=1 th% "2 \else \count@@@=\count@@@@ \count@@=\count@@@@ \divide\count@@ by \ten@ \multiply\count@@ by \ten@ \advance\count@@@ by -\count@@ \ifcase\count@@@ th\or st\or nd\or rd\else th\fi \fi} ** *\nordinal* and *\spordinal* are easy, since they are simply a number followed by the proper suffix, or the suffix superscripted: \C** \def\nordinal#1{\count@=#1\relax\number\count@\ordsuffix@} \def\spordinal#1{\count@=#1\relax\number\count@ $^{\text{\ordsuffix@}$} ** And *\ordinal* itself simply mimics *\cardinal*, using *\ordsuffix@* for numbers $\ge\hbox{100}$: \C** \def\ordinal#1{\count@=#1\relax \ifnum\count@>99 \number\count@\ordsuffix@ \else "2 \ifnum\count@=0 zeroth% \else "2 \ifnum\count@<\ten@ \ordnine@\count@ \else "2 \ifnum\count@<20 \advance\count@ by -\ten@ \ifcase\count@ tenth\or eleventh\or twelfth\or thirteenth\or fourteenth\or fifteenth\or sixteenth\or seventeenth\or eighteenth\or nineteenth\fi \else "2 \count@@=\count@ \divide\count@ by \ten@ \multiply\count@ by \ten@ \count@@@=\count@@ \advance\count@@@ by -\count@ \divide\count@ by \ten@ \ifcase\count@\or\or twent\or thirt\or fort\or fift\or sixt\or sevent\or eight\or ninet\fi \ifnum\count@@@=0 ieth\else y-\ordnine@\count@@@\fi \fi \fi \fi \fi} ** \chapter Inhibiting expansion\label{noexpands}\endchapter There are numerous situations where we want to suppress expansion, during both *\write*'s and *\edef*'s or *\xdef*'s, of the numbering control sequences *\arabic*, *\alph*, \dots\ as well as the font change control sequences *\rm*, *\it*, *\bf*, \dots. The latter includes *\smc*, which may not have been defined yet, so we take care of that first: \C** \font@\tensmc=cmcsc10 \textonlyfont@\smc\tensmc ** [For the details of *\font@* and *\textonlyfont@* see *amstex.doc* (and also page~\ref{FL}); basically, this is like saying *\font\tensmc=cmcsc10* and *\def\smc{\tensmc}*.] We introduce a new token list \C** \newtoks\noexpandtoks@ ** which will be a list of commands, and then *\noexpands@* will issue these commands by inserting this token list: \C** \noexpandtoks@={\let\arabic=\relax\let\alph=\relax \let\Alph=\relax\let\roman=\relax\let\Roman=\relax \let\fnsymbol=\relax\let\rm=\relax\let\it=\relax \let\bf=\relax\let\sl=\relax\let\smc=\relax \let\/=\relax\let\null=\relax} "slip \def\noexpands@{\the\noexpandtoks@} ** It is easy to define the construction *\Nonexpanding#1*, which is supposed to make *\let#1=\relax* also be executed: \footnote{In version~*1* of \lamstex@, this was called *\Noexpand*, but that seems too close to *\noexpand* for comfort.} \C** \def\Nonexpanding#1{\global\noexpandtoks@ =\expandafter{\the\noexpandtoks@\let#1=\relax}} ** *\let\/=\relax* was added to *\noexpandtoks@* because *\/* will often occur in ``style'' commands (pages~\ref{ic1}, \ref{ic2}, \ref{tagit}, et al\.) that appear in *\edef*'s or *\xdef*'s and *\write*'s\pagelabel{IC} (pages~\ref{necc}, \ref{GGGG}, et al\.) and *\/* is no longer a primitive in \amstex\ or \lamstex\ (page~\ref{itcorr}). Similarly *\let\null=\relax* was added because *\null* sometimes occurs in ``style'' commands (see page~\ref{NULLSTYLE}, for example). \chapter Invisibility\label{invis}\endchapter \section{Invisible constructions} All constructions that are supposed to be ``invisible'' (the most important examples being *\label* and *\pagelabel*) begin with *\prevanish@*. If we are in horizontal mode, this sets *\saveskip@* (a glue register declared in \amstex)\ to be the previous glue, and then removes that glue. Otherwise, it simply sets *\saveskip@* to *0pt*. \C** \def\prevanish@{\saveskip@=0pt \ifhmode\saveskip@=\lastskip\unskip\fi} ** ``Invisible'' constructions end with *\postvanish@*, which puts back the *\saveskip@* glue, if greater than zero; \footnote{We don't want to add *\hskip\saveskip@* when *\savekskip@* is zero, because that case can occur when there is {\it no\/} previous glue: adding *\hskip\saveskip@* might then allow a break where none was allowed before.} it must also look ahead to see if the next token is a space, and swallow up that space if the *\saveskip@* glue was greater than zero: \C** \def\postvanish@{\ifdim\saveskip@>0pt\hskip\saveskip@\fi \futurelet\next\postvanish@@} \def\postvanish@@{\def\next@.