def test_adding_empty(self):
ha1 = h1(None, "fixed_width", 10, adaptive=True)
ha1.fill_n(np.random.normal(100, 10, 1000))
ha2 = h1(None, "fixed_width", 10, adaptive=True)
ha3 = ha1 + ha2
assert ha1 == ha3
ha4 = ha2 + ha1
assert ha1 == ha4
def test_adding_empty(self):
ha1 = h1(None, "fixed_width", 10, adaptive=True)
ha1.fill_n(np.random.normal(100, 10, 1000))
ha2 = h1(None, "fixed_width", 10, adaptive=True)
ha3 = ha1 + ha2
assert ha1 == ha3
ha4 = ha2 + ha1
assert ha1 == ha4
def test_adding_full(self):
ha1 = h1(None, "fixed_width", 10, adaptive=True)
ha1.fill_n([1, 43, 23])
ha2 = h1(None, "fixed_width", 10, adaptive=True)
ha2.fill_n([23, 51])
ha3 = ha1 + ha2
ha4 = ha2 + ha1
assert np.array_equal(ha3.frequencies, [1, 0, 2, 0, 1, 1])
assert np.array_equal(ha3.numpy_bins, [0, 10, 20, 30, 40, 50, 60])
assert ha4 == ha3
def test_adding_full(self):
ha1 = h1(None, "fixed_width", 10, adaptive=True)
ha1.fill_n([1, 43, 23])
ha2 = h1(None, "fixed_width", 10, adaptive=True)
ha2.fill_n([23, 51])
ha3 = ha1 + ha2
ha4 = ha2 + ha1
assert np.array_equal(ha3.frequencies, [1, 0, 2, 0, 1, 1])
assert np.array_equal(ha3.numpy_bins, [0, 10, 20, 30, 40, 50, 60])
assert ha4 == ha3
def test_update(self):
example = h1(values)
example.dtype = np.int16
assert example.dtype == np.int16
assert example.frequencies.dtype == np.int16
example = h1(values, weights=[1, 2, 2.1, 3.2])
with pytest.raises(RuntimeError):
example.dtype = np.int16
example = h1(values, weights=[1, 2, 2, 3])
example.dtype = np.int16
assert example.dtype == np.int16
def test_update(self):
example = h1(values)
example.dtype = np.int16
assert example.dtype == np.int16
assert example.frequencies.dtype == np.int16
example = h1(values, weights=[1, 2, 2.1, 3.2])
with pytest.raises(RuntimeError):
example.dtype = np.int16
example = h1(values, weights=[1, 2, 2, 3])
example.dtype = np.int16
assert example.dtype == np.int16
def test_fill_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill(24)
assert h.bin_count == 1
assert np.array_equal(h.bin_left_edges, [20])
assert np.array_equal(h.bin_right_edges, [30])
assert np.array_equal(h.frequencies, [1])
def test_fill_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill(24)
assert h.bin_count == 1
assert np.array_equal(h.bin_left_edges, [20])
assert np.array_equal(h.bin_right_edges, [30])
assert np.array_equal(h.frequencies, [1])
def test_non_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill_n([4, 5, 11, 12])
h.fill_n([-13, 120])
assert h.bin_left_edges[0] == -20
assert h.bin_left_edges[-1] == 120
assert h.bin_count == 15
def test_coerce(self):
example = h1(values, dtype=np.int32)
example._coerce_dtype(np.int64)
assert example.dtype == np.int64
example._coerce_dtype(np.float)
assert example.dtype == np.float
example._coerce_dtype(np.int32)
assert example.dtype == np.float
def test_non_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill_n([4, 5, 11, 12])
h.fill_n([-13, 120])
assert h.bin_left_edges[0] == -20
assert h.bin_left_edges[-1] == 120
assert h.bin_count == 15
def test_coerce(self):
example = h1(values, dtype=np.int32)
example._coerce_dtype(np.int64)
assert example.dtype == np.int64
example._coerce_dtype(np.float)
assert example.dtype == np.float
example._coerce_dtype(np.int32)
assert example.dtype == np.float
def stair_histogram(path, width=1920, height=1280, bar=128, jitter=.1, **kwargs):
histogram = h1(None, "fixed_width", bar, range=(0, width))
data = np.random.normal(width, width, int(1000 / jitter))
histogram.fill_n(data)
dpi = 100
