def test_round_any():
x = 4.632
assert round_any(x, 1) == 5
assert round_any(x, 2) == 4
assert round_any(x, 3) == 6
assert round_any(x, 4) == 4
assert round_any(x, 5) == 5
assert round_any(x, 1.5) == 4.5
def test_round_any():
x = 4.632
assert round_any(x, 1) == 5
assert round_any(x, 2) == 4
assert round_any(x, 3) == 6
assert round_any(x, 4) == 4
assert round_any(x, 5) == 5
assert round_any(x, 1.5) == 4.5
# Maintains the same index
s = pd.Series([1.1, 2.2, 3.3], index=[3, 2, 1])
result = round_any(s, 2)
assert s.index.equals(result.index)
def test_round_any():
x = 4.632
assert round_any(x, 1) == 5
assert round_any(x, 2) == 4
assert round_any(x, 3) == 6
assert round_any(x, 4) == 4
assert round_any(x, 5) == 5
assert round_any(x, 1.5) == 4.5
# Maintains the same index
s = pd.Series([1.1, 2.2, 3.3], index=[3, 2, 1])
result = round_any(s, 2)
assert s.index.equals(result.index)
def fuzzybreaks(scale,
breaks=None,
boundary=None,
binwidth=None,
bins=30,
right=True):
"""
Compute fuzzy breaks
For a continuous scale, fuzzybreaks "preserve" the range of
the scale. The fuzzing is close to numerical roundoff and
is visually imperceptible.
Parameters
----------
scale : scale
Scale
breaks : array_like
Sequence of break points. If provided and the scale is not
discrete, they are returned.
boundary : float
First break. If `None` a suitable on is computed using
the range of the scale and the binwidth.
binwidth : float
Separation between the breaks
bins : int
Number of bins
right : bool
If `True` the right edges of the bins are part of the
bin. If `False` then the left edges of the bins are part
of the bin.
Returns
-------
out : array_like
"""
# Bins for categorical data should take the width
# of one level, and should show up centered over
# their tick marks. All other parameters are ignored.
if isinstance(scale, scale_discrete):
breaks = scale.get_breaks()
return -0.5 + np.arange(1, len(breaks) + 2)
else:
if breaks is not None:
breaks = scale.transform(breaks)
if breaks is not None:
return breaks
recompute_bins = binwidth is not None
srange = scale.limits
if binwidth is None or np.isnan(binwidth):
binwidth = (srange[1] - srange[0]) / bins
if boundary is None or np.isnan(boundary):
boundary = round_any(srange[0], binwidth, np.floor)
if recompute_bins:
bins = int(np.ceil((srange[1] - boundary) / binwidth))
# To minimise precision errors, we do not pass the boundary and
# binwidth into np.arange as params. The resulting breaks
# can then be adjusted with finer(epsilon based rather than
# some arbitrary small number) precision.
breaks = np.arange(boundary, srange[1] + binwidth, binwidth)
return _adjust_breaks(breaks, right)
def fuzzybreaks(scale, breaks=None, boundary=None,
binwidth=None, bins=30, right=True):
"""
Compute fuzzy breaks
For a continuous scale, fuzzybreaks "preserve" the range of
the scale. The fuzzing is close to numerical roundoff and
is visually imperceptible.
Parameters
----------
scale : scale
Scale
breaks : array_like
Sequence of break points. If provided and the scale is not
discrete, they are returned.
boundary : float
First break. If `None` a suitable on is computed using
the range of the scale and the binwidth.
binwidth : float
Separation between the breaks
bins : int
Number of bins
right : bool
If `True` the right edges of the bins are part of the
bin. If `False` then the left edges of the bins are part
of the bin.
Returns
-------
out : array_like
"""
# Bins for categorical data should take the width
# of one level, and should show up centered over
# their tick marks. All other parameters are ignored.
if isinstance(scale, scale_discrete):
breaks = scale.get_breaks()
return -0.5 + np.arange(1, len(breaks)+2)
else:
if breaks is not None:
breaks = scale.transform(breaks)
if breaks is not None:
return breaks
recompute_bins = binwidth is not None
srange = scale.limits
if binwidth is None or np.isnan(binwidth):
binwidth = (srange[1]-srange[0]) / bins
if boundary is None or np.isnan(boundary):
boundary = round_any(srange[0], binwidth, np.floor)
if recompute_bins:
bins = np.int(np.ceil((srange[1]-boundary)/binwidth))
# To minimise precision errors, we do not pass the boundary and
# binwidth into np.arange as params. The resulting breaks
# can then be adjusted with finer(epsilon based rather than
# some arbitrary small number) precision.
breaks = np.arange(boundary, srange[1]+binwidth, binwidth)
return _adjust_breaks(breaks, right)
