Debian Codegraph @

texmf-dist/source/latex/bnumexpr/bnumexpr.dtx

raw
1 % -*- coding: utf-8; time-stamp-format: "%02d-%02m-%:y at %02H:%02M:%02S %Z" ; time-stamp-active: nil -*-
2 %<*dtx>
3 \def\bnedtxtimestampaux #1<#2>{#2}
4 \edef\bnedtxtimestamp{\bnedtxtimestampaux Time-stamp: <27-09-2025 at 18:04:58 CEST>}
5 \catcode0 0 \catcode`\\ 12
6 ^^@iffalse
7 %</dtx>
8 %<*drv>
9 %% ---------------------------------------------------------------
10 \def\bnepackdate{2025/09/27}%
11 \def\bneversion {1.7b}%
12 %</drv>
13 %<*readme>--------------------------------------------------------
14 | Source: bnumexpr.dtx
15 | Version: v1.7b, 2025/09/27
16 | Author: Jean-François Burnol
17 | Info: Expressions with big integers
18 | License: LPPL 1.3c
19
20 Usage and aim
21 =============
22
23 The package provides `\bnumeval`.
24
25 To use with LaTeX, add `\usepackage{bnumexpr}` to the preamble.
26 To use with Plain TeX, execute `\input miniltx.tex` then
27 `\input bnumexpr.sty`.
28
29 `\bnumeval` is an expandable expression parser extending
30 LaTeX's `\inteval` (a wrapper of eTeX's `\numexpr`).
31
32 Compared to `\inteval`, the added functionalities are:
33
34 - support for arbitrarily big integers,
35 - `//` for floored division,
36 - `/:` for the associated remainder,
37 - `^` and `**` for powers,
38 - `!` for factorials,
39 - comma separated calculations,
40 - space character or underscore may serve to separate digits,
41 - `0b`, `0o` and `0x` are recognized as prefixes for binary, octal,
42 and hexadecimal inputs (in addition to the TeX prefixes `'` and `"`),
43 - letters used for hexadecimal input may be in lowercase,
44 - optional argument `[b]`, `[o]` or `[h]` for converting the output
45 to binary, octal, or hexadecimal (`[ha]` for lowercase hexadecimal),
46 - quasi complete customizability and extensibility of the syntax.
47
48 The last item means that not only is it possible to map any given
49 operation to a macro of one's own choosing, it is also possible to add
50 new infix or postfix operators to the syntax.
51
52 Installation
53 ============
54
55 The package is provided by the major LaTeX distributions.
56
57 For a manual installation obtain `bnumexpr.dtx`, from CTAN:
58
59 > <http://www.ctan.org/pkg/bnumexpr>
60
61 Run `"etex bnumexpr.dtx"` to extract in particular the style file
62 `bnumexpr.sty` (and other files inclusive of this `README.md`). Put it
63 where (La)TeX can find it and you are done.
64
65 To build the documentation check the instructions in extracted
66 file `bnumexpr.tex`.
67
68 License
69 =======
70
71 Copyright © 2014-2022, 2025 Jean-François Burnol
72
73 | This Work may be distributed and/or modified under the
74 | conditions of the LaTeX Project Public License 1.3c.
75 | This version of this license is in
76
77 > <http://www.latex-project.org/lppl/lppl-1-3c.txt>
78
79 | and version 1.3 or later is part of all distributions of
80 | LaTeX version 2005/12/01 or later.
81
82 This Work has the LPPL maintenance status "author-maintained".
83
84 The Author and Maintainer of this Work is Jean-François Burnol.
85
86 This Work consists of the main source file and its derived files
87
88 bnumexpr.dtx, bnumexpr.sty, bnumexpr.pdf, bnumexpr.tex,
89 bnumexprchanges.tex, README.md
90
91 %</readme>--------------------------------------------------------
92 %<*!readme>
93 %% ---------------------------------------------------------------
94 %% The bnumexpr package: Expressions with big integers
95 %% Copyright (C) 2014-2022, 2025 Jean-Francois Burnol
96 %%
97 %</!readme>
98 %<*changes>-------------------------------------------------------
99 \item[1.7b (2025/09/27)] Add to documentation an expandable macro
100 computing Fibonacci numbers and illustrate it with the computation
101 of $F_{10000}$.
102 \item[1.7a (2025/09/14)]\mbox{}\newline
103 \textbf{Bug fix:} inputs consisting exclusively of zeros and underscores
104 following an hexadecimal or other prefix caused a crash due to a |1.7|
105 regression. \emph{Yes, we have now added a long delayed test suite doing
106 more than checking only those examples as included in the documentation.}
107
108 \textbf{New feature:} if the parser expects a value and finds only
109 underscores (they are allowed as separators for blocks of digits) before
110 hitting an operator, an opening parenthesis (or |\bnumexpr| sub-expression),
111 a comma, or the end of the expression, it will consider it has fetched the
112 value zero. Formerly, for example expressions reduced to |_|, |__|, or
113 entirely empty, or for example |_^_| all reported errors.
114
115 \textbf{Other:} \cs{evaltohex} which signaled via an expandable
116 error its deprecation since |1.6| has been removed.
117 \item[1.7 (2025/09/13)]\mbox{}\newline
118 \textbf{Bug fix:} inputs with an underscore immediately after a
119 hexadecimal (or other) prefix caused a crash due to a |1.6| regression.
120
121 \textbf{New features:}
122 \begin{itemize}
123 \item Support for hexadecimal input using letters in lowercase.
124 \item Optional argument |[ha]| for lowercase hexadecimal output.
125 \end{itemize}
126 \item[1.6a (2025/09/07)]\mbox{}\newline
127 \textbf{Bug fix:} the |1.6| support for Babel-active
128 characters worked with |\bnumeval| (which is recommended interface)
129 but not with |\bnumexpr|.
130 \item[1.6 (2025/09/05)]\mbox{}\newline
131 \textbf{Breaking changes:}
132 \begin{itemize}[noitemsep]
133 \item Release |1.4n| or later of the \xintname bundle is required
134 (for those components actually used, which by default are
135 \xintkernelname, \xintcorename and \xintbinhexname).
136 \item \cs{evaltohex} is deprecated and causes
137 an error signaling it. \emph{It was removed at |1.7a|}. Use new |\bnumeval||[h]|.
138 \item \cs{bnumexprsetup} was deprecated at |1.5| and kept as alias of
139 |\bnumsetup|. It has now been removed.
140 \item \cs{bnumprintonetohex} and \cs{bnumhextodec}, which were documented
141 as customizable do not exist anymore. Check the documentation for
142 |\bnumprintonehex| and |\bnumsetup|'s key |hextodec|.
143 \item Under the |custom| option, not only \xintcorename but
144 also \xintbinhexname are not loaded.
145 Use |customcore| to avoid that. There is also |custombinhex|.
146 \end{itemize}
147
148 \textbf{Bug fix:}
149 An underscore |_| located in front of a number used to cause
150 an error. It is now ignored.
151
152 \textbf{New features:}
153 \begin{itemize}
154 \item |0b|, |0o| and |0x| are recognized as prefixes for binary, octal, and
155 hexadecimal inputs. And |'| is recognized as prefix for octal input, in
156 addition to |"| for hexadecimal.
157 \item |\bnumeval| accepts an optional argument |[b]| or |[o]| or
158 |[h]| for automatic conversion of the calculated value
159 (or comma separated values) to
160 respectively binary, octal, or hexadecimal.
161 \item Babel-active characters (such as |:| and |!| with French)
162 do not need any preventive measures anymore such as using
163 |\string!| in place of |!|.
164 \item |\bnumsetup| can now be used also to customize which macros
165 implement conversion from decimal to other bases.
166 \end{itemize}
167 The documentation was extensively revised and made more user-friendly.
168 \item[1.5 (2021/05/17)]
169 \begin{itemize}
170 \item \textbf{breaking change:} the power operators act now in a right
171 associative way; this has been announced at \xintexprname as a probable
172 future evolution, and is implemented in anticipation here now.
173 \item \textbf{fix two bugs} (imported from upstream \xintexprname) regarding
174 hexadecimal input: impossibility to use |"\foo| syntax (one had to do
175 |\expandafter"\foo| which is unexpected constraint; a very longstanding
176 \xintexprname bug) and issues with leading zeros (since \xintexprname |1.2m|).
177 \item renamed \cs{bnumexprsetup} into |\bnumsetup|; the former remains
178 available but is deprecated. [REMOVED AT |1.6|]
179 \item the customizability and extendibility is now total:
180 \begin{enumerate}
181 \item |\bnumprintone|, |\bnumprintonetohex|, |\bnumprintonesep|, |\bnumhextodec|,
182 \item |\bnumdefinfix| which allows to add extra infix operators,
183 \item |\bnumdefpostfix| which allows to add extra postfix operators.
184 \end{enumerate}
185 \item |\bnumsetup|, |\bnumdefinfix|, |\bnumdefpostfix| obey the
186 |\xintglobaldefstrue| and |\xintverbosetrue| settings.
187 \item documentation is extended, providing details regarding the precedence
188 model of the parser, as inherited from upstream \xintexprname; also an
189 example of usage of |\bnumsetup| is included on how to transform
190 |\bnumeval| into a calculator with fractions.
191 \end{itemize}
192
193 \item[1.4a (2021/05/13)]
194 \begin{itemize}
195 \item fix undefined control sequences errors encountered by the parser in
196 case of either extra or missing closing parenthesis (due to a problem
197 in technology transfer at |1.4| from upstream \xintexprname).
198 \item fix |\BNE_Op_opp| must now be \emph{f}-expandable (also caused as a
199 collateral to the technology transfer).
200 \item fix user documentation regarding the constraints applying to the user
201 replacement macros for the core algebra, as they have changed at |1.4|.
202 \end{itemize}
203
204 \item[1.4 (2021/05/12)]
205 \begin{itemize}
206 \item technology transfer from \xintexprname |1.4| of
207 |2020/01/31|. The |\expanded| primitive is now required (TeXLive
208 2019).
209 \item addition to the syntax of the |"| prefix for hexadecimal
210 input.
211 \item addition of \cs{evaltohex} which is like |\bnumeval| with an extra
212 conversion step to hexadecimal notation.
213 \end{itemize}
214
215 \item[1.2e (2019/01/08)]
216 Fixes a documentation glitch (extra braces when mentioning |\the\numexpr| or
217 |\thebnumexpr|).
218
219 \item[1.2d (2019/01/07)]
220 \begin{itemize}
221 \item requires \xintcorename |1.3d| or later (if not using option |custom|).
222 \item adds |\bnumeval|\marg{expression} user interface.
223 \end{itemize}
224
225 \item[1.2c (2017/12/05)] \textbf{Breaking changes:}
226 \begin{itemize}
227 \item requires \xintcorename |1.2p| or later (if not using option |custom|).
228 \item |divtrunc| key of \cs{bnumexprsetup} is renamed to |div|.
229 \item the |//| and |/:| operators are now by default associated to the
230 \emph{floored} division. This is to keep in sync with the change of
231 \xintcorename at |1.2p|.
232 \item for backwards compatibility, one
233 may add to existing document:\newline
234 \string\bnumexprsetup\{div=\string\xintiiDivTrunc,
235 mod=\string\xintiiModTrunc\}
236 \end{itemize}
237
238 \item[1.2b (2017/07/09)]
239 \begin{itemize}
240 \item the |_| may be used to separate visually blocks of digits in long
241 numbers.
242 \end{itemize}
243
244 \item[1.2a (2015/10/14)]
245 \begin{itemize}
246 \item requires \xintcorename |1.2| or later (if not using option |custom|).
247 \item additions to the syntax: factorial |!|, truncated division
248 |//|, its associated modulo |/:| and |**| as alternative to |^|.
249 \item all options removed except |custom|.
250 \item new command \cs{bnumexprsetup} which replaces the commands such as
251 |\bnumexprusesbigintcalc|.
252 \item the parser is no more limited to numbers with at most 5000 digits.
253 \end{itemize}
254
255 \item[1.1b (2014/10/28)]
256 \begin{itemize}
257 \item README converted to |markdown/pandoc| syntax,
258 \item the package now loads only |xintcore|, which belongs to |xint| bundle
259 version |1.1| and extracts from the earlier |xint| package the core arithmetic
260 operations as used by |bnumexpr|.
261 \end{itemize}
262
263 \item[1.1a (2014/09/22)]
264 \begin{itemize}
265 \item added |l3bigint| option to use experimental \LaTeX3
266 package of the same name,
267
268 \item added Changes and Readme sections to the documentation,
269
270 \item better |\BNE_protect| mechanism for use of
271 |\bnumexpr...\relax| inside an |\edef| (without |\bnethe|). Previous one,
272 inherited from |xintexpr.sty 1.09n|, assumed that the |\.=<digits>| dummy
273 control sequence encapsulating the computation result had \cs{relax}
274 meaning. But removing this assumption was only a matter of letting
275 |\BNE_protect| protect two, not one, tokens. This will be backported to
276 next version of \xintexprname, naturally (done with |xintexpr.sty 1.1|).
277 \end{itemize}
278
279 \item[1.1 (2014/09/21)] First release. This is down-scaled from the
280 (development version of) \xintexprname. Motivation came the previous day
281 from a chat with \textsc{Joseph Wright} over big int status in \LaTeX3.
282 The |\bnumexpr...\relax| parser can be used on top of big int macros of
283 one's choice. Functionalities limited to the basic operations. I leave
284 the power operator |^| as an option.
285 %</changes>-------------------------------------------------------
286 %<*drv>-----------------------------------------------------------
287 %% latex bnumexpr.tex (thrice) && dvipdfmx bnumexpr.dvi
288 %% to produce bnumexpr.pdf
289 %%
290 %% or pdflatex bnumexpr.tex (no need to change \Withdvipdfmx toggle,
291 %% pdf engine will be detected automatically)
292 %%
293 \NeedsTeXFormat{LaTeX2e}[2020/02/02]
294 \ProvidesFile{bnumexpr.tex}%
295 [\bnepackdate\space v\bneversion\space driver file for %
296 bnumexpr documentation (JFB)]%
297 \PassOptionsToClass{a4paper,fontsize=11pt}{scrartcl}
298 \chardef\Withdvipdfmx 1 %
299 \chardef\NoSourceCode 0 % replace 0 by 1 for not including source code
300 \input bnumexpr.dtx
301 %%% Local Variables:
302 %%% mode: latex
303 %%% End:
304 %</drv>-----------------------------------------------------------
305 %<*dtx>
306 ^^@fi ^^@catcode`\ 0 \catcode0 12
307 \chardef\noetex 0
308 \ifx\numexpr\undefined \chardef\noetex 1 \fi
309 \ifnum\noetex=1 \chardef\extractfiles 0 % extract files, then stop
310 \else
311 \ifx\ProvidesFile\undefined
312 \chardef\extractfiles 0 % no LaTeX2e; etex, ... on bnumexpr.dtx
313 \else % latex/pdflatex on bnumexpr.tex or on bnumexpr.dtx
314 \ifx\Withdvipdfmx\undefined
315 % latex run is on bnumexpr.dtx, we will extract all files
316 \chardef\extractfiles 1 % 1 = extract and typeset, 2=only typeset
317 \chardef\Withdvipdfmx 0 % 0 = pdflatex or latex+dvips, 1 = dvipdfmx
318 \chardef\NoSourceCode 0 % 0 = include source code, 1 = do not
319 \NeedsTeXFormat{LaTeX2e}[2020/02/02]%
320 \PassOptionsToClass{a4paper,fontsize=11pt}{scrartcl}%
321 \else % latex run is on bnumexpr.tex,
322 \chardef\extractfiles 2 % no extractions
323 \fi
324 \ProvidesFile{bnumexpr.dtx}[bundle source (\bnepackdate)]%
325 \fi
326 \fi
327 \ifnum\extractfiles<2 % extract files
328 \def\MessageDeFin{\newlinechar10 \let\Msg\message
329 \Msg{^^J}%
330 \Msg{********************************************************************^^J}%
331 \Msg{*^^J}%
332 \Msg{* To finish the installation you have to move the following^^J}%
333 \Msg{* files into a directory searched by TeX:^^J}%
334 \Msg{*^^J}%
335 \Msg{*\space\space\space\space bnumexpr.sty^^J}%
336 \Msg{*^^J}%
337 \Msg{* To produce the documentation run latex thrice on bnumexpr.tex^^J}%
338 \Msg{* then dvipdfmx on bnumexpr.dvi. (ignore the dvipdfmx warnings)^^J}%
339 \Msg{*^^J}%
340 \Msg{* Happy TeXing!^^J}%
341 \Msg{*^^J}%
342 \Msg{********************************************************************^^J}%
343 }%
344 \begingroup
345 \input docstrip.tex
346 \askforoverwritefalse
347 \generate{\nopreamble\nopostamble
348 \file{README.md}{\from{bnumexpr.dtx}{readme}}
349 \usepreamble\defaultpreamble
350 \usepostamble\defaultpostamble
351 \file{bnumexprchanges.tex}{\from{bnumexpr.dtx}{changes}}
352 \file{bnumexpr.tex}{\from{bnumexpr.dtx}{drv}}
353 \file{bnumexpr.sty}{\from{bnumexpr.dtx}{package}}}
354 \endgroup
355 \fi % end of file extraction
356 \ifnum\extractfiles=0
357 % direct tex/etex/xetex on bnumexpr.dtx, files now extracted, stop
358 \MessageDeFin\expandafter\end
359 \fi
360 % no use of docstrip to extract files if latex compilation was on bnumexpr.tex
361 \ifdefined\MessageDeFin\AtEndDocument{\MessageDeFin}\fi
362 %-------------------------------------------------------------------------------
363 \documentclass {scrartcl}
364 %%% START OF CUSTOM doc.sty LOADING (May 21, 2022)
365 \usepackage{doc}[=v2]
366 % As explained above I was formerly using scrdoc hence ltxdoc indirectly.
367 % Let's emulate here the little I appear to need from ltxdoc.cls and
368 % srcdoc.cls.
369 %
370 \AtBeginDocument{\MakeShortVerb{\|}}
371 \DeclareFontShape{OT1}{cmtt}{bx}{n}{<-> ssub * cmtt/m/n}{}
372 \DeclareFontFamily{OMS}{cmtt}{\skewchar\font 48} % '60
373 \DeclareFontShape{OMS}{cmtt}{m}{n}{<-> ssub * cmsy/m/n}{}
374 \DeclareFontShape{OMS}{cmtt}{bx}{n}{<-> ssub * cmsy/b/n}{}
375 \DeclareFontShape{OT1}{cmss}{m}{it}{<->ssub*cmss/m/sl}{}
376 \ifnum\NoSourceCode=1
377 \OnlyDescription
378 \fi
379 \CodelineNumbered
380 \DisableCrossrefs
381 % \setcounter{StandardModuleDepth}{1} % we don't use this mechanism currently
382 \def\cmd#1{\cs{\expandafter\cmd@to@cs\string#1}}
383 \def\cmd@to@cs#1#2{\char\number`#2\relax}
384 % This is hyperref compatible (contrarily to using \bslash)
385 % I need the \detokenize due to usage in implementation part with names
386 % containing the underscore for example
387 \DeclareRobustCommand\cs[1]{\texttt{\textbackslash\detokenize{#1}}}
388 \providecommand\marg[1]{%
389 {\ttfamily\char`\{}\meta{#1}{\ttfamily\char`\}}}
390 \providecommand\oarg[1]{%
391 {\ttfamily[}\meta{#1}{\ttfamily]}}
392 \providecommand\parg[1]{%
393 {\ttfamily(}\meta{#1}{\ttfamily)}}
394
395 % There is very little we seem to need from the scrdoc extras: page geometry
396 % is set by geometry package and a4paper option from bnumexpr.tex file. So it
397 % seems I only need the hologo loading:
398 \usepackage{hologo}
399 \DeclareRobustCommand*{\eTeX}{\hologo{eTeX}}%
400 %\DeclareRobustCommand*{\LuaTeX}{\hologo{LuaTeX}}%
401 %
402 %%% end of ltxdoc+srcdoc emulation.
403
404 \ifnum\NoSourceCode=1 \OnlyDescription\fi
405 \usepackage{ifpdf}
406 \ifpdf\chardef\Withdvipdfmx 0 \fi
407 \makeatletter
408 \ifnum\Withdvipdfmx=1
409 % I think those are mostly obsolete...
