lalsuite 7.26.2.dev20251218__cp314-cp314t-manylinux_2_28_x86_64.whl

lalpulsar/piecewise_model/estimating_knots.py

@@ -0,0 +1,833 @@
+# Copyright (C) 2019--2023 Benjamin Grace
+#
+# This program is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by the
+# Free Software Foundation; either version 2 of the License, or (at your
+# option) any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General
+# Public License for more details.
+#
+# You should have received a copy of the GNU General Public License along
+# with this program; if not, write to the Free Software Foundation, Inc.,
+# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+
+## \file
+## \ingroup lalpulsar_python_piecewise_model
+"""
+Construct a list of knots which gives the maximum allowed spacing between
+them while maintaining the desired accuracy.
+"""
+
+import ast
+import logging
+import os.path
+
+import numpy as np
+
+from . import basis_functions as bf
+from . import mols_for_gte as mols
+from . import errors
+from . import semicoherent_metric_methods as scmm
+
+knotarchivefile = None
+
+
+def setknotarchivepath(path):
+    global knotarchivefile
+    knotarchivefile = os.path.join(path, "KnotArchive")
+
+
+# In this module, we use the function, max(|f_GTE(t) - F_PP(t)|) - Delta f_i0,
+# to choose the spacing of our knots. The Delta f_i0 is the spacing between the frequency parameter given by the
+# metric, it is equal to sqrt(mu/g_ii). Ideally, we wish for the maximum difference between our piecewise model and
+# the GTE to be minimised and smaller than the spacing of our frequency parameters. When finding the maximum
+# difference between the GTE and our model, max(|f_GTE(t) - F_PP(t)|), we maximise over the value t which falls into
+# some interval [p_i, p_i+1]. If we make interval sizes small the difference will go to zero, however this would
+# dramatically increase the number of intervals we need, and hence most likely increase the template bank size.
+# However, if we make them too large the error becomes greater than our parameter spacing. As such, we wish to
+# find the values [p_i, p_i+1] such that the given function above is zero. This is the longest we can make a segment
+# before the error becomes greater than our parameter spacing. We start by assuming p_0 = 0 and then inductively
+# calculate all of the following knots.
+
+# Working in this module it is important that the knots function, p(i, ints, dur) in the basis_functions file is
+# written correctly. See that file for how it should be defined when using this file. As well as this, this module
+# requires methods written in the mols_for_gte, basis_functions and semicoherent_metric_methods files, all of which rely
+# on the p(i, ints, dur) method. This can be a problem, as for the methods in this module to work, p(i, ints,
+# dur) should simply extract an element from a 'knots list' that we build inductively in this module. Hence,
+# using the knots generated in this module can at times be tricky. The recommendation for using knots built by this
+# file is to either run this file for given model parameter and then use the output as the knots by copying and
+# pasting the output into the basis_functions file to be used as all knots, or in the module you wish to run, run
+# the GlobalVariableDeclarations module which initialises our knot list (by importing this module and using the
+# methods below) and then import the GlobalVariableDeclarations module into the module you wish to run. As an
+# example, if you want to run module A which requires knots, import the GlobalVariableDeclarations module into A,
+# making sure that in the GlobalVariableDecalartions module the knots list is initialised with one of the methods
+# here.
+#
+# Knots generated in this module are saved to the KnotArchive text file.
+
+# Another word of caution, some methods in this module are outdated, being replaced by more efficient methods later
+# in the module. These methods typically run much slower than those which they are replaced by. Further more, many of
+# these methods do not make use of the KnotArchive file which stores all generated values of knots. I have tried to
+# note these outdated methods were appropriate and add an explanation to why they have been rewritten further in the
+# module. They are kept in this module however for completeness and a reference in case they are needed in the
+# future
+
+# When using this module be sure to check that the method p and the knotslist parameter in the basis_functions module
+# are defined correctly or are being properly updated.
+
+# List that we append the maximum parameter spacing for each segment we build knots for in this module. Kept only for
+# inspection, not used in any methods except for printing.
