Why can't I apply this decoration to a curve produced using the “to” path operation?

Question

I have found that I cannot make decorations apply to (most) curves drawn using the "to" path operation. Here is a working example:

\documentclass[a4paper, reqno]{amsart}

\usepackage{tikz}
\usetikzlibrary{decorations.markings, topaths}

\makeatletter
\tikzset{nomorepostaction/.code = \let\tikz@postactions\pgfutil@empty}
\makeatother
\tikzset{bigdiagramarrow/.style = {
    decoration = {markings, mark = at position 1 with {\arrow[thick]{latex}}},
    postaction = {nomorepostaction, decorate}
}}

\begin{document}
    \begin{tikzpicture}
        \node [draw] (A) at (0, 0) {$A$};
        \node [draw] (B) at (0, 1) {$B$};
        \node [draw] (C) at (2, 1) {$C$};
        \draw [bigdiagramarrow] (A) -- (B);
        \draw [out = 0, in = 180, bigdiagramarrow] (A) to (C);
    \end{tikzpicture}
\end{document}

Here is what I get when I compile it:

As you can see, an arrowhead is drawn on the end of the path drawn using the -- path operation, but no arrowhead is drawn on the end of the path drawn using the to path operation.

Strangely, in the document in which I encountered this issue, I found that some curves created using the "to" path operation will display the decoration. To be more precise, 3 of the 12 curves in the document I am actually trying to typeset display the decoration. It is the same 3 every time I compile the document. I do not know what sets these 3 apart from the rest.

At first it appeared that this question might be relevant, but I do not appear to be encountering the same "gotcha" so far as I can see, since I am not passing options to the to path operation but rather to the \draw macro.

NB. The nomorepostaction key comes from Ulrich Schwarz's answer to this question.

Answer 1

I found two more or less appropriate ways to change the arrow size (which is not a solution of the actual problem at hand, namely that at position 1 is not calculated correct):

• less appropriate: 
  • New arrow head: _latex
  • New style: arrow line width that takes a length (the new \pgflinewidth)
  • and arrow ultra thin, arrow very thin, …, arrow ultra thick that set arrow line width
• more appropriate: 
  • Re-defining a macro that locally (!) sets a new line width for the arrows.
  • New style: arrow line width that takes a factor
  • and arrow ultra thin, arrow very thin, …, arrow ultra thick that set arrow line width

(They could even be combined.)

less appropriate:  New Arrow Head

I chose to define a new arrow _latex that is the same as latex only that \pgflinewidth is replaced by \pgflinewidth@arrow.

The length \pgflinewidth@arrow is per default \pgflinewidth so the arrow acts as usual unless the key arrow line width is used explicitly with an arbitrary length. (Sadly, thick and so on cannot be used, but arrow thick.)

Improvements could be made that one could use the already existing arrows without any change by finding a good hook to set \pgflinewidth locally.

For many other way to change the arrow size: Is it possible to change the size of an arrowhead in TikZ/PGF?

Code

\documentclass[tikz,border=2pt]{standalone}
\usepackage{tikz}
\usetikzlibrary{decorations.markings}

\makeatletter
\tikzset{nomorepostaction/.code = \let\tikz@postactions\pgfutil@empty}
\newdimen\pgflinewidth@arrow
\pgfarrowsdeclare{_latex}{_latex}
{
  \pgfutil@tempdima=0.28pt%
  \pgfutil@tempdimb=\pgflinewidth@arrow%
  \ifdim\pgfinnerlinewidth>0pt%
    \pgfmathsetlength\pgfutil@tempdimb{.6\pgflinewidth@arrow-.4*\pgfinnerlinewidth}%
  \fi%
  \advance\pgfutil@tempdima by.3\pgfutil@tempdimb%
  \pgfarrowsleftextend{+-1\pgfutil@tempdima}
  \pgfarrowsrightextend{+9\pgfutil@tempdima}
}
{
  \pgfutil@tempdima=0.28pt%
  \pgfutil@tempdimb=\pgflinewidth@arrow%
  \ifdim\pgfinnerlinewidth>0pt%
    \pgfmathsetlength\pgfutil@tempdimb{.6\pgflinewidth@arrow-.4*\pgfinnerlinewidth}%
  \fi%
  \advance\pgfutil@tempdima by.3\pgfutil@tempdimb%
  \pgfpathmoveto{\pgfqpoint{9\pgfutil@tempdima}{0pt}}
  \pgfpathcurveto
  {\pgfqpoint{6.3333\pgfutil@tempdima}{.5\pgfutil@tempdima}}
  {\pgfqpoint{2\pgfutil@tempdima}{2\pgfutil@tempdima}}
  {\pgfqpoint{-1\pgfutil@tempdima}{3.75\pgfutil@tempdima}}
  \pgfpathlineto{\pgfqpoint{-1\pgfutil@tempdima}{-3.75\pgfutil@tempdima}}
  \pgfpathcurveto
  {\pgfqpoint{2\pgfutil@tempdima}{-2\pgfutil@tempdima}}
  {\pgfqpoint{6.3333\pgfutil@tempdima}{-.5\pgfutil@tempdima}}
  {\pgfqpoint{9\pgfutil@tempdima}{0pt}}
  \pgfusepathqfill
}