410 \@for\@tempa:=hyperref,bookmark,graphicx,xcolor,pict2e\do
411 {\PassOptionsToPackage{dvipdfmx}\@tempa}
412 %
413 \PassOptionsToPackage{dvipdfm}{geometry}
414 \PassOptionsToPackage{bookmarks=true}{hyperref}
415 \PassOptionsToPackage{dvipdfmx-outline-open}{hyperref}
416 %\PassOptionsToPackage{dvipdfmx-outline-open}{bookmark}
417 %
418 \def\pgfsysdriver{pgfsys-dvipdfm.def}
419 \else
420 \PassOptionsToPackage{bookmarks=true}{hyperref}
421 \fi
422 \makeatother
423
424 \pagestyle{headings}
425
426 \usepackage[T1]{fontenc}
427
428 \usepackage[hscale=0.66,vscale=0.75]{geometry}
429
430 % requires newtxtt 1.05 or later
431 \usepackage[zerostyle=d,scaled=0.95,straightquotes]{newtxtt}
432 \renewcommand\familydefault\ttdefault
433 \usepackage[noendash]{mathastext}
434 \renewcommand\familydefault\sfdefault
435
436 \usepackage{graphicx}
437 \usepackage[dvipsnames]{xcolor}
438 \definecolor{joli}{RGB}{225,95,0}
439 \definecolor{JOLI}{RGB}{225,95,0}
440 \definecolor{BLUE}{RGB}{0,0,255}
441 %% \definecolor{niceone}{RGB}{38,128,192}% utilisé avant pour urlcolor
442 %% \colorlet{jfverbcolor}{yellow!5}
443 \colorlet{smallverbcolor}{magenta}% 1.6 release NavyBlue RoyalPurple
444 \colorlet{softwrapcolor}{blue}
445 \colorlet{digitscolor}{olive}% 1.6 {OrangeRed}
446
447 % transféré de xint-manual.tex (maintenant dans xint.dtx)
448 \DeclareFontFamily{U}{MdSymbolC}{}
449 \DeclareFontShape {U}{MdSymbolC}{m}{n}{<-> MdSymbolC-Regular}{}
450
451 \makeatletter
452 \newbox\cdbx@SoftWrapIcon
453 \def\cdbx@SetSoftWrapBox{%
454 \setbox\cdbx@SoftWrapIcon\hb@xt@\z@
455 {\hb@xt@\fontdimen2\font
456 {\hss{\color{softwrapcolor}\usefont{U}{MdSymbolC}{m}{n}\char"97}\hss}%
457 \hss}%
458 }
459 \makeatother
460
461 \usepackage[french, main=english]{babel}
462
463 \usepackage{hyperref}
464 \hypersetup{%
465 %linktoc=all,%
466 breaklinks=true,%
467 colorlinks=true,%
468 urlcolor=Cerulean,% 1.6 ProcessBlue,%SkyBlue
469 linkcolor=PineGreen,%
470 pdfauthor={Jean-François Burnol},%
471 pdftitle={The bnumexpr package},%
472 pdfsubject={Arithmetic with TeX},%
473 pdfkeywords={Expansion, arithmetic, TeX},%
474 pdfstartview=FitH,%
475 pdfpagemode=UseOutlines}
476 \usepackage{bookmark}
477
478 % importé de xint 2021/05/14
479 \newcommand\RaisedLabel[2][6]{%
480 \vspace*{-#1\baselineskip}%
481 \begingroup
482 \let\leavevmode\relax\phantomsection
483 \label{#2}%
484 \endgroup
485 \vspace*{#1\baselineskip}%
486 }
487
488 %---- \verb, and verbatim like `environments'. \MicroFont et \MacroFont
489 \makeatletter
490 \def\MicroFont {\ttfamily\cdbx@SetSoftWrapBox\color{smallverbcolor}}
491 %\def\MicroFont {\ttfamily \color[named]{OrangeRed}\cdbx@SetSoftWrapBox }
492
493 % \MacroFont est utilisé par macrocode, mais sa définition est figée dans
494 % \macro@font au \begin{document}
495
496 \def\MacroFont {\ttfamily \baselineskip12pt\relax }
497
498 %--- November 4, 2014 making quotes straight. (maintenu pour *)
499 %
500 % there is no hook in \macrocode after \dospecials etc. Thus I will need to
501 % take the risk that some future evolution of doc.sty (or perhaps scrdoc)
502 % invalidates the following.
503 %
504 % Actually, I should not at all rely on the doc class, I should do it all by
505 % myself.
506
507 \def\macrocode{\macro@code
508 \frenchspacing \@vobeyspaces
509 \makestarlowast % \makequotesstraight (obsolète)
510 \xmacro@code }
511
512 %--- lower * symbol in text
513
514 \def\lowast{\raisebox{-.25\height}{*}}
515 \catcode`* \active
516 \def\makestarlowast {\let*\lowast\catcode`\*\active}%
517 \catcode`* 12
518
519 \def\verb #1%
520 {%
521 \relax\leavevmode\null
522 \begingroup
523 \MicroFont
524 \def\@makeletter##1{\catcode`##1=11\relax}%
525 % \scantokens will have a result of inserting a space after cs's.
526 % hence the need to have the catcodes of things like _ right.
527 % I also need < for > for code comments
528 % \dospecials at begin document
529 % \do\ \do\\\do\{\do\}\do\$\do\&\do\#\do\^\do\_\do\%\do\~\do\|
530 \odef\dospecials{\dospecials\do\:\do\<\do\>\do\-\do\+}%
531 % naturally won't work in footnotes though.
532 % this code is truly not satisfying, but enough for my needs here.
533 \catcode`\\ 11 \catcode`## 11
534 % I don't do \catcode`\% 11 to avoid surprises in code comments
535 % if my |...| has a linebreak
536 \def\@jfverb ##1#1{\let\do\@makeletter \dospecials
537 % lowering a bit the *
538 \makestarlowast
539 %\let\do\do@noligs \verbatim@nolig@list % not needed here
540 \@vobeyspaces\everyeof{\noexpand}%
541 \expandafter\@@jfverb\scantokens{##1}\relax}%
542 \@jfverb
543 }%
544
545 % Pour v1.2 je laisse tomber les colorbox. À la place je change la couleur
546 % du texte. J'utilise le violet de xint. (mis dans \Microfont)
547 % \def\@@@jfverb #1{\ifcat\noexpand#1$% $
548 % \endgroup\else
549 % % \penalty\z@
550 % \discretionary{\copy\cdbx@SoftWrapIcon}{}{}%
551 % % \colorbox{jfverbcolor}{\strut #1}% unfortunately, impacts pdf size
552 % #1%
553 % \expandafter\@@@jfverb\fi }
554
555 \def\@@jfverb #1{\ifcat\noexpand#1\relax\endgroup\@\else
556 \discretionary{\copy\cdbx@SoftWrapIcon}{}{}%
557 #1\expandafter\@@jfverb\fi }
558
559 \makeatother
560
561 % Lundi 17 mai 2021
562 % J'ajoute un export automatique des exemples qui servira de poor man test
563 % suite
564 \newwrite\tests
565 \immediate\openout\tests=\jobname-out.txt
566
567 % mardi 13 octobre 2015
568 % le | n'est pas encore actif ici!
569 \catcode`| \active
570 % first one uses \bnumexpr deliberately with no \bnethe to test 1.4.
571 \def\bneshow #1%
572 {|\bnumexpr#1\relax|%
573 \edef\TEMP{\bnumexpr#1\relax}%
574 \immediate\write\tests{\detokenize{\bnumexpr#1\relax}^^J\TEMP}%
575 \[\color{digitscolor}\TEMP\]\ignorespaces}
576 \def\bneshowthe #1%
577 {|\bnethe\bnumexpr#1\relax|%
578 \edef\TEMP{\bnethe\bnumexpr#1\relax}%
579 \immediate\write\tests{\detokenize{\bnethe\bnumexpr#1\relax}^^J\TEMP}%
580 \[\color{digitscolor}\TEMP\]\ignorespaces}
581 \def\bneshoweval #1%
582 {|\bnumeval|\unskip|{{#1}}|% problème de grab de mon \verb d'où {{..}}
583 \edef\TEMP{\bnumeval{#1}}%
584 \immediate\write\tests{\detokenize{\bnumeval{#1}}^^J\TEMP}%
585 \[\color{digitscolor}\TEMP\]\ignorespaces}
586 % Ceci pour l'exemple avec " en output. Ici on utilise un display
587 % donc en mode math, mais mathastext ne s'occupe pas de "...
588 {\catcode`"\active\gdef"{\texttt{\string"}}}
589 \def\bneshowevaltohex #1%
590 {|\bnumeval|\unskip|[h]{#1}|%
591 \edef\TEMP{\bnumeval[h]{#1}}%
592 \immediate\write\tests{\detokenize{\bnumeval[h]{#1}}^^J\TEMP}%
593 \[\mathcode`"="8000 \color{digitscolor}\TEMP\]\ignorespaces}
594 \def\bneshowevaltolowhex #1%
595 {|\bnumeval|\unskip|[ha]{#1}|%
596 \edef\TEMP{\bnumeval[ha]{#1}}%
597 \immediate\write\tests{\detokenize{\bnumeval[ha]{#1}}^^J\TEMP}%
598 \[\mathcode`"="8000 \color{digitscolor}\TEMP\]\ignorespaces}
599 \def\bneshowevaltooct #1%
600 {|\bnumeval|\unskip|[o]{#1}|%
601 \edef\TEMP{\bnumeval[o]{#1}}%
602 \immediate\write\tests{\detokenize{\bnumeval[o]{#1}}^^J\TEMP}%
603 \[\color{digitscolor}\TEMP\]\ignorespaces}
604 \def\bneshowevaltobin #1%
605 {|\bnumeval|\unskip|[b]{#1}|%
606 \edef\TEMP{\bnumeval[b]{#1}}%
607 \immediate\write\tests{\detokenize{\bnumeval[b]{#1}}^^J\TEMP}%
608 \[\color{digitscolor}\TEMP\]\ignorespaces}
609 \catcode`| 12
610 % Mardi 14 juin 2022
611 \def\bnedotest #1{\oodef\x{#1}\immediate\write\tests{\detokenize{#1}^^J\x}}
612
613 % toujours pénible de devoir se battre avec les espaces verticaux de TeX
614 \AtBeginDocument{%
615 % \message{ICI below: \the\belowdisplayskip^^J%
616 % belowshort: \the\belowdisplayshortskip^^J%
617 % above: \the\abovedisplayskip^^J%
618 % aboveshor: \the\abovedisplayshortskip^^J%
619 % }%
620 \belowdisplayskip\belowdisplayshortskip
621 \abovedisplayskip\abovedisplayshortskip
622 }
623
624 \usepackage{xspace}
625
626 \def\bnumname
627 {\href{http://www.ctan.org/pkg/bnumexpr}{bnumexpr}\xspace }%
628
629 \def\bnumnameimp
630 {\texorpdfstring
631 {\hyperref[sec:bnumexprcode]{{\color{smallverbcolor}\ttzfamily bnumexpr}}}
632 {bnumexpr}%
633 \xspace }%
634
635 \def\bnumnameuserdoc
636 {\texorpdfstring
637 {\hyperref[titlepage]{{\color{PineGreen}\ttzfamily bnumexpr}}}
638 {bnumexpr}%
639 \xspace }%
640
641 \def\xintkernelname
642 {\href{http://www.ctan.org/pkg/xint}{xintkernel}\xspace }%
643 \def\xintcorename
644 {\href{http://www.ctan.org/pkg/xint}{xintcore}\xspace }%
645 \def\xintname
646 {\href{http://www.ctan.org/pkg/xint}{xint}\xspace }%
647 \def\xintfracname
648 {\href{http://www.ctan.org/pkg/xint}{xintfrac}\xspace }%
649 \def\xintexprname
650 {\href{http://www.ctan.org/pkg/xintexpr}{xintexpr}\xspace }%
651 \def\xintbinhexname
652 {\href{http://www.ctan.org/pkg/xint}{xintbinhex}\xspace }%
653
654 \DeclareRobustCommand\ctanpackage[1]{\href{https://ctan.org/pkg/#1}{#1}}
655
656 %\frenchspacing
657 \usepackage{bnumexpr}
658 \usepackage{xintfrac}
659 \usepackage{etoc}
660 \usepackage{centeredline}
661 \usepackage{framed}
662 \makeatletter\let\check@percent\relax\makeatother
663
664 % 2022/06/15 -- comment updated 2025/09/07
665 %
666 % latex builds at home during preparation for the release are done
667 % using --translate-file=natural.tcx for the benefit of \ALERTE.
668 % The shipped to CTAN dtx will not contain any \ALERTE execution
669 % hence can be build using latex normally.
670 %
671 % AT LONG LAST LATEX HAS MADE THE UTF-8 MACROS PROTECTED
672 % so many years waiting for that...
673 {\catcode27=12 \gdef\babar{^^[}}
674 \def\ALERTE{\edef\foo{\noexpand\AtEndDocument
675 {\immediate\write128{\babar[1;31mATTENTION METTRE À JOUR COMMENTAIRE EN LIGNE
676 \the\inputlineno\babar[0m}}}%
677 \foo}
678
679 % 2025/09/07
680 % https://tex.stackexchange.com/a/62032
681 % Et je mets [2020/02/02] dans les \NeedsTeXFormat de bnumexpr{.tex,.dtx},
682 % pour être sûr
683 \ExplSyntaxOn
684 \NewDocumentCommand{\getenv}{om}
685 {
686 \sys_get_shell:nnN { kpsewhich ~ --var-value ~ #2 } { } \l_tmpa_tl
687 \tl_trim_spaces:N \l_tmpa_tl
688 \IfNoValueTF { #1 }
689 {
690 \tl_use:N \l_tmpa_tl
691 }
692 {
693 \tl_set_eq:NN #1 \l_tmpa_tl
694 }
695 }
696 \ExplSyntaxOff
697 %
698 \getenv[\SHIPPINGNOW]{BNUMEXPR_RELEASING}
699 % \SHIPPINGNOW->\par si la variable n'est pas définie.
700 % L'expansion de \par débute par un \scan_stop: (au moins en 2025)
701 \if1\SHIPPINGNOW\def\ALERTE{\ERREUR}\fi
702
703 \usepackage[autolanguage,np]{numprint}
704 \usepackage{enumitem}
705
706 \newenvironment{TeXnote}{\par\footnotesize\medskip\noindent\textbf{\TeX-nical note:
707 }}{\par\medskip}
708
709 \begin{document}
710 \thispagestyle{empty}
711 \ttzfamily
712 \pdfbookmark[1]{Title page}{TOP}
713
714 {%
715 \normalfont\Large\parindent0pt \parfillskip 0pt\relax
716 \leftskip 2cm plus 1fil \rightskip 2cm plus 1fil
717 The \bnumname package\par
718 }
719
720 \RaisedLabel{titlepage}
721
722 {\centering
723 \textsc{Jean-François Burnol}\par
724 \footnotesize
725 jfbu (at) free (dot) fr\par
726 Package version: \bneversion\ (\bnepackdate)\par
727 {From source file \texttt{bnumexpr.dtx} of \texttt{\bnedtxtimestamp}}\par
728 }
729
730 \etocsetnexttocdepth{subsection}
731 \etocsettagdepth{code}{section}
732 \etocsettocstyle{}{}
733 \tableofcontents
734
735 \section{\cs{bnumeval}}
736 \label{sec:bnumexpr}
737 \label{sec:thebnumexpr}
738 \label{sec:bnumeval}
739
740 This \LaTeX\ package \bnumname provides |\bnumeval|, which is an expandable
741 parser of numerical expressions with big integers.
742
743 Recent \LaTeX\ has |\inteval|, which is a slim wrapper for \eTeX{}'s
744 \cs{numexpr} (embedded for twenty years in most \TeX-engines except
745 original Knuth's |tex|).
746 \begin{TeXnote}
747 More precisely \cs{inteval}\marg{expression} is equivalent (up to how \TeX{}
748 handles spaces located after in the source during tokenization, as
749 tokenization of control sequences such as \cs{relax} causes \TeX{} to ignore
750 space characters or end-of-line space after it) to:
751 \centeredline{\cs{the}\cs{numexpr}\meta{expression}\cs{relax}} In an analogous way
752 |\bnumeval|\marg{expression} has equivalent forms:
753 \centeredline{|\bnethe\bnumexpr|\meta{expression}\cs{relax}}
754 \centeredline{|\thebnumexpr|\meta{expression}\cs{relax}} For contexts where
755 the alternative forms may be useful, refer to the \autoref{sec:advanced}.
756 Everyday use needs only |\bnumeval|.
757 \end{TeXnote}
758
759 Here are the extra features from |\bnumeval| compared to \cs{inteval}:
760 \begin{enumerate}[noitemsep,label=\textbf{(\alph*)}]
761 \item It allows arbitrarily big integers, whereas \cs{inteval} is limited
762 to a maximal input equal to $\the\numexpr"7FFFFFFF\relax$
763 ($2^{31}-1$, or hexadecimal |7FFFFFFF|).
764 \item It recognizes |**| and |^| as infix operator for powers,
765 \item It recognizes |!| as postfix operator for the factorials,
766 \item The new operator |//| computes floored division with |/:| being the
767 operator for the associated remainder (the operator |/| computes
768 rounded division),
769 \item In addition to the \TeX{} prefixes |'| and |"| for octal and
770 hexadecimal, it recognizes |0b|, |0o| and |0x| for
771 binary, octal, and hexadecimal,
772 \item Letters in lowercase can be used for hexadecimal input,
773 \item The space character is ignored%
774 %
775 \footnote{It is not completely ignored, |\count37<space>| will automatically
776 be prefixed by \cs{number} and the space token delimits the integer indexing
777 the count register. Also, devious inputs using nested braces around spaces
778 may create unexpected internal situations and even break the parser.}
779 %
780 and can thus be used to separate in the
781 source blocks of digits for better readability of long numbers,
782 \item Also the underscore |_| may be used as visual digit
783 separator,
784 \item Braced material \string{...\string} encountered in the expression is automatically
785 unbraced,
786 \item Comma separated expressions are allowed,
787 \item Some idiosyncrasies of \cs{numexpr} such as |\inteval{-(1)}| causing an
788 error are avoided,
789 \item Syntax is fully customizable and extensible.
790 \end{enumerate}
791 %
792 Note the following about octal and binary inputs:
793 \begin{quote}
794 When parsing an octal number, if digits |8| or |9| are encountered, or when
795 parsing binary any digit larger than |1|, this does not cause an error but
796 ends the parsing and inserts a tacit multiplication: |\bnumeval{'118+3}|
797 computes $9\times 8 + 3$, hence obtains \bnumeval{'118+3}.
798 \end{quote}
799 %
800 Also note the following on precedences:
801 \begin{quote}
802 As is explained with complete details in \autoref{sssec:prec}, the
803 multiplication operator, as well as the division operators (including
804 modulo) are all at the same precedence level and behave in a
805 left-associative way. Example: |\bnumeval{10*10//3, 10*10/:3}|
806 evaluates to \textcolor{digitscolor}{33, 1}, not to
807 \textcolor{digitscolor}{30, 10}.
808
809 Also note that powers (which obviouly have
810 a higher precedence) work in a right-associative way so that
811 |\bnumeval{2^3^4}| will compute \textcolor{digitscolor}{$2^{81}$}, not
812 \textcolor{digitscolor}{$8^{4}$}.
813
814 The best among our users know how to set operator precedences to their own
815 liking, as they have read all the fine print in \autoref{ssec:expert}.
816 \end{quote}
817 %
818 Furthermore, |\bnumeval| recognizes an optional argument |[b]|, |[o]|, |[h]|,
819 or |[ha]| which says to have the calculation result (or comma separated results)
820 be converted to respectively binary, octal, hexadecimal (uppercase) or
821 lowercase hexadecimal digits.
822
823 \section{Dependencies}
824
825 \bnumname is a \LaTeX{} package but it can also be used with Plain \TeX,
826 thanks to \ctanpackage{miniltx}. Use for this
827 \cs{input} \ctanpackage{miniltx}|.tex|
828 and then \cs{input} \bnumname \unskip|.sty|. Do not use \cs{input} but only
829 \cs{usepackage} to load the package with \LaTeX.
830
831 Addition, subtraction, multiplication, division(s), modulo operator, powers,
832 and factorials are all by default executed by macros provided by the
833 \xintcorename package.
834
835 Conversions between decimal, binary, octal and
836 hexadecimal bases are done using the macros from the \xintbinhexname package.
837
838 |\bnumeval| is a scaled-down variant of |\xintiieval| from package
839 \xintexprname, lacking support for nested structures, functions, variables,
840 booleans, sequence generators, etc... .
841 The \xintexprname package is NOT loaded, only as said previously
842 \xintcorename and \xintbinhexname.
843
844 \begin{TeXnote}
845 Power users can use |\bnumsetup| to configure usage of alternative support
846 macros of their own choosing. Options can disable the loading of
847 \xintcorename and/or \xintbinhexname. But \xintkernelname is always loaded.
848 See \autoref{sec:advanced}. Expert users can even add new operators to the
849 syntax, even change the built-in ones or their precedences. See
850 \autoref{ssec:expert}.
851 \end{TeXnote}
852
853 \section{Examples}
854
855 Some of these examples use the ancient syntax
856 |\bnethe\bnumexpr...\relax| from the initial release (in 2014). The
857 |\bnethe| prefix converts from some private format (using braces and
858 other things). Some examples do not even have the |\bnethe| prefix to
859 |\bnumexpr| because it is allowed in typesetting context to omit it (but
860 in an \cs{edef} without it expansion gives the private format). For
861 details refer to \autoref{sec:advanced} on advanced topics.
862
863 Some further examples found in this documentation use the other ancient
864 syntax |\thebnumexpr...\relax| where |\thebnumexpr| is equivalent to
865 |\bnethe\bnumexpr|.
866
867 The recommended interface is |\bnumeval|, as it has optional
868 arguments to cause conversion to hexadecimal, octal or binary. They
869 have no equivalent with |\bnethe\bnumexpr| or with |\thebnumexpr|.
870
871 Some inputs are weird (such as the first one with three minus signs)
872 because they served originally to check the syntax.
873
874 % NOTE 2026/09/07. After the 1.6 blunder which shipped with \bnumexpr not
875 % accepting Babel active contrarily to what announcement (but \bnumeval was
876 % ok, and it is the recommened interface) and code comments said, I modified
877 % the preamble to add "french" as language in order to test it here. This
878 % means though that ! and : will be catcode active throughout the document
879 % (even though French is used only locally) so this means my unit testing here
880 % is not anymore one for the standard context with default catcodes but well.
881 %
882 % The \bneshow and other macros not only display the values but also save
883 % input and output to an auxiliary file which is tested via git diff to check
884 % for changes and detect regressions.
885 %
886 % Even with French active, one should beware that in vertical mode it
887 % self-silences itself, and one can be fooled to think no error happened, if
888 % the \bnumeval or \bnumexpr is first thing in the paragraph. But here the
889 % \bneshow and other macros will quit vertical mode prior to computing the
890 % expression, so that trap is avoided.