+parameterspacings = []
+
+# Returns the parameter spacing associated with the zeroth derivative at the lth knot. Note, if the knotslist variable
+# in the basis_functions is never updated, then this method will only work for l = 0
+def metricelementspacingonl(l, coeffs, mu):
+
+    metvalue = scmm.metricelementonl(l, l, l, 1, 1, 0, 0, coeffs)
+
+    stepsize = np.sqrt(mu / metvalue)
+
+    return stepsize
+
+
+# This method is an outdated method, replaced later in this module. Chronologiclly this was replaced first by the
+# knotatleff method and then finally by the knotatleffusinggivenknotlist method. No other methods in this module use
+# this method.
+#
+# Calcualtes the time (and hence knot) at which the function, max(|f_GTE(t) - F_PP(t)| - Delta f_i0 is zero. This is
+# achieved by sampling all times t between durmin and durmax and seeing if the appropriate value of that function is
+# within a given tolerance of zero. If not, we reuse this method with an extended duration of durmax to 2 * durmax. This
+# has the obvious flaw that if the correct knot value is below durmin, then this method will not be able to find it.
+# Because this method does not make use of faster algorithms, such as the bisection method which we use later on, it
+# runs very slowly.
+def knotatl(l, s, durmin, durmax, steps, f0, ngte, kgte, mu):
+    errorlists = []
+    paramspacinglist = []
+
+    durlist = np.linspace(durmin, durmax, steps)
+
+    for dur in durlist:
+        basiscoeffs = bf.allcoeffs(s)
+        modelparams = mols.solsbyint(mols.sols(basiscoeffs, 20, s, f0, ngte, kgte), s)
+        paramspacinglist.append(metricelementspacingonl(l, basiscoeffs, mu))
+
+        suberrorlist = []
+
+        for t in np.linspace(durmin, dur, steps):
+            suberrorlist.append(
+                mols.errorvalueatpoint(t, basiscoeffs, modelparams, f0, ngte, kgte)
+            )
+        errorlists.append(suberrorlist)
+
+    for i, dur in enumerate(durlist):
+
+        if np.max(errorlists[i]) - paramspacinglist[i] > 0:
+            return dur
+
+    logging.debug("No root found for specified duration, increasing duration span")
+    return knotatl(l, s, durmax, 2 * durmax, steps, f0, ngte, kgte)
+
+
+# This method is not used in our finalised knot generating method. It is however used in the buildcompleteknotlist
+# method. It was kept because the buildcompelteknotlist method (and this method) do not read or write to any files.
+# As such, if we want to do some experimentation with how we build our knots but not interfere and risk altering our
+# library of knots we can use these methods instead.
+#
+#
+# The same as the knotatl method but now done more efficiently by using a bisection method. Instead of sampling all
+# possible times/knots, we instead sample two possible durations, negdur and posdur. These knots should be chosen
+# such that negdur results in the function, max(|f_GTE(t) - F_PP(t)| - Delta f_i0, being negative and posdur such
+# that it is positive. We then select a time inbetween these two possible choices and calculated that function value.
+# If the value is within a given tolerance of zero we return that time as the new knot value. If not, we recursively use
+# this method by replacing negdur or posdur with this 'inbetween time' in accordance with the bisection method until we
+# find a time that is close enough to zero for us to be confident we can use that as the new knot value. This method
+# has a maximum recursion depth of 10.
+#
+# This method is somewhat sensitive to the initial choices of negdur and posdur. The checkfirst parameter is used
+# such that negdur and posdur have appropriate values. We must check that negdur and posdur result in negative and
+# positive values of the diff function respectively. In certain cases the ill-conditionedness of the basis functions
+# we calculate also influences our results. We check for this too in certain cases.
+#
+# The checkfirst value should always be used when knotatleff is first called and set to True. Once this check has
+# initially been cleared we do not need to check the negdur and posdur values again.
+#
+# In this method we refer to a boolean value fullMOLS, it is defined below.
+def knotatleff(
+    l,
+    s,
+    negdur,
+    posdur,
+    steps,
+    f0,
+    ngte,
+    kgte,
+    mu,
+    checkfirst=False,
+    recursiondepth=0,
+    ps=None,
+    negcounter=0,
+):
+    # logging.debug("Current recursion depth: " + str(counter))
+
+    knotnuma = len(bf.knotslist) - 1
+    knotnumb = knotnuma + 1
+
+    # As negdur is a candidate knot, it should be greater than the last knot we currently have in bf.knotslist.
