part of r5520 is bad for oceanography

Jeff,

http://matplotlib.svn.sourceforge.net/viewvc/matplotlib/trunk/toolkits/basemap/lib/mpl_toolkits/basemap/init.py?r1=5520&r2=5534

Lines 2311 and 2312 in changeset 5520 have bitten me badly, and I think they are ill-advised. They are based on the assumption that input to rotate_vector, which must be called before m_quiver, should be describing a rectangular array of points; but this is not the case for a typical oceanographic use-case, plotting current vectors as measured along a ship track. In that case one naturally has 4 1-D vectors of the same shape for U, V, X, and Y, and one does not want meshgrid to be called on X and Y. If you want automatic meshgrid, it should be called only if X and Y are 1-D *and* their lengths match the 2-D shape of U and V.

Eric

Eric Firing wrote:

Jeff,

matplotlib download | SourceForge.net

Lines 2311 and 2312 in changeset 5520 have bitten me badly, and I think

Oops! I meant 5534; 5520 was the predecessor.

Eric

Eric Firing wrote:

Jeff,

matplotlib download | SourceForge.net

Lines 2311 and 2312 in changeset 5520 have bitten me badly, and I think they are ill-advised. They are based on the assumption that input to rotate_vector, which must be called before m_quiver, should be describing a rectangular array of points; but this is not the case for a typical oceanographic use-case, plotting current vectors as measured along a ship track. In that case one naturally has 4 1-D vectors of the same shape for U, V, X, and Y, and one does not want meshgrid to be called on X and Y. If you want automatic meshgrid, it should be called only if X and Y are 1-D *and* their lengths match the 2-D shape of U and V.

Eric
  
Eric: Sorry about that. For some reason, I was thinking that rotate_vector only made sense for gridded data, which is obviously wrong. Fixed in r5960.

-Jeff

ยทยทยท

--
Jeffrey S. Whitaker Phone : (303)497-6313
NOAA/OAR/CDC R/PSD1 FAX : (303)497-6449
325 Broadway Boulder, CO, USA 80305-3328