891 \noindent\bneshow {--- 1 208 637 867 * (2 187 917 891 - 3 109 197 072)}%
892 \bneshow { ( 13_8089_1090 - 300_1890_2902 ) * ( 1083_1908_3901 - 109_8290_3890 )}%
893 \bneshoweval {92_874_927_979 ** 5 - 31_9792_7979 ** 6}%
894 \foreignlanguage{french}{%
895 \bneshowthe {10!, 20!, 30!}% broken at 1.6... fixed at 1.6a
896 Testing tacit multiplication elevated precedence\string:\newline
897 \bneshoweval {30!/(21*22*23*24*25)(26*27*28*29*30), 20!}% worked as expected at 1.6
898 \bneshoweval {13^50//12^50, 13^50/:12^50}% idem
899 }%
900 \bneshoweval {13^50/12^50, 12^50}%
901 \bneshoweval {(1^10+2^10+3^10+4^10+5^10+6^10+7^10+8^10+9^10)^3}%
902 \bneshoweval {100! /: 10^50}%
903 %
904 Let's check hexadecimal input:\newline
905 \bneshoweval {"0010 * "0100 * 0x1000 * 0xA0000, 16^(1+2+3+4)*10}%
906 \bneshoweval {"000_abc_def, 0xABCDEF, "abcdef + "543210 + 1, 0x10^6}%
907 And also hexadecimal output:\newline
908 \bneshowevaltohex{"_7f_fff_fff+1, 0x_0400^3, "aBcDeF*"0000fedcba,
909 1234}
910 And also in lowercase\newline
911 \bneshowevaltolowhex{"abcdef, "ABCDEF, "999_999_999, 16^10-1, 167772160}
912 %
913 Let's make a few checks of octal and binary:\newline
914 \bneshowevaltooct {'75316420 * 0o44445555}
915 \bneshowevaltobin {'75316420 * 0o44445555}
916 \bneshowevaltobin {0xFFFF, 0o77, 0b1000001^3}
917 %
918 We end with some strange non-recommended things to check details
919 of how the parser expands the input:\newline
920 \bneshoweval {"0000\bnumeval[h]{00000012345678}FFFF, 000012345679*16**4-1}
921 %
922 \bneshowevaltooct{0b000\bnumeval[b]{'123456}, 0x\bnumeval[h]{0o00000123456}}
923
924 \section{Customizing how output is ``printed out''}
925
926 \subsection{Printing big numbers}
927 \label{sec:printing}
928 \TeX{} and \LaTeX{} will not split long numbers at the end of lines. I personally
929 often use helper macros (not in the package) of the following type:
930
931 \begin{verbatim}
932 \def\allowsplits #1{\ifx #1\relax \else #1\hskip 0pt plus 1pt\relax
933 \expandafter\allowsplits\fi}%
934 \def\printnumber #1{\expandafter\allowsplits \romannumeral-`0#1\relax }%
935 \end{verbatim}
936
937 \def\allowsplits #1{\ifx #1\relax \else #1\hskip 0pt plus 1pt\relax
938 \expandafter\allowsplits\fi}%
939 \def\printnumber #1{\expandafter\allowsplits \romannumeral-`0#1\relax }%
940
941 Here is an example of use and its output:
942 \begin{verbatim}
943 \noindent|\bnumeval{1000!} =|
944 \textcolor{digitscolor}{\printnumber{\bnumeval{1000!}}}
945 \end{verbatim}
946 \noindent|\bnumeval{1000!} =|\phantom{0}%
947 \textcolor{digitscolor}{\printnumber{\bnumeval{1000!}}}
948
949 \begin{TeXnote}
950 |\bnumeval| is \emph{f}-expandable, so the |\romannumeral-`0| as used
951 here in |\printnumber| causes its full expansion (even if for example
952 the output contains multiple values, separated by commas). So then
953 |\printnumber|'s auxiliary can simply loop over the tokens.
954 \end{TeXnote}
955
956 \begin{TeXnote}
957 Note that inside math mode, the inserted \cs{hskip}'s have no effect.
958 There should be some \cs{allowbreak}'s. By the way, we allow some
959 stretch so that line endings match the actual linewidth.
960 \end{TeXnote}
961
962 \subsection{\cs{bnumprintone}, \cs{bnumprintonesep}}
963
964 The output values are each fetched to
965 |\bnumprintone| and separated by |\bnumprintonesep|.
966
967 Here are the default definitions (or rather some quasi equivalents in \LaTeX's
968 lingua):
969 \begin{verbatim}
970 \newcommand{\bnumprintone}[1]{#1}
971 \newcommand{\bnumprintonesep}{, }
972 \end{verbatim}
973
974 In other terms |\bnumprintone| produces its argument ``as is'', and multiple
975 values get separated by a comma and a space.
976
977 Let's say you want the output to be boxed. Doing
978 |\fbox{\bnumeval{...}}| will make one single frame even in case
979 of multiple values. Redefining |\bnumprintone|
980 is the way to go:
981 \begingroup\RenewDocumentCommand{\bnumprintone}{m}{\fbox{#1}}
982 \begin{verbatim}
983 \RenewDocumentCommand{\bnumprintone}{m}{\fbox{#1}}
984 \bnumeval{2^10, 3^10, 5^10, 7^10}
985 \end{verbatim}
986 \textcolor{digitscolor}{\bnumeval{2^10, 3^10, 5^10, 7^10}}
987 \endgroup
988
989 It is important to have used |\RenewDocumentCommand| and not \cs{renewcommand}
990 here, because |\bnumprintone| and |\bnumprintonesep| have to be compatible
991 with expansion only context.
992 \begin{TeXnote}
993 That means that |\bnumprintone| in an \cs{edef} should not give rise to any
994 \cs{newcommand}, lower level \cs{def}, count or dimen assignments, etc....
995
996 This constraint is due to the fact that |\bnumeval| wraps the final
997 print-out inside of \cs{expanded}, for \TeX nical reasons.
998
999 The simplest way for |\bnumprintone| (considering that
1000 its argument will already have been fully expanded to digit tokens) and
1001 |\bnumprintonesep| to be ``safe'' is that they do not expand at all in
1002 \cs{edef}. This is the case if they are defined using
1003 \cs{RenewDocumentCommand}.
1004 %
1005 With an older \LaTeX{}, or Plain \eTeX{}
1006 (but having some \cs{fbox} at our disposal), we would have used here
1007 |\protected\def\bnumprintone#1{\fbox{#1}}|.
1008 \end{TeXnote}
1009
1010 A more common use case will be to have the outputs be typeset according to the
1011 conventions of the document language. This is easily done redefining
1012 |\bnumprintone| in terms of (for example) the \cs{np} macro of package
1013 \ctanpackage{numprint}.
1014 \begingroup\RenewDocumentCommand{\bnumprintone}{m}{\np{#1}}
1015 \renewcommand{\bnumprintonesep}{ --- }
1016 \begin{verbatim}
1017 \RenewDocumentCommand{\bnumprintone}{m}{\np{#1}}
1018 \renewcommand{\bnumprintonesep}{ --- }
1019 \bnumeval{2^10, 3^10, 5^10, 7^10}
1020 \end{verbatim}
1021 \textcolor{digitscolor}{\bnumeval{2^10, 3^10, 5^10, 7^10}}
1022 \endgroup
1023
1024 \begin{TeXnote}
1025 Usage of \cs{RenewDocumentCommand} for |\bnumprintonesep| was not
1026 needed here, obviously its expansion could cause no trouble.
1027 \end{TeXnote}
1028
1029 Let's give another use case. Assume you are computing in one go multiple
1030 large values, too large to fit on a line. The simple-minded |\printnumber| of
1031 the previous section will (due to some \TeX nicality) swallow the spaces
1032 injected by |\bnumprintonesep|. To fix this, the simplest is to redefine
1033 |\bnumprintone| to execute |\printnumber|:
1034 \begingroup\let\bnumprintone\printnumber
1035 \begin{verbatim}
1036 \renewcommand{\bnumprintone}[1]{\printnumber{#1}}
1037 \bnumeval{2^100, 3^100, 5^100, 7^100}
1038 \end{verbatim}
1039 % je ne peux pas utiliser \bneshoweval car il passe ne mode mathématique
1040 \textcolor{digitscolor}{\bnumeval{2^100, 3^100, 5^100}}
1041 \endgroup
1042
1043 \begin{TeXnote}
1044 Our |\printnumber| belongs to this family of macros causing no damage
1045 if expanding in an \cs{edef}. So, it was not needed to use
1046 \cs{RenewDocumentCommand}.
1047 \end{TeXnote}
1048
1049 \subsection[\cs{bnumprintonehex}, \cs{bnumprintonelowerhex},
1050 \cs{bnumprintoneoct},\texorpdfstring{\newline}{}
1051 \cs{bnumprintonebin}]{\cs{bnumprintonehex}, \cs{bnumprintonelowerhex},
1052 \cs{bnumprintoneoct}, \cs{bnumprintonebin}}
1053
1054 When |\bnumeval| is exerted with |[h]|, |[ha]|, |[o]| or |[b]|
1055 it does not use |\bnumprintone| but one of |\bnumprintonehex|,
1056 |\bnumprintonelowerhex|,
1057 |\bnumprintoneoct| or |\bnumprintonebin|.
1058 The same |\bnumprintonesep| is used as with decimal numbers.
1059
1060 The default definitions are as for |\bnumprintone| to ``print as is''.
1061
1062 \begingroup
1063 \def\bnumprintonehex#1{0x#1}
1064 \def\bnumprintonelowerhex#1{"#1}
1065 To give an example of a custom
1066 definition, one may want hexadecimal to use the |0x| prefix with
1067 uppercase output or the |"| prefix with lowercase output. This is
1068 very easy:
1069 \begin{verbatim}
1070 \renewcommand{\bnumprintonehex}[1]{0x#1}
1071 \renewcommand{\bnumprintonelowerhex}[1]{"#1}
1072 \end{verbatim}
1073 \bneshowevaltohex{7^30, 13^20, 20!}
1074 \bneshowevaltolowhex{7^30, 13^20, 20!}
1075 \endgroup
1076
1077 \begin{TeXnote}
1078 It was unneeded to use \cs{RenewDocumentCommand} here because prefixing
1079 with |0x| is obviously compatible with expansion-only context.
1080 \end{TeXnote}
1081
1082 \section{Babel-active characters are not a problem!}
1083
1084 Some languages use active characters with PDF\LaTeX. For example the
1085 |babel-french| module turns the colon |:| and the exclamation mark |!|
1086 into active characters (whose expansions would cause |\bnumeval| to
1087 crash). It used to be necessary to take preventive measures such as
1088 either turning the activation off altogether or use in the input
1089 |/\string:| and |\string!| as clumsy replacements of |/:| and |!|.
1090
1091 Those troubled times are gone! With release |1.6| they will work fine
1092 as is in |\bnumeval|. The same applies to all other characters if
1093 babel-active. There are miracles sometimes!
1094
1095 Warning: characters made active otherwise still need the \cs{string} or
1096 other workaround to be usable as operators in the syntax.
1097
1098 \section{Fine print (not needed to read this for regular use)}
1099 \label{sec:advanced}
1100
1101 \footnotesize
1102
1103 \subsection{Adding support for binomial coefficients}
1104 \label{ssec:binomial}
1105
1106 As will be documented in the section for expert users, it is possible
1107 to extend the syntax with one's own operators.
1108
1109 Let's turn the semicolon into an operator which computes binomial
1110 coefficients: |a;b| will evaluate to ``|a| choose |b|''. The precedence will
1111 be chosen stronger than addition and multiplication but less than powers.
1112 This only needs adding those two lines to the preamble:
1113 \begin{verbatim}
1114 \usepackage{xint}% as xintcore does not have \xintiiBinomial
1115 \bnumdefinfix{;}{\xintiiBinomial}{15}{15}
1116 \end{verbatim}
1117 \foreignlanguage{french}{%
1118 We can now test it:\newline
1119 \bnumdefinfix{\string;}{\xintiiBinomial}{15}{15}%
1120 \bneshoweval{100;50}
1121 \bneshoweval{3^5;2^5}
1122 }
1123
1124 \subsection{An expandable macro for computing large Fibonacci numbers}
1125 \label{ssec:fibo}
1126
1127 % added 2025/09/22
1128 We are going to define an expandable macro which takes on input a positive
1129 integer $N$ (we did not bother about case of zero or negative input) and
1130 computes the $N$\textup{th} Fibonacci number $F_N$.
1131
1132 The algorithm to compute $F_N$ is based upon the quadratic relations which
1133 allow to go from $(F_n,F_{n-1})$ to either $(F_{2n}, F_{2n-1})$ (operation of
1134 type A) or to $(F_{2n+1}, F_{2n})$ (operation of type B). We first convert
1135 $N$ into binary base, and read the digits from most significant to least
1136 significant applying each time either type A or type B. Each one of type A or
1137 B will cost only 3 multiplications and a bunch of additions, which of course
1138 are done using |\bnumexpr|. In total the algorithm costs (because $3\log_2
1139 10\approx 10$) about $10L$ multiplications (doubling is not counted as a
1140 multiplication) where $L$ is the number of decimal digits of $N$. Here is the
1141 code:
1142 \begin{verbatim}
1143 \makeatletter
1144 \newcommand\Fibonacci[1]{%
1145 \expandafter\Fibo@start
1146 \romannumeral0\xintdectobin{\the\numexpr#1\relax};%
1147 }
1148 % To chain \bnumexpr's easily we use the once-expanded form which is
1149 % \romannumeral0\bnumexpro. We could also have used \expanded and \bnumeval
1150 % calls, that would have been simpler but old habits die hard.
1151 %
1152 % Gobble the first digit which is 1 (this is where we could have tested for a
1153 % negative or zero input). Prepare encapsulated 1 and 0. Then repeatedly
1154 % dispatch according to binary digits (which are strategically located at end
1155 % of input stream to optimize), or the semicolon which signals completion.
1156 %
1157 % We use \csname-governed branching (because it is cool). We could have
1158 % used some \if test or even an \ifcase with an {-1} at end, or any other
1159 % of the many possibilities opened up by TeX syntax.
1160 \def\Fibo@start 1#1{%
1161 \csname Fibo@#1\expandafter\endcsname
1162 \romannumeral0\bnumexpro 1\expandafter\relax\expandafter;%
1163 \romannumeral0\bnumexpro 0\relax;%
1164 }
1165 \@namedef{Fibo@0}#1;#2;#3{%
1166 \csname Fibo@#3\expandafter\endcsname
1167 \romannumeral0\bnumexpro (#1+2*#2)*#1\expandafter\relax\expandafter;%
1168 \romannumeral0\bnumexpro #1*#1+#2*#2\relax;%
1169 }
1170 \@namedef{Fibo@1}#1;#2;#3{%
1171 \csname Fibo@#3\expandafter\endcsname
1172 \romannumeral0\bnumexpro 2*(#1+#2)*#1+#2*#2\expandafter\relax\expandafter;%
1173 \romannumeral0\bnumexpro (#1+2*#2)*#1\relax;%
1174 }
1175 % #1 is F_N (and #2 is F_{N-1}) but in need of unpacking.
1176 \@namedef{Fibo@;}#1;#2;{\bnethe#1}
1177 \makeatother
1178 \end{verbatim}
1179
1180 \makeatletter
1181 \newcommand\Fibonacci[1]{%
1182 \expandafter\Fibo@start
1183 \romannumeral0\xintdectobin{\the\numexpr#1\relax};%
1184 }
1185 \def\Fibo@start 1#1{%
1186 \csname Fibo@#1\expandafter\endcsname
1187 \romannumeral0\bnumexpro 1\expandafter\relax\expandafter;%
1188 \romannumeral0\bnumexpro 0\relax;%
1189 }
1190 \@namedef{Fibo@0}#1;#2;#3{%
1191 \csname Fibo@#3\expandafter\endcsname
1192 \romannumeral0\bnumexpro (#1+2*#2)*#1\expandafter\relax\expandafter;%
1193 \romannumeral0\bnumexpro #1*#1+#2*#2\relax;%
1194 }
1195 \@namedef{Fibo@1}#1;#2;#3{%
1196 \csname Fibo@#3\expandafter\endcsname
1197 \romannumeral0\bnumexpro 2*(#1+#2)*#1+#2*#2\expandafter\relax\expandafter;%
1198 \romannumeral0\bnumexpro (#1+2*#2)*#1\relax;%
1199 }
1200 \@namedef{Fibo@;}#1;#2;{\bnethe#1}
1201 \makeatother
1202
1203 We can now compute for example $F_{10}$, $F_{100}$, $F_{1000}$ and $F_{10000}$:
1204 \begin{verbatim}
1205 \noindent\textcolor{cyan}{$F_{10}={}$}%
1206 \textcolor{digitscolor}{\Fibonacci{10}}\newline
1207 \textcolor{cyan}{$F_{100}={}$}%
1208 \textcolor{digitscolor}{\Fibonacci{100}}\newline
1209 \textcolor{cyan}{$F_{1000}={}$}%
1210 \textcolor{digitscolor}{\printnumber{\Fibonacci{1000}}}\newline
1211 \textcolor{cyan}{$F_{10000} ={}$}%
1212 \textcolor{digitscolor}{\printnumber{\Fibonacci{10000}}}
1213 \end{verbatim}
1214 \noindent\textcolor{cyan}{$F_{10}={}$}%
1215 \textcolor{digitscolor}{\Fibonacci{10}}\newline
1216 \textcolor{cyan}{$F_{100}={}$}%
1217 \textcolor{digitscolor}{\Fibonacci{100}}\newline
1218 \textcolor{cyan}{$F_{1000}={}$}%
1219 \textcolor{digitscolor}{\printnumber{\Fibonacci{1000}}}\newline
1220 \textcolor{cyan}{$F_{10000} ={}$}%
1221 \textcolor{digitscolor}{\printnumber{\Fibonacci{10000}}}
1222 \begin{TeXnote}
1223 We did not use math mode because our |\printnumber| is a bit dumb and does
1224 not work in math mode. See the comment at the end of \autoref{sec:printing}.
1225 \end{TeXnote}
1226 \begin{TeXnote}
1227 In the code above we used |;| as delimiter. But the macro definitions are
1228 located in the body of this document, not in the preamble, and |;| is
1229 catcode active here due to French being among the languages used by this
1230 \LaTeX{} document. This is not a problem in itself for the uses of the
1231 semicolon as delimiter for macro definitions. It would be a problem if
1232 the semicolon had been used to terminate a \cs{numexpr} (we would then
1233 have injected a \cs{relax}). One could also be
1234 very worried seeing |\@namedef{Fibo@;}| where |;| will expand in
1235 |\csname...\endcsname|, but \ctanpackage{babel} has a built-in protection
1236 for such things. This all being said, it is safer to have such definitions
1237 located in the preamble of the \LaTeX{} document. But in next section we
1238 make a modification which can only be done after the original |\Fibonacci|
1239 has already been used (as it changes its interface), so putting all things
1240 into the preamble is not always logical nor convenient.
1241 \end{TeXnote}
1242
1243 \subsection{More fun with Fibonacci numbers}
1244
1245 % added 2025/09/27
1246 Our |\Fibonacci|\marg{N} macro expands to compute $F_N$. But what if we want
1247 a whole sequence $F_N, F_{N+1}, \dots$? Then using directly the additive
1248 recurrence $F_{n+2}=F_{n+1}+F_n$ will be more efficient than computing again
1249 and again via our fancy multiplicative algorithm new values. So it is a bit
1250 sad that expansion of |\Fibonacci|\marg{N} actually computed both of $F_N$ and
1251 $F_{N-1}$ but discarded the latter. We will modify its ending to keep also
1252 $F_{N-1}$. And we switch to a down-to-earth style based on assignments. The
1253 \xintname documentation provides a variant construct which obtains expandably
1254 a sequence of braced items |{F_a}{F_{a+1}}...{F_b}|, but for this, expandable
1255 loops from \ctanpackage{xinttools} or from \LaTeX3 are the appropriate tools,
1256 and we then need extra tools to handle such output.
1257 \begin{verbatim}
1258 \makeatletter
1259 % reuse \Fibonacci's earlier meaning, then make \Fibonacci undefined
1260 % because \Fibo@; will be modified.
1261 \let\Fibonacci@\Fibonacci
1262 \let\Fibonacci\undefined
1263 \newcommand\SetToFibonacciPair[3]{\Fibonacci@{#1}{#2}{#3}}
1264 \@namedef{Fibo@;}#1;#2;#3#4{\def#3{#1}\def#4{#2}}
1265 \makeatother
1266 \end{verbatim}
1267 \makeatletter
1268 \let\Fibonacci@\Fibonacci
1269 \let\Fibonacci\undefined
1270 \newcommand\SetToFibonacciPair[3]{\Fibonacci@{#1}{#2}{#3}}
1271 \@namedef{Fibo@;}#1;#2;#3#4{\def#3{#1}\def#4{#2}}
1272 \makeatother
1273 \SetToFibonacciPair{1001}{\FibB}{\FibA}
1274
1275 We add the above code to our document and now execute
1276 %
1277 \centeredline{|\SetToFibonacciPair{1001}{\FibB}{\FibA}|}
1278 %
1279 Now |\FibB| holds $F_{1001}$ and |\FibA| holds $F_{1000}$ (both being
1280 |\bnumexpr|-encapsulated and needing |\bnethe| to get converted to explicit
1281 digits). We can use LaTeX's |\loop| to organize the iterative computation and
1282 printing of new values. But we first abstract the ``printing'' into a
1283 \LaTeX{} ``command''.
1284 \begin{verbatim}
1285 \newcommand\PrintOneFib[2]{%
1286 \noindent\textcolor{cyan}{$F_{#1}={}$}\textcolor{digitscolor}{\printnumber{\bnethe#2}}%
1287 }
1288 \def\myN{1000}
1289 \loop
1290 \PrintOneFib{\myN}{\FibA}\newline
1291 \edef\FibC{\bnumexpr \FibB + \FibA\relax}% compute and assign a new Fibonacci number
1292 \let\FibA\FibB \let\FibB\FibC % update the storage macros
1293 \ifnum\myN<1010 % terminate the loop when F_{1010} has been printed
1294 \edef\myN{\the\numexpr\myN+1}%
1295 \repeat
1296 \end{verbatim}
1297 \newcommand\PrintOneFib[2]{%
1298 \noindent\textcolor{cyan}{$F_{#1}={}$}%
1299 \textcolor{digitscolor}{\printnumber{\bnethe#2}}%
1300 }
1301 \def\myN{1000}
1302 \loop
1303 \PrintOneFib{\myN}{\FibA}\newline
1304 \edef\FibC{\bnumexpr \FibB + \FibA\relax}%
1305 \let\FibA\FibB
1306 \let\FibB\FibC
1307 \ifnum\myN<1010
1308 \edef\myN{\the\numexpr\myN+1}%
1309 \repeat
1310
1311 At this stage we have printed up to $F_{1010}$. On exit of the loop |\FibA|
1312 holds $F_{1011}$ and |\FibB| holds $F_{1012}$. We print them out too:
1313 \begin{verbatim}
1314 \PrintOneFib{\the\numexpr\myN+1}{\FibA}\newline
1315 \PrintOneFib{\the\numexpr\myN+2}{\FibB}\par
1316 \end{verbatim}
1317 \PrintOneFib{\the\numexpr\myN+1}{\FibA}\newline
1318 \PrintOneFib{\the\numexpr\myN+2}{\FibB}\par
1319
1320 \makeatletter
1321 \@namedef{Fibo@;}#1;#2;{\bnethe#1}
1322 \ifnum\pdfstrcmp{\bnethe\FibA}{\Fibonacci@{1011}}=0 \else\ERROR\fi
1323 \ifnum\pdfstrcmp{\bnethe\FibB}{\Fibonacci@{1012}}=0 \else\ERROR\fi
1324 \makeatother
1325
1326 \subsection{The \cs{bnumsetup} command}
1327
1328 Package \bnumname needs that some \emph{big integer engine} provides the macros doing
1329 the actual computations.