+    if bf.knotslist[l] >= negdur:
+        logging.debug(
+            "negdur parameter not within appropriate range, adjusting with larger value"
+        )
+        logging.debug(
+            "negdur = " + str(negdur) + ", previous knot = " + str(bf.knotslist[l])
+        )
+        return knotatleff(
+            l, s, bf.knotslist[l] + 1, posdur, steps, f0, ngte, kgte, mu, True, ps=ps
+        )
+
+    if checkfirst:
+        bf.knotslist.append(negdur)
+        basiscoeffsneg = bf.allcoeffs(s)
+
+        # Can't remember why but this sometimes throws a TypeError. I suspect it probably happens when negdur is too
+        # large and the sample points in the sampling_methods module are chosen poorly. Corrected for by using a
+        # smaller negdur value. Don't ever remember seeing this error for the posdur parameter though
+        try:
+            # logging.debug("Prev knot: " + str(ps))
+            if fullMOLS:
+                modelparamsneg = mols.solsbyint(
+                    mols.sols(basiscoeffsneg, 20, s, f0, ngte, kgte), s
+                )
+            else:
+                modelparamsneg = mols.solsbyint(
+                    mols.solsbetweenknots(
+                        knotnuma, knotnumb, basiscoeffsneg, 20, s, f0, ngte, kgte
+                    ),
+                    s,
+                )
+        except TypeError as error:
+            logging.debug(error)
+            logging.debug("Reducing negdur duration")
+            diff = negdur - ps
+            bf.knotslist.pop()
+            return knotatleff(
+                l, s, ps + diff / 10, posdur, steps, f0, ngte, kgte, mu, True, 0, ps=ps
+            )
+
+        # Calculating the diff function value for negdur and posdur
+        paramspacingneg = metricelementspacingonl(l, basiscoeffsneg, mu)
+
+        errorlistneg = []
+
+        for t in np.linspace(bf.knotslist[-1], negdur, steps):
+            errorlistneg.append(
+                mols.errorvalueatpoint(
+                    t, basiscoeffsneg, modelparamsneg, f0, ngte, kgte
+                )
+            )
+
+        bf.knotslist.pop()
+
+        bf.knotslist.append(posdur)
+        basiscoeffspos = bf.allcoeffs(s)
+        if fullMOLS:
+            modelparamspos = mols.solsbyint(
+                mols.sols(basiscoeffspos, 20, s, f0, ngte, kgte), s
+            )
+        else:
+            modelparamspos = mols.solsbyint(
+                mols.solsbetweenknots(
+                    knotnuma, knotnumb, basiscoeffspos, 20, s, f0, ngte, kgte
+                ),
+                s,
+            )
+
+        paramspacingpos = metricelementspacingonl(l, basiscoeffspos, mu)
+
+        errorlistpos = []
+
+        for t in np.linspace(bf.knotslist[-1], posdur, steps):
+            errorlistpos.append(
+                mols.errorvalueatpoint(
+                    t, basiscoeffspos, modelparamspos, f0, ngte, kgte
+                )
+            )
+
+        bf.knotslist.pop()
+
+        maxerrorneg = np.max(errorlistneg)
+        maxerrorpos = np.max(errorlistpos)
+
+        logging.debug("Neg and pos durs: " + str([negdur, posdur]))
+        logging.debug("Max errors: " + str([maxerrorneg, maxerrorpos]))
+        logging.debug("Metric values: " + str([paramspacingneg, paramspacingpos]))
+        logging.debug()
+
+        diffvalneg = maxerrorneg - paramspacingneg
+        diffvalpos = maxerrorpos - paramspacingpos
+
+        # Checking if posdur and negdur have the appropriate signs for the bisection method and readjusting if not
+        if not (diffvalneg < 0 and diffvalpos > 0):
+            # logging.debug("Diff vals are: " + str([diffvalneg, diffvalpos]))
+            if diffvalpos < 0:
+                # logging.debug("Negative pos val")
+                diff = posdur - ps
+                return knotatleff(
+                    l, s, negdur, ps + 2 * diff, steps, f0, ngte, kgte, mu, True, ps=ps
+                )
+            if diffvalneg > 0:
+                # logging.debug("Positive neg val")
+                diff = negdur - ps
+
+                return knotatleff(
+                    l,