fig, ax = plt.subplots(figsize=(width / dpi, height / dpi))
histogram.plot("fill", ax=ax, color="red", **kwargs)
ax.set_axis_off()
save_only_axis(ax, path, dpi=dpi)
def test_scalar_arithmetic(self):
example = h1(values, dtype=np.int32)
assert (example / 3).dtype == np.float
assert (example * 3).dtype == np.int32
assert (example * 3.1).dtype == np.float
with pytest.raises(TypeError):
example * complex(4, 5)
def stair_histogram(path, width=1920, height=1280, bar=128, jitter=.1, **kwargs):
histogram = h1(None, "fixed_width", bar, range=(0, width))
data = np.random.normal(width, width, int(1000 / jitter))
histogram.fill_n(data)
dpi = 100
fig, ax = plt.subplots(figsize=(width / dpi, height / dpi))
histogram.plot("fill", ax=ax, color="red", **kwargs)
ax.set_axis_off()
save_only_axis(ax, path, dpi=dpi)
def test_scalar_arithmetic(self):
example = h1(values, dtype=np.int32)
assert (example / 3).dtype == np.float
assert (example * 3).dtype == np.int32
assert (example * 3.1).dtype == np.float
with pytest.raises(TypeError):
example * complex(4, 5)
def test_create_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
assert h.bin_count == 0
assert h.total == 0
assert np.allclose(h.bin_widths, 10)
assert np.isnan(h.mean())
assert np.isnan(h.std())
assert h.overflow == 0
assert h.underflow == 0
def test_create_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
assert h.bin_count == 0
assert h.total == 0
assert np.allclose(h.bin_widths, 10)
assert np.isnan(h.mean())
assert np.isnan(h.std())
assert h.overflow == 0
assert h.underflow == 0
def create_h1(cursor, *args, **kwargs):
axis_names = _get_axis_names(cursor)
if len(axis_names) != 1:
raise RuntimeError("Invalid number of columns: {0}".format(len(axis_names)))
kwargs["axis_name"] = kwargs.get("axis_name", axis_names[0])
if kwargs.get("adaptive", False):
h = h1(None, *args, **kwargs)
for row in cursor:
h << row[0]
return h
else:
raise NotImplementedError()
def compute_pdfs(fname, d_arr):
data = np.load(fname % ('data', 'y'), mmap_mode='r')
pdfs = np.zeros((d_arr.size, 3, 2, 100))
for i in range(d_arr.size):
d = d_arr[i]
ix = data[:,d:,...,0] - data[:,:-d,...,0]
iy = data[:,:,d:,:,1] - data[:,:,:-d,:,1]
iz = data[:,...,d:,2] - data[:,...,:-d,2]
for j, i_arr in enumerate([ix, iy, iz]):
tmp = i_arr / i_arr.std()
h = h1(tmp[np.abs(tmp)<20], bins=100)
h.normalize(True)
pdfs[i,j,0] = h.bin_centers
pdfs[i,j,1] = h.densities
np.save(fname % ('pdfs', 'y'), pdfs)
def compute_pdfs(samples, name):
fname = name % ('pdfs', 'y')
if os.path.isfile(fname):
pdfs = np.load(fname)
else:
displ = np.array([2, 4, 8, 16, 32, 64])
pdfs = np.zeros((displ.size, 2, 100))
for i in range(displ.size):
incr = samples[:, displ[i]:] - samples[:, :-displ[i]]
incr /= incr.std()
h = h1(incr[np.abs(incr) < 20], bins=100)
h.normalize(True)
pdfs[i, 0] = h.bin_centers
pdfs[i, 1] = h.densities
np.save(fname, pdfs)
return pdfs
def test_hist_arithmetic(self):
example = h1(values, dtype=np.int32)
example2 = example.copy()
example2.dtype = np.float
example2 *= 1.01
example3 = example.copy()
example3.dtype = np.int64
assert (example + example2).dtype == np.float
assert (example2 + example).dtype == np.float
assert (example + example3).dtype == np.int64
assert (example3 - example).dtype == np.int64
example += example2
assert example.dtype == np.float
def test_hist_arithmetic(self):
example = h1(values, dtype=np.int32)
example2 = example.copy()
example2.dtype = np.float
example2 *= 1.01
example3 = example.copy()
example3.dtype = np.int64
assert (example + example2).dtype == np.float
