Updates to Var and Axis classes:
- Vars now have a 'units' attribute
- Vars and Axes now have a 'plotatts' dictionary, which contains all plotting
parameters. Some possible keys are:
plottitle - The title to use for plots of this variable (or the axis
label, if this is an Axis)
plotname - A shorter title, used for axis labels.
plotunits - The units of the variable (overrides the 'units' attribute
of the variable).
plotscale - The type of scaling to apply to the values, either 'linear'
or 'log'.
plotorder - The direction to plot the axis coordinates. 1 = away from
origin, -1 = towards origin.
formatstr - A special string for formatting the values on axis ticks
and plot titles.
scalefactor - unit conversion, to get values in the plotunits
offset - unit conversion, to get values in the plotunits
- Auto-completion of axis & variable members of Var and Dataset objects
(works in ipython). These members are also listed in dir() calls now.
Updates to the plot module:
- Added 'every' keyword to plotquiver() to thin the vector field, implemented
lblx, lbly kwargs.
- In plotsigmask(), added "alpha" kwarg to control alpha value of mask
- Keyword arguments "scaleAx", "scaleX", "scaleY" control scaling the values
by one of the axis coordinates. This preserves integrals under a
logarithmic axis.
- added log10-scaling of 2D field values with new kwarg 'logVal'
- apply scalefactor and offset from plotatts, if they're defined. Only
triggers if var.plotatts['plotunits'] is different from var.units.
- disable anti-aliasing in 2D plots. In some versions of matplotlib (e.g.
the one shipped with Ubuntu 11.10), antialiasing causes annoying gaps
between polygons in various plots (contourf, pcolor, etc.).
- Created a more modularized set of plotting routines. This gives greater
control over the construction of the plots, e.g. for publication. The old
'plotvar' routine is still available for cases where the automatic
formatting is good enough.
Updates to the stats module:
- Added some t test code:
difference(X, Y) - Calculates difference between the mean values of X
and Y averaged over the dimensions specified by axes.
Returns X - Y, p values, confidence intervals, and
degrees of freedom.
isnonzero(X) - Determines if X is non-zero, assuming X is normally
distributed. Returns mean of X along axes, p value,
and confidence interval.
- Updated regress() to work on arbitrary axes
- can now specify which axes to correlate()
- correlate() now works with NaN values.
Updates to the timeutils module:
- Lag variable for lagged correlations
- Moved _uniquify, modify, reltime, delta, wrapdate, and date_diff from
timeaxis module to timeutils module.
- New operator removeleapyears() to convert a standard time axis to a 365-day
or 360-day axis.
- New operator splittimeaxis() to split a time axis into 2 dimensions (e.g.
year, day-of-year).
- New function conform_values() to make 2 time axes have the same units and
start date.
- New operator jointimeaxes() to convert a 2D time representation (e.g. year,
day-of-year) to a single time axis.
Other updates:
- Interpolation now is nan-aware and can extrapolate linearly. By default,
no extrapolation is done (gives NaNs outside the data range).
- Improved compatibility with newer versions of the progressbar package.
- In concat module, allow the concatenation axis to be specified by the user,
so it doesn't have to be guessed.
- Added argmax() operator (works on one dimension)
- In composite operator, events that extend past the edge of the data now are
filled with NaNs
- Improved handling of endpoints in the 'smooth' operator.