+                    s,
+                    ps + diff / 10,
+                    posdur,
+                    steps,
+                    f0,
+                    ngte,
+                    kgte,
+                    mu,
+                    True,
+                    ps=ps,
+                    negcounter=negcounter + 1,
+                )
+
+    # Calculating the diff function value for the time between negdur and posdur
+    halfdur = (posdur + negdur) / 2
+    bf.knotslist.append(halfdur)
+    basiscoeffs = bf.allcoeffs(s)
+    if fullMOLS:
+        modelparams = mols.solsbyint(mols.sols(basiscoeffs, 20, s, f0, ngte, kgte), s)
+    else:
+        modelparams = mols.solsbyint(
+            mols.solsbetweenknots(
+                knotnuma, knotnumb, basiscoeffs, 20, s, f0, ngte, kgte
+            ),
+            s,
+        )
+    paramspacing = metricelementspacingonl(l, basiscoeffs, mu)
+
+    errorlist = []
+
+    for t in np.linspace(bf.knotslist[-1], halfdur, steps):
+        errorlist.append(
+            mols.errorvalueatpoint(t, basiscoeffs, modelparams, f0, ngte, kgte)
+        )
+
+    bf.knotslist.pop()
+
+    maxerror = np.max(errorlist)
+
+    diffval = maxerror - paramspacing
+
+    # Either return halfdur as our new knot value or recursing
+    if -(2**-10) < diffval <= 0:
+        parameterspacings.append(paramspacing)
+        return halfdur
+    elif recursiondepth > 10:
+        logging.debug(
+            "Recursion depth of 10 reached, terminating and returning current knot value"
+        )
+        logging.debug("Function value for this knot is " + str(diffval))
+        parameterspacings.append(paramspacing)
+        return halfdur
+    else:
+        if diffval < 0:
+            return knotatleff(
+                l,
+                s,
+                halfdur,
+                posdur,
+                steps,
+                f0,
+                ngte,
+                kgte,
+                mu,
+                False,
+                recursiondepth + 1,
+            )
+        elif diffval > 0:
+            return knotatleff(
+                l,
+                s,
+                negdur,
+                halfdur,
+                steps,
+                f0,
+                ngte,
+                kgte,
+                mu,
+                False,
+                recursiondepth + 1,
+            )
+
+
+# Same as the knotatleff method but instead of refering to the bf.knotslist parameter we instead include this list of
+# previous knots as the parameter knotslist. The motivation behind this is that previous methods used to generate
+# knots had to start from an initial knot p_0 = 0. This was inconvenient if we instead wanted to calculate p_100 and
+# had already calculated p_99. Instead we carry all generated knots as a parameter in this function. We of course do
+# not need to carry the entire knot list for the model specifications, only the last knot in that list,
+# but we do this anyway.
+#
+# The logic in this method is identical to that of the knotatleff method with only the addition of the knotslist
+# parameter. Read that method's description for more details about the process of this method. This method has
+# a maximum recursion depth of 10.
+#
+# The fullMOLS parameter allows us to decide when considering the worse case scenario whether we want to consider the
+# accuracy of the entire MOLS solution or just the MOLS solution which belongs to the segment we are currently trying
+# to determine the knot for. Typically, if we consider the MOLS solution for the full signal, the knots are spread
+# further apart.
+fullMOLS = True
+
+
+def knotatleffusinggivenknotlist(
+    l,
+    s,
+    negdur,
+    posdur,
+    steps,
+    f0,
+    ngte,
+    kgte,
+    knotslist,
+    mu,
+    checkfirst=False,
+    counter=0,
+    ps=None,
+    negcounter=0,
+    prevdiffvalneg=np.inf,
+):
+    # logging.debug("Current recursion depth: " + str(counter))
+
+    knotnuma = len(knotslist) - 1
+    knotnumb = knotnuma + 1
+
+    # As negdur is a candidate knot, it should be greater than the last knot we currently have in bf.knotslist.
