This is the README file for the 'loops' package.
AUTHOR
Ahmed Musa
VERSION
Version 1.2, 2013/02/10.
LOCATION ON CTAN
macros/latex/contrib/loops/
LICENSE
Copyright (c) 2012-2013 Ahmed Musa
This software is author-maintained. Permission is granted to copy,
distribute and/or modify this software under the terms of the
LaTeX Project Public License, version 1.3 or higher. This software
is provided 'as it is', without warranty of any kind, either
expressed or implied, including, but not limited to, the implied
warranties of merchantability and fitness for any particular purpose.
NOTE
Non-LaTeX users can use the \newforeach macro (and some other
looping macros) by loading the file skeyval-for.tex. These exclude
\foreachfox, which are available only in loops.sty.
SUMMARY
The 'loops' package provides efficient looping macros for processing
both csv (separated-values) and nsv/tsv (non-separated values) lists.
csv lists having arbitrary parsers, and even active parsers, can
be processed with the tools of this package.
All our loops can be safely nested. Breaking the inner loop doesn't
affect the continuation of the outer loop.
The looping macros include:
\skvrecurse:
\skvrecurse{
}\while<\if...>\fi
A variant of D.E. Knuth's \loop macro.
\@tempcnta\z@
\skvrecurse{
\advance\@tempcnta\@ne
\@tempcntb\z@
\skvrecurse
\advance\@tempcntb\@ne
\typeout{Doing \the\@tempcnta,\the\@tempcntb}
\while
\ifnum\@tempcntb<3\relax
\fi
}
\while
\ifnum\@tempcnta<4\relax
\fi
\carlisleloop:
\carlisleloop{}<\if...>\repeat
A nesting version of D.E. Knuth's \loop macro,
adapted from the original macros of David Carlisle.
\skvnewcounts{rowcnt,colcnt,maxrow,maxcol}
\maxrow4 \maxcol3\relax
\def\entries{}
\let\xp\skvexpandonce
\def\generate{%
\advance\colcnt\@ne
\edef\entries{\xp\entries\the\numexpr\rowcnt*\colcnt\relax}%
\ifnum\colcnt<\numexpr\maxcol+1\relax\relax
\edef\entries{\xp\entries\noexpand&}%
}
\def\generateentries{%
\carlisleloop\a
\advance\rowcnt\@ne\colcnt\z@
\carlisleloop\b
\generate
\repeat\b
\ifnum\rowcnt<\numexpr\maxrow+1\relax\relax
\edef\entries{\xp\entries\noexpand\cr}%
\repeat\a
\edef\entries{\xp\entries\noexpand\crcr}%
}
\begin{document}
\generateentries
$$\vbox{\halign{&\ \hfil#\hfil\strut\cr\entries}}$$
\end{document}
\akterloop:
\akterloop{}<\if...>\repeat
A generalized and naturally nesting version of D.E. Knuth's
\loop macro. No nesting brace is required, unlike in \skvrecurse.
No is required, unlike in \carlisleloop.
\def\generateentries{%
\akterloop
\advance\rowcnt\@ne\colcnt\z@
\akterloop
\generate
\repeat
\ifnum\rowcnt<\numexpr\maxrow+1\relax\relax
\edef\entries{\xp\entries\noexpand\cr}%
\repeat
\edef\entries{\xp\entries\noexpand\crcr}%
}
\begin{document}
\generateentries
$$\vbox{\halign{&\ \hfil#\hfil\strut\cr\entries}}$$
\end{document}
\skvtfor:
\skvtfor:={}\do{}
\skvtfor*:={}\do{}
A variant of LaTeX kernel's \@tfor that has a different
approach to breaking the loop prematurely.
\skvnewregister\count\nr
\skvtfor\x:=a \ifx {\def\x#1{#1}} \fi\do{%
\nr\z@
\skvtfor\y:=12345\do{%
\typeout{Doing: \detokenize\expandafter{\x}--\y}%
\advance\nr\@ne
\ifnum\nr>\tw@\skvbreakloop\fi
}%
}
\ayeloop:
\ayeloop{}\do{<1.parameter.callback>}
\ayeloop*{}\do{<1.parameter.callback>}
A variant of \skvtfor that uses parameter characters
directly, instead of a holder macro.
