cite.sty

%     C I T E . S T Y
%
%     version 5.3  (Sep 2010)
%
%     Compressed, sorted lists of on-line or superscript numerical citations.
%     see also drftcite.sty (And the stub overcite.sty)
%
%     Copyright (C) 1989-2010 by Donald Arseneau
%     These macros may be freely used, transmitted, reproduced, or modified
%     provided that this notice is left intact.
%
%     Instructions follow \endinput.
%  ------------------------------------
% First, ensure that some catcodes have the expected values
\edef\citenum{% to restore funny codes
  \catcode\string`\string ` \the\catcode\string`\`
  \catcode\string`\string ' \the\catcode\string`\'
  \catcode\string`\string = \the\catcode\string`\=
  \catcode\string`\string _ \the\catcode\string`\_
  \catcode\string`\string : \the\catcode\string`\:}
\catcode\string`\` 12
\catcode`\' 12
\catcode`\= 12
\catcode`\_ 8
\catcode`\: 12

%   Prepare for optional variations:
%   [ verbose, nospace, space, ref, nosort, noadjust, superscript, nomove ],
%   \citeform,\citeleft,\citeright,\citemid,\citepunct,\citedash
%
%   Set defaults first:

%   [ on the left.  Option [ref] does: [Ref. 12, note]
\providecommand\citeleft{[}

%   ] on the right:
\providecommand\citeright{]}

%   , (comma space) before note
\providecommand\citemid{,\penalty\citemidpenalty\ }

%   , (comma thin-space) between entries; [nospace] eliminates the space
\providecommand\citepunct{,\penalty\citepunctpenalty%
        \hskip.13emplus.1emminus.1em\relax}%

%   -- (endash) designating range of numbers:
% (using \hbox avoids easy \exhyphenpenalty breaks)
\providecommand{\citedash}{\hbox{--}\penalty\citepunctpenalty}

% Default line-breaking penalties. Use \mathchardef instead of count registers
\mathchardef\citeprepenalty=\@highpenalty
\mathchardef\citemidpenalty=\@medpenalty
\mathchardef\citepunctpenalty=\@m

%   Each number left as-is:
\providecommand\citeform{}

%   punctuation characters to move for overcite
\providecommand{\CiteMoveChars}{.,:;}

%   font selection for superscript numbers
\providecommand\OverciteFont{\fontsize\sf@size\baselineskip\selectfont}


%   Do not repeat warnings.  [verbose] reverses
\let\oc@verbo\relax

% Default is to move punctuation:
\def\oc@movep#1{\futurelet\@tempb\@citey}

%----------------------
% \citen uses \@nocite to ignore spaces after commas, and write the aux file
% \citation. \citen then loops over the citation tags, using \@make@cite@list
% to make a sorted list of numbers.  Finally, \citen executes \@citelist to
% compress ranges of numbers and print the list. \citen can be used by itself
% to give citation numbers without the brackets and other formatting; e.g.,
% "See also ref.~\citen{junk}."
%  Make internal version called \@cite@n just in case packages put hooks in
%  \citen
%
\DeclareRobustCommand\citen{\@cite@n}
\def\@cite@n#1{%
 \begingroup
  \let\@safe@activesfalse\@empty
  \@nocite{#1}% ignores spaces, writes to .aux file, returns #1 in \@no@sparg
  \@tempcntb\m@ne    % \@tempcntb tracks highest number
  \let\@celt\delimiter % an unexpandable, but identifiable, token
  \def\@cite@list{}% % empty list to start
  \let\@citea\@empty % no punctuation preceding first
  \@for \@citeb:=\@no@sparg\do{\@make@cite@list}% make a sorted list of numbers
  % After sorted citelist is made, execute it to compress citation ranges.
  \@tempcnta\m@ne    % no previous number
  \mathchardef\@cite@incr\z@ % no previous sequence
  \let\@h@ld\@empty  % nothing held from list yet
  \let\@celt\@compress@cite \@cite@list % output number list with compression
  \@h@ld % output anything held over
 \endgroup
 \@restore@auxhandle
 }