+    bf.knotslist = knotslist
+    if knotslist[l] >= negdur:
+        logging.debug(
+            "negdur parameter not within appropriate range, adjusting with larger value"
+        )
+        logging.debug(
+            "negdur = " + str(negdur) + ", previous knot = " + str(knotslist[l])
+        )
+        return knotatleffusinggivenknotlist(
+            l,
+            s,
+            knotslist[l] + 1,
+            posdur,
+            steps,
+            f0,
+            ngte,
+            kgte,
+            knotslist,
+            mu,
+            True,
+            ps=ps,
+        )
+
+    if checkfirst:
+        bf.knotslist.append(negdur)
+        basiscoeffsneg = bf.allcoeffs(s)
+
+        # Can't remember why but this sometimes throws a TypeError. I suspect it probably happens when negdur is too
+        # large and the sample points in the sampling_methods module are chosen poorly. Corrected for by using a
+        # smaller negdur value. Don't ever remember seeing this error for the posdur parameter though
+        try:
+            if fullMOLS:
+                modelparamsneg = mols.solsbyint(
+                    mols.sols(basiscoeffsneg, 20, s, f0, ngte, kgte), s
+                )
+            else:
+                modelparamsneg = mols.solsbyint(
+                    mols.solsbetweenknots(
+                        knotnuma, knotnumb, basiscoeffsneg, 20, s, f0, ngte, kgte
+                    ),
+                    s,
+                )
+        except TypeError as error:
+            logging.debug(error)
+            logging.debug("Reducing negdur duration")
+            bf.knotslist.pop()
+            diff = negdur - ps
+            return knotatleffusinggivenknotlist(
+                l,
+                s,
+                ps + diff / 10,
+                posdur,
+                steps,
+                f0,
+                ngte,
+                kgte,
+                knotslist,
+                mu,
+                True,
+                0,
+                ps=ps,
+            )
+
+        # Calculating the diff function value for negdur and posdur
+        paramspacingneg = metricelementspacingonl(l, basiscoeffsneg, mu)
+
+        errorlistneg = []
+
+        for t in np.linspace(knotslist[-1], negdur, steps):
+            errorlistneg.append(
+                mols.errorvalueatpoint(
+                    t, basiscoeffsneg, modelparamsneg, f0, ngte, kgte
+                )
+            )
+
+        bf.knotslist.pop()
+
+        bf.knotslist.append(posdur)
+        basiscoeffspos = bf.allcoeffs(s)
+
+        if fullMOLS:
+            modelparamspos = mols.solsbyint(
+                mols.sols(basiscoeffspos, 20, s, f0, ngte, kgte), s
+            )
+        else:
+            modelparamspos = mols.solsbyint(
+                mols.solsbetweenknots(
+                    knotnuma, knotnumb, basiscoeffspos, 20, s, f0, ngte, kgte
+                ),
+                s,
+            )
+
+        paramspacingpos = metricelementspacingonl(l, basiscoeffspos, mu)
+
+        errorlistpos = []
+
+        for t in np.linspace(knotslist[-1], posdur, steps):
+            errorlistpos.append(
+                mols.errorvalueatpoint(
+                    t, basiscoeffspos, modelparamspos, f0, ngte, kgte
+                )
+            )
+
+        bf.knotslist.pop()
+
+        maxerrorneg = np.max(errorlistneg)
+        maxerrorpos = np.max(errorlistpos)
+
+        diffvalneg = maxerrorneg - paramspacingneg
+        diffvalpos = maxerrorpos - paramspacingpos
+
+        # Checking if posdur and negdur have the appropriate signs for the bisection method and readjusting if not
+        if not (diffvalneg < 0 and diffvalpos > 0):
+            if diffvalpos < 0:
+                diff = posdur - ps
+                return knotatleffusinggivenknotlist(
+                    l,
+                    s,
+                    negdur,
+                    ps + 2 * diff,
+                    steps,
+                    f0,
+                    ngte,
+                    kgte,
+                    knotslist,
+                    mu,
+                    True,
+                    ps=ps,
+                    prevdiffvalneg=diffvalneg,
+                )
+            if diffvalneg > 0:
+                diff = negdur - ps
+
+                if diffvalneg > prevdiffvalneg:
+                    return knotatleffusinggivenknotlist(
+                        l,
+                        s,
+                        ps + 2 * diff,
+                        posdur,
+                        steps,
+                        f0,
+                        ngte,
+                        kgte,
+                        knotslist,
+                        mu,
+                        True,
