Hello all, As part of a wider plan to examine implementing siunitx in LaTeX3, I have just uploaded a generic number parser (numparse3) to CTAN. It is also available via SVN at http://numparse3.berlios.de I've tried to examine how a general module can return information to a module higher up the chain. At present, I've implemented this using a plist (for data) and an integer (for error codes). I'd be interested to hear how people see this approach. I'm hoping to move on to a generalised formatting system for numbers, to take output of numparse3 and format it. First, I've got a few new ideas for keys3 to implement, though. Once all of that is done, I'm hoping to look at the unit engine of siunitx in LaTeX3, and also consider numbers in tables. The later still hinges on what is going to happen about handling tables in LaTeX3. I'm happy to examine porting the array package, but before I sit down to do that I'd be keen to know how people see this. I imagine that for the user, the current \begin{tabular}{<letters>} approach is likely to be maintained. Speaking from the siunitx experience, I'd certainly like to make this a bit more robust (for example, I have issues because I want optional arguments to work with the <letter>). The alternative approach is to have something like: \begin{tabular}{ {<keyval list>} {<keyval list>} ... with one list of options for each column generated. Personally, I suspect that it a bit clumsy. Any feedback on numparse3 (or indeed my other LaTeX3 experiments) is always welcome. -- Joseph Wright