Currently ipython caches references to matplotlib objects, which exacerbates the likelihood of running out of memory when interactively plotting large python arrays. This seems to be a common pitfall. The advice is usually to disable ipython's caching when using matplotlib in this way. It would be great if ipython had the option to selectively avoid caching references to objects in the matplotlib module. As this problem presumably also occurs with other modules, one might envisage allowing specification on a namespace basis for the caching to avoid.
This is a dupe of (wishlist) issue 257
preserve volume over application restarts. closes #258