% For each citation, check if it is defined. If so, then extract plain
% value to \@B@citeB (without hyperlink info). Then, 
% If it is a pure number, add it to cite list
% Otherwise, try extracting prefix and suffix characters.
%
\def\@make@cite@list{%
 \expandafter\ifx\csname b@\@citeb\@extra@b@citeb 
      \endcsname\relax % undefined: output ? and warning
    \@citea {\bfseries ?}\let\@citea\citepunct \G@refundefinedtrue
    \@warning {Citation `\@citeb' on page \thepage\space undefined}%
 %%   \oc@verbo \global\@namedef{b@\@citeb\@extra@b@citeb}{?}%  ???
 \else %  defined               % remove previous line to repeat warnings
    \@cite@nonhyper@sanitize
    \@addto@cite@list
 \fi}

\def\@nonhyper@@link [#1]#2#3#4{#4}
\def\@cite@nonhyper@sanitize{\begingroup
    \let\hyper@@link\@nonhyper@@link
    \protected@xdef\@B@citeB{\csname b@\@citeb\@extra@b@citeb \endcsname}%
  \endgroup}

\def\@cite@out#1{\citeform{\csname #1\endcsname}}

% Add entry to the list of citations.  This default definition sorts pure
% numbers as well as numbers with other single-character tags.  There 
% is presently no other definition than this default, but features may 
% be added later.
%
\def\@addto@cite@list{% 
 \@cite@posnumtest\@B@citeB
   {\@addnumto@cite@list\@B@citeB}%  a positive number, put in list
   {\@cite@combo@num}% not a pure positive number, test for combo forms
}

% With this \@cite@combo@num we delve into handling of numbers combined
% with non-numeric tags.  The specific command name \@cite@combo@num can
% serve as a hook for redefinition, perhaps to give simple non-sorting 
% for anything not a pure number, or to attempt even more complicated
% sorting, say dictionary sorting of textual citations.  The following
% definition leads down the road of sorting mostly-numbers but with 
% optional single-character prefix and/or suffix. 

\def\@cite@combo@num{\expandafter\@cite@try@combo\@B@citeB\delimiter}

% First of many stages for sorting numbers with prefix/suffix characters.
% Test for a leading token of category letter or other (appropriate for
% all combination types).
%
\def\@cite@try@combo{%
  \@if@printable@char{\@cite@try@prefix}{\@cite@gobbledump@now}}

% First token is good, so test for a character prefix before a number.
% Process first token, either a first digit or a prefix
%
\def\@cite@try@prefix#1{% #1 is first character of citation
  \@cite@posnumtest{#1}% a digit else prefix
   {\@cite@add@letnumD {\z@}#1}% no prefix character (use zero)
   {\@cite@add@letnumC {`#1}}% prefix char; use the char code
}

% Examine character after prefix to ensure it is a number.  First must
% ensure it is a plain character token
%
\def\@cite@add@letnumC#1{\@if@printable@char%
   {\@cite@add@letnumD{#1}}% continue with prefix (perhaps zero)
   {\@cite@gobbledump@now}% else abandon fancy processing
}

% Save prefix (if any) numerically in \@tempcnta, test next character for being
% a digit, then collect main number
%
\def\@cite@add@letnumD#1#2{% #1 = numeric code for prefix, #2 = next char
  \@tempcnta=#1\multiply\@tempcnta 16384 %
  \@cite@posnumtest{#2}% if next char is a digit, continue with number:
   {\afterassignment\@cite@add@letnumE \advance\@tempcnta #2}%
   {\@cite@gobbledump@now}% No number so output citation
}

% Have collected number.  Now look for a non-number suffix or separator.
%
\gdef\@cite@add@letnumE{%
  \multiply\@tempcnta\@cclvi
  \@if@printable@char{% a suffix or separator given
      \@cite@add@letnumF
    }{% else, maybe nothing remains
      \ifx\@let@token\delimiter % use number, and remove trailing \delimiter
        \@citeaddcnta \expandafter\@gobble
      \else % non-printable char found, so abandon fancy processing
        \expandafter\@cite@gobbledump@now
      \fi
    }}

%  Have everything up to a suffix or separator character.  Check 
%  following to see which. Three possibilites are (1) noting =>
%  a suffix; (2) number => separator-number; (3) other => garbage.
%
\def\@cite@add@letnumF#1#2\delimiter{% #1 = suffix/separator #2=rest
  \advance\@tempcnta`#1\relax
  \@cite@posnumtest{#2}{\@cite@add@numsepnum{#2}}% handle as num sep num
    {% else...
     \ifx\delimiter#2\delimiter % nothing left, so #1 is a suffix
       \@citeaddcnta
     \else % some non-number; dump it
       \@cite@dump@now
     \fi
}}

% Handle citation as number separator number.
% Yes, there is a bug that the list 1.1,1.258,1.515 will be compressed as
% 1.1-1.515; so sue me.