{}% \ifx\next\space@\ifdim\saveskip@>0pt\def\next@. {}\fi\fi \next@.} ** Here we use the method of page~\ref{swallow}. Note that we {\it don't\/} want to use *\FNSS@* (section~\Sref{FIX}) in this special situation, since we want to eliminate the space only when *\saveskip@* is greater than zero. [It is important to note that *\saveskip@* appears only in these definitions, except for certain \amstex\ constructions that will not ever appear within an ``invisible'' construction. So we don't have to worry about the value of *\saveskip@* being clobbered before *\postvanish@* is applied.] These constructions allow us to define the general construction to make anything ``invisible'': \C** \def\invisible#1{\prevanish@\ignorespaces#1\unskip \postvanish@} ** (Notice that, because of the *\ignorespaces* and *\unskip*, the remark in the \lamstex~Manual on page~32, lines~6--8, is incorrect.) We will need a list, *\vanishlist@*, of all invisible constructions; this will be a list of the sort discussed in section~\Sref{LISTS}. We initialize it as: \C** \def\vanishlist@{\\\invisible} ** \section{Special considerations for invisible constructions\label{ZZZ}} If we have a construction like ** \par \invisible{...}" Some text ** then the *\prevanish@* in *\invisible* sets *\saveskip@* to *0pt*. Consequently, the *\postvanish@* will {\it not\/} delete the following space token that precedes `*Some text*'. But this space token is ignored in vertical mode ({\it The \tex book}, page~282), so we don't get an extra space before ``Some text''. The situation is quite different, however, if we have ** \par \noindent \invisible{...}" Some text ** because now the space token is not ignored, since it is encountered in horizontal mode; consequently, we will get an extra space before ``Some text''. In \amstex\ and \lamstex@, the combination *\par\noindent* has the special abbreviation *\flushpar*. We can avoid this difficulty with invisible constructions by redefining *\flushpar* as ** \def\flushpar{\par\noindent\futurelet\next\pretendspace@} ** where *\pretendspace@* simply inserts *\hskip-1pt\hskip1pt* (preceded by *\nobreak* just as a precaution) when *\next* is something in *\vanishlist@*: ** \def\pretendspace@{\ismember@\vanishlist@\next \iftest@\nobreak\hskip-1pt\hskip1pt\fi} ** [A precautionary *\relax* after the *\hskip1pt* is not needed here---the *\fi* will stop the scanning of *\hskip*.] As a result of this definition, ** \flushpar\invisible{...}" "0 ** becomes ** \par\noindent\nobreak\hskip-1pt\hskip1pt\invisible{...}" "0 ** (and similarly for *\flushpar\label{...}*, and any other ``invisible'' constructions that we will eventually define, and add to *\vanishlist@*). Consequently, the *\prevanish@* removes the *\hskip1pt* and then sets\linebreak *\saveskip@* to *1pt*. Then the *\postvanish@* adds back the *\hskip1pt* once again, canceling out the *\hskip-1pt*. Moreover, since *\saveskip@* is now positive, the space token after the *\invisible* will be thrown away, so that *\invisible{...}* will really be invisible. \footnote{If some one has typed *\define\foo{\invisible{...}}* and then used *\flushpar\foo*, this presents no problem, since in this case no space after *\foo* will appear. (But, of course, *\define\foo/{\invisible{...}}* would make `*\flushpar\foo/" *' behave incorrectly; this didn't seem worth worrying about!)} Note, by the way, that *\ismember@* was defined in such a way that the test *\ismember@\vanishlist@\next* sets *\iftest@* to be true when *\next* has been *\let* equal to a control sequence in *\vanishlist@* (page~\ref{ISFL}). [In the above definitions, it might seem that we could simply add the\linebreak *\hskip-1pt\hskip1pt* in all cases. But that wouldn't quite work, because the next construction might begin with an *\unskip* (e.g., *\linebreak* or *\dots*). Aside from such a case, however, the *\hskip-1pt\hskip1pt* doesn't do any harm.