algorithmic + babel with Hebrew conflict: Numbering missing

Question

If I use algorithmic with babel, having both Hebrew and English, the line numbers go missing. Using MikTeX 2.9.

\documentclass{article}
\usepackage[hebrew,english]{babel}
\usepackage{algorithm}
\usepackage{algorithmic}

\begin{document}
\begin{algorithm}
\caption{A caption}
\begin{algorithmic}[1]
\STATE{A line which should be numbered.}
\end{algorithmic}
\end{algorithm}
\end{document}

Answer 1

Joseph Wright has already mentioned the cause of the problem and a possible solution. I found another approach in the page Hebrew in LaTeX. Under the title algorithmics you can find a modified version of algorithmic.sty. You can download this file and save it (where TeX can find it) under a different name (heb-algorithmic.sty, for example) and then load the modified version on the preamble of your document:

\usepackage{heb-algorithmic}

I did a test with the following file:

\documentclass{article}
\usepackage[hebrew,english]{babel}
\usepackage{heb-algorithmic}

\begin{document}

\begin{algorithmic}[1]
\STATE{A line which should be numbered.}
\end{algorithmic}

\end{document}

and everything worked as expected; the line numbers were back.

Answer 2

You can cut the minimal example down to

\documentclass{article}
\usepackage[hebrew,english]{babel}
\usepackage{algorithmic}

\begin{document}

\begin{algorithmic}[1]
\STATE{A line which should be numbered.}
\end{algorithmic}

\end{document}

i.e. the algorithm package is nothing to do with this. The fundamental problem seems to be that the hebrew option for babel redefines \@arabic such that it is not expandable. That makes some internal testing fail, and so no number is printed. Restoring the original definition sorts this out

\documentclass{article}
\usepackage[hebrew,english]{babel}
\usepackage{algorithmic}

\begin{document}

\makeatletter
\def\@arabic#1{\number#1}
\makeatother

\begin{algorithmic}[1]
\STATE{A line which should be numbered.}
\end{algorithmic}

\end{document}

Whether this will be okay in the real case, I don't know (I'm sure that there is a reason for the change).