\def\@cite@add@numsepnum#1{% #1 = last number
  \ifnum\@tempcnta<262144 % OK numeric range
    \multiply\@tempcnta 4096
    \advance\@tempcnta #1 % num,sep,num have maximum numbers:  1023, 255, 4095
    \@citeaddcnta
  \else % out of range, treat as raw string
    \@cite@dump@now
  \fi}

% This is our bail-out when the citation cannot be processed as
% [prefix]number[suffix] or number[sep]number: it outputs the citation 
% immediately (unsorted) and consumes tokens to the \delimiter tag used 
% as an end-marker
%
\def\@cite@gobbledump@now#1\delimiter{\@cite@dump@now}%

\def\@cite@dump@now{%
  \@citea \@cite@out{b@\@citeb\@extra@b@citeb}\let\@citea\citepunct}

% add an entry to the sorted list, using its sort-number \@tempcnta, and
% also saving the plain-text value \@B@citeB as well as the csname 
% b@\@citeb\@extra@b@citeb. (The \@B@citeB is actually not used, unless 
% somebody extends the definitions.)
\def\@citeaddcnta{%
   \ifnum \@tempcnta>\@tempcntb % new highest, add to end (efficiently)
      \edef\@cite@list{\@cite@list 
        \@celt{\number\@tempcnta}{\@B@citeB}{b@\@citeb\@extra@b@citeb}}%
      \@tempcntb\@tempcnta
   \else % other sortable value: insert appropriately
      \edef\@cite@list{\expandafter\@sort@celt\@cite@list \@gobble.\@gobble.}%
   \fi
  }

% add pure numeric entry to cite list, with sorting
\def\@addnumto@cite@list#1{%
   \@tempcnta#1\relax
   \multiply\@tempcnta\@cclvi
   \@citeaddcnta}

% \@sort@celt inserts number (\@tempcnta) into list of \@celt{num}{text}{tag} 
% (#1{#2}{#3}{#4})
% \@celt must not be expandable, and the arguments must not be fragile. 
% List should end with four vanishing tokens.
%
\def\@sort@celt#1#2#3#4{\ifx \@celt #1% parameters are \@celt{num}{text}{tag}
   \ifnum #2<\@tempcnta % number goes later in list
      \@celt{#2}{#3}{#4}%
      \expandafter\expandafter\expandafter\@sort@celt % continue
   \else % number goes here
      \@celt{\number\@tempcnta}{\@B@citeB}{b@\@citeb\@extra@b@citeb}%
      \@celt{#2}{#3}{#4}% stop comparing
\fi\fi}

% Check if each number follows previous and can be put in a range. 
% Since there are suffix characters allowed, there are two kinds of 
% ranges: ranges of consecutive pure numbers with no (or same) 
% suffix, or ranges of the same number with consecutive suffix
% characters.
%
\def\@compress@cite#1#2#3{%% This is executed for each number
  \ifnum\@cite@incr=\z@   % no consecutives pending. Try both types of sequence
    \advance\@tempcnta\@cclvi % Now \@tempcnta has incremented number
    \ifnum #1=\@tempcnta    % Start a sequence of consecutive numbers
       \expandafter\def\expandafter\@h@ld\expandafter{\@citea\@cite@out{#3}}%
       \mathchardef\@cite@incr=\@cclvi
    \else % next try increment of suffix
       \advance\@tempcnta-\@cclv % Now \@tempcnta has incremented suffix
       \ifnum #1=\@tempcnta      % Start a sequence of suffix increments
         \expandafter\def\expandafter\@h@ld\expandafter{\@citea\@cite@out{#3}}%
         \mathchardef\@cite@incr=\@ne
       \else % it is no type of sequence -- emit number (nothing is held)
         \@citea \@cite@out{#3}%
       \fi
    \fi
  \else % a sequence is running
    \advance\@tempcnta\@cite@incr % Now \@tempcnta is next in sequence
    \ifnum #1=\@tempcnta   % Number follows previous--hold on to it
       \def\@h@ld{\citedash \@cite@out{#3}}%
    \else   %  non-successor -- dump what's held and do this one
       \@h@ld \@citea \@cite@out{#3}%
       \let\@h@ld\@empty
       \mathchardef\@cite@incr=\z@
    \fi
  \fi
  \@tempcnta#1\let\@citea\citepunct
}

