Color series using method `last` never cycle

[muzimuzhi]

The name of \colorseriescycle and the phrase vector \emph{last} used in the doc for color series may make users misunderstand that such a series cycle.

In the following example, user may expect foo!![4] == base == {.3,.3,.3}, but actually it's {.6,.6,.6}

\documentclass{article}
\usepackage{xcolor}

\begin{document}
\definecolorseries{foo}{rgb}{last}[rgb]{.3,.3,.3}[rgb]{.5,.5,.5}
\resetcolorseries[2]{foo} % step = .1,.1,.1

\ttfamily
\begin{testcolors}[rgb]
  \testcolor{foo!![0]}
  \testcolor{foo!![1]}
  \testcolor{foo!![2]}
  \testcolor{foo!![3]}
\end{testcolors}
\end{document}

The relevant doc indicates this is the expected behavior, not a bug.

xcolor/xcolor.dtx

Lines 1267 to 1281 in 1924d5a

	% This is the general scheme:
	% \begin{gather}
	% \emph{color}_1 := \emph{base},\qquad
	% \emph{color}_{n+1} := U\bigl(\emph{color}_{n} + \emph{step}\bigr)
	% \end{gather}
	% for $n=1,2,\dots$, where $U$ maps arbitrary real $m$-vectors into the unit $m$-cube:
	% \begin{gather}
	% U(x_1,\dots,x_m) = (u(x_1),\dots,u(x_m)),\qquad
	% u(x) =
	% \begin{cases}
	% 1 & \text{if $x=1$}\\
	% x-[x]& \text{if $x\neq1$}\\
	% \end{cases}\label{eq.U}
	% \end{gather}
	% Thus, every step of the algorithm yields a valid color with parameters from the interval $[0,1]$.

Perhaps xcolor could add some words to the doc?

[muzimuzhi]

The relevant doc indicates this is the expected behavior, not a bug.

That's why tabu did a manual cycling for its \taburowcolors.

[u-fischer]

It does cycle. But it leaves out the base color:

\documentclass{article}
\usepackage{xcolor}

\begin{document}
\definecolorseries{foo}{rgb}{last}{blue}{red}
\resetcolorseries[4]{foo}

\ttfamily

\begin{testcolors}[rgb]
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  \testcolor{foo!!+}
  
\end{testcolors}
\end{document}

The same happens with absolute index numbers. This looks like a bug to me.

[zauguin]

It's not a bug, it's working as documented. Every component individually wraps around, so after 1 0 0 we get to 1.25 0 -0.25 which gets remapped to allowed color values by the documented U function from the manual to 0.25 0 0.75. This just happens to repeat the second value since the base and last values in your example only contain 0 and 1 and therefore are exactly at the points where it gets wrapped.

[u-fischer]

@zauguin yes, I just saw it too. We should document this better.