\pagelabel{NOHARM}] Since some users might type *\noindent* instead of *\flushpar*, we might as well add the *\futurelet\next\pretendspace@* to *\noindent* also. \C** \let\noindent@=\noindent "8NOIN"8 \def\noindent{\par\noindent@\futurelet\next\pretendspace@} "slip \def\pretendspace@{\ismember@\vanishlist@\next \iftest@\nobreak\hskip-1pt\hskip1pt\fi} "slip \let\flushpar=\noindent ** \smallskip *\pretendspace@* will be needed at several other points in \lamstex\ (sections~\Sref{pretend1}, \Sref{item}, et~ al.). \medskip In *amstexl.tex* we delete the definition of *\flushpar*, since it will be replaced with this new definition. \chapter Special considerations for \CS{everypar}\label{EVPAR}\endchapter Numerous \lamstex\ constructions use *\noindent@* to start an unindented paragraph. If some ** \everypar={...} ** has been specified in the document, these unindented paragraphs would also start with the *\everypar* tokens, which is normally not desired. Although this might be regarded as a rather paranoid concern, since *\everypar*'s should presumably be used only within some region of the document that contains only text, \lamstex\ contains a special construction to deal with this problem. First we introduce a new token list \C** \newtoks\everypartoks@ ** and then we define \C** \def\noindent@@{\par \everypartoks@=\expandafter{\the\everypar}\everypar={}% \noindent@\everypar=\expandafter{\the\everypartoks@}} ** Thus, *\noindent@@* \list \item ends the previous paragraph, \item stores the current value of *\everypar* in *\everypartoks@*, \item sets *\everypar* to be empty, and \item starts an unindented paragraph, which \item resets *\everypar* to its original value for the {\it next\/} paragraph. \endlist Note that the original value of *\everypar* will {\it not\/} be inserted before the *\noindent@*'ed paragraph, because it gets the value *{}* that was current when the *\noindent@* was encountered. In \lamstex@, *\noindent@@* will usually be used instead of *\noindent*, with a *\futurelet\next\pretendspace@* added when an invisible construction might follow. \chapter \CS{page}\label{PG}\endchapter In \lamstex\ the control sequence *\page* can be manipulated like *\tag*, *\claim*, etc. Thus, we can use *\Reset\page*, *\newpre\page*, \dots . But we want *\page* by itself to give an error message: \C** \def\page{\Err@{\noexpand\page has no meaning by itself}} ** (Again, see section~\Sref{SACN} for the use of *\noexpand*.) As we will see in Chapter~\ref{LBL1}, associated with the \lamstex\ construction *\tag* we have \setbox0\hbox{the counter associated with *\tag*}% \setbox1\hbox{the ``pre'' material for *\tag*}% \setbox2\hbox{the ``post'' material for *\tag*}% \setbox3\hbox{the style for *\tag*}% \setbox4\hbox{the numbering style for *\tag*}% \setbox5\hbox{the font for *\tag*}% ** \tag@C "box0 \tag@P "box1 \tag@Q "box2 \tag@S "box3 \tag@N "box4 \tag@F "box5 ** Likewise, *\claim* and all other \lamstex\ constructions that can be given a \ have similar counters and control sequences associated with them. At present, we simply want to consider the counter and control sequences associated to *\page*. For *\page@C* we just use *plain* \tex's *\pageno* \C** \let\page@C=\pageno ** and then we introduce default values (*\empty* is defined in *plain.tex* by *\def\empty{}*, so *\let\page@P=\empty* is just a briefer way of saying *\def\page@P{}*): \C** \let\page@P=\empty \let\page@Q=\empty "2 \def\page@S#1{#1\/} "8ic1"8 \def\page@F{\rm} \def\page@N{\arabic} % cannot be \let ** The *\/* in *\page@S* might be useful if *\page@F* is ever chosen to be a slanted font. We want to have *\def\page@F{\rm}* rather than *\let\page@F=\rm*, because *\rm* may actually change definitions.\pagelabel{pagef} For example, ** \fontstyle\page{...} ** expands out (page~\ref{FS}) to ** {\page@F...} ** and if we are in 9-point type at the time, we would expect to get 9-point roman type, not the 10-point roman type that is in effect at the time that *\page@F* is specified. And, as we will see later (page~\ref{SUB}), it is even more critical that we have *\def\page@N{\arabic}* rather than *\let\page@N=\arabic*. \chapter Indexing\label{INDX}\endchapter The indexing macros were placed next, because they use the fact that *""* is active in \lamstex@, and we would like to get this declared soon, so that any *""* appearing in other macros will refer to this active *""*. Some of the methods used here will also be crucial in Chapters~\ref{HLS} and ~\ref{PI}. It should be noted that in version~*1* of \lamstex@, the index entries were written to one file, with the extension~ *.ndx*, while the corresponding page numbers were written to another file, with the extension~*.npg*.