% Make \cite choose superscript or normal

\DeclareRobustCommand{\cite}{%
  \@ifnextchar[{\@tempswatrue\@citex}{\@tempswafalse\@citex[]}}

% Do \cite command on line.
%
\def\@citex[#1]#2{\@cite{\@cite@n{#2}}{#1}}

\def\@cite#1#2{\leavevmode \cite@adjust
  \citeleft{#1\if@tempswa\@safe@activesfalse\citemid{#2}\fi
  \spacefactor\@m % punctuation in note doesn't affect outside
  }\citeright
 \@restore@auxhandle}

%  Put a penalty before the citation, and adjust the spacing: if no space
%  already or if there is extra space due to some punctuation, then change
%  to one inter-word space.
%
\def\cite@adjust{\begingroup%
  \@tempskipa\lastskip \edef\@tempa{\the\@tempskipa}\unskip
  \ifnum\lastpenalty=\z@ \penalty\citeprepenalty \fi
  \ifx\@tempa\@zero@skip \spacefactor1001 \fi % if no space before, set flag
  \ifnum\spacefactor>\@m \ \else \hskip\@tempskipa \fi
  \endgroup}


\edef\@zero@skip{\the\z@skip}

%  Superscript cite, with no optional note.  Check for punctuation first.
%
\def\@citew#1{\begingroup \leavevmode
  \@if@fillglue \lastskip \relax \unskip
  \def\@tempa{\@tempcnta\spacefactor
     \/% this allows the last word to be hyphenated, and it looks better.
     \@citess{\@cite@n{#1}}\spacefactor\@tempcnta
     \endgroup \@restore@auxhandle}%
  \oc@movep\relax}% check for following punctuation (depending on options)

%  Move trailing punctuation before the citation:
%
\def\@citey{\let\@tempc\@tempa
   % Watch for double periods and suppress them
   \ifx\@tempb.\ifnum\spacefactor<\@bigSfactor\else
     \@citeundouble
   \fi\fi
   % Move other punctuation
   \expandafter\@citepc\CiteMoveChars\delimiter
   \@tempc}%

% This is in a separate macro in case the next "character" (token)
% is \if or \fi, etc.
\def\@citeundouble{% Suppress doubling of periods
   \let\@tempb\relax \let\@tempc\oc@movep
}

\def\@citepc#1{%
   \ifx\@tempb#1\@empty #1\let\@tempc\oc@movep \fi
   \ifx\delimiter#1\else \expandafter\@citepc\fi}

%  Replacement for \@cite which defines the formatting normally done
%  around the citation list.  This uses superscripts with no brackets.
%  HOWEVER, trailing punctuation has already been moved over.  The
%  format for cites with note is given by \@cite.  Redefine \@cite and/
%  or \@citex to get different appearance.  I don't use \textsuperscript
%  because it is defined BADLY in compatibility mode.

\def\@citess#1{\mbox{$\m@th^{\hbox{\OverciteFont{#1}}}$}}

% \nocite: This is changed to ignore *ALL* spaces and be robust.  The
% parameter list, with spaces removed, is `returned' in \@no@sparg, which
% is used by \@cite@n (\citen).
%
\DeclareRobustCommand\nocite[1]{%
 \@bsphack \@nocite{#1}%
 \@for \@citeb:=\@no@sparg\do{\@ifundefined{b@\@citeb\@extra@b@citeb}%
    {\G@refundefinedtrue\@warning{Citation `\@citeb' undefined}%
    %%\oc@verbo \global\@namedef{b@\@citeb\@extra@b@citeb}{?}
    }{}}%
 \@esphack}

\def\@nocite#1{\begingroup\let\protect\string% normalize active chars
 \xdef\@no@sparg{\expandafter\@ignsp#1 \: }\endgroup% and remove ALL spaces
 \if@filesw \immediate\write\@newciteauxhandle % =\@auxout, except with multibib
    {\string\citation {\@no@sparg}}\fi
 }

% for ignoring *ALL* spaces in the input.  This presumes there are no
% \outer tokens and no \if-\fi constructs in the parameter.  Spaces inside
% braces are retained.
%
\def\@ignsp#1 {\ifx\:#1\@empty\else #1\expandafter\@ignsp\fi}

% \@if@fillglue{glue}{true}{false}
\begingroup
 \catcode`F=12 \catcode`I=12\catcode`L=12
 \lowercase{\endgroup
 \def\@if@fillglue#1{%
  \begingroup \skip@#1\relax
  \expandafter\endgroup\expandafter 
  \@is@fil@ \the\skip@ \relax\@firstoftwo FIL\relax\@secondoftwo\@nil}
 \def\@is@fil@ #1FIL#2\relax#3#4\@nil{#3}
}