\ayeloop abcd\do{%
\def\nr{0}%
\ayeloop 12345\do{%
\typeout{Doing #1--##1}%
\skvpushnumber\nr
\ifnum\nr>3\relax\skvbreakloop\fi
}%
}
\sifakaloop:
\sifakaloop{}{<1.parameter.callback>}
\sifakaloop*{}{<1.parameter.callback>}
\sifakaloop is as \ayeloop but will first normalize the list
using \skvtsvnormalize before parsing. Unlike all the
other nsv/tsv loops, it preserves outer braces in the
arguments. Hence, it is costlier. You can insert \listbreaker
as a token in or to terminate the list prematurely.
\skvfor:
\skvfor{}<1.parameter.callback.macro>
\skvfor is an expandable comma loop. The list is not
normalized prior to parsing, and it is not possible to
terminate the loop prematurely (ie, before exhausting
the list).
\newcount\nr\nr=\tw@
\def\do#1{%
\let\noexpand#1%
\expandafter\noexpand\csname\skvremovescape{#1}@%
\romannumeral\nr\endcsname
}
\edef\x{\skvfor{\cmda,\cmdb}\do}
\show\x -> \let\cmda\cmda@ii \let\cmdb\cmdb@ii
\skvgenloop:
\skvgenloop{}{}{<1.parameter.callback>}
\skvgenloop is non-expandable but it takes any arbitrary
list separator (parser). In the case of non-separated (nsv/tsv)
lists, can be empty. The list is not normalized prior to
parsing, and it is not possible to terminate the loop prematurely
(ie, before exhausting the list).
Here, <1.parameter.callback> is not a macro, unlike the
<1.parameter.callback.macro> for \skvfor.
\newcount\nr\nr=\tw@
\def\stack{}
\skvgenloop{,}{\cmda,\cmdb}{%
\edef\stack{%
\unexpanded\expandafter{\stack}%
\let\noexpand#1%
\expandafter\noexpand\csname\skvremovescape{#1}@%
\romannumeral\nr\endcsname
}%
}
\show\stack
\skvcommaloop, \skvecommaloop, \skvcommaparse, \skvkvparse:
\skvcommaloop{}{}
\skvcommaloop*{}{}
All these have the same syntax. \skvecommaloop is as
\skvcommaloop but will execute once when
is empty. This is needed for processing empty key families.
\skvcommaparse is as \skvcommaloop but will first normalize
using \skvcsvnormalize. \skvkvparse is as \skvcommaloop
but will first normalize using \skvkvnormalize.
\skvkvparse is meant for processing key-value lists.
\skvdolist and \skvparselist:
\skvparselist{}{}{}
\skvparselist*{}{}{}
These can process lists with arbitrary parsers/list
separators.The only difference between \skvdolist and
\skvparselist is that \skvparselist will first normalize
the list before parsing.
\cicadaloop:
\cicadaloop[]{}<\if...>\fi{}
\cicadaloop*[]{}<\if...>\fi{}
\cicadaloop will terminate prematurely on the current depth
whenever the condition specified by <\if...> is true.
is the regular code, to be executed
for every item of the list before the loop is terminated.
Some internal macros are accessible by the user: \iflastcicada,
\currentcicada, \nextcicada, \lastcicada, \cicadacount.
\cicadacount is depth-dependent.
You can insert \listbreaker as an item in or
to terminate the list prematurely. \listpauser isn't recognized by
\cicadaloop.
\let\romn\romannumeral
\@tempcnta\z@
\def\blist{{X};{Y};Z;\fi}
\def\stack{}
\cicadaloop{{a},{b},c,\if}\if01\fi{%
\advance\@tempcnta\@ne
\@tempcntb\z@
\@tempswafalse
\cicadaloop*[;]\blist\if@tempswa\fi{%
\advance\@tempcntb\@ne
\typeout{Doing \romn\@tempcnta,\romn\@tempcntb:
\detokenize{#1,##1}}%
\edef\stack{%
\unexpanded\expandafter{\stack}\noexpand\do
{\romn\@tempcnta,\romn\@tempcntb}{\unexpanded{#1;##1}}%
}%
\ifnum\@tempcntb>\@ne
\@tempswatrue
\skvcsedef{cmd@\romn\cicadaloopdepth}{\lastcicada,\cicadacount}%
\fi
}%
}
\show\stack
\newforeach:
\newforeach[]{}
{}{}{}
\newforeach[]{}
{}{}{}
The star (*) form of \newforeach is equivalent to declaring
the option 'list is a macro' in .
\newforeach is a more versatile and robust version of the popular
\foreach macro of the PGF bundle. The macro accepts arbitrary list
parsers and sub-parsers, including active list separators.