\pagelabel{OLDV} That has all changed, however, and now the entry and the page number are written together to the *.ndx* file. Similarly, as we will see in Chapter~\ref{HLS}, heading levels will be written together with their page numbers in the *.toc* file, and as we'll see in Chapter~\ref{PI}, Figures, Tables, etc., will be written together with their page numbers in one file. \section{The {\tt.ndx} file} We will need a flag, \C** \newif\ifindexing@ ** to tell whether an index file is being made. *\indexfile* will (globally) set the flag *\ifindexing@* to be true the first time it is used; it will also test this flag when called, so that if it is called twice it will do nothing at all the second time. The first time *\indexfile* is called, it should create a new output stream, ** \newwrite\ndx@ ** associated with the file *\jobname.ndx* (where *\jobname* will be `*foo*' when \tex\ is processing *foo.tex*). Instead of using *\newwrite*, we will just write the code for it instead, substituting *\alloc@@* (page~\ref{NEWALLOC}) for *\alloc@*: \C** \def\indexfile{\ifindexing@\else "8INDEXFILE"8 \alloc@@7\write\chardef\sixt@@n\ndx@ \immediate\openout\ndx@=\jobname.ndx \global\indexing@true\fi} ** Then (compare section~\Sref{NEWNEW}), since we used *\alloc@@* rather then *\alloc@*, nothing will be written to the *.log* file, even though *\indexfile* is used after *\alloc@* itself has been redefined at the end of *lamstex.tex*. \section{\CS{indexproofing}} We will need an insertion class, called *\margin@*, for index entries that are to appear in the margin if *\indexproofing* has been specified. So we would like to say `*\newinsert\margin@*'. But *\newinsert* is defined differently than all other *\new...* constructions in *plain*, and it will write something to the *.log* file, despite our redefinition of *\alloc@*. So we instead simply restate everything from *plain* in the definition of *\newinsert* except for the *\wlog* part: \C** \global\advance\insc@unt\m@ne \ch@ck0\insc@unt\count \ch@ck1\insc@unt\dimen \ch@ck2\insc@unt\skip \ch@ck4\insc@unt\box \allocationnumber\insc@unt \global\chardef\margin@\allocationnumber ** Notice that although this takes up a lot of space in the file, it takes up hardly any space within \tex\ itself, just like *\newinsert\margin@*. We put no limit on the number of marginal notes on a page, and they take up no space (compare {\it The \tex book}, page~415): \C** \dimen\margin@=\maxdimen \count\margin@=0 \skip\margin@=0pt ** The flag *\ifindexproofing@* will tell us whether *\indexproofing* (and\slash or *\noindexproofing*) appears: \C** \newif\ifindexproofing@ \def\indexproofing{\indexproofing@true} \def\noindexproofing{\indexproofing@false} ** \section{Converting tokens to type {\tt 12}\label{UNMACRO}} If a control sequence *\controlseq* has been defined by \setbox0\hbox{\} \setbox1\hbox{\} ** \def\controlseq"box0{"box1} ** (see {\it The \tex book}, page 203, for terminology), then the \tex\ primitive ** \meaning\controlseq ** expands to \setbox0\hbox{\} \setbox1\hbox{\} ** macro:"box0->"box1 ** where all non-space tokens are of type~*12*. The construction *\unmacro@* is used to store the \ in *\macpar@* and the \ in *\macdef@*: \C** \def\unmacro@#1:#2->#3\unmacro@{\def\macpar@{#2}% \def\macdef@{#3}} ** In particular, if we ** \def\foo{#1} ** where *#1* is any text with balanced braces, and we do ** \expandafter\unmacro@\meaning\foo\unmacro@ ** then \footnote{There is no problem with *:* being part of the syntax of *\unmacro@* even in a file where *:* has been made active (compare section~\Sref{ALF}), because *\unmacro@* will always be used like this, to work on some value of *\meaning*.