% Test if next token is a char of "printable" categories other or letter or 
% active.  Syntax:
%  \@if@printable@char {do if printable}{do if not printable}<char>
%
\def\@if@printable@char#1#2{%
  \def\reserved@a{#1}%
  \def\reserved@b{#2}%
  \futurelet\@let@token\@test@print@char
}

% Note side-effect of redefining \reserved@a and \reserved@b
\def\@test@print@char{%
 \ifnum 
    \ifcat\noexpand\@let@token A1\fi
    \ifcat\noexpand\@let@token 11\fi
    \ifcat\noexpand\@let@token \noexpand~1\fi%
  0>\z@
    \expandafter\reserved@a \else
    \expandafter\reserved@b \fi
}

% Test for a pure positive number: {possible number}{true}{false}
\def\@cite@posnumtest#1{%
  \ifcat _\ifnum\z@<0#1_\else A\fi
    \expandafter\@firstoftwo \else \expandafter\@secondoftwo \fi
}

\let\nocitecount\relax  % in case \nocitecount was used for drftcite

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%                     option processing

\DeclareOption{verbose}{\def\oc@verbo#1#2#3#4{}}
\DeclareOption{nospace}{\def\citepunct{,\penalty\citepunctpenalty}}
\DeclareOption{space}{\def\citepunct{,\penalty\citepunctpenalty\ }}
\DeclareOption{nobreak}{% no line-breaks
 \mathchardef\citeprepenalty=\@M
 \mathchardef\citemidpenalty=\@M
 \mathchardef\citepunctpenalty=\@M
}
\DeclareOption{ref}{\def\citeleft{[Ref.\penalty\@M\ }}
%  To disable sorting [nosort], the redefinition depends on whether [nocompress]
%  was also selected, so just set a flag first.
\DeclareOption{nosort}{\let\@citeaddcnta\@empty}
\DeclareOption{sort}{}% default!
%  Likewise set flag for [nocompress]
\DeclareOption{nocompress}{\let\@compress@cite\@empty}
\DeclareOption{compress}{}% default
\DeclareOption{nomove}{\def\oc@movep{\@tempa}\let\@citey\oc@movep}
\DeclareOption{move}{}% default
\DeclareOption{super}{\ExecuteOptions{superscript}}
\DeclareOption{superscript}{%
  \DeclareRobustCommand{\cite}{%
    \@ifnextchar[{\@tempswatrue\@citex}{\@tempswafalse\@citew}}
}
\DeclareOption{noadjust}{\let\cite@adjust\@empty}% Don't change spaces
\DeclareOption{adjust}{}% adjust space before [ ]
\DeclareOption{biblabel}{\def\@biblabel#1{\@citess{#1}\kern-\labelsep\,}}
\ProvidesPackage{cite}[2010/09/10 \space v 5.3]
\ProcessOptions

\ifx\@citey\oc@movep\else % we are moving punctuation; must ensure sfcodes
  \mathchardef\@bigSfactor3000
  \expandafter\def\expandafter\frenchspacing\expandafter{\frenchspacing
    \mathchardef\@bigSfactor1001
    \sfcode`\.\@bigSfactor \sfcode`\?\@bigSfactor \sfcode`\!\@bigSfactor }%
  \ifnum\sfcode`\.=\@m \frenchspacing \fi
\fi

%  make redefinitions to handle [nosort] [nocompress] and their combination
\ifx\@compress@cite\@empty
   \ifx\@citeaddcnta\@empty
      % [nosort,nocompress] -- short-circuit much processing
      \def\@addto@cite@list{\@cite@dump@now}
   \else
      % [sort,nocompress]
      \def\@compress@cite#1#2#3{%  % This is executed for each number
        \@h@ld \@citea \@cite@out{#3}%
        \let\@h@ld\@empty \let\@citea\citepunct
      }
   \fi
\else %  
   \ifx\@citeaddcnta\@empty % [nosort,compress]
     %  nosort: always add to end of list, but still calculate
     %  sort-order number (\@tempcnta) because it may be used for
     %  collapsing consecutive numbers.
     \def\@citeaddcnta{%
       \edef\@cite@list{\@cite@list 
        \@celt{\number\@tempcnta}{\@B@citeB}{b@\@citeb\@extra@b@citeb}}%
     }
   \fi
\fi

