From 20d1eb1d5619b74c2e88f7436b2e8add47db8ce0 Mon Sep 17 00:00:00 2001
From: Jean-Baptiste Bayle <j2b.bayle@gmail.com>
Date: Mon, 1 Feb 2021 21:51:22 +0100
Subject: [PATCH] Remove unnecessary function in dsp.py
---
lisainstrument/dsp.py | 85 ++-----------------------------------------
1 file changed, 3 insertions(+), 82 deletions(-)
diff --git a/lisainstrument/dsp.py b/lisainstrument/dsp.py
index c3506ae..18791c1 100644
--- a/lisainstrument/dsp.py
+++ b/lisainstrument/dsp.py
@@ -95,85 +95,6 @@ def lagrange_coeffs(eps, num_tabs, p): # pylint: disable=invalid-name
return coeffs.T
-def time_shift(data, shift, order=31):
- """Shift data in time by tau
-
- Args:
- data (numpy.ndarray): data to be shited
- tau (numpy.ndarray): amount of time each data point is to be shifted
- (must be of same dimension as data)
- fs (double): sampling frequency in (Hz)
- order (int): interpolation order
- """
- logging.debug("Time shifting data '%s' by '%s' (order=%d)", data, shift, order)
-
- if numpy.isscalar(data):
- logging.debug("Data is a constant scalar and cannot be time shifted")
- return data
- if numpy.all(shift == 0):
- logging.debug("Time shift is vanishing and cannot be applied")
- return data
-
- data = numpy.asarray(data)
- mode = "timeseries" if isinstance(shift, numpy.ndarray) else "constant"
- logging.debug("Using mode '%s'", mode)
-
- if mode == "timeseries" and data.size != shift.size:
- raise ValueError(f"`data` and `tau` must be of the same size (got {data.size}, {shift.size})")
-
- num_tabs = order + 1
- p = num_tabs // 2 # pylint: disable=invalid-name
- size = data.size
-
- def lagrange_coeffs(eps):
- """Calculate coefficients for lagrange interpolation"""
- coeffs = numpy.zeros([num_tabs, eps.size])
-
- if p > 1:
- factor = numpy.ones(eps.size, dtype=numpy.float64)
- factor *= eps * (1 - eps)
-
- for j in range(1, p):
- factor *= (-1) * (1 - j / p) / (1 + j / p)
- coeffs[p - 1 - j] = factor / (j + eps)
- coeffs[p + j] = factor / (j + 1 - eps)
-
- coeffs[p - 1] = 1 - eps
- coeffs[p] = eps
-
- for j in range(2, p):
- coeffs *= 1 - (eps / j)**2
-
- coeffs *= (1 + eps) * (1 - eps / p)
- else:
- coeffs[p - 1] = 1 - eps
- coeffs[p] = eps
-
- return coeffs.T
-
- k = numpy.floor(shift).astype(int)
- eps = shift - k
- coeffs = lagrange_coeffs(eps)
- logging.debug("Using Lagrange coefficiens '%s'", coeffs)
-
- if mode == "timeseries":
- logging.debug("Computing Lagrange matrix")
- indices = numpy.arange(size)
- i_min = numpy.min(k - (p - 1) + indices)
- i_max = numpy.max(k + p + indices + 1)
- csr_ind = numpy.tile(numpy.arange(num_tabs), size) + numpy.repeat(k + indices, num_tabs) - (p - 1)
- csr_ptr = num_tabs * numpy.arange(size + 1)
- mat = scipy.sparse.csr_matrix((numpy.ravel(coeffs), csr_ind - i_min, csr_ptr), shape=(size, i_max - i_min))
- logging.debug("Padding data (left=%d, right=%d)", max(0, -i_min), max(0, i_max - size))
- data_padded = numpy.pad(data[max(0, i_min):min(size, i_max)], (max(0, -i_min), max(0, i_max - size)))
- logging.debug("Computing matrix-vector product")
- shifted = mat.dot(data_padded)
- elif mode == "constant":
- i_min = k - (p - 1)
- i_max = k + p + size
- logging.debug("Padding data (left=%d, right=%d)", max(0, -i_min), max(0, i_max - size))
- data_padded = numpy.pad(data[max(0, i_min):min(size, i_max)], (max(0, -i_min), max(0, i_max - size)))
- logging.debug("Computing correlation product")
- shifted = numpy.correlate(data_padded, coeffs[0], mode="valid")
-
- return shifted
+def identity(x):
+ """Identity function, no operations."""
+ return x
--
GitLab