1330
1331 By default, it loads package \xintcorename (a subset of \xintexprname) and
1332 package \xintbinhexname.
1333 \begin{verbatim}
1334 \usepackage{xintcore}
1335 \usepackage{xintbinhex}
1336 \end{verbatim}
1337
1338 It then uses |\bnumsetup| in the following way (the final comma is optional,
1339 and spaces around equal signs also; there can also be spaces before the commas
1340 but the author dislikes such style a lot so they are not used here):
1341 \begin{verbatim}
1342 \bnumsetup{%
1343 add = \xintiiAdd, sub = \xintiiSub, opp = \xintiiOpp,
1344 mul = \xintiiMul, pow = \xintiiPow, fac = \xintiiFac,
1345 div=\xintiiDivFloor, mod=\xintiiMod, divround=\xintiiDivRound,
1346 hextodec=\xintHexToDec, octtodec=\xintOctToDec, bintodec=\xintBinToDec,
1347 dectohex=\xintDecToHex, dectooct=\xintDecToOct, dectobin=\xintDecToBin,
1348 }%
1349 \end{verbatim}
1350
1351 One can use |\bnumsetup| to map one, some, or all keys to macros of one's
1352 own choosing. Of course it is then up to user to load the suitable packages.
1353
1354 If one has alternatives for all of the above \xintcorename macros, so that
1355 this package is not needed at all, one can pass option |customcore| to
1356 \bnumname at loading time:
1357 \centeredline{|\usepackage[customcore]|\{\bnumname\}} This tells to not load
1358 \xintcorename.
1359
1360 Similarly there is an option |custombinhex| to not load \xintbinhexname.
1361 Make sure then to provide suitable replacements to all base conversion macros!
1362
1363 Option |custom| means doing both of |customcore| and |custombinhex|.
1364 Even under this option package \xintkernelname will always be loaded.
1365
1366 Here are the conditions that the custom macros must obey:
1367 \begin{enumerate}[noitemsep]
1368 \item They all must be \emph{f}-expandable, with some exceptions
1369 \begin{enumerate}[noitemsep]
1370 \item The macro for computing factorials only has to be \emph{x}-expandable.
1371 \item If |\bnumprintonehex| (or |\bnumprintoneoct|, or |\bnumprintonebin|)
1372 is redeinfed as a \cs{protected}, then conversion to hexadecimal
1373 (resp. octal, resp. binary) only has to be \emph{x}-expandable, except
1374 that
1375 |[ha]| option will break if the |dectohex| macro is not \emph{f}-expandable.
1376 \end{enumerate}
1377 Note that any \emph{x}-expandable macro can be wrapped into an
1378 \emph{f}-expandable one, using \cs{expanded}.
1379 \item It is sufficient for them to be able to handle arguments in \emph{raw
1380 normalized form}, i.e., sequences of explicit decimal (or hexadecimal for
1381 the macro associated with key |hextodec|) digits, no leading zeros, with at
1382 most one minus sign and no plus sign.
1383 \item Their output format is limited only by the fact that it should be
1384 acceptable input to all the other operators, as well as to the user optional
1385 re-definition of |\bnumprintone|. If one cares about hexadecimal (et al.)
1386 output one must ensure the macros output format is suitable input for those
1387 macros actually doing the conversion from decimal to other bases.
1388 \item \emph{Important}: hence if only some macros among those associated to
1389 operators (i.e. those by default originating in \xintcorename), or to
1390 conversions into decimal, are custom, their output \textbf{must} be produced
1391 in \emph{raw normalized form}, as this is the format required by the
1392 \xintcorename macros and by the \xintbinhexname macros converting from
1393 decimal to other bases. However if one does not care about producing output
1394 in binary, octal or hexadecimal (as is the case in the next section),
1395 and if one has replaced all \xintcorename macros, the output format can be
1396 as one likes.
1397 \item If |dectohex| key is used and the associated custom macro produces
1398 \emph{lowercase} hexadecimal, then |\bnumeval[h]{...}| produces as
1399 instructed hexadecimal in lowercase. Mind that the |[ha]| option is now
1400 broken because its internals expect the input to be in uppercase
1401 hexadecimal, and this input is computed using the macro assigned to the
1402 |dectohex| key. This is of course unimportant because |[h]| now has
1403 replaced |[ha]|, but I figured I could mention it here.
1404 \end{enumerate}
1405
1406 \subsection{Let's handle fractions!}
1407
1408 I will show how to transform |\bnumeval| into a calculator with fractions! We
1409 will use the \xintfracname macros, but coerce them into always producing
1410 fractions in lowest terms (except for powers). For optimization we use
1411 the |[0]| post-fix which speeds-up the input parsing by the \xintfracname
1412 macros. We remove it on output via a custom |\bnumprintone|.
1413
1414 Note that the |/| operator is associated to |divround| key but of course here
1415 the used macro will simply do an exact division of fractions, not a rounded-to-%
1416 an integer division. This is the whole point of using a macro of our own
1417 choosing!
1418 \begin{verbatim}
1419 \usepackage{xintfrac}
1420 \newcommand\myIrrAdd[2]{\xintIrr{\xintAdd{#1}{#2}}[0]}
1421 \newcommand\myIrrSub[2]{\xintIrr{\xintSub{#1}{#2}}[0]}
1422 \newcommand\myIrrMul[2]{\xintIrr{\xintMul{#1}{#2}}[0]}
1423 \newcommand\myDiv[2]{\xintIrr{\xintDiv{#1}{#2}}[0]}
1424 \newcommand\myDivFloor[2]{\xintDivFloor{#1}{#2}[0]}
1425 \newcommand\myMod[2]{\xintIrr{\xintMod{#1}{#2}}[0]}
1426 \newcommand\myPow[2]{\xintPow{#1}{#2}}% will have already postfix [0]
1427 \newcommand\myFac[1]{\xintFac{#1}}% will have already postfix [0]
1428 \makeatletter
1429 \def\myRemovePostFix#1{\@myRemovePostFix#1[0]\relax}%
1430 \def\@myRemovePostFix#1[0]#2\relax{#1}
1431 \makeatother
1432 \let\bnumprintone\myRemovePostFix
1433 \bnumsetup{add=\myIrrAdd, sub=\myIrrSub, mul=\myIrrMul,
1434 divround=\myDiv, div=\myDivFloor,
1435 mod=\myMod, pow=\myPow, fac=\myFac}%
1436 \end{verbatim}
1437 \begingroup
1438 \newcommand\myIrrAdd[2]{\xintIrr{\xintAdd{#1}{#2}}[0]}
1439 \newcommand\myIrrSub[2]{\xintIrr{\xintSub{#1}{#2}}[0]}
1440 \newcommand\myIrrMul[2]{\xintIrr{\xintMul{#1}{#2}}[0]}
1441 \newcommand\myDiv[2]{\xintIrr{\xintDiv{#1}{#2}}[0]}
1442 \newcommand\myDivFloor[2]{\xintDivFloor{#1}{#2}[0]}
1443 \newcommand\myMod[2]{\xintIrr{\xintMod{#1}{#2}}[0]}
1444 \newcommand\myPow[2]{\xintPow{#1}{#2}}% will have already postfix [0]
1445 \newcommand\myFac[1]{\xintFac{#1}}% will have already postfix [0]
1446 \makeatletter
1447 \def\myRemovePostFix#1{\@myRemovePostFix#1[0]\relax}%
1448 \def\@myRemovePostFix#1[0]#2\relax{#1}
1449 \makeatother
1450 \let\bnumprintone\myRemovePostFix
1451 \bnumsetup{add=\myIrrAdd, sub=\myIrrSub, mul=\myIrrMul,
1452 divround=\myDiv, div=\myDivFloor,
1453 mod=\myMod, pow=\myPow, fac=\myFac}%
1454 \bneshoweval{1000000*(1/100+1/2^7-20/5^4)/(1/3-5/7+9/11)^2}
1455 \bneshoweval{(1-1/2)(1-1/3)(1-1/4)(1-1/5)(1-1/6)(1-1/7)}
1456 \bneshoweval{(1-1/3+1/9-1/27-1/81+1/243-1/729+1/2187)^5}
1457 \bneshoweval{(1+1/10)^10 /: (1-1/10)^10}
1458 \bneshoweval{2^-3^4}
1459 \endgroup
1460
1461 Computations with fractions quickly give birth to big results, see
1462 \autoref{sec:printing} on how to modify |\bnumprintone| to coerce \TeX\ into
1463 wrapping numbers too long for the available width.
1464
1465 \subsection{For the expert user: expression syntax and its customizability}
1466 \label{ssec:expert}
1467
1468 \subsubsection{Significant differences between \cs{bnumexpr} and \cs{numexpr}}
1469 \label{sec:differences}
1470
1471 Apart from the extension to big integers and the added operators, there
1472 are a number of important differences between |\bnumexpr| and
1473 |\numexpr|:
1474
1475 \begin{enumerate}[noitemsep]
1476 \item Contrarily to \cs{numexpr}, the |\bnumexpr| parser stops only after having
1477 found (and swallowed) a mandatory ending \cs{relax} token (it can arise from expansion),
1478
1479 \item In particular note that spaces between digits do not stop |\bnumexpr|, in contrast
1480 with \cs{numexpr}:
1481
1482 |\the\numexpr 3 5+79\relax| expands (in one step) to \expandafter|\the\numexpr 3 5+79\relax|
1483
1484 |\thebnumexpr 3 5+79\relax| expands (in two steps) to
1485 \expandafter\expandafter\expandafter|\thebnumexpr 3 5+79\relax|
1486
1487 \item With |\edef\myvariable{\bnumexpr 1+2\relax}|, the computation is
1488 done at time of the \cs{edef}. It prepares |\myvariable| as a
1489 self-contained pre-computed unit which is recognized as such when
1490 inserted in a \bnumname expressions. It triggers tacit
1491 multiplication: |7\myvariable| is like |7*\myvariable|. This is
1492 different from what would happen if we had used
1493 |\edef\myvariable{\bnethe\bnumexpr...}| which would simply have
1494 |\myvariable| expand to digit tokens so |7\myvariable| then constructs
1495 a number with |7| as first digit.
1496
1497 \begin{quote}
1498 Let's give an example. Note that \cs{edef} has the effect of
1499 pre-evaluating. With \cs{def} the outputs would be the same, but the
1500 computations would be delayed to |\bnumeval| execution.
1501
1502 % attention mon \verb incapable de gérer %
1503 \noindent|\edef\x{\bnumexpr 3^10\relax}|%
1504 \% precomputes but keeps private format\newline
1505 \edef\x{\bnumexpr 3^10\relax}\bneshoweval{10000\x}
1506
1507 \noindent|\edef\y{\bnethe\bnumexpr 3^10\relax}|%
1508 \% evaluates to explicit digits\newline
1509 \edef\y{\bnethe\bnumexpr 3^10\relax}\bneshoweval{10000\y}
1510
1511 In the example with |\x|, tacit multiplication applied, whereas in the
1512 example with |\y| it is as if the digits had been input by hand in place
1513 of |\y|.
1514 %
1515 Note that the tacit multiplication behaves as expected relative to
1516 powers:\newline \bneshoweval{10^10\x} And we certainly do no want to try
1517 |10^10\y| which is like |10^10|\expandafter|\y|.
1518 \end{quote}
1519
1520 There is no analog with \cs{numexpr}:
1521 \begin{enumerate}[noitemsep]
1522 \item |\edef\foo{\numexpr1+2\relax}| will define |\foo| as
1523 |\numexpr1+2\relax| where the calculation is not yet done.
1524 \item Inserting the |\foo| as is in the document text causes an error.
1525 \item Trying |\the\numexpr 7\foo\relax| with such a |\foo| causes an
1526 error. One must use the multiplication sign |*| explicitly.
1527 \end{enumerate}
1528
1529 \item Expressions may be comma separated. On input, spaces are ignored,
1530 and on output the values are comma separated with a space after
1531 each comma,
1532
1533 \item |\thebnumexpr -(1+1)\relax| is legal contrarily to |\the\numexpr
1534 -(1+1)\relax| which errors (also with |-(1)| but |-(1+1)| has a more
1535 legitimate use of parentheses), and one has to employ workarounds such as
1536 |\the\numexpr 0-(1+1)\relax|,
1537
1538 \item |\the\numexpr 2+-(1+1)\relax| (or with |-+|) errors (but |+-| and |-+|
1539 if not followed by parentheses do work), whereas |\bnumexpr| has no such
1540 counter-intuitive limitation,
1541
1542 \item |\the\numexpr 2\cnta\relax| is illegal (with \cs{cnta} a
1543 \cs{count}-variable) (note that |\the\dimexpr 2\mydimen\relax| or
1544 |\the\dimexpr2\dimen0\relax| do work; even |\the\dimexpr 1\count0\relax|
1545 does work, but |\the\dimexpr \count0\relax| errors!). On the other hand
1546 |\thebnumexpr 2\cnta\relax| is perfectly legal and does the tacit
1547 multiplication; also |\thebnumexpr 2\mydimen\relax| works, with |\mydimen|
1548 getting replaced by its value in scaled points,
1549
1550 \item |\the\numexpr3(4+5)\relax| stops at the parenthesis, i.e.\@ it expands to
1551 |3(4+5)\relax|, whereas of course |\thebnumexpr 3(4+5)\relax| does the
1552 computation and expands to
1553 |27|,
1554
1555 \item |\the\numexpr(4+5)3\relax| expands to |93\relax| whereas
1556 |\thebnumexpr(4+5)3\relax| does the computation and expands to |27|,
1557
1558 \item |\thebnumexpr (4+5)\count0\relax| expands to nine time the current
1559 value of |\count0| register, but with |\the\numexpr| there would be a
1560 \emph{Missing number} error,
1561
1562 \item Generally speaking |\numexpr| does not implement the tacit
1563 multiplication which |\bnumexpr| does in front of parentheses and
1564 sub |\numexpr...\relax| expression: or after parentheses, or in-between
1565 two parenthesized expressions such as happens with |(1+2)(3+4)|,
1566
1567 \item The underscore |_| is accepted within the digits composing a number and
1568 is silently ignored by |\bnumexpr|.
1569
1570 \end{enumerate}
1571
1572 Regarding constructs such as |\edef\myvariable{\bnumexpr 1+2\relax}|, it
1573 was explained |\myvariable| behaves then in a special way in another
1574 |\bnumexpr| expression (or |\bnumeval|). It is also worth mentioning
1575 that it can be used directly in the typesetting stream. But if written
1576 to an external file it will expand to some internal format which is not
1577 documented as it may vary in future.
1578
1579 One can NOT use a |\myvariable| as above in an \cs{ifnum} test,
1580 even if representing a single small integer. It will work with syntax
1581 such as |\ifnum\bnethe\myvariable=7 ...|.
1582
1583 A point of note is that |\bnethe\myvariable| or
1584 |\bnethe\bnumexpr...\relax| expand to explicit digits so (assuming here
1585 there no other comma separated value computed),
1586 \begin{verbatim}
1587 \ifnum 3>\bnethe\bnumexpr...\relax
1588 ...
1589 \fi
1590 \end{verbatim}
1591 is dangerous, because the integer is not properly terminated.
1592 Here one could reverse the order, but the simplest way is simply
1593 to use |\bnumeval|:
1594 \begin{verbatim}
1595 \ifnum 3>\bnumeval{...}
1596 ...
1597 \fi
1598 \end{verbatim}
1599 Now, the end of line space injected by \TeX{} will terminate the integer
1600 and make the \cs{ifnum} test safe.
1601
1602 \subsubsection{Expression syntax}
1603 The implemented syntax is the expected one with infix operators and
1604 parentheses, the recognized operators being |+|, |-|, |*|, |/| (rounded
1605 division), |^| (power), |**| (power), |//| (by default floored division), |/:|
1606 (the associated modulo) and |!| (factorial).
1607
1608 One can input hexadecimal numbers as familiar from the \TeX\ number
1609 assignments syntax, i.e.\@ using the |"| prefix. But also lowercase letters
1610 |abcdef| are accepted in addition to uppercase |ABCDEF| (feature added at
1611 |1.7|). Release |1.6| added support for the |0x| prefix. It also added
1612 support for octal input via either |'| or |0o| prefixes, and for binary input
1613 via |0b| prefix.
1614
1615 Commas separating multiple expressions are allowed. The whole expression is
1616 handled token by token, any component (digit, operator, parentheses... even
1617 the ending \cs{relax}) may arise on the spot from macro expansions. The
1618 underscore |_| can be used to separate digits in long numbers, for readability
1619 of the input.
1620
1621 The precedence rules are as expected and detailed in the next section.
1622 Operators on the same level of precedence (like |*|, |/|, |//|, |/:|) behave
1623 in a left-associative way, and these examples behave as e.g.\@ with Python
1624 analogous operators:\newline \bneshoweval{100//3*4, 100*4//3, 100/:3*4,
1625 100*4/:3, 100//3/:5}
1626
1627 At |1.5| a change was made to the power operators which became
1628 right-associative. Again, this matches the behaviour e.g.\@ of Python:\newline
1629 \bneshoweval{2^3^4, 2^(3^4)}
1630
1631 It is possible to customize completely the behaviour of the parser, in two ways:
1632 \begin{itemize}
1633 \item via |\bnumsetup| which has a simple interface to replace the macros
1634 associated with |+|, |-|, |*|, |/|, |//|, |/:|, |**|, |^| and |!| by
1635 custom macros,
1636 \item or even more completely via |\bnumdefinfix| and |\bnumdefpostfix| which
1637 allow to add new operators to the syntax! They can also be used to replace
1638 the built-in ones or modify their precedence levels.
1639 \end{itemize}
1640
1641 \subsubsection{Precedences}
1642 \label{sssec:prec}
1643
1644 The parser implements precedence rules based on concepts which are summarized
1645 below. I am providing them for users who will use the customizing macros.
1646 \begin{itemize}
1647 \item an infix operator has two associated precedence levels, say |L| and
1648 |R|,
1649 \item the parser proceeds from left to right, pausing each time it has found a
1650 new number and an operator following it,
1651 \item the parser compares the left-precedence |L| of the new found operator to
1652 the right-precedence |R_last| of the last delayed operation (which already
1653 has one argument and would like to know if it can use the new found one): if
1654 |L| is at most equal to it, the delayed operation is now executed, else the
1655 new-found operation is kept around to be executed first, once it will have
1656 gathered its arguments, of which only one is known at this stage.
1657 \end{itemize}
1658
1659 Although there is thus internally all the needed room for sophistication, the
1660 implemented table of precedences simply puts all of multiplication and
1661 division related operations at the same level, which means that left
1662 associativity will apply with these operators. I could see that Python behaves
1663 the same way for its analogous operators.
1664
1665 Here is the default table of precedences as implemented by the package:
1666 \begin{center}
1667 \DeleteShortVerb{\|}%
1668 \begin{tabular}{|c|c|c|}
1669 \multicolumn3{c}{Table of precedences}\\\hline
1670 \MakeShortVerb{\|}%
1671 operator& left &right\\\hline
1672 |+|,|-|& 12&12\\
1673 |*|,|/|,|//|,|/:|& 14&14\\
1674 tacit |*|&16&14\\
1675 |**|, |^|& 18&17\\
1676 |!| &20&n/a\\\hline
1677 \end{tabular}
1678 \end{center}
1679 Tacit multiplication applies in front of parentheses, and after them, also in
1680 front of count variables or registers. As shown in the table it has an
1681 elevated precedence compared to multiplication explicitly induced by |*|, so
1682 |100/4(9)| is computed as |100/36| and not as |25*9|:\newline
1683 \bneshoweval{100/4(9), (100/4)9, 1000 // (100/4) 9 (1+1) * 13}
1684 More generally |A/B(C)(D)(E)*F| will compute |(A/(B*C*D*E))*F|.%
1685 %
1686 \footnote{The |B(C)(D)(E)| product will be computed as |B*(C*(D*E))| because
1687 the right-precedence of tacit multiplication is |14| but its left-precedence
1688 is |16|, creating right associativity. As the underlying mathematical
1689 operation is associative this is irrelevant to final result.}
1690
1691 The unary |-|, as prefix, has a special behaviour: after an infix operator it
1692 will acquire a right-precedence which is the minimum of |12| (i.e.\@ the
1693 precedence of addition and subtraction) and of the right-precedence of the
1694 infix operator. For example |2^-3^4| will be parsed as |2^(-(3^4))|, raising
1695 an error because the parser is by default integer only, but see the section
1696 about |\bnumsetup| which explains how to let |\bnumeval| compute fractions!
1697
1698
1699 \subsubsection{\cs{bnumdefinfix}}
1700
1701 It is possible to define infix binary operators of one's own choosing.%
1702 %
1703 \footnote{The effect of |\bnumdefinfix| is global if under |\xintglobaldefstrue| setting.}
1704
1705 For an example see the \autoref{ssec:binomial} on adding |;| as operator
1706 computing binomial coefficients. Other examples will also be given here.
1707
1708 The syntax is
1709 \centeredline{|\bnumdefinfix|\marg{operator}\marg{\string\macro}\marg{L-prec}\marg{R-prec}}
1710 \begin{description}
1711 \item[\marg{operator}] The characters for the operator, they may be letters or
1712 non-letters. Digits are not allowed to be first or last in \meta{operator}.
