I made a small change to the array creation function so that if you explicitly specify dtype=object, then the logic for determining the shape of the array is changed.
The new logic for constructing dtype=object arrays from Python sequences is that the shape is determined by nested lists (and only lists). The nesting must be consistent or else the elements of the last conforming lists are assumed to be objects. Here are some examples: array([[1,2],3,4],dtype=object) returns a 1-d array of shape (3,) array([[1,2],[3,4]],dtype=object) returns a 2-d array of shape (2,2) array([(1,2),(3,4)],dtype=object) returns a 1-d array of shape (2,) array([],dtype=object) returns a 0-d array of shape () array([[],[],[]],dtype=object) returns a 1-d array of shape (3,) array([[3,4],[5,6],None],dtype=object) returns a 1-d array of shape (3,) You have to actually specify dtype=object to get this, otherwise the old code-path will be taken. -Travis ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Numpy-discussion mailing list Numpy-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/numpy-discussion