assert (example2 + example).dtype == np.float
assert (example + example3).dtype == np.int64
assert (example3 - example).dtype == np.int64
example += example2
assert example.dtype == np.float
def test_fill_non_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill(4)
h.fill(-14)
assert h.bin_count == 3
assert h.total == 2
assert np.array_equal(h.bin_left_edges, [-20, -10, 0])
assert np.array_equal(h.bin_right_edges, [-10, 0, 10])
assert np.array_equal(h.frequencies, [1, 0, 1])
h.fill(-14)
assert h.bin_count == 3
assert h.total == 3
h.fill(14)
assert h.bin_count == 4
assert h.total == 4
assert np.array_equal(h.frequencies, [2, 0, 1, 1])
def test_fill_non_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill(4)
h.fill(-14)
assert h.bin_count == 3
assert h.total == 2
assert np.array_equal(h.bin_left_edges, [-20, -10, 0])
assert np.array_equal(h.bin_right_edges, [-10, 0, 10])
assert np.array_equal(h.frequencies, [1, 0, 1])
h.fill(-14)
assert h.bin_count == 3
assert h.total == 3
h.fill(14)
assert h.bin_count == 4
assert h.total == 4
assert np.array_equal(h.frequencies, [2, 0, 1, 1])
def compute_stats(fname):
data = np.load(fname % ('data', 'y'), mmap_mode='r')
num = data.shape[1]
d_arr = np.arange(1, num-1, 1)
# p_arr = np.arange(1, 13)
p_arr = np.array([1, 2])
# energy spectrum
E = np.zeros((num//2-1, 3))
f = np.fft.fftfreq(num, 2./num)[1:num//2]
for i in range(3):
reshaped_data = np.reshape(np.moveaxis(data, 1+i, 0), (num, -1))
cov = np.mean(np.cov(reshaped_data, bias=True), axis=-1)
E[:,i] = np.abs(np.fft.fft(cov)[1:num//2])
np.savez(fname % ('energy', 'z'), f=f, E=E)
del f; del E
# increment moments and pdf
moments = np.zeros((d_arr.size, 3, p_arr.size))
pdfs = np.zeros((d_arr.size, 100, 2, 3))
for i in range(d_arr.size):
d = d_arr[i]
ix = data[:,d:,...,0] - data[:,:-d,...,0]
iy = data[:,:,d:,:,1] - data[:,:,:-d,:,1]
iz = data[:,...,d:,2] - data[:,...,:-d,2]
for j, i_arr in enumerate([ix, iy, iz]):
h = h1(i_arr/np.expand_dims(i_arr.std(axis=1+j), 1+j), bins=100)
h.normalize(True)
pdfs[i,:,0,j] = h.bin_centers
pdfs[i,:,1,j] = h.frequencies
for k in range(p_arr.size):
moments[i,j,k] = np.mean(np.mean(np.abs(i_arr)**p_arr[k], axis=1+j))
np.savez(fname % ('moments', 'z'), d=d_arr, m=moments, p=pdfs)
del d_arr; del moments; del pdfs
def test_empty_exact(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill_n([10])
assert np.array_equal(h.bin_left_edges, [10])
assert np.array_equal(h.frequencies, [1])
assert h.total == 1
def test_empty_exact(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill_n([10])
assert np.array_equal(h.bin_left_edges, [10])
assert np.array_equal(h.frequencies, [1])
assert h.total == 1
def test_copy(self):
example = h1(values, dtype=np.int32)
assert example.dtype == np.int32
assert example.copy().dtype == np.int32
def test_with_weights(self):
example = h1(values, weights=[1, 2, 2.1, 3.2])
assert example.dtype == np.float
def test_2(self):
data = np.random.rand(100)
hh = h1(data, 120)
hha = h1(data, 60)
hhb = hh.merge_bins(2, inplace=False)
assert hha == hhb
def test_empty(self):
example = h1(None, "fixed_width", 10, adaptive=True)
assert example.dtype == np.int64
from physt import h1
import matplotlib.pyplot as plt
# Create the sample
heights = [
160, 155, 156, 198, 177, 168, 191, 183, 184, 179, 178, 172, 173, 175, 172,
177, 176, 175, 174, 173, 174, 175, 177, 169, 168, 164, 175, 188, 178, 174,
173, 181, 185, 166, 162, 163, 171, 165, 180, 189, 166, 163, 172, 173, 174,
183, 184, 161, 162, 168, 169, 174, 176, 170, 169, 165