1713 The following characters are not allowed at all: |\|,
1714 \textcolor{smallverbcolor}{\{}, \textcolor{smallverbcolor}{\}}, and
1715 {\catcode`\% 12 |%|}.
1716 Spaces get removed.
1717
1718 \begin{quote}
1719 \textbf{Warning:} \meta{operator} will be expanded in an \cs{edef} and then
1720 further processed. If it is (or contains) a Babel-active character chances
1721 are that the further processing by |\bnumdefinfix| will give breakage.
1722 There will be no problem if |\bnumdefinfix| is used in the document
1723 preamble. Else, use |\string|, for example:
1724 |\bnumdefinfix{\string;}{\xintiiBinomial}{15}{15}|.
1725 \end{quote}
1726
1727 The underscore and the hash-tag both are allowed, but come with their
1728 specific provisos:
1729
1730 \begin{quote}
1731 The |_| character can be used, but not as first character of the operator,
1732 as it would be mis-construed on usage as part of the previous number, and
1733 ignored as such. Although |\bnumdefinfix{_}...| will not complain, it
1734 will remain ineffective. There must be some other character first.
1735 \end{quote}
1736
1737 \begin{quote}
1738 The |#| character can be used as an operator name or a character in such a
1739 name but the definition with |\bnumdefinfix| must be done either with
1740 |\string#| or with |####| (or even double that if the definition is done
1741 inside the replacement text of a macro; doubling also applies with the
1742 |\string| method, one will need |\string##| there) or while having
1743 temporarily set the catcode to for example letter or other.
1744
1745 Usage can be done without any special precaution (but as usual |#|
1746 may need doubling if in the replacement body of some macro definition).
1747 \end{quote}
1748
1749 \item[\marg{\string\macro}] The expandable macro (expecting two mandatory
1750 arguments) which is to assign to the infix operator. This macro must be
1751 \emph{f}-expandable. Also it must (if the default package configuration is
1752 not modified for the core operators) produce integers in the ``strict''
1753 format which is expected by the \xintcorename macros for arithmetic: no
1754 leading zeros, at most one minus sign, no plus sign, no spaces.
1755 \item[\marg{L-prec}] An integer, minimal |4|, maximal |22|, which governs
1756 the left-\hskip0pt precedence of the infix operator.
1757 \item[\marg{R-prec}] An integer, minimal |4|, maximal |22|, which governs the
1758 right-\hskip0pt precedence of the infix operator.
1759 \end{description}
1760 Generally, the two precedences are set to the same value.
1761
1762 Once a multi-character operator is defined, the first characters of its name
1763 can be used if no ambiguity. In case of ambiguity, it is the earliest defined
1764 shortcut which prevails, except for the full name. So for example if |$abc|
1765 operator is defined, and |$ab| is defined next, then |$| and |$a| will still
1766 serve as shortcuts to the original |$abc|, but |$ab| will refer to the newly
1767 defined operator.
1768
1769 Fully qualified names are never ambiguous, and a shortcut once defined will
1770 change meaning only under two circumstances:
1771 \begin{itemize}
1772 \item it is re-defined as the full name of a new operator,
1773 \item the original operator to which the shortcut refers is defined again; then
1774 the shortcut is automatically updated to point to the new meaning.
1775 \end{itemize}
1776 \begin{verbatim}
1777 \def\equals#1#2{\ifnum\pdfstrcmp{#1}{#2}=0 \expandafter1\else
1778 \expandafter0\fi}
1779 % or:
1780 \def\equals#1#2{\expanded{\ifnum\pdfstrcmp{#1}{#2}=0 1\else0\fi}}
1781 \def\differ#1#2{\expanded{\ifnum\pdfstrcmp{#1}{#2}=0 0\else1\fi}}
1782 \bnumdefinfix{==}{\equals}{10}{10}
1783 \bnumdefinfix{!=}{\differ}{10}{10}
1784 \bnumdefinfix{times}{\xintiiMul}{14}{14}
1785 \bnumdefinfix{++}{\xintiiAdd}{19}{19}
1786 \end{verbatim}
1787 %\def\equals#1#2{\ifnum\pdfstrcmp{#1}{#2}=0 \expandafter1\else\expandafter0\fi}
1788 \def\equals#1#2{\expanded{\ifnum\pdfstrcmp{#1}{#2}=0 1\else0\fi}}
1789 \def\differ#1#2{\expanded{\ifnum\pdfstrcmp{#1}{#2}=0 0\else1\fi}}
1790 \bnumdefinfix{==}{\equals}{10}{10}
1791 \bnumdefinfix{!=}{\differ}{10}{10}
1792 \bnumdefinfix{times}{\xintiiMul}{14}{14}
1793 \bnumdefinfix{++}{\xintiiAdd}{19}{19}
1794 %
1795 \bneshoweval{2 + 3! = 5, 2 + (3!) == 8} Notice in the |2+3! = 5| example that
1796 the existence of |!=| prevails on applying the factorial, so this is test
1797 whether |2+3| and |5| differ; it is not a matter of precedence here, but of
1798 input parsing ignoring spaces. And |2+3! == 8| would create an error as after
1799 having found the |!=| operator and now expecting a digit (as there is no |!==|
1800 operator) the parser would find an unexpected |=| and report an error. Hence
1801 the usage of parentheses in the input.%
1802 %
1803 \footnote{With \xintexprname, whose \cs{xinteval} has a |!=| operator,
1804 |2+3!==8| is interpreted automatically as |2+(3!)==8|, thanks to internal
1805 work-around added at |1.4g|. This has not been backported to \bnumname as
1806 it does not per default support operators such as |!=| or |==| and only has
1807 generic support for adding multi-character operators.
1808
1809 Regarding |2 + 3! = 5|, trying to let this be interpreted as |2+(3!)=5|
1810 makes sense only if a |=| operator has been defined. If no |!=| operator
1811 exists, the magic will be automatic. If however both |=| and |!=| exist,
1812 then it would need special overhead to the parser dealings when finding |!|
1813 to avoid the |!=| interpretation. One could imagine distinguishing |! =|
1814 from |!=| but the swallowing of spaces is deeply coded in the parser. As
1815 \bnumname default infix operators do not include one starting with |!|, it is not
1816 worth it to include in the package extra overhead to solve such issues when
1817 extending the syntax. At the level of \xintexprname, there is no issue
1818 because there is no |=| operator.}
1819 % RAPPEL j'ai dû utiliser à un moment \pagebreak[3] ici. Encore un
1820 % comportement inattendu et sous-optimal de TeX/LaTeX.
1821 \newline
1822 \noindent\bneshoweval{2^5 == 4 times 8, 11 t 14}
1823 \bneshoweval{100 ++ -10 ^ 3, (100 - 10)^3, 2 ^ 5 ++ 3, 2^(5+3)}
1824
1825
1826 \subsubsection{\cs{bnumdefpostfix}}
1827
1828 It is possible to define postfix unary operators of one's own choosing.%
1829 %
1830 \footnote{The effect of |\bnumdefpostfix| is global if under |\xintglobaldefstrue| setting.}
1831 %
1832 The syntax is
1833 \centeredline{|\bnumdefpostfix|\marg{operator}\marg{\string\macro}\marg{L-prec}}
1834
1835 \begin{description}
1836 \item[\marg{operator}] The characters for the operator name: same
1837 conditions as for |\bnumdefinfix|. Postfix and infix operators share the
1838 same name-space, regarding abbreviated names.
1839 \begin{quote}
1840 \textbf{Warning:} \meta{operator} will be expanded in an \cs{edef} and then
1841 further processed. If it is (or contains) a Babel-active character chances
1842 are that the further processing by |\bnumdefpostfix| will give breakage.
1843 There will be no problem if |\bnumdefpostfix| is used in the document
1844 preamble.
1845 \end{quote}
1846
1847 \item[\marg{\string\macro}] The one argument expandable macro to assign to
1848 the postfix operator. This macro only needs to be \emph{x}-expandable.
1849 \item[\marg{L-prec}] An integer, minimal |4|, maximal |22|, which governs the
1850 left-\hskip0pt precedence of the infix operator.
1851 \end{description}
1852
1853 Examples below which use the maximal precedence are typical of what is
1854 expected of a ``function'' (and I even used |.len()| notation with parentheses
1855 in one example, the parentheses are part of the postfix operator name). And
1856 indeed such postfix operators are thus a way to implement functions in
1857 disguise, circumventing the fact that the \bnumname parser will never be
1858 extended to work with functional syntax (for this, see \xintexprname). With
1859 the convention (followed in some examples) that such postfix operators start
1860 with a full stop, but never contain another one, we can chain simply by using
1861 concatenation (no need for parentheses), as there will be no
1862 ambiguity.
1863 \begin{verbatim}
1864 \usepackage{xint}% for \xintiiSum, \xintiiSqrt
1865 \def\myRev#1{\xintNum{\xintReverseOrder{#1}}}% reverse and trim leading zeros
1866 \bnumdefpostfix{$}{\myRev}{22}% the $ will have top precedence
1867 \bnumdefpostfix{:}{\myRev}{4}% the : will have lowest precedence
1868 \bnumdefpostfix{::}{\xintiiSqr}{4}% the :: is a completely different operator
1869 \bnumdefpostfix{.len()}{\xintLength}{22}% () for fun but a single . will be enough!
1870 \bnumdefpostfix{.sumdigits}{\xintiiSum}{22}% .s will abbreviate
1871 \bnumdefpostfix{.sqrt}{\xintiiSqrt}{22}% .sq will be unambiguous (but confusing)
1872 \bnumdefpostfix{.rep}{\xintReplicate3}{22}% .r will be unambiguous
1873 \end{verbatim}
1874 \def\myRev#1{\xintNum{\xintReverseOrder{#1}}}% reverse digits and trim leading zeros
1875 \bnumdefpostfix{$}{\myRev}{22}% $
1876 \bnumdefpostfix{:}{\myRev}{4}%
1877 \bnumdefpostfix{::}{\xintiiSqr}{4}%
1878 \bnumdefpostfix{.len()}{\xintLength}{22}%
1879 \bnumdefpostfix{.sumdigits}{\xintiiSum}{22}%
1880 \bnumdefpostfix{.sqrt}{\xintiiSqrt}{22}%
1881 \bnumdefpostfix{.rep}{\xintReplicate3}{22}%
1882 \bneshoweval{(2^31).len(), (2^31)., 2^31$, 2^31:, (2^31)$}
1883 \bneshoweval{(2^31).sqrt, 100000000.sq.sq}
1884 \bneshoweval{(2^31).sumdigits, 123456789.s, 123456789.s.s, 123456789.s.s.s}
1885 \bneshoweval{10^10+10000+2000+300+40+5:}
1886 \bneshoweval{1+2+3+4+5+6+7+8+9+10 :: +1 :: *2 :: :: :}
1887 \bneshoweval{123456789.r}\par\vspace*{-\baselineskip}% Toujours LaTeX avec ses
1888 % espaces verticaux TROP GRANDS DEPUIS 30 ANS
1889 \begin{verbatim}
1890 \bnumdefpostfix{.rep}{\xintReplicate5}{22}% .rep modified --> .r too
1891 \end{verbatim}
1892 \bnumdefpostfix{.rep}{\xintReplicate5}{22}%
1893 \bneshoweval{123456789.r}
1894
1895
1896 \normalsize
1897
1898 \section{Changes}
1899
1900 \begin{description}
1901 \input bnumexprchanges.tex
1902 \end{description}
1903
1904 \section{License}
1905
1906 \begingroup\makeatletter
1907 \baselineskip10pt
1908 \begin{verbatim}
1909 Copyright © 2014-2022, 2025 Jean-François Burnol
1910
1911 | This Work may be distributed and/or modified under the
1912 | conditions of the LaTeX Project Public License 1.3c.
1913 | This version of this license is in
1914
1915 > <http://www.latex-project.org/lppl/lppl-1-3c.txt>
1916
1917 | and version 1.3 or later is part of all distributions of
1918 | LaTeX version 2005/12/01 or later.
1919
1920 This Work has the LPPL maintenance status "author-maintained".
1921
1922 The Author and Maintainer of this Work is Jean-François Burnol.
1923
1924 This Work consists of the main source file and its derived files
1925
1926 bnumexpr.dtx, bnumexpr.sty, bnumexpr.pdf, bnumexpr.tex,
1927 bnumexprchanges.tex, README.md
1928 \end{verbatim}
1929 \endgroup
1930
1931 \normalsize
1932 \StopEventually{\end{document}\endinput}
1933
1934 % workaround
1935 % https://github.com/latex3/latex2e/issues/563
1936 \makeatletter
1937 \AddToHook{env/macrocode/begin}{\partopsep0pt\relax}
1938 \AddToHook{env/macrocode/after}{\@nobreakfalse}
1939 \makeatother
1940 \newgeometry{hscale=0.75,vscale=0.75}% ATTENTION \newgeometry fait
1941 % un reset de vscale si on ne le
1942 % précise pas ici !!!
1943
1944 \MakePercentIgnore
1945 \AddToHook{env/macrocode/begin}{\bfseries}
1946 %
1947 % \catcode`\<=0 \catcode`\>=11 \catcode`\*=11 \catcode`\/=11
1948 % \let</dtx>\relax
1949 % \def<*package>{\catcode`\<=12 \catcode`\>=12 \catcode`\*=12 \catcode`\/=12 }
1950 %</dtx>
1951 %<*package>
1952 % \etocdepthtag.toc{code}
1953 % \section{Commented source code}
1954 % \label{sec:bnumexprcode}
1955 % \etocdefaultlines
1956 % \etocignoredepthtags
1957 % \etocsetnexttocdepth{subsection}
1958 % \localtableofcontents
1959 % \etocmarkbothnouc{\bnumnameuserdoc \hyperref[sec:bnumexprcode]{implementation}}
1960 %
1961 % At |1.7|, support for lowercase hexadecimal was added.
1962 %
1963 % At |1.6|, |\bnumeval| requires the |1.4n| release of \xintcorename and
1964 % \xintbinhexname (or at least of \xintkernelname if option |custom| is used).
1965 % It adds |0b|, |0o|, |'|, and |0x| to the syntax, and
1966 % admits optional parameters |[b]|, |[o]|, and |[h]| to produce the
1967 % output converted to binary, octal, or hexadecimal.
1968 %
1969 % It is amusing that implementing the support for the optional argument
1970 % had the un-anticipated corollary that Babel active characters (such as
1971 % |!| with French) are auto-taming. See the code comments.
1972 %
1973 % A problem with |_| if upfront in numbers was fixed.
1974 %
1975 % There was some refactoring, relative to extending |\bnumsetup| with new keys
1976 % related to base conversion macros and this lead to the removal of
1977 % \cs{bnumprintonetohex} and \cs{bnumhextodec}.
1978 %
1979 % At |1.5|, right-associativity was enforced for powers in anticipation of
1980 % upstream \xintexprname |1.4g 2021/05/25|, and the customizability and
1981 % extendibility of the package is made total via added |\bnumdefinfix| and
1982 % |\bnumdefpostfix|.
1983 %
1984 %
1985 % \footnotesize
1986 %
1987 % Older comments at time of |1.4| and |1.4a| releases:
1988 %
1989 % I transferred mid-May 2021 from \xintexprname its \cs{expanded} based
1990 % infra-structure from its own |1.4| release of January 2020 and bumped
1991 % version to |1.4|. Also I added support for hexadecimal input and output, via
1992 % \xintbinhexname.
1993 %
1994 % A few comments added here at |1.4a|:
1995 % \begin{itemize}
1996 % \item It looked a bit costly and probably would have been mostly useless to
1997 % end users to integrate in \bnumname support for nested structures via
1998 % square brackets [, ], which is in \xintexprname since its January 2020
1999 % |1.4| release. But some of the related architecture remains here; we
2000 % could make some gains probably but diverging from upstream code
2001 % would make maintenance a nightmare.
2002 % \item Formerly, the |\csname...\endcsname| encapsulation technique had the
2003 % after-effect to allow the macros supporting the infix operators to be only
2004 % \emph{x}-expandable. At |1.4|, I could have still allowed support macros
2005 % being only \emph{x}-expandable, but, keeping in sync with upstream, I have
2006 % used only a \cs{romannumeral} trigger and did not insert an \cs{expanded}, so
2007 % now the support macros must be \emph{f}-expandable. The |1.4a| release
2008 % fixes the related user documentation of |\bnumsetup| which was not
2009 % updated at |1.4|. The support macro for the factorial however needs only
2010 % be \emph{x}-expandable.
2011 % \item Also, I simply do not understand why the legacy (|1.2e|) user documentation
2012 % said that the support macros were supposed to \emph{f}-expand their arguments, as
2013 % they are used only with arguments being explicit digit tokens (and
2014 % optional minus sign).
2015 % \item The |\bnumexpr\relax| syntax creating an empty ople is by itself now
2016 % legal, and can be injected (comma separated) in an expression, keeping it
2017 % invariant, however |\bnumeval{}| ends in a
2018 % |Paragraph ended before \BNE_print_c was complete|
2019 % error because |\BNEprint| makes the tacit requirement
2020 % that the 1D ople to output has at least one item.
2021 % \end{itemize}
2022 %
2023 % \normalsize
2024 % \subsection{Package identification}
2025 % \begin{macrocode}
2026 \NeedsTeXFormat{LaTeX2e}%
2027 \ProvidesPackage{bnumexpr}[2025/09/27 v1.7b Expressions with big integers (JFB)]%
2028 % \end{macrocode}
2029 % \subsection{Load \xintkernelname}
2030 % At |1.6|, in order to make the base conversion macros also customizable,
2031 % hence not mandate loading of \xintbinhexname, we only load unconditionally
2032 % \xintkernelname.
2033 %
2034 % We then switch to the familiar catcode regime of the \xintexprname sources.
2035 % \begin{macrocode}
2036 \RequirePackage{xintkernel}[2025/09/05]%
2037 % \end{macrocode}
2038 % \subsection{Save catcode regime and switch to our own}
2039 % \begin{macrocode}
2040 \edef\BNErestorecatcodesendinput{\XINTrestorecatcodes\noexpand\endinput}%
2041 \XINTsetcatcodes%
2042 % \end{macrocode}
2043 % \subsection{Load optionally \xintcorename and \xintbinhexname}
2044 % |1.6| adds |customcore| as alias of legacy |custom|. It adds |custombinhex|
2045 % to add possibility of not loading \xintbinhexname either.
2046 % Option |custom| now means both of |customcore| and |custombinhex|.
2047 %
2048 % But who on Earth isn't going to use with delight both my \xintcorename and
2049 % \xintbinhexname?
2050 % \begin{macrocode}
2051 \def\BNE_tmpa{1}\def\BNE_tmpb{1}%
2052 \DeclareOption{custom}{\def\BNE_tmpa{0}\def\BNE_tmpb{0}}%
2053 \DeclareOption{customcore}{\def\BNE_tmpa{0}}%
2054 \DeclareOption{custombinhex}{\def\BNE_tmpb{0}}%
2055 \ProcessOptions\relax
2056 \if1\BNE_tmpa\RequirePackage{xintcore}[2025/09/05]\fi
2057 \if1\BNE_tmpb\RequirePackage{xintbinhex}[2025/09/05]\fi
2058 % \end{macrocode}
2059 % \subsection{\cs{bnumsetup}}
2060 % |\bnumsetup| is the new name at |1.5| of \cs{bnumexprsetup}.
2061 % The old name was kept as an alias at |1.5|, and deleted at |1.6|.
2062 %
2063 % Note that a final comma will cause no harm.
2064 % \begin{macrocode}
2065 \catcode`! 3
2066 \def\bnumsetup #1{\BNE_parsekeys #1,=!,}%
2067 \def\BNE_parsekeys #1=#2#3,%
2068 {%
2069 \ifx!#2\expandafter\BNE_parsedone\fi
2070 \XINT_global
2071 \expandafter
2072 \let\csname BNE_Op_\xint_zapspaces #1 \xint_gobble_i\endcsname%
2073 =#2%
2074 \ifxintverbose
2075 \PackageInfo{bnumexpr}{assigned
2076 \ifxintglobaldefs globally \fi
2077 \string#2 to \xint_zapspaces #1 \xint_gobble_i\MessageBreak
2078 % \end{macrocode}
2079 % Workaround for the space inserted by \cs{on@line}.
2080 % \begin{macrocode}
2081 \expandafter\xint_firstofone}%
2082 \fi
2083 \BNE_parsekeys
2084 }%
2085 \def\BNE_parsedone #1\BNE_parsekeys {}%
2086 \catcode`! 12
2087 % \end{macrocode}
2088 % Final comma and spaces are only to check if it does work. But I will
2089 % NOT insert spaces before commas, even though they are allowed!
2090 %
2091 % |1.6| also handles base conversion macros here. Prior to |1.6| this
2092 % |\bnumsetup| configuration was not executed if package received option
2093 % |custom| (now |customcore|). But as the user is then responsible for
2094 % redefining all keys, why bother.
2095 % \begin{macrocode}
2096 \bnumsetup{%
2097 add = \xintiiAdd, sub = \xintiiSub, opp = \xintiiOpp,
2098 mul = \xintiiMul, pow = \xintiiPow, fac = \xintiiFac,
2099 div = \xintiiDivFloor, mod = \xintiiMod, divround = \xintiiDivRound,
2100 hextodec=\xintHexToDec, octtodec=\xintOctToDec, bintodec=\xintBinToDec,
2101 dectohex=\xintDecToHex, dectooct=\xintDecToOct, dectobin=\xintDecToBin,
2102 }%
2103 % \end{macrocode}
2104 % By the way the keys should have been |Add|, |Sub|, \dots, not |add|, |sub|,
2105 % \dots, so internally |\BNE_Op_Add| etc\dots\ would have been the macros
2106 % defined by |\bnumsetup| and used in the code, not |\BNE_Op_add| (et al.)
2107 % whose casing does not match my naming conventions.
2108 %
2109 % \subsection{Some extra constants needed for user defined precedences}
2110 %
2111 % For the mechanism of |\bnumdefinfix| we need precedence levels to be
2112 % available as \cs{chardef}'s. \xintkernelname already provides 0-10, 12, 14,
2113 % 16, 17, 18, 20, 22.