When needed, parameters corresponding to the holder macros can
be used directly in the callback. \newforeach does everything
that PGF's \foreach does. It doesn't automatically scope
calculations as \foreach does. Automatically creating local
groups for computations has been a problem for some users
of \foreach. Instead, \newforeach uses stacks to save and
restore the values of holder macros and some other internal
parameters/quantities.
There is a long list of options/keys in , available for
\newforeach and \foreachfox. For example, key 'reverse list'
allows users to automatically reverse their lists prior to processing.
Also, there are many internal macros that can be accessed by the user.
A few of them are
\breaknewforeach (=\skvbreakloop),
\breakallforeachloops, \foreachcurrentitem,
\foreachnextitem, \foreachitemcount, \foreachprevitem,
\prependtobeginforeach, \appendtobeginforeach (=\atbeginforeach),
\prependtoendforeach, \appendtoendforeach (=\atendforeach),
\ifforeachlastitem, \foreachnestdepth, \foreachlistremainder,
\skvforeachonlyinitially.
\foreachitemcount counts the items as the loop progresses.
It is depth dependent, ie, it can be called to access the
number of items processed on each depth of nesting of
\newforeach. On any depth, the user can say, eg,
\ifnum\foreachitemcount>10\relax\breaknewforeach\fi
or
\ifnum\foreachitemcount>10\relax\breaknewforeach\fi
You can insert \listbreaker as an item in or
to terminate the list prematurely. \listpauser will pause
the processing for user action.
\newforeach found a bug in PGF's \foreach, as reported at
. However, \newforeach itself
can't be said to be bug free.
Maybe you fancy the following syntactic sugar:
\begingroup
\catcode`\,=13
\gdef\alist{1,2,...,5,7,8,...,12}
\endgroup
\parindent-20pt
\begin{tikzpicture}
\draw[step=.5cm,blue!65,very thin] (0,0) grid (13,6);
\newforeach \x [
expand list once,
count in=\xc all \x satisfying \ifnum\x>5\fi
] in \alist {
\newforeach [
parser={;},
subparser=:,
evaluate=\y as \ye using \numexpr\y*10
]
\y:\z in {1:red; 2:green; 3:blue; 4:brown; 5:purple} do {
\draw [fill=\z\ifnum\x>5!\ye\fi] (\x,\y) +(-.5,-.5)
rectangle ++(.5,.5);
\draw (\x,\y) node {\x,\y};
}
}
\global\let\xc\xc
\end{tikzpicture}
\show\xc
\begin{document}
\begin{tikzpicture}[every node/.style=draw]
\node (A) at (0,0) {A};
\node (B) at (1,0) {B};
\node (C) at (2,0) {C};
\node (D) at (3,0) {D};
\newforeach \n [remember=\n as \lastn initially A] in {B,C,D}
\draw [->] (\lastn) -- (\n);
\end{tikzpicture}
\foreachfox [remember=#1 as \x initially A] in {B,...,H} {%
$\overrightarrow{\x#1}$\ifforeachlastitem.\else,\space\fi
}
\foreachfox [list type=tsv,remember=#1 as \x initially A] {B(C1)DE[F1]GH} {%
$\overrightarrow{\x#1}$\ifforeachlastitem.\else,\space\fi
}
\end{document}
% Example credit: Tom Bombadil
\usetikzlibrary{arrows}
\parindent-40pt
\begin{tikzpicture}
% You may fix \pgfmathsetseed to get a constant output:
% \pgfmathsetseed{42}
\newforeach \y [count=\yc] in {-0.5,-1,...,-2.5} {%
\node[right] at (15.2,\y+0.1) {\scriptsize Subject \yc};
\newforeach \x in {1,...,300}{%
\pgfmathtruncatemacro{\drawbool}{rand-0.7 > 0 ? 1 : 0}%
\ifnum\drawbool=1\relax
\fill (\x/20,\y) rectangle (\x/20+0.05,\y+0.3);
\fi
}%
}%
\draw[-stealth] (0,0) -- (15.5,0);
\node[right] at (15.5,0.2) {t in ms};
\newforeach \x [
pgf evaluate=\x as \xv using int(20*\x)
] in {0,...,15} {%
\draw (\x,-0.05) -- (\x,0.05) node[above] {\xv};
}
\end{tikzpicture}
Here, it is possible to use #1, ##1, ##2 in the callback to
refer to \x, y, \z, respectively. This relieves the user of the
need to manually expand the holder macros \x,\y,\z in the callbacks,
as I have seen some TikZ users do.
Here is an example that uses parameters.
\newforeach will not accept \skvifstreq or \skvifstrcmpTF because
they don't start with '\if'. So skeyval package cast \skvifstreq to
\ifforeachstrcmp.