%  Compatability with chapterbib (see use of \@extra@b@citeb above and in chapterbib)
\@ifundefined{@extra@b@citeb}{\def\@extra@b@citeb{}}{}

%  Compatability with multibib (see use of \@newciteauxhandle) (Yes, this is
%  overly messy, but I asked for it...  I can't have multibib putting junk after 
%  the cite command because it hides following punctuation, but then I have
%  to restore the ordinary meaning of \@newciteauxhandle = \@auxout.)
\providecommand\@newciteauxhandle{\@auxout}
\AtBeginDocument{\@ifundefined{newcites}{\global\let\@restore@auxhandle\relax}{}}
\def\@restore@auxhandle{\def\@newciteauxhandle{\@auxout}}

%  compatability with backref: prevent it from redefining \@citex
%  in the wrong way (ignoring \@citew and \citen.  I install hook in
%  \@nocite so it applies to \cite, \citen, and \nocite.
%
\AtBeginDocument{\@ifundefined{Hy@backout}{}{%
  \@ifundefined{BRorg@citex}{}{\global\let\@citex\BRorg@citex}%
  \global\let\BR@citex\@citex
  \global\let\@citeorg@nocite\@nocite % use my own hook -> into \@nocite
  \gdef\@nocite#1{\@citeorg@nocite{#1}\Hy@backout{#1}}%
}}

%  compatability with ooold LaTeX
\@ifundefined{G@refundefinedtrue}{\let\G@refundefinedtrue\relax}{}

%  compatability with babel:  Prevent it from redefining \@citex
\@ifundefined{@safe@activesfalse}{\let\@safe@activesfalse\relax}{}
\@ifundefined{bbl@cite@choice}{}{\@ifundefined{org@@citex}{}%
  {\let\org@@citex\@citex}}% Prevent stomping by babel

\citenum % execute restore-catcodes

% Aliases:
\def\citenum{\citen}
\def\citeonline{\citen}

\endinput
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


                   CITE.STY

Modify LaTeX's normal citation mechanism for improved handling of numeric
citations, behaving as follows:

o Put a comma and a small space between each citation number. The option
  [nospace] removes that space, and the option [space] replaces it with
  an ordinary inter-word space.

o Compress lists of three or more consecutive numbers to one number range
  which can be split, with difficulty, after the dash.  All numbers should
  be greater than zero.  E.g., if you used to get the (nonsense) list 
  [7,5,6,?,4,9,8,Einstein,6], then this style will give [?,Einstein,4-6,6-9].
  Compression of ranges is disabled by the [nocompress] package option.

o Sort citations into ascending order (this is the default, but may also
  be declared with the package option [sort]). The [nosort] package option
  turns off sorting.  Sortable citations must fit one of these forms:
    1.  <number>
    2.  <optional-char><number><optional-char>
    3.  <number><separator-char><number>
  Forms 1 and 2 are really the same, and they mix well, but form 3 is
  separate and if used simultaneously with form 1 the citations become,
  ummm, mixed.  Non-sortable forms (those not listed) are printed before
  all sortable forms.  Here <number> means a positive integer (natural 
  number) less than some limit (different for each form), <optional-char>
  is a single printable character (or nothing), and <separator-char> is also
  a single printable character.

o Sorting of citations with prefix and/or suffix characters is done so so 
  different prefixes are grouped separately, and suffixes form sub-lists 
  for the same number.  Compression knows about suffixes, so you can get
  lists like [18a-18c,19] or [A2,Q1,Q3-Q5].

o Allow, but discourage, line breaks within the group of citations (after
  dashes, and after punctuation). Penalties are \citepunctpenalty and
  \citemidpenalty. 

o Put a high-penalty breakpoint (value \citeprepenalty) before the citation 
  (unless there is a different penalty specified there).  Also, adjust the 
  spacing: if there is no space or if there is extra space due to some 
  punctuation, then change to one inter-word space. E.g.,   
  A space will be inserted here\cite{Larry,Curly,Moe}.

o All breaks can be forbidden with the [nobreak] package option.  They can 
  be adjusted independently by setting the parameters \citeprepenalty, 
  \citemidpenalty, and \citepunctpenalty.  Use \mathchardef to change these
  penalty values!  E.g.,  \mathchardef\citeprepenalty=9999 (Yes, that is 
  obscure but I don't want to use up counters or to pretend they are counters.)

o With package option [superscript] (or [super] for short), display citation
  numbers as superscripts (unless they have optional notes, causing them to
  be printed on-line with brackets).  Superscripted citations follow these
  additional rules:

-> Superscript citations use THE SAME INPUT FORMAT as ordinary citations; this
   style will ignore spaces before the citation, and move trailing punctuation
   before the superscript citation.  For example, "information \cite{source};"
   ignores the space before \cite and puts the semicolon before the number,
   just as if you had typed "information;$^{12}$".  You may switch off movement
   with the [nomove] package option (only relevant with [superscript]).