} *\macdef@* will consist of *#1* {\it with all non-space tokens converted to type~*12*}. Some information is lost in the process: multiple spaces in *#1* coalesce to single spaces in *\macdef@*, control words in *#1* are followed by spaces in *\macdef@* even if no spaces appear after them in *#1*, and line breaks in *#1* simply become spaces in *\macdef@*. So this method is not particularly useful for literal mode, especially since it cannot be applied at all unless *#1* has balanced braces. Nevertheless, it alleviates considerably the problems that arise when we want to *\write* the string *#1* to a file without having control sequences expanded. We simply have to write *\macdef@* instead! More precisely, for some output stream, like *\ndx@*, instead of using an\linebreak *\immediate\write* like ** \immediate\write\ndx@{#1} ** we can ** \def\next@{#1} \expandafter\unmacro@\meaning\next@\unmacro@ \immediate\write\ndx@{\macdef@} ** For delayed *\write*'s we have to be more careful, since *\macdef@* may have been redefined by the time the *\write* occurs. Instead of ** \write\ndx@{\macdef@} ** we must use ** \edef\next@{\write\ndx@{\macdef@}} "8WRITENDX"8 \next@ ** The *\nxd@* is not expanded in this *\edef*, since it was created with *\chardef*; such control sequences aren't expanded in *\edef*'s. \pagelabel{CHARDEFNE} Consequently, the *\edef* simply makes *\next@* mean \setbox0\hbox{\} ** \write\ndx@{"box0} ** so that *\next@* then produces this *\write*. Notice that it is irrelevant that we are writing type~*12* tokens to the *.ndx* file: once they are written to that file their category codes are completely irrelevant\pagelabel{wcodes}---if \tex\ reads this file later, they will simply be given the category codes that are in force at the time. \section{The \CS{starparts\@} and \CS{windex\@} routines} In version *1* of \lamstex@, only invisible entries could have *"1* optional entries, but now even visible entries can have them. We will use a construction *\starparts@#1* that determines if *#1* contains a *"1* and defines \setbox0\hbox{ to be } \setbox1\hbox{all of *#1*} \setbox2\hbox{the part of *#1* before the first *"1* (or all of *#1* if there is none)} \setbox3\hbox{the part of *#1* after the first *"1* (or empty if there is none)} ** "hskip-20pt\stari@ "copy0"0 "box1 "hskip-20pt\starii@ "copy0"0 "box2 "hskip-20pt\stariii@ "copy0"0 "box3 ** We begin by choosing the values that will hold when no *"1* appears: ** "hskip-20pt\def\starparts@#1{\def\stari@{#1}\def\starii@{#1}\let\stariii@=\empty . . . ** Then we perform a test that sets *\iftest@* to be true if *"1* appears in *#1* and false if it doesn't (compare the definition of *\tagin@* in section~\Sref{TESTS}): ** \test@false \def\next@##1"1##2##3\next@{\ifx\starparts@##2\test@false \else\test@true\fi} \next@#1"1\starparts@\next@ ** If no *"1* appears we are done; otherwise we will have to call another routine that separates the two parts: \C** \def\starparts@#1{\def\stari@{#1}\def\starii@{#1}% \let\stariii@=\empty \test@false \def\next@##1"1##2##3\next@{\ifx\starparts@##2\test@false \else\test@true\fi}% \next@#1"1\starparts@\next@ \iftest@\def\next@{\starparts@@#1\starparts@@}% \else\let\next@=\relax\fi\next@} "slip \def\starparts@@#1"1#2\starparts@@{\def\starii@{#1}% \def\stariii@{"1#2}} ** \medskip Once our *""...""* and *""""...""* constructions, to be defined in section~\Sref{MKI}, have used *\starparts@* to determine *\stari@*, *\starii@*, and *\stariii@*, we will use *\starii@* to typeset `*...*' in the case of a visible index entry, and then we will use the ``write@-index'' routine *\windex@*. When we are making an *.ndx* file, this routine will first ** \expandafter\unmacro@\meaning\stari@\unmacro@ ** to convert `*...*' to type~*12* tokens. Then ** \edef\macdef@{\string""\macdef@\string""} ** will add *""* at each end, for the sake of the *index* program; *\string""* is needed since *""* will be active. Then we will use the *\edef* of page~\ref{WRITENDX} to write these tokens to the *.ndx* file. This will be followed by the page number. Actually, instead of writing just the page number, we write four groups, the first containing the page number, and the next three containing the page numbering style, the ``pre-page'' material, and the ``post-page'' material (Chapter~\ref{PG}), ** \write\ndx@{{\number\pageno}{\page@N}{\page@P}{\page@Q}} ** This allows the *index* program to deal with all sorts of special page numbering possibilities. In addition, when *\ifindexproofing@* has been set true, we want to ** \insert\margin@{\hbox{\rm\vrule \height9pt \depth2pt \width0pt ... ** where the *\hbox* begins with a ``strut'', designed to keep