Normally, breaking an inner loop doesn't affect the continuation of
outer loops, but \breakallforeachloops will break all current inner
and outer loops.
\newforeach \x [
item counter = \xc,
exit when = \ifforeachstrcmp{#1}{b}\then\fi
] in {a,b,c} do {%
\newforeach [
count in = \yc all \y satisfying \ifnum\y>2\fi,
loop stopper = \ifnum\y>3\fi
] \y in {1,2,3,4,5} {%
\typeout{Doing items: #1, ##1}%
%\ifnum\y=3 \breakallforeachloops\fi
}%
}
\begin{document}
Numbers: {\tt\string\xc}: \xc, {\tt\string\yc}: \yc
\end{document}
Because of 'continue={\ifnum\y=3\fi}' item number 3 will not appear
in the log file from the following loop:
\newforeach [
item counter = \xc,
exit when = \ifforeachstrcmp{#1}{c}\then\fi
] \x in {a,...,e} do {%
\newforeach \y [
count in = \yc all \y satisfying \ifnum\y>2\fi,
continue = {\ifnum\y=3\fi}
] in {1,...,10} {%
\typeout{Doing items: #1, ##1}%
% \ifnum\y=5 \breakallforeachloops\fi
}%
}
\begin{document}
\begin{tikzpicture}[scale=2]
\let\pgftrun\pgfmathtruncatemacro
\newforeach [count=\ic] \x/\y in {1/.5,1/1,0/1,0/0,1/0,1/-1,0/-1,0/-.5,0/0}
\coordinate (p\ic) at (\x,\y);
\draw[line width=1pt,blue] (p1) \newforeach \p in {2,...,8} {-- (p\p)};
\newforeach \q in {1,...,7} {%
\pgftrun{\qa}{\q+1}\pgftrun{\qb}{\q+2}\pgftrun{\ind}{2*\q-1}%
\pgftrun{\next}{2*\q}%
\coordinate (n\ind) at (barycentric cs:p\q=0.5,p\qa=0.5);
\coordinate (n\next) at (barycentric cs:p\q=0.125,p\qa=0.75,p\qb=0.125);
}
\draw[line width=2pt,red] (n1)\newforeach \q in {2,...,13}{--(n\q)};
\end{tikzpicture}
\end{document}
% A complete graph.
% Example credit:
% Quintin Jean-Noël
%
\usepackage[active,tightpage]{preview}
\PreviewEnvironment{tikzpicture}
\setlength\PreviewBorder{5pt}%
\usetikzlibrary[topaths]
\newcount\mycount
\begin{document}
\begin{tikzpicture}[transform shape]
\newforeach \nr in {1,...,8}{
\mycount=\numexpr(\nr-1)*45\relax
\node[draw,circle,inner sep=0.25cm] (N-\nr) at
(\the\mycount:5.4cm) {};
}
\newforeach \nr in {9,...,16}{
\mycount=\numexpr(\nr-1)*45+22\relax
\node[draw,circle,inner sep=0.25cm] (N-\nr) at
(\the\mycount:5.4cm) {};
}
\newforeach \nr in {1,...,15}{
\mycount=\numexpr\nr+1\relax
\newforeach \nra in {\the\mycount,...,16}{
\path (N-\nr) edge[->,bend right=3] (N-\nra)
edge[<-,bend left=3](N-\nra);
}
}
\end{tikzpicture}
\foreachfox:
\foreachfox[]{}{}
\foreachfox*[]{}{}
\foreachfox does everything that \newforeach does, but, in addition,
it can process nsv/tsv (non-separated, tokenwise lists). It can
auto-complete ellipsis lists (ie, lists with '...'), but this
feature is available only for csv lists and not for tsv lists.
1. One important characteristic of \foreachfox is that it doesn't use
holder macros (eg, \x/\y), but instead it uses parameter
characters directly. So it doesn't accept/expect holder macros.
2. That means that if the argument isn't simple (ie, if it isn't #1),
then it has to be specified as the value of the key/option 'arg'
(eg, arg=#1/#2).
3. Note that \foreachfox doesn't expect any optional 'in' or 'do' in the
argument list.
4. Unlike \newforeach, \foreachfox doesn't recognize the semicolon
(;) as a callback terminator. In fact, the callback for \foreachfox
should ideally be always enclosed in balanced braces. PGF
often uses the semicolon as the callback terminator. \newforeach
supports that feature.