+                        ps=ps,
+                        negcounter=negcounter + 1,
+                        prevdiffvalneg=diffvalneg,
+                    )
+
+                # if negcounter > 5 or (diff/10) < 1:
+                #    logging.debug("Too many negdur reductions")
+                #    logging.debug()
+                #    raise errors.TooManyNegdurReductions
+
+                return knotatleffusinggivenknotlist(
+                    l,
+                    s,
+                    ps + diff / 10,
+                    posdur,
+                    steps,
+                    f0,
+                    ngte,
+                    kgte,
+                    knotslist,
+                    mu,
+                    True,
+                    ps=ps,
+                    negcounter=negcounter + 1,
+                    prevdiffvalneg=diffvalneg,
+                )
+
+    # Calculating the diff function value for the time between negdur and posdur
+    halfdur = (posdur + negdur) / 2
+    bf.knotslist.append(halfdur)
+    basiscoeffs = bf.allcoeffs(s)
+
+    if fullMOLS:
+        modelparams = mols.solsbyint(mols.sols(basiscoeffs, 20, s, f0, ngte, kgte), s)
+    else:
+        modelparams = mols.solsbyint(
+            mols.solsbetweenknots(
+                knotnuma, knotnumb, basiscoeffs, 20, s, f0, ngte, kgte
+            ),
+            s,
+        )
+
+    paramspacing = metricelementspacingonl(l, basiscoeffs, mu)
+
+    errorlist = []
+
+    for t in np.linspace(knotslist[-1], halfdur, steps):
+        errorlist.append(
+            mols.errorvalueatpoint(t, basiscoeffs, modelparams, f0, ngte, kgte)
+        )
+    bf.knotslist.pop()
+    maxerror = np.max(errorlist)
+
+    diffval = maxerror - paramspacing
+
+    # Either return halfdur as our new knot value or recurse
+    if -(2**-10) < diffval <= 0:
+        parameterspacings.append(paramspacing)
+        return halfdur
+    elif counter > 10:
+        logging.debug(
+            "Recursion depth of 10 reached, terminating and returning current knot value"
+        )
+        logging.debug("Function value for this knot is " + str(diffval))
+        parameterspacings.append(paramspacing)
+        return halfdur
+    else:
+        if diffval < 0:
+            return knotatleffusinggivenknotlist(
+                l,
+                s,
+                halfdur,
+                posdur,
+                steps,
+                f0,
+                ngte,
+                kgte,
+                knotslist,
+                mu,
+                counter=counter + 1,
+            )
+        elif diffval > 0:
+            return knotatleffusinggivenknotlist(
+                l,
+                s,
+                negdur,
+                halfdur,
+                steps,
+                f0,
+                ngte,
+                kgte,
+                knotslist,
+                mu,
+                counter=counter + 1,
+            )
+
+
+# knotlist = [0, 2.107421875, 4.18438720703125]
+# ints = len(knotlist)
+# logging.debug(knotatleffusinggivenknotlist(ints - 1, 1, ints, 1, 10, 30, knotlist, checkfirst=True, ps=knotlist[-1]))
+
+# Calculates and adds the next knot to the KnotArchive file with the appropriate specifications by using the
+# knotatleffusinggivenknotlist method
+def addnextknottofile(f0, nmax, kgte, s, mu, steps=30):
+    spindownspecification = [f0, nmax, kgte, s, mu, fullMOLS]
+
+    knotarchive = open(knotarchivefile, "r")
+    alllines = knotarchive.readlines()
+    knotarchive.close()
+
+    foundspecifications = False
+    newknot = 0
+
+    for i, line in enumerate(alllines):
+        thisline = ast.literal_eval(line)
+
+        if thisline[0] == spindownspecification:
+            foundspecifications = True
+
+            knotslist = thisline[1]
+            ps = thisline[1][-1]
+            ints = len(knotslist)
+
+            if ps == 0:
+                negdurincrement = 1.1
+                while True:
+                    try:
+                        newknot = knotatleffusinggivenknotlist(
+                            ints - 1,
+                            s,
+                            negdurincrement,
+                            10 * negdurincrement,
+                            steps,
+                            f0,
+                            nmax,
+                            kgte,
+                            knotslist,
+                            mu,
+                            True,
+                            ps=ps,
+                        )
+
+                        break
+                    except errors.TooManyNegdurReductions:
+                        negdurincrement *= 10
+            else:
+                negdurincrement = 0
+                while True:
+                    ordmag = np.floor(np.log10(ps)) + negdurincrement
+                    try:
+                        newknot = knotatleffusinggivenknotlist(
+                            ints - 1,
+                            s,
+                            ps + 10**ordmag,
+                            ps + 10 ** (ordmag + 0.5),
+                            steps,
+                            f0,
+                            nmax,
+                            kgte,
+                            knotslist,
+                            mu,
+                            True,
+                            ps=ps,
+                        )
+                        break
+                    except errors.TooManyNegdurReductions:
+                        negdurincrement += 0.5
+            thisline[1].append(newknot)
+            updatedline = str(thisline) + "\n"
+            alllines[i] = updatedline
+            logging.info("Knotnum and knot val: " + str(ints) + ", " + str(newknot))
+            break
+
+    if foundspecifications:
+        knotarchive = open(knotarchivefile, "w")
+        knotarchive.writelines(alllines)
+        knotarchive.close()
+    else:
+        newline = [spindownspecification, [0]]
+
+        knotarchive = open(knotarchivefile, "a")
+        knotarchive.write(str(newline) + "\n")
+        knotarchive.close()
+
+        addnextknottofile(f0, nmax, kgte, s, mu, steps=steps)
+
+    return newknot
+
+
+# Calculates and writes all knots up the knotnum th knot to the KnotArchive file. E.g. If we want to know the first 10
+# knots for certain model specifications, if we run this method with knotnum = 10, if the KnotArchive file does not
+# already have those 10 knot values, this method will calculate and add those knot values to that file until all knots
+# up to the 10th knot are present
+def addknottoknotnum(s, knotnum, f0, nmax, kgte, mu, steps=30):
+    spindownspecifications = [f0, nmax, kgte, s, mu, fullMOLS]
+
+    knotarchive = open(knotarchivefile, "r")
+    alllines = knotarchive.readlines()
+    knotarchive.close()
+
+    currentknotnum = 0
+
+    for i, line in enumerate(alllines):
+        thisline = ast.literal_eval(line)
+
+        if thisline[0] == spindownspecifications:
+            currentknotnum = len(thisline[1]) - 1
+            break
+
+    while currentknotnum <= knotnum:
+        addnextknottofile(f0, nmax, kgte, s, mu, steps=steps)
+        currentknotnum += 1
+
+
+# Ad the above method but now instead of calculating up to a given knotnumber instead calculates all knots up to a given
+# signal duration.
+def addknottodur(s, dur, f0, nmax, kgte, mu, steps=30):
+    spindownspecifications = [f0, nmax, kgte, s, mu, fullMOLS]
+
+    knotarchive = open(knotarchivefile, "r")
+    alllines = knotarchive.readlines()
+    knotarchive.close()
+
+    currentdur = 0
+
+    for i, line in enumerate(alllines):
+        thisline = ast.literal_eval(line)
+
+        if thisline[0] == spindownspecifications:
+            currentdur = thisline[1][-1]
+            break
+
+    while currentdur < dur:
+        currentdur = addnextknottofile(f0, nmax, kgte, s, mu, steps=steps)
+
+
+# Calculates and returns all knots either up to the time dur or given knot number from the KnotArchive file. By using
+# this method we no longer need to always recalculate our knots, hopefully speeding up some of our other modules.
+def allidealisedknots(s, dur, steps, f0, nmax, kgte, mu, knotnum=0):
+    spindownspecifications = [f0, nmax, kgte, s, mu, fullMOLS]
+
+    if os.path.exists(knotarchivefile):
+        pass
+    else:
+        with open(knotarchivefile, "w") as _:
+            pass
+
+    if knotnum != 0:
+        addknottoknotnum(s, knotnum, f0, nmax, kgte, mu, steps=steps)
+    else:
+        addknottodur(s, dur, f0, nmax, kgte, mu, steps=steps)
+
+    knotarchive = open(knotarchivefile, "r")
+    alllines = knotarchive.readlines()
+    knotarchive.close()
+
+    for i, line in enumerate(alllines):
+        thisline = ast.literal_eval(line)
+
+        if thisline[0] == spindownspecifications:
+            bf.knotslist = thisline[1]
+
+            knotsbelowdur = []
+
+            if knotnum != 0:
+                knotsbelowdur = thisline[1][0 : knotnum + 1]
+            else:
+                # In case the knots stored in KnotArchive go well past the specified duration, we make sure we only set the knots to go to the duration we require
+                for i, knot in enumerate(bf.knotslist):
+                    if knot < dur:
+                        knotsbelowdur.append(knot)
+                    else:
+                        knotsbelowdur.append(dur)
+                        break
+
+            bf.knotslist = knotsbelowdur
+
+            return bf.knotslist