-> The punctuation characters that will migrate before the superscript are
   listed in the macro \CiteMoveChars, which you can redefine.  The default 
   set of characters is ".,;:";  Perhaps ! and ? should be included too, but
   they weren't listed in the APS style manual I looked at, and I agree with 
   that design choice because they put too much visual separation between the 
   cite and what it applies to.  Feel free to redefine \CiteMoveChars.  Quotes
   were listed as coming before the cite notation, but they should be typed 
   before the \cite command in any case because both on-line and superscript
   cites come after what is quoted (when citing a quotation). This gives one
   difficulty -- punctuation following quotes won't migrate inside the 
   quotation: e.g., "``Transition State Theory''\cite{Eyring}." gives out
   ``Transition State Theory''.$^8$, but you may want the period inside the
   quotes, thus: ``Transition State Theory.''$^8$.

-> Doubling of periods (.., ?., !.) is checked for and suppressed. The spacing
   after the citation is set according to the final punctuation mark moved.
   There is a problem with double periods after a capitalized abbreviation
   or directly after \@ : Both of "N.A.S.A. \cite{space}." and "et al.\@
   \cite{many}." will give doubled periods.  These can be fixed as follows:
   "N.S.A\@. \cite{space}." and "et al.\ \cite{many}.". The NSA example 
   gives the wrong spacing when there is no citation.  Sorry.  Use "\ " after
   abbreviations like et al. to get the right spacing within a sentence whether
   or not a citation follows.

-> Remember, these rules regarding punctuation only apply when the [super]
   or [superscript] option was given (or overcite.sty used) and the [nomove]
   option was NOT given.

o Define \citen to get just the numbers without the brackets or superscript
  and extra formatting.  Aliases are \citenum and \citeonline for easy
  conversion to other citation packages.

o All of \nocite, \cite, and \citen ignore spaces in the input tags.

Although each \cite command sorts its numbers, better compression into
ranges can usually be achieved by carefully selecting the order of the
\bibitem entries or the order of initial citations when using BibTeX.
Having the entries pre-sorted will also save processing time, especially
for long lists of numbers.