2114 %
2115 % Left levels need to be represented by one token; right levels are
2116 % hard-coded into |checkp_<op>| macros and could have been there explicit
2117 % digit tokens but we will use the |\xint_c_...| \cs{char}-tokens.
2118 % \begin{macrocode}
2119 \chardef\xint_c_xi 11
2120 \chardef\xint_c_xiii 13
2121 \chardef\xint_c_xv 15
2122 \chardef\xint_c_xix 19
2123 \chardef\xint_c_xxi 21
2124 % \end{macrocode}
2125 % \subsection{\cs{bnumexpr}, \cs{bnethe}, \cs{bnumeval}}
2126 %
2127 % |\XINTfstop| has to be the same as defined in \xintexprname, in order for a
2128 % subexpression |\xintiiexpr...\relax| to get recognized in |\bnumeval| or
2129 % conversely for |\bnumexpr...\relax| to possibly serve inside an
2130 % |\xinteval|. But why use \bnumname then? Besides a sub \xintexprname-ession
2131 % will break |\bnumeval| if it is anything else than a 1D flat sequence. And
2132 % even then it can work only if internal storage format are kept in sync.
2133 %
2134 % |1.6| deprecates \cs{evaltohex} (and |1.7a| removed it) and has
2135 % |\bnumeval||[h]| as equivalent replacement.
2136 %
2137 % The \cs{protected} |\BNEprint| will survive to |\bnumexpr| being expanded in
2138 % a \cs{write} or \cs{edef}. But its expansion will be forced by the
2139 % \cs{expanded} from |\bnethe|.
2140 %
2141 % I now really dislike |\thebnumexpr| macro name and at some point had
2142 % replaced it with |\bnumtheexpr| but this got reverted.
2143 %
2144 % |1.6a| uses the strange |\csname| in place of directly |\BNE_wrap|
2145 % in order to fix the |1.6| blunder which had done a similar thing,
2146 % but too late. This is to tame Babel active characters. The |\bnumeval|
2147 % was OK, though. Sadly the blunder was first done in \xintexprname,
2148 % after I had reverted perfectly valid implementation there, having
2149 % thought I could apply a shortcut, which was simply a brain fault.
2150 % And I backported it here... alas...
2151 % \begin{macrocode}
2152 \def\XINTfstop {\noexpand\XINTfstop}%
2153 \def\bnumexpr {\romannumeral0\bnumexpro}%
2154 \def\bnumexpro {\csname BNE_wrap\expandafter\endcsname
2155 \romannumeral0\BNE_bareeval}%
2156 % \end{macrocode}
2157 % While preparing |1.6| I wondered why the ``|.|'' after |\BNEprint| in
2158 % |\BNE_wrap| which is then gobbled by |\BNEprint|. It was clear it came from
2159 % \xintexprname, but why was it kept here?
2160 %
2161 % The reason is to support having a sub |\bnumexpr...\relax| inside
2162 % |\bnumeval| or |\xinteval|. Indeed such a sub-expression is identified via
2163 % the presence of the |\XINTfstop| after its expansion, and the code inside
2164 % \bnumname handling this is inherited from \xintexprname, so it expects the
2165 % structure |\XINTfstop| then a ``print'' macro, then possibly some stuff
2166 % delimited by a full stop (this is related to the implementation of the
2167 % optional arguments of |\xintfloateval| and |\xintieval|).
2168 %
2169 % As we keep this stuff handled the same way we must inject the seemingly
2170 % silly full stop here for |\bnumexpr...\relax| (or a macro defined from it
2171 % via an \cs{edef}) to be usable inside |\bnumval| or another
2172 % |\bnumexpr...\relax|.
2173 % \begin{macrocode}
2174 \def\BNE_wrap {\XINTfstop\BNEprint.}%
2175 % \end{macrocode}
2176 % It is important to keep in mind that |#1| has the structure
2177 % \{|{...}{...}...{...}|\} with an external brace pair, which here gets
2178 % removed. ^^A attention que mon |...| va retirer une paire d'accolades.
2179 % In the replacement the external \{|...|\} are for \cs{expanded}.
2180 %
2181 % We also define a non \cs{protected} variant without the strange extra full
2182 % stop, it will serve for |\bnumeval| (and |\thebnumexpr|) and thus does not
2183 % need it.
2184 % \begin{macrocode}
2185 \protected\def\BNEprint.#1{{\BNE_unpack#1.}}%
2186 \def\BNEprint_#1{{\BNE_unpack#1.}}%
2187 % \end{macrocode}
2188 % |\bnethe| removes the |\XINTfstop| and activates the printing via
2189 % |\BNEprint|.
2190 %
2191 % Attention that prior to |1.6| |\bnethe| grabbed a |#1|, hence would
2192 % work to print a braced |\bnumexpr...\relax|, but I don't see the reason
2193 % for doing that. Removed.
2194 %
2195 % |1.6a| modifies |\thebnumexpr| here for the Babel active thing.
2196 % \begin{macrocode}
2197 \def\bnethe{\expanded\expandafter\xint_gobble_i\romannumeral`&&@}%
2198 \def\thebnumexpr{\expanded\csname BNEprint_\expandafter\endcsname
2199 \romannumeral0\BNE_bareeval}%
2200 % \end{macrocode}
2201 % At |1.6| after implementing the |[h]| optional argument of |\bnumeval|,
2202 % there was the un-anticipated result that this tamed Babel active characters.
2203 % This is explained by the expansion happening while a \cs{csname} is not yet
2204 % closed. And by the fact that during its expansion |\bnumeval| does not
2205 % use delimited macros, for example to fetch up to a closing parenthesis.
2206 %
2207 % There used to be here a |\BNE_start| but it got replaced by its expansion.
2208 %
2209 % The |\BNE_check| is defined in the section \hyperref[ssec:span]{``Expansion
2210 % spanning''}.
2211 %
2212 % Prior to |1.6| |\BNE_bareeval| was named \cs{bnebareeval}, but this was
2213 % outside of the package namespace (it should have been \cs{bnumbareeval}, or
2214 % \cs{bnumexprbareeval}). Upstream has |\xintbareeval| without underscores
2215 % for legacy reasons.
2216 % \begin{macrocode}
2217 \def\BNE_bareeval{\expandafter\BNE_check\romannumeral`&&@\BNE_getnext}%
2218 % \end{macrocode}
2219 % These next are not \cs{protected} because they are only used with
2220 % |\bnumeval|, there is no analog of the private format which |\bnumexpr|
2221 % expands to. This also spares us having to define macros with names which
2222 % can be written to an external file and re-read using the standard catcodes.
2223 %
2224 % MEMO: |\BNEprint_| (with the trailing underscore) will be used in case of
2225 % absence of optional argument and has been defined already. It is important
2226 % for compatibility with the others here that it did not use the strange full
2227 % stop in its parameter pattern. It is also important that it is \emph{not}
2228 % |\protected|, as we want |\bnumeval| to expand fully in an |\edef|.
2229 % \begin{macrocode}
2230 \expandafter\def\csname BNEprint_[h]\endcsname#1{{\BNE_unpack_tohex#1.}}%
2231 \expandafter\def\csname BNEprint_[ha]\endcsname#1{{\BNE_unpack_tolowhex#1.}}%
2232 \expandafter\def\csname BNEprint_[o]\endcsname#1{{\BNE_unpack_tooct#1.}}%
2233 \expandafter\def\csname BNEprint_[b]\endcsname#1{{\BNE_unpack_tobin#1.}}%
2234 \expandafter\let\csname BNEprint_[]\endcsname\BNEprint_
2235 % \end{macrocode}
2236 % |[b]|, |[o]| and |[h]| added at |1.6|.
2237 %
2238 % We break here the \xintname legacy pattern to always use |\romannumeral0|
2239 % trigger first, via a CamelCase/lowercase pair of macros.
2240 %
2241 % The |\expanded| will \emph{not} control the actual computation, only launch
2242 % it via propagation across the |\csname...\endcsname| to a |\romannumeral0|
2243 % for the \emph{f}-expansion of |\BNE_bareeval|. After computation is done
2244 % the |\BNEprint_[...]| will present the braced result (maybe containing
2245 % multiple items) to this |\expanded|, and the unpacking is thus done via an
2246 % \cs{edef}-like expansion. This does not mean that all helpers doing
2247 % conversion to the binary bases only have to be \emph{x}-expandable, because
2248 % their full expansion must be achieved prior to the ``printone'' macros
2249 % activate. If the latter are |\protected|, then conversions do only need
2250 % \emph{x}-expandability. Except for the support of the |[ha]| option which
2251 % needs |\BNE_Op_dectohex| to be \emph{f}-expandable as the helper doing
2252 % conversion to lowercase must act on a fully prepared uppercase hexadecimal
2253 % input.
2254 % \begin{macrocode}
2255 \def\bnumeval #1#{\expanded\bnumeval_a{#1}}%
2256 \def\bnumeval_a#1#2{%
2257 \csname BNEprint_\xint_zapspaces #1 \xint_gobble_i\expandafter
2258 \endcsname\romannumeral0\BNE_bareeval#2\relax
2259 }%
2260 % \end{macrocode}
2261 % This code is more compact at |1.6| than at |1.5|. Various renamings at
2262 % |1.7| and addition of the |[ha]| optional argument.
2263 % \begin{macrocode}
2264 \def\BNE_unpack#1{%
2265 \bnumprintone{#1}\expandafter\BNE_unpack_a\string
2266 }%
2267 \def\BNE_unpack_a#1{%
2268 \if#1.\BNE_allitemsdone\fi\bnumprintonesep
2269 \expandafter\BNE_unpack\expandafter{\iffalse}\fi
2270 }%
2271 \def\BNE_allitemsdone\fi#1\fi{\fi}%
2272 % \end{macrocode}
2273 % There is a breaking change at |1.6| as formerly there was a
2274 % \cs{bnumprintonetohex}. Now, the decimal to hexadecimal conversion is done
2275 % always, and the customizable wrapper was thus renamed to |\bnumprintonehex|.
2276 % \begin{macrocode}
2277 \def\BNE_unpack_tohex#1{%
2278 \expandafter\bnumprintonehex
2279 \expandafter{\romannumeral`&&@\BNE_Op_dectohex{#1}}%
2280 \expandafter\BNE_unpack_tohex_a\string
2281 }%
2282 \def\BNE_unpack_tohex_a#1{%
2283 \if#1.\BNE_allitemsdone\fi\bnumprintonesep
2284 \expandafter\BNE_unpack_tohex\expandafter{\iffalse}\fi
2285 }%
2286 % \end{macrocode}
2287 % Conversion to lowercase hexadecimal added at |1.7|. By the way |[ha]|
2288 % option is broken (but also is not needed) if the |\BNE_Op_dectohex| is
2289 % customized to produce lowercase, not uppercase as |\xintiiDecToHex|.
2290 %
2291 % As we inject ending pattern already here, to skip one ``grab argument''
2292 % step, we do not provide a wrapper which could be configured via
2293 % |\bnumsetup| with something such as |dectolowhex=\macro|.
2294 % \begin{macrocode}
2295 \def\BNE_unpack_tolowhex#1{%
2296 \expandafter\bnumprintonelowerhex
2297 \expanded{{\BNE_dectolowhex_a{#1}\xint_bye23456789\xint_bye\endcsname}}%
2298 \expandafter\BNE_unpack_tolowhex_a\string
2299 }%
2300 \def\BNE_unpack_tolowhex_a#1{%
2301 \if#1.\BNE_allitemsdone\fi\bnumprintonesep
2302 \expandafter\BNE_unpack_tolowhex\expandafter{\iffalse}\fi
2303 }%
2304 % \end{macrocode}
2305 % So the |dectohex| must be \emph{f}-expandable: even with a
2306 % |\bnumprintonelowerhex| redefined |\protected|, the conversion
2307 % to lowercase needs the |\BNE_Op_dectohex| action to be complete
2308 % before |\BNE_dectolowhex_b| kicks in.
2309 % \begin{macrocode}
2310 \def\BNE_dectolowhex_a{%
2311 \expandafter\BNE_dectolowhex_b\romannumeral`&&@\BNE_Op_dectohex
2312 }%
2313 % \end{macrocode}
2314 % When preparing the |1.7| release I hesitated about testing the token to
2315 % check if a digit or a letter (which could have been done on the basis of the
2316 % catcode only, but would have then put a constraint on the catcodes used by
2317 % the |\BNE_Op_dectohex| macro), and let digits go through ``as is''. But I
2318 % was not willing to think too much and the advantage of the approach chosen
2319 % is the elegance of its termination. I did not do comparative efficiency
2320 % tests, which would have cost me at least one or two hours in setting them up
2321 % and comparing with alternative implementation, besides I would have had to
2322 % decide if I base decision on behavior of very long inputs or of the short
2323 % ones. All of that for something perhaps nobody on Earth will ever use
2324 % anyhow.
2325 % \begin{macrocode}
2326 \def\BNE_dectolowhex_b #1#2#3#4#5#6#7#8#9{%
2327 \csname BNE_lower #1\endcsname
2328 \csname BNE_lower #2\endcsname
2329 \csname BNE_lower #3\endcsname
2330 \csname BNE_lower #4\endcsname
2331 \csname BNE_lower #5\endcsname
2332 \csname BNE_lower #6\endcsname
2333 \csname BNE_lower #7\endcsname
2334 \csname BNE_lower #8\endcsname
2335 \csname BNE_lower #9\endcsname
2336 \BNE_dectolowhex_b
2337 }%
2338 \expandafter\let\csname BNE_lower \endcsname\empty
2339 \expandafter\def\csname BNE_lower 0\endcsname{0}%
2340 \expandafter\def\csname BNE_lower 1\endcsname{1}%
2341 \expandafter\def\csname BNE_lower 2\endcsname{2}%
2342 \expandafter\def\csname BNE_lower 3\endcsname{3}%
2343 \expandafter\def\csname BNE_lower 4\endcsname{4}%
2344 \expandafter\def\csname BNE_lower 5\endcsname{5}%
2345 \expandafter\def\csname BNE_lower 6\endcsname{6}%
2346 \expandafter\def\csname BNE_lower 7\endcsname{7}%
2347 \expandafter\def\csname BNE_lower 8\endcsname{8}%
2348 \expandafter\def\csname BNE_lower 9\endcsname{9}%
2349 \expandafter\def\csname BNE_lower A\endcsname{a}%
2350 \expandafter\def\csname BNE_lower B\endcsname{b}%
2351 \expandafter\def\csname BNE_lower C\endcsname{c}%
2352 \expandafter\def\csname BNE_lower D\endcsname{d}%
2353 \expandafter\def\csname BNE_lower E\endcsname{e}%
2354 \expandafter\def\csname BNE_lower F\endcsname{f}%
2355 % \end{macrocode}
2356 % Octal and binary added at |1.6|.
2357 % \begin{macrocode}
2358 \def\BNE_unpack_tooct#1{%
2359 \expandafter\bnumprintoneoct
2360 \expandafter{\romannumeral`&&@\BNE_Op_dectooct{#1}}%
2361 \expandafter\BNE_unpack_tooct_a\string
2362 }%
2363 \def\BNE_unpack_tooct_a#1{%
2364 \if#1.\BNE_allitemsdone\fi\bnumprintonesep
2365 \expandafter\BNE_unpack_tooct\expandafter{\iffalse}\fi
2366 }%
2367 \def\BNE_unpack_tobin#1{%
2368 \expandafter\bnumprintonebin
2369 \expandafter{\romannumeral`&&@\BNE_Op_dectobin{#1}}%
2370 \expandafter\BNE_unpack_tobin_a\string
2371 }%
2372 \def\BNE_unpack_tobin_a#1{%
2373 \if#1.\BNE_allitemsdone\fi\bnumprintonesep
2374 \expandafter\BNE_unpack_tobin\expandafter{\iffalse}\fi
2375 }%
2376 \let\bnumprintone \xint_firstofone
2377 \let\bnumprintonehex\xint_firstofone
2378 \let\bnumprintonelowerhex\xint_firstofone
2379 \let\bnumprintoneoct\xint_firstofone
2380 \let\bnumprintonebin\xint_firstofone
2381 \def\bnumprintonesep{, }%
2382 % \end{macrocode}
2383 % \subsection{\cs{BNE_getnext}}
2384 % The upstream |\BNE_put_op_first| has a string of included \cs{expandafter},
2385 % which was imported here at |1.4| and |1.4a| but they serve nothing in our
2386 % context. Removed this useless overhead at |1.5|.
2387 %
2388 % This |\BNE_getnext| token is injected via "start" macros associated to
2389 % operators or similar syntax elements, as will be seen later on. It tries
2390 % to get next operand.
2391 % \begin{macrocode}
2392 \def\BNE_getnext #1%
2393 {%
2394 \expandafter\BNE_put_op_first\romannumeral`&&@%
2395 \expandafter\BNE_getnext_a\romannumeral`&&@#1%
2396 }%
2397 \def\BNE_put_op_first #1#2#3{#2#3{#1}}%
2398 \def\BNE_getnext_a #1%
2399 {%
2400 \ifx\relax #1\xint_dothis\BNE_foundprematureend\fi
2401 \ifx\XINTfstop#1\xint_dothis\BNE_subexpr\fi
2402 \ifcat\relax#1\xint_dothis\BNE_countetc\fi
2403 \xint_orthat{}\BNE_getnextfork #1%
2404 }%
2405 % \end{macrocode}
2406 % This is executed when the parser was trying to find an operand but ended up
2407 % hitting the \cs{relax} end-marker. This happens if the expression is empty
2408 % for example. It trickled down here from \xintexprname, but |\xinteval| has
2409 % a notion of empty value. This is not the case of the \bnumname parser. So
2410 % let's at |1.7a| allow |\bnumeval{}| or |\bnumeval{_}|. Indeed
2411 % |\bnumeval{"}| and |\bnumeval{"_}| which use an other codebranch worked
2412 % already. This goes via replacing |{{}}| ^^A attention my verb
2413 % by |{{0}}|.
2414 % \begin{macrocode}
2415 \def\BNE_foundprematureend\BNE_getnextfork #1{{0}\xint_c_\relax}%
2416 \def\BNE_subexpr #1.#2%
2417 {%
2418 \expanded{\unexpanded{{#2}}\expandafter}\romannumeral`&&@\BNE_getop
2419 }%
2420 % \end{macrocode}
2421 % At |1.6| this also filters for \cs{catcode} (as per \xintname |1.4g|
2422 % |2021/05/25|).
2423 % \begin{macrocode}
2424 \def\BNE_countetc\BNE_getnextfork#1%
2425 {%
2426 \if0\ifx\count#11\fi
2427 \ifx\numexpr#11\fi
2428 \ifx\catcode#11\fi
2429 \ifx\dimen#11\fi
2430 \ifx\dimexpr#11\fi
2431 \ifx\skip#11\fi
2432 \ifx\glueexpr#11\fi
2433 \ifx\fontdimen#11\fi
2434 \ifx\ht#11\fi
2435 \ifx\dp#11\fi
2436 \ifx\wd#11\fi
2437 \ifx\fontcharht#11\fi
2438 \ifx\fontcharwd#11\fi
2439 \ifx\fontchardp#11\fi
2440 \ifx\fontcharic#11\fi
2441 0\expandafter\BNE_fetch_as_number\fi
2442 \expandafter\BNE_getnext_a\number #1%
2443 }%
2444 \def\BNE_fetch_as_number
2445 \expandafter\BNE_getnext_a\number #1%
2446 {%
2447 \expanded{{{\number#1}}\expandafter}\romannumeral`&&@\BNE_getop
2448 }%
2449 % \end{macrocode}
2450 % In the case of hitting a |(|, previous release inserted directly a
2451 % |\BNE_oparen|. But the expansion architecture imported from upstream
2452 % |\xintiiexpr| has been refactored, and the |..._oparen| meaning and
2453 % usage evolved. We stick with |{}\xint_c_ii^v (| from upstream.
2454 %
2455 % Also, at |1.6|, slight refactoring to handle digit tokens and opening
2456 % parenthesis a bit faster (but this is only first token...); and to ignore an
2457 % underscore as first character (rather than raise an error in this case).
2458 %
2459 % This merges former |\BNE_getnextfork| and |\BNE_scan_number|.
2460 % \begin{macrocode}
2461 \def\BNE_getnextfork #1{%
2462 \if#1-\xint_dothis {{}{}-}\fi
2463 \if#1(\xint_dothis {{}\xint_c_ii^v (}\fi
2464 \ifnum\xint_c_ix<1\string#1 \xint_dothis {\BNE_startint#1}\fi
2465 \xint_orthat {\BNE_getnextfork_a #1}%
2466 }%
2467 % \end{macrocode}
2468 % Prior to |1.7a|, the |_| was handled as is a |+| encountered while looking
2469 % for a value, i.e.\@ was simply ignored and expansion continued with
2470 % |\BNE_get_next_a|. This made |_^| raise an ``unexpected token'' error when
2471 % hitting the |^|. But in contrast |"_^| does work without complaining and is
2472 % the same as |0^|. So at |1.7a| we jump to |\BNE_startint|. Mind that the
2473 % latter differs from |\BNE_startoct| and |\BNE_starthex| and expects to grab
2474 % an already expanded argument, and naturally we inject a zero. As a
2475 % corollary of this |1.7a| change, also |_x|, |_o| and |_b| can serve as
2476 % prefixes for hexadecimal, octal and binary.
2477 % \begin{macrocode}
2478 \def\BNE_getnextfork_a #1{%
2479 \if#1_\xint_dothis {\BNE_startint 0}\fi
2480 \if#1+\xint_dothis \BNE_getnext_a \fi
2481 \if#1'\xint_dothis \BNE_startoct\fi
2482 \if#1"\xint_dothis \BNE_starthex\fi
2483 \xint_orthat {\BNE_unexpected #1}%
2484 }%
2485 % \end{macrocode}
2486 % If user employs |\bnumdefinfix| with |\string#|, and then tries |100##3|,
2487 % the first |#| will be interpreted as operator (assuming no
2488 % operator starting with |##| has actually been defined) and the error
2489 % "message" (which is not using \cs{message} or a \cs{write})
2490 % will then be
2491 % \centeredline{|! xint error: Unexpected token `##'. Ignoring.|}
2492 % because the parser is actually looking for a digit but finds the second |#|,
2493 % and TeX displays it doubled. This is doubly
2494 % confusing, but well, let's not dwell on that.