baselines of successive entries 11~points apart (see page~\ref{HWD} for the use of *\height*, \dots). \footnote{Perhaps \pagelabel{STRUTINSERT} this is a good place to mention something that tends to be obscured in discussions about struts. Most \tex nicians are familiar with struts as the device that allows one to place one *\vbox* above another and still have the proper space between the bottom baseline of the top *\vbox* and top baseline of the bottom *\vbox*. The problem here is that \tex\ will normally insert only *\lineskip* space between the two boxes, since the baseline of the second box is so far from the baseline of the first. But the situation with regard to *\footnote*'s, or members of other insertion classes, is really quite different: \tex\ inserts {\it no interline glue whatsoever\/} between two different members of an insertion class ({\it The \tex book}, page~125). Thus, even single line footnotes will be spaced incorrectly without struts! (Struts are discussed further in section~\Sref{VF}.)} This *\hbox* should contain all material before any *"1* in `*...*' typeset in *\rm*, but all material after the first *"1* should be converted to type~*12* tokens, and typeset in the *\tt* font, \footnote{Other styles (compare Part~\ref{SFPART}) may use smaller print for these side notes.} since it contains things like *"1e\it*, which are not supposed to be acted upon, but merely convey information to the *index* program. This is all accomplished with the following code: \C** \def\windex@{\ifindexing@ \expandafter\unmacro@\meaning\stari@\unmacro@ \edef\macdef@{\string""\macdef@\string""}% \edef\next@{\write\ndx@{\macdef@}}\next@ \write\ndx@{{\number\pageno}{\page@N}{\page@P}{\page@Q}}% \fi \ifindexproofing@ \ifx\stariii@\empty\else \expandafter\unmacro@\meaning\stariii@\unmacro@\fi \insert\margin@{\hbox{\rm\vrule \height9pt \depth2pt \width0pt \starii@ \ifx\stariii@\empty\else\tt\macdef@\fi}}% \fi} ** At the time that our *\write* is performed, we will want *\noexpands@* to be in effect, partly to prevent expansion of any font control sequences that might appear in *\page@P* and *\page@Q*, but mainly because we want to be sure that *\page@N* isn't expanded during the *\write*, since the *index* program expects to see a numbering control sequence in this second group. But there's no point putting ** {\noexpands@ \write\npg@{{\number\pageno}{\page@N}{\page@P}{\page@Q}}} ** in our definition, because this delayed *\write* is simply added to the main vertical list and does not take place until a *\shipout*. Instead, we will have to be careful to specify *\noexpands@* during any *\shipout* (section~\Sref{SO}).\pagelabel{SUB1} Nevertheless, this is an appropriate time to discuss the problems that would be encountered if expansion were not prohibited. Expansion would clearly cause problems if *\page@N* is defined as *\alph* or *\Alph*, but, in fact, for this particular *\write*, expansion would be a problem even if *\page@N* is defined as *\arabic*, because *\page@N* appears in a group by itself---\tex\ would complain during the *\write* that ** ! Argument of \page@N has an extra }. ** And here comes the most subtle point of all: If we *\let\page@N=\arabic*, then our *\write* would not put *\arabic* in place of *\page@N*, even if *\noexpands@* is in effect when it takes place, and we would get the very same error message.\pagelabel{SUB} That is because *\page@N* would have the {\it original\/} meaning of *\arabic*---although *\noexpands@* says `*\let\arabic=\relax*', our *\page@N* would not be this *\arabic*! On the other hand, when we *\def\page@N{\arabic}*, the *\write* first expands *\page@N* to (the current) *\arabic*, and then doesn't expand {\it this\/} *\arabic* further. \section{Indexing\label{MKI}} Now we are finally ready to make *""* active, \C** \catcode`\""=\active ** and define the action of *""*. As we have already noted (page~\ref{SW}), the combination *@""* will still work when~ *""*~ is active. First of all, *""* will have to look ahead to see if it is followed by another *""*, because this indicates an invisible entry: \C** \def""{\futurelet\next\quote@} \def\quote@{\ifx\next""\expandafter\quote@@\else \expandafter\quote@@@\fi} ** Note that we are