\tikzset{
    bigdiagramarrow/.style = {
        -_latex,
        arrow thick
    },
    arrow line width/.code={
        \pgfmathsetlength\pgflinewidth@arrow{#1}
    },
    arrow line width/.initial=\pgflinewidth,
    arrow ultra thin/.style={arrow line width=0.1pt},
    arrow very thin/.style={arrow line width=0.2pt},
    arrow thin/.style={arrow line width=0.4pt},
    arrow semithick/.style={arrow line width=0.6pt},
    arrow thick/.style={arrow line width=0.8pt},
    arrow very thick/.style={arrow line width=1.2pt},
    arrow ultra thick/.style={arrow line width=1.6pt},
}
\makeatother
\begin{document}
    \begin{tikzpicture}
        \node [draw] (A) at (0, 0) {$A$};
        \node [draw] (B) at (0, 1) {$B$};
        \node [draw] (C) at (2, 1) {$C$};
        \draw [bigdiagramarrow] (A) to (B);
        \draw [out = 0, in = 180][bigdiagramarrow] (A) to (C);
    \end{tikzpicture}
\end{document}

Output

more appropriate: 

Code

\documentclass[tikz,border=2pt]{standalone}
\usepackage{tikz}
\usetikzlibrary{decorations.markings}

\makeatletter
\tikzset{nomorepostaction/.code = \let\tikz@postactions\pgfutil@empty}
\def\pgf@arrowset#1#2{%
  \pgfutil@ifundefined{pgf@arrow@code@#2}
  {\PackageError{pgf}{Arrow #1 type ``#2'' unknown}{}}
  {%
    \edef\pgf@arrow@temp{#2}%
    \ifx\pgf@arrow@temp\pgfutil@empty% ah. clear!
      \expandafter\let\csname pgf@#1arrow\endcsname=\pgfutil@empty%
      \expandafter\let\csname pgf@shorten@#1\endcsname=\pgfutil@empty%
    \else%
      %\pgf@arrow@prepare{#2}%
      \expandafter\def\csname pgf@#1arrow\endcsname{\multiply\pgflinewidth by \pgflinewidth@arrow\relax\pgf@arrow@call{#2}}%
      \expandafter\edef\csname pgf@shorten@#1\endcsname{\multiply\pgflinewidth by \noexpand\pgflinewidth@arrow\relax\expandafter\noexpand\csname pgf@arrow@right@#2\endcsname}%
    \fi%
  }%
}
\tikzset{
    bigdiagramarrow/.style = {
        -latex,
        arrow thick
    },
    arrow line width/.code={
        \pgfmathsetmacro\pgflinewidth@arrow{#1}
    },
    arrow line width=1,
    arrow ultra thin/.style={arrow line width=0.25},
    arrow very thin/.style={arrow line width=0.5},
    arrow thin/.style={arrow line width=1},
    arrow semithick/.style={arrow line width=1.5},
    arrow thick/.style={arrow line width=2},
    arrow very thick/.style={arrow line width=3},
    arrow ultra thick/.style={arrow line width=4},
}
\makeatother
\begin{document}
    \begin{tikzpicture}
        \node [draw] (A) at (0, 0) {$A$};
        \node [draw] (B) at (0, 1) {$B$};
        \node [draw] (C) at (2, 1) {$C$};
        \draw [bigdiagramarrow] (A) to (B);
        \draw [out = 0, in = 180][bigdiagramarrow] (A) to (C);
    \end{tikzpicture}
\end{document}

Output

Answer 2

Why to adapt a simple way when there are complex methods?

\documentclass{standalone}
\usepackage{tikz}
\begin{document}
    \begin{tikzpicture}
        \node [draw] (A) at (0, 0) {$A$};
        \node [draw] (B) at (0, 1) {$B$};
        \node [draw] (C) at (2, 1) {$C$};
        \draw [-latex] (A) -- (B);
        \draw [out = 0, in = 180, -latex] (A) to (C);
    \end{tikzpicture}
\end{document}