2495 %
2496 % |\BNE_unexpected| replaced here |\BNE_notadigit| at |1.6|.
2497 % \begin{macrocode}
2498 \def\BNE_unexpected#1%
2499 {%
2500 \XINT_expandableerror{Unexpected token `#1'. Ignoring.}\BNE_getnext_a
2501 }%
2502 % \end{macrocode}
2503 % \subsection{Parsing decimal, hexadecimal, octal, and binary}
2504 %
2505 % Somewhat refactored at |1.6| compared to upstream |1.4m|, on the occasion of
2506 % adding support for prefixes beyond only |"| as earlier.
2507 %
2508 % Ironically, |1.6| fixed the case of an underscore |_| as first character in
2509 % decimal input, but caused a regression for |"| prefixed hexadecimal, which
2510 % ceased to allow a |_| right after the |"|. Fixed at |1.7|.
2511 %
2512 % But |1.7| introduced a regression causing |"0| input to now break... pfff...
2513 % hence |1.7a|. A secondary (longstanding) issue is that the test suite was
2514 % limited to examples as shown in the docs (which used a special mark-up to
2515 % check automatically for regressions), but this way made inconvenient to test
2516 % for special inputs such as an hexadecimal prefix only followed with zeros
2517 % and underscores, which is what got broken in a too speedy refactoring at
2518 % |1.7|. The development sources now have a better way of checking against
2519 % regression prior to releases.
2520 %
2521 % \subsubsection{Prefix dispatch}
2522 % \begin{macrocode}
2523 \def\BNE_startint #1%
2524 {%
2525 \if #10\expandafter\BNE_scanint_gobz_a\else\expandafter\BNE_scanint_a\fi #1%
2526 }%
2527 \def\BNE_wrapint_after{\iffalse{{{\fi}}}}%
2528 \def\BNE_scanint_gobz_a #1#2{%
2529 \expandafter\BNE_scanint_gobz_b\romannumeral`&&@#2%
2530 }%
2531 % \end{macrocode}
2532 % It is important in case of |x|, |o|, or |b| to jump to
2533 % |\BNE_starthex| (et al.) and not for example to |\BNE_scanhex_a|
2534 % because the latter expects an \emph{f}-expansion to have been
2535 % applied already to what comes next (this comment is half-obsolete
2536 % at |1.7| which has no |\BNE_scanhex_a| anymore).
2537 %
2538 % Besides, we do want to trim out
2539 % leading zeroes after the |0b|, |0o|, or |0x| prefix: although the
2540 % macros of \xintbinhexname do accept leading zeros on input, they may
2541 % then produce decimal output with leading zeros, and the ``|ii|''
2542 % macros of \xintcorename consider that an input is vanishing as soon
2543 % as the first digit is |0|.
2544 % \begin{macrocode}
2545 \def\BNE_scanint_gobz_b #1%
2546 {%
2547 \ifx b#1\xint_dothis \BNE_startbin \fi
2548 \ifx o#1\xint_dothis \BNE_startoct \fi
2549 \ifx x#1\xint_dothis \BNE_starthex \fi
2550 \xint_orthat {\BNE_scanint_gobz_c #1}%
2551 }%
2552 \def\BNE_scanint_gobz_c #1%
2553 {%
2554 \expandafter{\expanded{{\iffalse}}}\fi
2555 \BNE_scanint_gobz_main#1%
2556 }%
2557 % \end{macrocode}
2558 % \subsubsection{Decimal}
2559 % No |\BNE_wrapint_before|, it has been inlined here at |1.7| (and in
2560 % |\BNE_scanint_gobz_c|).
2561 % \begin{macrocode}
2562 \def\BNE_scanint_a #1#2{%
2563 \expandafter{\expanded{{\iffalse}}}\fi #1%
2564 \expandafter\BNE_scanint_main\romannumeral`&&@#2%
2565 }%
2566 \def\BNE_scanint_main #1%
2567 {%
2568 \ifcat \relax #1\expandafter\BNE_scanint_hit_cs \fi
2569 \ifnum\xint_c_ix<1\string#1 \else\expandafter\BNE_scanint_checkagain\fi
2570 #1\BNE_scanint_again
2571 }%
2572 \def\BNE_scanint_again #1%
2573 {%
2574 \expandafter\BNE_scanint_main\romannumeral`&&@#1%
2575 }%
2576 % \end{macrocode}
2577 % Upstream (at |1.4f|) has |_getop| here, but let's jump directly to |BNE_getop_a|.
2578 % \begin{macrocode}
2579 \def\BNE_scanint_hit_cs \ifnum#1\fi#2\BNE_scanint_again
2580 {%
2581 \expandafter\BNE_wrapint_after\romannumeral`&&@\BNE_getop_a#2%
2582 }%
2583 \def\BNE_scanint_checkagain #1\BNE_scanint_again
2584 {%
2585 \if _#1\BNE_scanint_checkagain_skip\fi
2586 \expandafter\BNE_wrapint_after\romannumeral`&&@\BNE_getop_a#1%
2587 }%
2588 % \end{macrocode}
2589 % |#1| is \cs{fi}.
2590 % \begin{macrocode}
2591 \def\BNE_scanint_checkagain_skip#1#2\BNE_getop_a#3{#1\BNE_scanint_again}%
2592 \def\BNE_scanint_gobz_main #1%
2593 {%
2594 \ifcat \relax #1\expandafter\BNE_scanint_gobz_hit_cs\fi
2595 \ifnum\xint_c_x<1\string#1 \else\expandafter\BNE_scanint_gobz_checkagain\fi
2596 #1\BNE_scanint_again
2597 }%
2598 % \end{macrocode}
2599 % Upstream (at |1.4f|) has |_getop| here, but let's jump directly to
2600 % |BNE_getop_a|. The |#2| has been grabbed already and f-expanded.
2601 % But this means one less brace-stripping.
2602 % \begin{macrocode}
2603 \def\BNE_scanint_gobz_hit_cs\ifnum#1\fi#2\BNE_scanint_again
2604 {%
2605 0\expandafter\BNE_wrapint_after\romannumeral`&&@\BNE_getop_a#2%
2606 }%
2607 % \end{macrocode}
2608 % Fix at |1.6| for when an underscore is used as first character followed by
2609 % digits. No need to worry about being very efficient here.
2610 % \begin{macrocode}
2611 \def\BNE_scanint_gobz_checkagain #1\BNE_scanint_again
2612 {%
2613 \if _#1\xint_dothis\BNE_scanint_gobz_again\fi
2614 \if 0#1\xint_dothis\BNE_scanint_gobz_again\fi
2615 \xint_orthat
2616 {0\expandafter\BNE_wrapint_after\romannumeral`&&@\BNE_getop_a#1}%
2617 }%
2618 \def\BNE_scanint_gobz_again #1%
2619 {%
2620 \expandafter\BNE_scanint_gobz_main\romannumeral`&&@#1%
2621 }%
2622 % \end{macrocode}
2623 % \subsubsection{Hexadecimal}
2624 %
2625 % At |1.6| the code here is refactored to follow closely the |scanint| one,
2626 % rather than downscaling upstream \xintexprname which also has to handle
2627 % fractional input. This avoids gathering the hexadecimal digits then
2628 % grabbing them again as a whole via a delimited macro.
2629 %
2630 % Removing leading zeros and leading underscores
2631 % is the job of |\BNE_scanhex_gobz_main|.
2632 %
2633 % Things could be done better if |\BNE_Op_hextodec| was only required to be
2634 % \emph{x}-expandable. Already now we could inline the |\BNE_wraphex_before|
2635 % sparing its use of |\iffalse..\fi| to move closing braces to the one after
2636 % |\expanded|. But we would still have three |\expandafter|'s in total.
2637 % \begin{macrocode}
2638 \def\BNE_starthex #1%
2639 {%
2640 \expandafter\BNE_wraphex_before
2641 \expanded{{\iffalse}}\fi
2642 \expandafter\BNE_scanhex_gobz_main\romannumeral`&&@#1%
2643 }%
2644 \def\BNE_wraphex_before{\expandafter{\expandafter{%
2645 \romannumeral`&&@\iffalse}}\fi\BNE_Op_hextodec}%
2646 \def\BNE_wraphex_after{\iffalse{{{{\fi}}}}}%
2647 % \end{macrocode}
2648 % At |1.6| we apply exact same scheme as for the |scanint| code. The sole
2649 % difference is the more complicated test for recognizing a digit.
2650 %
2651 % At |1.7| the code evolved to support |a..f| as hexadecimal input.
2652 % As |\bnumeval| supports no functions or user variables, there is
2653 % no breaking change with tacit multiplication as would be the case
2654 % for |\xinteval| if we did the same addition to it.
2655 %
2656 % The expansion is |\expanded|-governed which spares us quite
2657 % some annoyances with exiting from \TeX{} nested conditionals.
2658 %
2659 % Hesitation whether to check for lowercase hexadecimal
2660 % before or after checking for the underscore.
2661 % \begin{macrocode}
2662 \def\BNE_scanhex_main #1%
2663 {%
2664 \ifcat \relax #1\BNE_scanhex_done_hit_cs #1\fi
2665 \if\ifnum`#1>`/
2666 \ifnum`#1>`9
2667 \ifnum`#1>`@
2668 \ifnum`#1>`F
2669 0\else1\fi\else0\fi\else1\fi\else0\fi 1%
2670 #1%
2671 \else
2672 \if _#1\else
2673 \if\ifnum`#1>``
2674 \ifnum`#1>`f 0\else1\fi\else0\fi 1%
2675 \csname BNE_upper #1\endcsname
2676 \else
2677 \BNE_scanhex_done #1%
2678 \fi
2679 \fi
2680 \fi
2681 \BNE_scanhex_again
2682 }%
2683 \expandafter\def\csname BNE_upper a\endcsname{A}%
2684 \expandafter\def\csname BNE_upper b\endcsname{B}%
2685 \expandafter\def\csname BNE_upper c\endcsname{C}%
2686 \expandafter\def\csname BNE_upper d\endcsname{D}%
2687 \expandafter\def\csname BNE_upper e\endcsname{E}%
2688 \expandafter\def\csname BNE_upper f\endcsname{F}%
2689 % \end{macrocode}
2690 % |#2| is \cs{fi}.
2691 % \begin{macrocode}
2692 \def\BNE_scanhex_done_hit_cs #1#2#3\BNE_scanhex_again
2693 {%
2694 #2\expandafter\BNE_wraphex_after\romannumeral`&&@\BNE_getop_a#1%
2695 }%
2696 % \end{macrocode}
2697 % |#2| is \cs{fi}\cs{fi}\cs{fi} or \cs{fi}\cs{fi}\cs{fi}\cs{fi} (if called
2698 % from the |gobz| variant).
2699 % \begin{macrocode}
2700 \def\BNE_scanhex_done #1#2\BNE_scanhex_again
2701 {%
2702 #2\expandafter\BNE_wraphex_after\romannumeral`&&@\BNE_getop_a#1%
2703 }%
2704 \def\BNE_scanhex_again #1%
2705 {%
2706 \expandafter\BNE_scanhex_main\romannumeral`&&@#1%
2707 }%
2708 % \end{macrocode}
2709 % Initial scanning which skips leading zeros and underscores. Priority in
2710 % efficiency is when there are none. Slight refactoring to check for lowercase
2711 % hexadecimal earlier at |1.7a|. But the |\BNE_scanhex_main| will then again
2712 % check for underscores prior to them.
2713 %
2714 % Also |1.7a| fixes the unfortunate |1.7| regression when only zeros (or
2715 % underscores) are found after the hexadecimal (or other) prefixes.
2716 % \begin{macrocode}
2717 \def\BNE_scanhex_gobz_main #1%
2718 {%
2719 \ifcat \relax #1%
2720 0\BNE_scanhex_done_hit_cs #1\fi
2721 \if\ifnum`#1>`0
2722 \ifnum`#1>`9
2723 \ifnum`#1>`@
2724 \ifnum`#1>`F
2725 0\else1\fi\else0\fi\else1\fi\else0\fi 1%
2726 #1%
2727 \else
2728 \if\ifnum`#1>`` \ifnum`#1>`f 0\else1\fi\else0\fi 1%
2729 \csname BNE_upper #1\endcsname
2730 \else
2731 \if 0#1\BNE_scanhex_gobzero\else
2732 \if _#1\BNE_scanhex_gobunderscore\else
2733 0\BNE_scanhex_done #1%
2734 \fi\fi
2735 \fi
2736 \fi
2737 \BNE_scanhex_again
2738 }%
2739 \def\BNE_scanhex_gobzero #1\BNE_scanhex_again #2%
2740 {%
2741 \fi\fi\fi
2742 \expandafter\BNE_scanhex_gobz_main\romannumeral`&&@#2%
2743 }%
2744 \def\BNE_scanhex_gobunderscore #1\BNE_scanhex_again #2%
2745 {%
2746 \fi\fi\fi\fi
2747 \expandafter\BNE_scanhex_gobz_main\romannumeral`&&@#2%
2748 }%
2749 % \end{macrocode}
2750 % \subsubsection{Octal}
2751 %
2752 % Added at |1.6|. Leading
2753 % zeros are removed.
2754 %
2755 % At |1.7| the code for hexadecimal was a bit refactored and
2756 % the one here was changed to follow same pattern.
2757 % \begin{macrocode}
2758 \def\BNE_startoct #1%
2759 {%
2760 \expandafter\BNE_wrapoct_before
2761 \expanded{{\iffalse}}\fi
2762 \expandafter\BNE_scanoct_gobz_main\romannumeral`&&@#1%
2763 }%
2764 \def\BNE_wrapoct_before{\expandafter{\expandafter{%
2765 \romannumeral`&&@\iffalse}}\fi\BNE_Op_octtodec}%
2766 \def\BNE_wrapoct_after{\iffalse{{{{\fi}}}}}%
2767 \def\BNE_scanoct_main #1%
2768 {%
2769 \ifcat \relax #1\expandafter\BNE_scanoct_done_hit_cs #1\fi
2770 \if\ifnum`#1>`/ \ifnum`#1>`7 0\else1\fi\else0\fi 1%
2771 #1%
2772 \else
2773 \if _#1\else
2774 \BNE_scanoct_done #1%
2775 \fi
2776 \fi
2777 \BNE_scanoct_again
2778 }%
2779 % \end{macrocode}
2780 % |#2| is \cs{fi}.
2781 % \begin{macrocode}
2782 \def\BNE_scanoct_done_hit_cs #1#2#3\BNE_scanoct_again
2783 {%
2784 #2\expandafter\BNE_wrapoct_after\romannumeral`&&@\BNE_getop_a#1%
2785 }%
2786 % \end{macrocode}
2787 % |#2| is \cs{fi}\cs{fi} or \cs{fi}\cs{fi}\cs{fi} (if called
2788 % from the |gobz| variant).
2789 % \begin{macrocode}
2790 \def\BNE_scanoct_done #1#2\BNE_scanoct_again
2791 {%
2792 #2\expandafter\BNE_wrapoct_after\romannumeral`&&@\BNE_getop_a#1%
2793 }%
2794 \def\BNE_scanoct_again #1%
2795 {%
2796 \expandafter\BNE_scanoct_main\romannumeral`&&@#1%
2797 }%
2798 \def\BNE_scanoct_gobz_main #1%
2799 {%
2800 \ifcat \relax #1%
2801 0\BNE_scanoct_done_hit_cs #1\fi
2802 \if\ifnum`#1>`0 \ifnum`#1>`7 0\else1\fi\else0\fi 1%
2803 #1%
2804 \else
2805 \if 0#1\BNE_scanoct_gobzero\else
2806 \if _#1\BNE_scanoct_gobunderscore\else
2807 0\BNE_scanoct_done #1%
2808 \fi\fi
2809 \fi
2810 \BNE_scanoct_again
2811 }%
2812 \def\BNE_scanoct_gobzero #1\BNE_scanoct_again #2%
2813 {%
2814 \fi\fi
2815 \expandafter\BNE_scanoct_gobz_main\romannumeral`&&@#2%
2816 }%
2817 \def\BNE_scanoct_gobunderscore #1\BNE_scanoct_again #2%
2818 {%
2819 \fi\fi\fi
2820 \expandafter\BNE_scanoct_gobz_main\romannumeral`&&@#2%
2821 }%
2822 % \end{macrocode}
2823 % \subsubsection{Binary}
2824 %
2825 % Added at |1.6|. Same code skeleton as for octal and hexadecimal.
2826 % \begin{macrocode}
2827 \def\BNE_startbin #1%
2828 {%
2829 \expandafter\BNE_wrapbin_before
2830 \expanded{{\iffalse}}\fi
2831 \expandafter\BNE_scanbin_gobz_main\romannumeral`&&@#1%
2832 }%
2833 \def\BNE_wrapbin_before{\expandafter{\expandafter{%
2834 \romannumeral`&&@\iffalse}}\fi\BNE_Op_bintodec}%
2835 \def\BNE_wrapbin_after{\iffalse{{{{\fi}}}}}%
2836 \def\BNE_scanbin_main #1%
2837 {%
2838 \ifcat \relax #1\expandafter\BNE_scanbin_done_hit_cs #1\fi
2839 \if \if0#11\else\if1#11\else0\fi\fi 1%
2840 #1%
2841 \else
2842 \if _#1\else
2843 \BNE_scanbin_done #1%
2844 \fi
2845 \fi
2846 \BNE_scanbin_again
2847 }%
2848 % \end{macrocode}
2849 % |#2| is \cs{fi}.
2850 % \begin{macrocode}
2851 \def\BNE_scanbin_done_hit_cs #1#2#3\BNE_scanbin_again
2852 {%
2853 #2\expandafter\BNE_wrapbin_after\romannumeral`&&@\BNE_getop_a#1%
2854 }%
2855 % \end{macrocode}
2856 % |#2| is \cs{fi}\cs{fi} or \cs{fi}\cs{fi}\cs{fi} (if called
2857 % from the |gobz| variant).
2858 % \begin{macrocode}
2859 \def\BNE_scanbin_done #1#2\BNE_scanbin_again
2860 {%
2861 #2\expandafter\BNE_wrapbin_after\romannumeral`&&@\BNE_getop_a#1%
2862 }%
2863 \def\BNE_scanbin_again #1%
2864 {%
2865 \expandafter\BNE_scanbin_main\romannumeral`&&@#1%
2866 }%
2867 \def\BNE_scanbin_gobz_main #1%
2868 {%
2869 \ifcat \relax #1%
2870 0\BNE_scanbin_done_hit_cs #1\fi
2871 \if \if0#11\else\if1#11\else0\fi\fi 1%
2872 #1%
2873 \else
2874 \if 0#1\BNE_scanbin_gobzero\else
2875 \if _#1\BNE_scanbin_gobunderscore\else
2876 0\BNE_scanbin_done #1%
2877 \fi\fi
2878 \fi
2879 \BNE_scanbin_again
2880 }%
2881 \def\BNE_scanbin_gobzero #1\BNE_scanbin_again #2%
2882 {%
2883 \fi\fi
2884 \expandafter\BNE_scanbin_gobz_main\romannumeral`&&@#2%
2885 }%
2886 \def\BNE_scanbin_gobunderscore #1\BNE_scanbin_again #2%
2887 {%
2888 \fi\fi\fi
2889 \expandafter\BNE_scanbin_gobz_main\romannumeral`&&@#2%
2890 }%
2891 % \end{macrocode}
2892 % \subsection{\cs{BNE_getop}}
2893 % The upstream analog to |\BNE_getop_a| applies \cs{string} to |#1| in its
2894 % thirdofthree branch before handing over to analog of |\BNE_scanop_a|, but I
2895 % see no reason for doing it here (and I do have to check if upstream has
2896 % any valid reason to do it). Removed. First branch was a |\BNE_foundend|,
2897 % used only here, and expanding to |\xint_c_\relax|, let's move the
2898 % |#1| (which will be \cs{relax}) last and simply insert |\xint_c_|.
2899 %
2900 % The |_scanop| macros have been refactored at upstream and here |1.5|.
2901 % \begin{macrocode}
2902 \def\BNE_getop #1%
2903 {%
2904 \expandafter\BNE_getop_a\romannumeral`&&@#1%
2905 }%
2906 \catcode`* 11
2907 \def\BNE_getop_a #1%
2908 {%
2909 \ifx \relax #1\xint_dothis\xint_firstofthree\fi
2910 \ifcat \relax #1\xint_dothis\xint_secondofthree\fi
2911 \ifnum\xint_c_ix<1\string#1 \xint_dothis\xint_secondofthree\fi
2912 \if (#1\xint_dothis \xint_secondofthree\fi %)
2913 \xint_orthat \xint_thirdofthree
2914 \xint_c_
2915 {\BNE_prec_tacit *}%
2916 \BNE_scanop_a
2917 #1%
2918 }%
2919 \catcode`* 12
2920 \def\BNE_scanop_a #1#2%
2921 {%
2922 \expandafter\BNE_scanop_b\expandafter#1\romannumeral`&&@#2%
2923 }%
2924 \def\BNE_scanop_b #1#2%
2925 {%
2926 \unless\ifcat#2\relax
2927 \ifcsname BNE_itself_#1#2\endcsname
2928 \BNE_scanop_c
2929 \fi\fi
2930 \BNE_foundop_a #1#2%
2931 }%
2932 \def\BNE_scanop_c #1#2#3#4#5% #1#2=\fi\fi
2933 {%
2934 #1#2%
2935 \expandafter\BNE_scanop_d\csname BNE_itself_#4#5\expandafter\endcsname
2936 \romannumeral`&&@%
2937 }%
2938 \def\BNE_scanop_d #1#2%
2939 {%
2940 \unless\ifcat#2\relax
2941 \ifcsname BNE_itself_#1#2\endcsname
2942 \BNE_scanop_c
2943 \fi\fi
2944 \BNE_foundop #1#2%
2945 }%
2946 % \end{macrocode}
2947 % If a postfix say |?s| is defined and |?r| is encountered the |?| will have
2948 % been interpreted as a shortcut to |?s| and then the |r| will be found with
2949 % the parser (after having executed the already found postfix) now looking for
2950 % another operator so the error message will be |Operator? (got `r')| which is
2951 % doubly confusing... well, let's not dwell on that.