using the ``K-method'' here (see section~\Sref{K}). *\quote@@@*, the result when *""* isn't followed by another *""*, so that we have a visible index entry, is simply defined as \C** \def\quote@@@#1""{\starparts@{#1}\starii@\windex@} ** Thus, after *\starparts@* defines *\stari@*, *\starii@*, and *\stariii@*, we typeset *\starii@*---the part before any *"1*---% and then apply the ``write-index'' routine *\windex@*. We should note that as a consequence of the definition of *\quote@@@#1*, the indexed word *#1* may be followed by an *\insert* and/or a ``whatsit'', namely, the *\write* produced by *\windex@*. But either an *\insert* or a ``whatsit'' can appear {\it after\/} a word without suppressing hyphenation---% see {\it The \tex book}, third paragraph from the bottom on page~454. (This should be compared to the *\makexref* macro on page~424 of {\it The \tex book}, where the *\insert* appears {\it before\/} the word, and therefore suppresses hyphenation of the word.) Similarly, as we will see in a moment, an invisible index entry simply supplies an *\insert* and/or a *\write*. The warning on page~100 of the \lamstex~Manual is therefore inaccurate: an invisible entry will suppress hyphenation of a word only if it immediately precedes it, not if it follows it. Similarly, the warnings on pages~31 and~33 are inaccurate; only a *\label* or *\pagelabel* immediately preceding a word will interfere with its hyphenation---in particular, the example given on page~33 won't interfere with hyphenation. \medskip *\quote@@* is not that much different from *\quote@*, except that we want it to swallow the next *""*, and begin with *\prevanish@*, so that it will be invisible: \C** \def\quote@@""#1""{\prevanish@\starparts@{#1}\windex@ \futurelet\next\quote@@@@} ** The *\futurelet\next\quote@@@@* is needed to see whether yet another *""* occurs after the third *""* that caused all this to happen. If a fourth *""* didn't occur, we insert the *\postvanish@*, and if a fourth *""* did occur, we simply swallow it up, and then insert the *\postvanish@*: \C** \def\quote@@@@{\ifx\next""\def\next@""{\postvanish@}\else \let\next@=\postvanish@\fi\next@} ** \section{Changes to the \lamstex\ Manual}Because of changes in the indexing macros, almost every caveat on page~101 of the \lamstex~Manual is wrong. The first paragraph is wrong: index entries within heading levels {\it will\/} show up in the margins when *\indexproofing* has been specified. (Of course, one had better not type something like ** \HL1 ""Disappearing"" words\endHL ** since *""Disappearing""* would then be interpreted as a ``quoted'' number for *\HL1*\,! Something like *\HL1{}""Disappearing""* is needed.) The third paragraph is wrong, because only the parts after the first *"1* will be typeset in the typewrite font, with characters of type~*12*. The fourth paragraph is wrong: *\""* can be used within an invisible entry. The fifth paragraph is wrong: invisible entries can now appear anywhere. It is true, as the final paragraph claims, that index entries in *\footnote*'s won't appear in the margin (they are *\insert*'s within an *\insert*, and won't migrate out). However, no special efforts are required in the *\footnote* macros to get indexing to work within *\footnote*. \section{Invisibility} After all this, we want to add *""* to *\vanishlist@*: \C** \rightadd@""\to\vanishlist@ ** This probably looks wrong, since it is only the double mark *""""* that indicates an invisible entry, but ** \nobreak\hskip-1pt\hskip1pt""..."" ** and ** \nobreak\hskip-1pt\hskip1pt""""..."" ** will both work out just right: before the visible index entry *""...""* the *\hskip-1pt\hskip1pt* will simply be irrelevant, while before an invisible index entry *""""...""* it provides the right clues for the *\prevanish@* called by *\quote@@* (compare page~\ref{NOHARM}). \section{Other delimiters for index entries\label{OTHERD}}The use of *""* as a delimiter for index entries conflicts with its use in German styles (this will probably remain true even when the international font layouts are in use, although then the *""* will presumably no longer be active for German). However, it is not very hard to set up other delimiters for this purpose. For example, suppose we want to use *<...