Customization:
~~~~~~~~~~~~~~
There are several options for \usepackage{cite}, some already mentioned.

 [superscript] use superscrpts for cites without optional notes
 [super]       alias for [superscript] (like natbib)
 [verbose]     UNUSED NOW! (do repeat duplicate warnings)
 [ref]         uses the format "[Ref.~12, given note]" (useful with 
               the superscript option)
 [nospace]     eliminates the spaces after commas in the number list
 [space]       uses a full inter-word space after the commas
 [nobreak]     eliminate all line-breaks
 [nosort]      prevents sorting of the numbers (default is to sort, and the...
 [sort]        option is provided for completeness).
 [nomove]      prevents moving the superscript cite after punctuation.
 [move]        is the default
 [noadjust]    disables `smart' handling of space before a cite
 [adjust]      is the default
 [nocompress]  inhibit compression of consecutive numbers into ranges
 [compress]    is the default
 [biblabel]    define the bibliography label as a superscript

If your citations are not numeric, then you should probably not use 
cite.sty, but if you must, then at least use the [nosort,nocompress] 
options.

There are several commands that you may redefine (using \renewcommand
or \def) to change the formatting of citation lists:

command          function                    default
----------       -----------------------     ----------------------------
\citeform        reformats each number       nothing
\citepunct       printed between numbers     comma + penalty + thin space
\citeleft        left delimiter of list      [
\citeright       right delimeter of list     ]
\citemid         printed before note         comma + penalty + space
\citedash        used in a compressed range  endash + penalty
\CiteMoveChars   charcters that move         .,:;
\OverciteFont    font sel. for superscripts  \fontsize{\sf@size}...

The left/mid/right commands don't affect the formatting of superscript
citations.  You may use \renewcommand to change any of these.  Remember,
these commands are extensions made by this package; they are not regular
LaTeX.  Some examples of changes:

1: \renewcommand\citeform[1]{\romannumeral 0#1}} % roman numerals i,vi
2: \renewcommand\citeform[1]{(#1)} % parenthesized numbers (1)-(5),(9)
3: \renewcommand\citeform{\thechapter.}  % by chapter: ^{2.18-2.21}
4: \renewcommand\citepunct{,} % no space and no breaks at commas
5: \renewcommand\citemid{; }  % semicolon before optional note
6: \renewcommand\citeleft{(}  % parentheses around list with note
   \renewcommand\citeright{)} % parentheses around list with note

The appearance of the whole citation list is governed by \@cite, (for full-
sized cites) and \@citess (for superscripts).  For more extensive changes 
to the formatting, redefine these.  For example, to get brackets around the 
list of superscript numbers you can do:

   \def\@citess#1{\textsuperscript{[#1]}}

after \makeatletter.

Related Note:  The superscript option does not affect the numbering format
of the bibliography; the "[12]" style is still the default.  To get
superscripts in the bibliography (at any time) you can define

   \renewcommand\@biblabel[1]{\textsuperscript{#1}}

Aw, OK, for your convenience, there is the [biblabel] package option that
just performs this definition (sort of).

Line breaking can be turned off using the [nobreak] option.  It can be
controlled more precisely by changing three numeric values, assigned 
with \mathchardef, for controlling the line-break penalties:

\citeprepenalty   penalty before cite         default \@highpenalty
\citemidpenalty   penalty used in \citemid    default \@medpenalty
\citepunctpenalty penalty used in \citepunct  default 1000

(Use \mathchardef assignments like \mathchardef\citemidpenalty=900.
These were chosen so as to not waste registers.)  Alternatively, the 
commands \citemid, \citedash, and \citepunct can be redefined to use 
different penalty parameters, or none at all.




% Version 1991: Ignore spaces after commas in the parameter list. Move most of
% \citen into \@cmpresscites for speed. Give the proper \spacefactor afterwards.
% Version 1992: make \citepunct hold the punctuation between numbers (for ease
% of changing).  Add \/ to allow hyphenation of previous word, and look better
% in italics.
% 1992a: Make it work with NFSS.  (Thank you C. Hamlin and Rainer Schoepf)
%
% Version 3.0 (1992):  Rewrite, including sorting.  Make entries like "4th"
% be treated properly as text.
% 3.1: Bug fixes (and Joerg-Martin Schwarz also convinced me to use \ifcat)
% 3.2: NFSS support was wrong--added \reset@font.  Suppress repetitions of
%      warnings.  Include \@extra@b@citeb hook.
% 3.3: Handle LaTeX2e options. Introduce various customization hooks.
% 3.4: Heuristics to avoid removing \hspace glue before on-line \cite.
%      Make \nocite ignore spaces in list, simplify. Aliases for \citen.
%      Compatability with amsmath (which defines \over).
% 3.5: Replace \reset@font with \selectfont so italics are preserved
%      Include \G@refundefinedtrue.  Fix cite-with-note bug (Lars Engebretsen).
% 3.6: Add nosort option.
% 3.7: Add nomove option; catcode preservation and global \@no@sparg for
%      french.sty; warnings in \nocite.
% 3.8: \citedash hook, fix token look-ahead (Heiko Selber), noadjust, babel.
% 3.9: More babel-compatibility hacks. Punctuation move with \frencspacing.
% 4.0: Combine overcite with cite: [superscript] option.  Also add [nocompress]
%      option and \CiteMoveChars; multibib hooks.
% 4.01 \bf -> \bfseries
% 4.02 Bury undouble action in a separate macro to avoid extra \fi error.
% 5.0  Hyperref and backref compatability! Penalty parameters and [nobreak]. 
%      Letter prefix and suffix sorting. Stop suppressing multiple warnings.
% 5.1  Fix a missing "b@" (disappearing named cites), fix nosort
% 5.2  More robust treatment of non-numbers
% 5.3  Handle sort/compress of compound citation numbers (number by chapter)
%      such as 3.18 or 5-3.  Note that these compounds cannot have prefix or
%      suffix letters (not enough bits in the maximum TeX number).
%
% TODO: other sorting, like dictionary or roman numeral
% TODO: create special "final punct" that could be ", and " and likewise
%       a "single punct" that could be " and " 
%
% Send problem reports to asnd@triumf.ca

Test file integrity:  ASCII 32-57, 58-126:  !"#$%&'()*+,-./0123456789
:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~