]
hist = h1(heights, 10) # <--- get the histogram data
hist << 190 # <--- add a forgotten value
hist.plot(show_values=True)
def test_mean_weights(self):
hw = physt.h1(values, weights=weights)
assert hw.mean() == 2.8
def test_mean_no_weights(self):
h = physt.h1(values)
assert h.mean() == 2.5
def test_empty(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill_n([4, 5, 11, 12])
assert np.array_equal(h.bin_left_edges, [0, 10])
assert h.total == 4
assert h.mean() == 8.0
def test_empty_right_edge(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill_n([4, 4, 10])
assert np.array_equal(h.bin_left_edges, [0, 10])
assert h.total == 3
assert h.mean() == 6.0
def test_with_weights(self):
h = h1(None, "fixed_width", 10, adaptive=True)
h.fill_n([4, 5, 6, 12], [1, 1, 2, 3])
assert np.array_equal(h.frequencies, [4, 3])
assert np.array_equal(h.errors2, [6, 9])
assert np.array_equal(h.numpy_bins, [0, 10, 20])
def test_with_incorrect_weights(self):
h = h1(None, "fixed_width", 10, adaptive=True)
with pytest.raises(RuntimeError):
h.fill_n([0, 1], [2, 3, 4])
with pytest.raises(RuntimeError):
h.fill_n([0, 1, 2, 3], [2, 3, 4])
def test_2(self):
data = np.random.rand(100)
hh = h1(data, 120)
hha = h1(data, 60)
hhb = hh.merge_bins(2, inplace=False)
assert hha == hhb
def make_big_h1(data):
# 0.3.26 The following creates 2 additional copies!!!
# 0.3.25 The following creates 5 additional copies!!!
h = physt.h1(data)
def test_simple(self):
example = h1(values)
assert example.dtype == np.int64
def test_simple(self):
example = h1(values)
assert example.dtype == np.int64
def test_with_weights(self):
example = h1(values, weights=[1, 2, 2.1, 3.2])
assert example.dtype == np.float
def test_explicit(self):
example = h1(values, dtype=float)
assert example.dtype == float
with pytest.raises(RuntimeError):
example = h1(values, weights=[1, 2, 2.1, 3.2], dtype=int)
def test_explicit(self):
example = h1(values, dtype=float)
assert example.dtype == float
with pytest.raises(RuntimeError):
example = h1(values, weights=[1, 2, 2.1, 3.2], dtype=int)
def test_copy(self):
example = h1(values, dtype=np.int32)
assert example.dtype == np.int32
assert example.copy().dtype == np.int32
def test_multiplication(self):
ha1 = h1(None, "fixed_width", 10, adaptive=True)
ha1.fill_n([1, 43, 23])
ha1 *= 2
ha1.fill_n([-2])
assert np.array_equal(ha1.frequencies, [1, 2, 0, 2, 0, 2])
def test_stats_filled_in(self):
h = physt.h1(values)
assert h._stats["sum"] == 10
assert h._stats["sum2"] == 30
def test_with_incorrect_weights(self):
h = h1(None, "fixed_width", 10, adaptive=True)
with pytest.raises(RuntimeError):
h.fill_n([0, 1], [2, 3, 4])
with pytest.raises(RuntimeError):
h.fill_n([0, 1, 2, 3], [2, 3, 4])
def test_empty(self):
example = h1(None, "fixed_width", 10, adaptive=True)
assert example.dtype == np.int64
def test_mean_no_weights(self):
h = physt.h1(values)
assert h.mean() == 2.5
def test_stats_filled_in(self):
h = physt.h1(values)
assert h._stats["sum"] == 10
assert h._stats["sum2"] == 30
def test_std_no_weights(self):
h = physt.h1(values)
assert np.allclose(h.std(), np.sqrt(5/4))
def test_std_no_weights(self):
h = physt.h1(values)
assert np.allclose(h.std(), np.sqrt(5 / 4))
def test_mean_weights(self):
hw = physt.h1(values, weights=weights)
assert hw.mean() == 2.8
def test_std_weights(self):
hw = physt.h1(values, weights=weights)
assert np.allclose(hw.std(), np.sqrt(6.8 / 5))
def test_std_weights(self):
hw = physt.h1(values, weights=weights)
assert np.allclose(hw.std(), np.sqrt(6.8 / 5))
