pt_blockmedian.py: rolled back changes to help with merge of tyler's …

…branch

pt_blockmedian.py

@@ -24,11 +24,11 @@ def pt_blockmedian(xx, yy, zz, delta, xy0=[0.,0.], return_index=False, index_onl
     xr=np.floor((x-xy0[0])/delta)
     yr=np.floor((y-xy0[1])/delta)
     xyind=xr*yscale+(yr-np.min(yr))
-    sorted_ind=np.argsort(xyind+(z-np.nanmin(z))/zscale)
-    xs=x[sorted_ind]
-    ys=y[sorted_ind]
-    zs=z[sorted_ind]
-    xyind=xyind[sorted_ind]
+    ind=np.argsort(xyind+(z-np.nanmin(z))/zscale)
+    xs=x[ind]
+    ys=y[ind]
+    zs=z[ind]
+    xyind=xyind[ind]
     ux, ix=np.unique(xyind, return_index=True)
     if not index_only:
         xm=np.zeros_like(ux)+np.NaN
@@ -51,7 +51,7 @@ def pt_blockmedian(xx, yy, zz, delta, xy0=[0.,0.], return_index=False, index_onl
                 ym[count]=(ys[iM[0]]+ys[iM[1]])/2.
                 zm[count]=(zs[iM[0]]+zs[iM[1]])/2.
             if return_index:
-                ind[count,:]=sorted_ind[iM]
+                ind[count,:]=iM
         else:
             # odd case: iM ends in 0.5.  if ni=3, iM=3/2-1 = 0.5, want 1
             # if ni=1, iM=-0.5, want 0
@@ -61,7 +61,7 @@ def pt_blockmedian(xx, yy, zz, delta, xy0=[0.,0.], return_index=False, index_onl
                 ym[count]=ys[iM]
                 zm[count]=zs[iM]
             if return_index:
-                ind[count,:]=sorted_ind[iM]
+                ind[count,:]=iM
         #plt.figure(1); plt.clf(); plt.subplot(211); plt.cla(); plt.scatter(xs[ii]-xm[count], ys[ii]-ym[count], c=zs[ii]); plt.axis('equal'); plt.subplot(212); plt.plot(zs[ii]); 
         #plt.pause(1)
         #print(count)
@@ -72,32 +72,3 @@ def pt_blockmedian(xx, yy, zz, delta, xy0=[0.,0.], return_index=False, index_onl
     else:
         return xm, ym, zm
 
-
-def __main__():
-    scale=10
-    N=200000
-    delta=2
-    x=np.random.rand(N)*scale
-    y=np.random.rand(N)*scale
-    z=np.random.rand(N)*scale
-
-    xm, ym, zm0, ii=pt_blockmedian(x, y, z, delta, return_index=True)
-
-    zm=0.5*(z[ii[:,0]]+z[ii[:,1]])
-
-    xi=np.floor(x/delta)*delta
-    yi=np.floor(y/delta)*delta
-    if False:
-        for x0 in np.unique(xi):
-            for y0 in np.unique(yi):
-                els=(xi==x0) & (yi==y0)
-                if ~np.any(els):
-                    continue
-                zmi=np.median(z[els])
-
-                this=(np.floor(xm/delta)*delta == x0) & (np.floor(ym/delta)*delta == y0)
-                print(f"({x0}, {y0}: {zmi-zm[this]}?")
-
-if __name__=='__main__':
-    __main__()
-