>* delimiters, so that ** Beauty<> is . ** will produce a index entry for *truth*, and an invisible index entry for *beauty*. For this, we could ** \catcode`\<=\active \let<="" % we might as well continue using \quote@, ... "slip % \def""{...} if we have new definitions for German, or % \catcode`\""=12 if "" should no longer be active "slip \def\windex@{\ifindexing@ \expandafter\unmacro@\meaning\nextii@\unmacro@ \xdef\nextii@{\string<\macdef@>}% . . . \fi} "slip \def\quote@{\ifx\next<\expandafter\quote@@\else \expandafter\quote@@@\fi} \def\quote@@@#1>{\starparts@{#1}\starii@\windex@} \def\quote@@<#1>{\prevanish\starparts@{#1}\windex@ \futurelet\next\quote@@@@} \def\quote@@@@{\ifx\next>\def\next@>{\postvanish@}\else \let\next@=\postvanish@\fi\next@} ** The new version of the *index* program (see Chapter~\ref{IPROG}) now accepts any delimiters in the *.ndx* file. However, as Chapter~\ref{IPROG} points out, for German alphabetization we would probably want some modifications to deal with words with umlauts. We should probably also remove *""* from *\vanishlist@*. There is no general mechanism for removing something from *\vanishlist@*. However, since we know that *\vanishlist@* will be of the form ** \\\invisible\\"" ... ** we can ** \def\next@\\\invisible\\""#1\next@{\def\vanishlist@{#1}} \expandafter\next@\vanishlist@\next@ ** \small For consistency, it would probably be better to choose, once and for all, index entry delimiters that could be used in all cases; *<...>* don't satisfy that requirement, since Scandinavian keyboards have letters instead of *<* and *>*. There aren't too many possibilities left, however! The only reasonable candidates are *_* and *|* (although *+* and *=* would also be possible, if we insisted that people never used them outside of math mode), and ** Beauty||beauty|| is |truth|, truth beauty ** doesn't look too bad. If only *|* weren't used for something else in German styles!) \endsmall \section{{\tt\bs idefine} and {\tt\bs iabbrev}\label{IDAA}} A construction like \setbox0\hbox{\}\setbox1\hbox{\} ** \idefine\cs"copy0{"copy1} ** has to ** \define\cs"copy0{"copy1} ** and also send this definition off to the *.ndx* file. *\idefine* first stores its argument, *\cs* say, in *\next@*, and also stores *\noexpand\cs* in *\nextii@*, for later use: %%%%%%% ** \def\next@{#1}\def\nextii@{\noexpand#1}% ** Then we will apply the construction *\idefine@* once we have suitably swallowed up the \ and \: \C** \def\idefine#1{\def\next@{#1}\def\nextii@{\noexpand#1}% \afterassignment\idefine@\def\nextiii@} ** Here the *\def\nextiii@* will cause the following \ and \ to be digested into a definition of *\nextiii@*, after which assignment we will apply *\idefine@*. Since *\nextiii@* now has the definition that we want for *\cs*, the first thing *\idefine@* must do is to ** \let\cs=\nextiii@ ** Since *\next@* was *\def*'ed to be *\cs*, we can do this with ** \expandafter\let\next@=\nextiii@ ** Then, if we are indexing, we need to recover the \ and \ for *\cs*, which is now that for *\nextiii@*. So we use ** \expandafter\unmacro@\meaning\nextiii@\unmacro@ ** [This construction doesn't work if `*->*' appears in the \ of a definition, so let's hope no one ever makes such a definition.] Now we want to write \setbox0\hbox{\}\setbox1\hbox{\} ** \define\cs"box0{"box1} ** to the *.ndx* file. Since *\nextii@* was defined as *\noexpand\cs*, we can do this with ** \immediate\write\ndx@{\noexpand\define \nextii@\macpar@{\macdef@}} ** ---note that the *\write* will first expand *\nextii@* to *\noexpand\cs*, and then replace this with *\cs*, unexpanded. Since all tokens in *\macpar@* and *\macdef@* are type~12, we don't have to worry about their expansion (and note the remark on page~\ref{wcodes}). Thus, the definition of *\idefine@* reads: \C** \def\idefine@{\ifindexing@ \expandafter\let\next@=\nextiii@ \expandafter\unmacro@\meaning\nextiii@\unmacro@ \immediate\write\ndx@{\noexpand\define\nextii@ \macpar@{\macdef@}}\fi} ** *\iabbrev* is simpler. Recall that *\iabbrev* must be used in the form ** \iabbrev"1\cs{...} ** so we define \C** \def\iabbrev"1#1#2{\ifindexing@ \toks@={#2}% \immediate\write\ndx@ {\noexpand\abbrev"1\noexpand#1{\the\toks@}}\fi} **