def diffkeys_explore(self, key1, key2, dump_sample=0):
'''return list of difference observations comparing two fields'''
out = []
vals = self.fields(key1, key2)
notequal = vals[0] != vals[1]
diffcnt = notequal.sum()
if diffcnt == 0:
return out
tcodes = [get_format(val) for val in vals]
tchars = [tcode[0] for tcode in tcodes]
# are all values within floating point fuzz?
if not tchars.count('a') and na.allclose(*vals):
out.append('no differences using allclose()')
# real differences exist
out.append('%s values differ (%1.2f%% of %s)' %
(diffcnt, 100.0 * diffcnt / len(self), len(self)))
if tcodes[0] != tcodes[1]:
out.append('field types differ %s' % str(tuple(tcodes)))
# skip detail if any field is alpha type
if tchars.count('a'):
return out
# differences as different types?
for typestr in ('Int', 'Bool'):
tvals = [v.astype(typestr) for v in vals]
if na.allclose(*tvals):
out.append('field values match as type(%r)' % typestr)
# extract differences and examine in greater detail
dvals = [any_compress(notequal, val) for val in vals]
nzmask = [(dval != 0) for dval in dvals]
if (nzmask[0] != nzmask[1]).sum() == 0:
# all zeros match, compare the nonzero values
nzvals = [any_compress(nzmask[0], dval) for dval in dvals]
ratio = nzvals[1].astype('Float') / nzvals[0].astype('Float')
factor = ratio.mean()
if na.allclose(ratio, factor):
out.append('field values differ by constant factor: %f' %
factor)
delta = dvals[0] - dvals[1]
dmin, dmax = delta.min(), delta.max()
out.append('difference mean=%f range=%f (%s to %s)' %
(delta.mean(), dmax - dmin, dmin, dmax))
if dump_sample:
tmp = UniTable()
tmp['_idx_'] = diffidx = na.nonzero(notequal)[0]
for key in (key1, key2):
tmp[key] = self.field(key)[notequal]
if len(tmp) > dump_sample:
tmp.resize(dump_sample)
out.extend(str(tmp).split('\n'))
return out
def diffkeys_explore(self,key1,key2,dump_sample=0):
'''return list of difference observations comparing two fields'''
out = []
vals = self.fields(key1,key2)
notequal = vals[0] != vals[1]
diffcnt = notequal.sum()
if diffcnt == 0:
return out
tcodes = [get_format(val) for val in vals]
tchars = [tcode[0] for tcode in tcodes]
# are all values within floating point fuzz?
if not tchars.count('a') and na.allclose(*vals):
out.append('no differences using allclose()')
# real differences exist
out.append('%s values differ (%1.2f%% of %s)' % (
diffcnt,100.0*diffcnt/len(self),len(self)))
if tcodes[0] != tcodes[1]:
out.append('field types differ %s' % str(tuple(tcodes)))
# skip detail if any field is alpha type
if tchars.count('a'):
return out
# differences as different types?
for typestr in ('Int','Bool'):
tvals = [v.astype(typestr) for v in vals]
if na.allclose(*tvals):
out.append('field values match as type(%r)' % typestr)
# extract differences and examine in greater detail
dvals = [any_compress(notequal,val) for val in vals]
nzmask = [(dval != 0) for dval in dvals]
if (nzmask[0] != nzmask[1]).sum() == 0:
# all zeros match, compare the nonzero values
nzvals = [any_compress(nzmask[0],dval) for dval in dvals]
ratio = nzvals[1].astype('Float') / nzvals[0].astype('Float')
factor = ratio.mean()
if na.allclose(ratio,factor):
out.append('field values differ by constant factor: %f' % factor)
delta = dvals[0] - dvals[1]
dmin,dmax = delta.min(),delta.max()
out.append('difference mean=%f range=%f (%s to %s)' % (
delta.mean(),dmax-dmin,dmin,dmax))
if dump_sample:
tmp = UniTable()
tmp['_idx_'] = diffidx = na.nonzero(notequal)[0]
for key in (key1,key2):
tmp[key] = self.field(key)[notequal]
if len(tmp) > dump_sample:
tmp.resize(dump_sample)
out.extend(str(tmp).split('\n'))
return out
def compress(self,mask):
'''return new table containing only rows where mask is nonzero'''
out = self._new_hook()
for key in self._keys:
out[key] = any_compress(mask,self._data[key])
out._postinit_hook()
return out
def compress(self, mask):
'''return new table containing only rows where mask is nonzero'''
out = self._new_hook()
for key in self._keys:
out[key] = any_compress(mask, self._data[key])
out._postinit_hook()
return out