The following macros are available to the user:
\breakforeachfox, \foreachnestdepth, \foreachitemcount, \foreachprevitem,
\foreachcurrentitem, \foreachnextitem, \ifforeachlastitem, \foreachlistremainder,
\prependtobeginforeach, \appendtobeginforeach (=\atbeginforeachfox),
\prependtoendforeach, \appendtoendforeach (=\atendforeachfox).
Examples:
\foreachfox [
remember=#1 as \x initially A
] {B,...,H} {%
$\overrightarrow{\x#1}$\ifforeachlastitem.\else,\space\fi
}
This is equivalent to:
\foreachfox [
list type=tsv,
remember=#1 as \x initially A
] in {BCDEFGH} {%
$\overrightarrow{\x#1}$\ifforeachlastitem.\else,\space\fi
}
\usepackage{multicol}
\begin{document}
\begin{multicols}{2}
\foreachfox [count=\nchar] {a,...,f} {
\foreachfox [
evaluate=##1 as \xe using \numexpr##1*20,
evaluate=##1 as \hx using \pgfmathparse{##1 mm*5},
evaluate=##1 as \vx using \pgfmathparse{##1 mm*1.5},
count in= \ci all ##1 satisfying \ifnum##1<4\fi initially 2,
count in= \di all ##1 satisfying \ifnum##1>2\fi,
] {1,...,6} {
\endgraf
Doing: #1, ##1, \nchar, \ci, \di
\endgraf
\ifnum\nchar>3\relax
\vskip-\vx pt\relax
\fi
\hskip\hx pt\relax
\tikz[rounded corners,ultra thick]{
\shade[ball color=red!\xe!blue] (0,0) circle (.25cm);
\node[white,font=\large] (0,0) {#1};
}%
}%
}
\end{multicols}
\end{document}
% Example credit: Heiko Oberdiek.
\makeatletter
\let\do\newdimen\do\DimWD\do\DimHT\do\DimDP
\newcommand*{\DimBox}[1]{%
\begingroup
\sbox0{\makebox[\DimWD]{#1}}%
\ht0=\DimHT\dp0=\DimDP\usebox\z@
\endgroup
}
\newcommand*{\DimMeasure}[3]{%
\node at (0,0) {%
\let\do\global
\do\DimWD\z@\do\DimHT\z@\do\DimDP\z@
\foreachfox* [arg=##1/##2,parser={#1}] #2 {%
\sbox0{#3}%
\ifdim\wd0>\DimWD\global\DimWD=\wd0\fi
\ifdim\ht0>\DimHT\global\DimHT=\ht0\fi
\ifdim\dp0>\DimDP\global\DimDP=\dp0\fi
}%
};%
}
\makeatother
\begin{document}
\def\yellowlist{a/-1,-2;b/1,-2;c/2,-1;d/2,1;e/1,2;f/-1,2;g/-2,1;h/-2,-1}
\def\bluelist{a/b,b/c,c/d,d/e,e/f,f/g,g/h,h/a}
\begin{tikzpicture}[scale=1.2,auto=left,every node/.style={circle,thick}]
\DimMeasure{;}{\yellowlist}{#1}
\foreachfox* [arg=#1/#2,parser={;},count=\xc] \yellowlist {
\node (#1) at (#2) [fill=blue!20,draw=yellow] {\DimBox{#1}};
}
\DimMeasure{,}{\bluelist}{#1--#2}
\foreachfox* [arg=#1/#2] \bluelist {
\draw [->] (#1) -- (#2)
node [midway,fill=red!20,draw=blue]{\DimBox{#1--#2}};
}
\end{tikzpicture}
\end{document}
\foreachfox [
item counter = \xc,
exit when = \ifforeachstrcmp{#1}{c}\then\fi
] {a,b,c,d} {%
\foreachfox [
count in = \yc all ##1 satisfying \ifnum##1>2\fi,
ignore callback = {\ifnum##1=3\fi}
] {1,...,10} {%
\typeout{Doing items: #1, ##1}%
\ifnum##1=8 \breakforeachfox\fi
}%
}
%\show\xc
\show\yc
* Unfortunately, for drawing instructions like
\draw[blue] (p) \newforeach \p in {1,...,8}{...};
(ie, when \draw precedes \newforeach), both \newforeach and \foreachfox
will not work, because of the way TikZ hardwires \foreach for this type
of task. The tack that loops package resorts to in this situation is
simply to replace \newforeach and \foreachfox with \foreach. The fact
here is that in this case, both \newforeach and \foreachfox will not work
but leave everything for \foreach. This can be a problem when \newforeach
or \foreachfox is called to avoid a known bug in \foreach, such as at
% End of readme file of loops package