I simply use f(x)
(inside a formula, that is, between $...$
or \[...\]
or any other mathematical construct, which shall be implicit in what follows) and consider all the other proposed usages wrong. One might argue about \mathop{\kern0pt f}(x)
so a thin space would be added in front of the f
if preceded by certain kinds of atoms (what happens for \sin
and \log
). A definition should be
\newcommand{\fn}[1]{\mathop{\kern0pt #1}\nolimits}
and \fn{f}(x)
would give the desired result.
However, the following example shows that it is suboptimal: there is no reason for the thin space.
\documentclass{article}
\newcommand{\fn}[1]{\mathop{\kern0pt #1}\nolimits}
\begin{document}
$g(x)f(x)$
$\fn{g}(x)\fn{f}(x)$
$f(x)\ne f'(x)$
$\fn{f}(x)\ne \fn{f}'(x)$
\end{document}

$f(x)=$
to define and then refer to it as$f$
. But sometimes, depending on the size of arguments, I do$f \bigl( \bigr)$
.f(x)
and consider all the other proposed usages wrong.\mathop{f}(x)
is absolutely wrong because the leterf
is vertically positioned not by baseline but centered by math axis. Try\mathop{g}(x)
where the wrong result is more visible. Of course, the vertically positioning by math axis can be cancelled by adding something second to the\mathop
parameter (\kern0pt
in egreg's example) but normalf(x)
is the best.