2952 %
2953 % Update 2021/05/22, I have changed the message, as part of a systematic
2954 % removal of |I<something>| invites, in part because |xint 1.4g| changed its
2955 % expandable error method and now has a nice message saying \xintname will try
2956 % to recover by itself. And now I have about 55 characters available for the
2957 % message.
2958 % \begin{macrocode}
2959 \def\BNE_foundop_a #1%
2960 {%
2961 \ifcsname BNE_precedence_#1\endcsname
2962 \csname BNE_precedence_#1\expandafter\endcsname
2963 \expandafter #1%
2964 \else
2965 \expandafter\BNE_getop_a\romannumeral`&&@%
2966 \xint_afterfi{\XINT_expandableerror
2967 {Expected an operator but got `#1'. Ignoring.}}%
2968 \fi
2969 }%
2970 \def\BNE_foundop #1{\csname BNE_precedence_#1\endcsname #1}%
2971 % \end{macrocode}
2972 % \subsection{Expansion spanning; opening and closing parentheses}
2973 % \label{ssec:span}
2974 % There was refactoring of expandable error messages at \xintname |1.4g| and I
2975 % can now use up to 55 characters, but should not really invite user to
2976 % |I|nsert something as it does not fit well with generic message saying
2977 % \xintname will go ahead "hoping repair was complete".
2978 %
2979 % At |1.6|, we removed the |\BNE_start|, current |\BNE_bareeval| has its
2980 % meaning rather than expanding to it as formerly.
2981 %
2982 % Also here macros are defined one by one so that it is easier to understand
2983 % what is happening. Formerly |\BNE_tmpa| defined all of them in one go
2984 % (as is still the case in upstream \xintexprname).
2985 % \begin{macrocode}
2986 \def\BNE_tmpa#1{%
2987 \def\BNE_check##1%
2988 {%
2989 \xint_UDsignfork
2990 ##1{\expandafter\BNE_checkp\romannumeral`&&@#1}%
2991 -{\BNE_checkp##1}%
2992 \krof
2993 }%
2994 }\expandafter\BNE_tmpa\csname BNE_op_-xii\endcsname
2995 \def\BNE_tmpa#1{%
2996 \def\BNE_checkp##1##2%
2997 {%
2998 \ifcase ##1%
2999 \expandafter\BNE_done
3000 \or\expandafter#1%
3001 \else
3002 \expandafter\BNE_checkp
3003 \romannumeral`&&@\csname BNE_op_##2\expandafter\endcsname
3004 \fi
3005 }%
3006 }\expandafter\BNE_tmpa\csname BNE_extra_)\endcsname
3007 \expandafter\def\csname BNE_extra_)\endcsname{%
3008 \XINT_expandableerror
3009 {An extra ) was removed. Hit <return>, fingers crossed.}%
3010 \expandafter\BNE_check\romannumeral`&&@\expandafter\BNE_put_op_first
3011 \romannumeral`&&@\BNE_getop_legacy
3012 }%
3013 \let\BNE_done\space
3014 \def\BNE_getop_legacy #1%
3015 {%
3016 \expanded{\unexpanded{{#1}}\expandafter}\romannumeral`&&@\BNE_getop
3017 }%
3018 % \end{macrocode}
3019 % Code style left untouched at |1.6|.
3020 % \begin{macrocode}
3021 \catcode`) 11
3022 \def\BNE_tmpa #1#2#3#4#5#6%
3023 {%
3024 \def #1##1% op_(
3025 {%
3026 \expandafter #4\romannumeral`&&@\BNE_getnext
3027 }%
3028 \def #2##1% op_)
3029 {%
3030 \expanded{\unexpanded{\BNE_put_op_first{##1}}\expandafter}%
3031 \romannumeral`&&@\BNE_getop
3032 }%
3033 \def #3% oparen
3034 {%
3035 \expandafter #4\romannumeral`&&@\BNE_getnext
3036 }%
3037 \def #4##1% check-
3038 {%
3039 \xint_UDsignfork
3040 ##1{\expandafter#5\romannumeral`&&@#6}%
3041 -{#5##1}%
3042 \krof
3043 }%
3044 \def #5##1##2% checkp
3045 {%
3046 \ifcase ##1\expandafter\BNE_missing_)
3047 \or \csname BNE_op_##2\expandafter\endcsname
3048 \else
3049 \expandafter #5\romannumeral`&&@\csname BNE_op_##2\expandafter\endcsname
3050 \fi
3051 }%
3052 }%
3053 \expandafter\BNE_tmpa
3054 \csname BNE_op_(\expandafter\endcsname
3055 \csname BNE_op_)\expandafter\endcsname
3056 \csname BNE_oparen\expandafter\endcsname
3057 \csname BNE_check-_)\expandafter\endcsname
3058 \csname BNE_checkp_)\expandafter\endcsname
3059 \csname BNE_op_-xii\endcsname
3060 \let\BNE_precedence_)\xint_c_i
3061 \def\BNE_missing_)
3062 {\XINT_expandableerror{Missing ). Hit <return> to proceed.}%
3063 \xint_c_ \BNE_done }%
3064 \catcode`) 12
3065 % \end{macrocode}
3066 % \subsection{The comma as binary operator}
3067 % At |1.4|, it is simply a union operator for 1D oples. Inserting directly
3068 % here a |<comma><space>| separator (as in earlier releases) in accumulated
3069 % result would avoid having to do it on output but to the cost of diverging
3070 % from \xintexprname upstream code, and to have to let the ``|unpack|''
3071 % (new name at |1.7|)
3072 % routines handle comma separated values rather than braced values.
3073 %
3074 % Note bene: contiguous commas |,,| will cause the parser to raise an
3075 % ``unexpected token'' error on the second comma, contrarily to |\xinteval|
3076 % which silently ignores it since its |1.4| release (see the first subsection
3077 % about basic terminology in \emph{Oples and nutples: the 1.4 terminology} in
3078 % |xint.pdf|).
3079 %
3080 % It does work to use |\bnumdefinfix{,,}{\macro}{N}{N}|. Then |,,| acts as a
3081 % custom infix operator without breaking the use of a single comma to
3082 % separate inputs.
3083 % \begin{macrocode}
3084 \def\BNE_tmpa #1#2#3#4#5%
3085 {%
3086 \def #1##1% \BNE_op_,
3087 {%
3088 \expanded{\unexpanded{#2{##1}}\expandafter}%
3089 \romannumeral`&&@\expandafter#3\romannumeral`&&@\BNE_getnext
3090 }%
3091 \def #2##1##2##3##4{##2##3{##1##4}}% \BNE_exec_,
3092 \def #3##1% \BNE_check-_,
3093 {%
3094 \xint_UDsignfork
3095 ##1{\expandafter#4\romannumeral`&&@#5}%
3096 -{#4##1}%
3097 \krof
3098 }%
3099 \def #4##1##2% \BNE_checkp_,
3100 {%
3101 \ifnum ##1>\xint_c_iii
3102 \expandafter#4%
3103 \romannumeral`&&@\csname BNE_op_##2\expandafter\endcsname
3104 \else
3105 \expandafter##1\expandafter##2%
3106 \fi
3107 }%
3108 }%
3109 \expandafter\BNE_tmpa
3110 \csname BNE_op_,\expandafter\endcsname
3111 \csname BNE_exec_,\expandafter\endcsname
3112 \csname BNE_check-_,\expandafter\endcsname
3113 \csname BNE_checkp_,\expandafter\endcsname
3114 \csname BNE_op_-xii\endcsname
3115 \expandafter\let\csname BNE_precedence_,\endcsname\xint_c_iii
3116 % \end{macrocode}
3117 % \subsection{The minus as prefix operator of variable precedence level}
3118 % This |\BNE_Op_opp| caused trouble at |1.4| as it must be \emph{f}-expandable,
3119 % whereas earlier it expanded inside |\csname...\endcsname| context, so I
3120 % could define it as \centeredline{|\if-#1\else\if0#10\else-#1\fi\fi|} where
3121 % |#1| was the
3122 % first token of unbraced argument but this meant at |1.4| an added
3123 % |\xint_firstofone| here. Well let's return to sanity at |1.4a| and not add
3124 % the |\xint_firstofone| and simply default |\BNE_Op_opp| to |\xintiiOpp|,
3125 % which it should have been all along! And on this occasion let's trim user
3126 % documentation of complications.
3127 %
3128 % The package used to need to define unary minus operator with precedences 12,
3129 % 14, and 18. It also defined it at level 16 but this was unneedeed actually,
3130 % no operator possibly generating usage of an |op_-xvi|.
3131 %
3132 % At |1.5| the right precedence of powers was lowered to 17, so we now need
3133 % here only 12, 14, and 17.
3134 %
3135 % Due to |\bnumdefinfix| it is needed to support also, perhaps, the other
3136 % levels 13, 15, 16, 18, .... This will be done only if necessary and is the
3137 % reason why the macros |\BNE_defminus_a| and |\BNE_defminus_b| are given
3138 % permanent names. In fact it is now |\BNE_defbin_b| which will decide to invoke
3139 % or not the |\BNE_defminus_a|, and we activate it here only for the base
3140 % precedence 12.
3141 %
3142 % The |\XINT_global|'s are absent from upstream \xintexprname as it does not
3143 % incorporate yet some analog to |\bnumdefinfix/\bnumdefpostfix|.
3144 % \begin{macrocode}
3145 \def\BNE_defminus_b #1#2#3#4#5%
3146 {%
3147 \XINT_global\def #1% \BNE_op_-<level>
3148 {%
3149 \expandafter #2\romannumeral`&&@\expandafter#3%
3150 \romannumeral`&&@\BNE_getnext
3151 }%
3152 \XINT_global\def #2##1##2##3% \BNE_exec_-<level>
3153 {%
3154 \expandafter ##1\expandafter ##2\expandafter
3155 {\expandafter{\romannumeral`&&@\BNE_Op_opp##3}}%
3156 }%
3157 \XINT_global\def #3##1% \BNE_check-_-<level>
3158 {%
3159 \xint_UDsignfork
3160 ##1{\expandafter #4\romannumeral`&&@#1}%
3161 -{#4##1}%
3162 \krof
3163 }%
3164 \XINT_global\def #4##1##2% \BNE_checkp_-<level>
3165 {%
3166 \ifnum ##1>#5%
3167 \expandafter #4%
3168 \romannumeral`&&@\csname BNE_op_##2\expandafter\endcsname
3169 \else
3170 \expandafter ##1\expandafter ##2%
3171 \fi
3172 }%
3173 }%
3174 \def\BNE_defminus_a #1%
3175 {%
3176 \expandafter\BNE_defminus_b
3177 \csname BNE_op_-#1\expandafter\endcsname
3178 \csname BNE_exec_-#1\expandafter\endcsname
3179 \csname BNE_check-_-#1\expandafter\endcsname
3180 \csname BNE_checkp_-#1\expandafter\endcsname
3181 \csname xint_c_#1\endcsname
3182 }%
3183 \BNE_defminus_a {xii}%
3184 % \end{macrocode}
3185 % \subsection{The infix operators.}
3186 % I could have at the |1.4| refactoring injected usage of \cs{expanded} here,
3187 % but kept in sync with upstream \xintexprname code. Any \emph{x}-expandable
3188 % macro can easily be converted into an \emph{f}-expandable one using
3189 % \cs{expanded}, so this is no serious limitation.
3190 %
3191 % Macro names are somewhat bad and there is much risk of confusion in future
3192 % maintenance of |\BNE_Op_| prefix (used for |\BNE_Op_add| etc...; besides
3193 % this should have been |\BNE_Op_Add|) and |\BNE_op_| prefix (used for
3194 % |\BNE_op_+| etc...).
3195 %
3196 % At |1.5| decision is made to anticipate the announced upstream change to let
3197 % the power operators be right associative, matching Python behaviour. This
3198 % change is simply implemented by hardcoding in |\BNE_checkp_<op>| the right
3199 % precedence which so far, for such operators, had been identical with the
3200 % left precedence (upstream has examples of direct coding without
3201 % formalization). In fact the right precedence existed already as argument to
3202 % |\BNE_defbin_b| as the precedence to assign to unary minus following |<op>|.
3203 %
3204 % Note1: although it is easy to change the left precedence at user level, the
3205 % right precedence is now more inaccessible. But on the other hand \bnumname
3206 % provides |\bnumdefinfix| so all is customizable at user level.
3207 %
3208 % Note2: Tacit multiplication is not really a separate operator, it is
3209 % the |*| with an elevated left precedence, which costs nothing to create and
3210 % this precedence is stored in chardef token |\BNE_prec_tacit|.
3211 %
3212 % Compared to upstream, we use here numbers as arguments to |\BNE_defbin_b|,
3213 % and convert to roman numerals internally, also the operator macro is passed
3214 % as a control sequence not as its name (and |#6| and |#7| are permuted in
3215 % |\BNE_defbin_c|).
3216 % \begin{macrocode}
3217 \def\BNE_defbin_c #1#2#3#4#5#6#7%
3218 {%
3219 \XINT_global\def #1##1% \BNE_op_<op>
3220 {%
3221 \expanded{\unexpanded{#2{##1}}\expandafter}%
3222 \romannumeral`&&@\expandafter#3\romannumeral`&&@\BNE_getnext
3223 }%
3224 \XINT_global\def #2##1##2##3##4% \BNE_exec_<op>
3225 {%
3226 \expandafter##2\expandafter##3\expandafter
3227 {\expandafter{\romannumeral`&&@#7##1##4}}%
3228 }%
3229 \XINT_global\def #3##1% \BNE_check-_<op>
3230 {%
3231 \xint_UDsignfork
3232 ##1{\expandafter#4\romannumeral`&&@#5}%
3233 -{#4##1}%
3234 \krof
3235 }%
3236 \XINT_global\def #4##1##2% \BNE_checkp_<op>
3237 {%
3238 \ifnum ##1>#6%
3239 \expandafter#4%
3240 \romannumeral`&&@\csname BNE_op_##2\expandafter\endcsname
3241 \else
3242 \expandafter ##1\expandafter ##2%
3243 \fi
3244 }%
3245 }%
3246 \def\BNE_defbin_b #1#2#3#4%
3247 {%
3248 \expandafter\BNE_defbin_c
3249 \csname BNE_op_#1\expandafter\endcsname
3250 \csname BNE_exec_#1\expandafter\endcsname
3251 \csname BNE_check-_#1\expandafter\endcsname
3252 \csname BNE_checkp_#1\expandafter\endcsname
3253 \csname BNE_op_-\romannumeral\ifnum#3>12 #3\else 12\fi
3254 \expandafter\endcsname
3255 \csname xint_c_\romannumeral#3\endcsname #4%
3256 \XINT_global
3257 \expandafter
3258 \let\csname BNE_precedence_#1\expandafter\endcsname
3259 \csname xint_c_\romannumeral#2\endcsname
3260 \unless
3261 \ifcsname BNE_exec_-\romannumeral\ifnum#3>12 #3\else 12\fi\endcsname
3262 % \end{macrocode}
3263 % This will execute only for |#3>12| as |\BNE_exec_-xii| exists.
3264 % \begin{macrocode}
3265 \expandafter\BNE_defminus_a\expandafter{\romannumeral#3}%
3266 \fi
3267 }%
3268 \BNE_defbin_b + {12} {12} \BNE_Op_add
3269 \BNE_defbin_b - {12} {12} \BNE_Op_sub
3270 \BNE_defbin_b * {14} {14} \BNE_Op_mul
3271 \BNE_defbin_b / {14} {14} \BNE_Op_divround
3272 \BNE_defbin_b {//} {14} {14} \BNE_Op_div
3273 \BNE_defbin_b {/:} {14} {14} \BNE_Op_mod
3274 \BNE_defbin_b ^ {18} {17} \BNE_Op_pow
3275 % \end{macrocode}
3276 % \xintexprname uses shortcut
3277 % \centeredline{|\expandafter\def\csname XINT_expr_itself_**\endcsname {^}|}
3278 % But doing it would mean that any redefinition of |^| propagates to |**|.
3279 % And it
3280 % creates a special case which would need consideration by
3281 % |\BNE_dotheitselves|, or special restrictions to add to user documentation.
3282 % Better to simply handle |**| as a full operator.
3283 % \begin{macrocode}
3284 \BNE_defbin_b {**} {18} {17} \BNE_Op_pow
3285 \expandafter\def\csname BNE_itself_**\endcsname {**}%
3286 \expandafter\def\csname BNE_itself_//\endcsname {//}%
3287 \expandafter\def\csname BNE_itself_/:\endcsname {/:}%
3288 \let\BNE_prec_tacit\xint_c_xvi
3289 % \end{macrocode}
3290 % \subsection{Extending the syntax: \cs{bnumdefinfix}, \cs{bnumdefpostfix}}
3291 % \subsubsection{\cs{bnumdefinfix}}
3292 % |#1| gives the operator characters, |#2| the associated macro, |#3| its
3293 % left-precedence and |#4| its right precedence (as integers).
3294 %
3295 % The "itself" definitions are done in such a way that unambiguous abbreviations
3296 % work; but in case of ambiguity the first defined operator is used.
3297 %
3298 % However, if for example operator |$a| was defined after |$ab|, then although
3299 % |$| will use |$ab| which was defined first, |$a| will use as expected the
3300 % second defined operator.
3301 %
3302 % The mismatch |\BNE_defminus_a| vs |\BNE_defbin_b| is inherited from
3303 % upstream, I keep it to simplify maintenance.
3304 % \begin{macrocode}
3305 \def\bnumdefinfix #1#2#3#4%
3306 {%
3307 \edef\BNE_tmpa{#1}%
3308 \edef\BNE_tmpa{\xint_zapspaces_o\BNE_tmpa}%
3309 \edef\BNE_tmpL{\the\numexpr#3\relax}%
3310 \edef\BNE_tmpL{\ifnum\BNE_tmpL<4 4\else\ifnum\BNE_tmpL<23 \BNE_tmpL\else 22\fi\fi}%
3311 \edef\BNE_tmpR{\the\numexpr#4\relax}%
3312 \edef\BNE_tmpR{\ifnum\BNE_tmpR<4 4\else\ifnum\BNE_tmpR<23 \BNE_tmpR\else 22\fi\fi}%
3313 \BNE_defbin_b \BNE_tmpa\BNE_tmpL\BNE_tmpR #2%
3314 \expandafter\BNE_dotheitselves\BNE_tmpa\relax
3315 \ifxintverbose
3316 \PackageInfo{bnumexpr}{infix operator \BNE_tmpa\space
3317 \ifxintglobaldefs globally \fi
3318 does
3319 \unexpanded{#2}\MessageBreak with precedences \BNE_tmpL, \BNE_tmpR;}%
3320 \fi
3321 }%
3322 \def\BNE_dotheitselves#1#2%
3323 {%
3324 \if#2\relax\expandafter\xint_gobble_ii
3325 \else
3326 \XINT_global
3327 \expandafter\edef\csname BNE_itself_#1#2\endcsname{#1#2}%
3328 \unless\ifcsname BNE_precedence_#1\endcsname
3329 \XINT_global
3330 \expandafter\edef\csname BNE_precedence_#1\endcsname
3331 {\csname BNE_precedence_\BNE_tmpa\endcsname}%
3332 \XINT_global
3333 \expandafter\odef\csname BNE_op_#1\endcsname
3334 {\csname BNE_op_\BNE_tmpa\endcsname}%
3335 \fi
3336 \fi
3337 \BNE_dotheitselves{#1#2}%
3338 }%
3339 % \end{macrocode}
3340 % \subsubsection{\cs{bnumdefpostfix}}
3341 % Support macros for postfix operators only need to be \emph{x}-expandable.
3342 % \begin{macrocode}
3343 \def\bnumdefpostfix #1#2#3%
3344 {%
3345 \edef\BNE_tmpa{#1}%
3346 \edef\BNE_tmpa{\xint_zapspaces_o\BNE_tmpa}%
3347 \edef\BNE_tmpL{\the\numexpr#3\relax}%
3348 \edef\BNE_tmpL{\ifnum\BNE_tmpL<4 4\else\ifnum\BNE_tmpL<23 \BNE_tmpL\else 22\fi\fi}%
3349 \XINT_global
3350 \expandafter\let\csname BNE_precedence_\BNE_tmpa\expandafter\endcsname
3351 \csname xint_c_\romannumeral\BNE_tmpL\endcsname
3352 \XINT_global
3353 \expandafter\def\csname BNE_op_\BNE_tmpa\endcsname ##1%
3354 {%
3355 \expandafter\BNE_put_op_first
3356 \expanded{{{#2##1}}\expandafter}\romannumeral`&&@\BNE_getop
3357 }%
3358 \expandafter\BNE_dotheitselves\BNE_tmpa\relax
3359 \ifxintverbose
3360 \PackageInfo{bnumexpr}{postfix operator \BNE_tmpa\space
3361 \ifxintglobaldefs globally \fi
3362 does \unexpanded{#2}\MessageBreak
3363 with precedence \BNE_tmpL;}%
3364 \fi
3365 }%
3366 % \end{macrocode}
3367 % \subsection{! as postfix factorial operator}
3368 % \begin{macrocode}
3369 \bnumdefpostfix{!}{\BNE_Op_fac}{20}%
3370 % \end{macrocode}
3371 % \subsection{Cleanup}
3372 % \begin{macrocode}
3373 \let\BNE_tmpa\relax \let\BNE_tmpb\relax \let\BNE_tmpc\relax
3374 \let\BNE_tmpR\relax \let\BNE_tmpL\relax
3375 \BNErestorecatcodesendinput%
3376 % \end{macrocode}
3377 % \MakePercentComment
3378 %</package>
3379 %<*dtx>
3380 \CheckSum {1670}%
3381 \makeatletter\check@checksum\makeatother%
3382 \Finale%
3383 %% End of file xint.dtx