pyerrors.correlators

   1import warnings
   2from itertools import permutations
   3import numpy as np
   4import autograd.numpy as anp
   5import matplotlib.pyplot as plt
   6import scipy.linalg
   7from .obs import Obs, reweight, correlate, CObs
   8from .misc import dump_object, _assert_equal_properties
   9from .fits import least_squares
  10from .roots import find_root
  11
  12
  13class Corr:
  14    """The class for a correlator (time dependent sequence of pe.Obs).
  15
  16    Everything, this class does, can be achieved using lists or arrays of Obs.
  17    But it is simply more convenient to have a dedicated object for correlators.
  18    One often wants to add or multiply correlators of the same length at every timeslice and it is inconvenient
  19    to iterate over all timeslices for every operation. This is especially true, when dealing with matrices.
  20
  21    The correlator can have two types of content: An Obs at every timeslice OR a GEVP
  22    matrix at every timeslice. Other dependency (eg. spatial) are not supported.
  23
  24    """
  25
  26    __slots__ = ["content", "N", "T", "tag", "prange"]
  27
  28    def __init__(self, data_input, padding=[0, 0], prange=None):
  29        """ Initialize a Corr object.
  30
  31        Parameters
  32        ----------
  33        data_input : list or array
  34            list of Obs or list of arrays of Obs or array of Corrs
  35        padding : list, optional
  36            List with two entries where the first labels the padding
  37            at the front of the correlator and the second the padding
  38            at the back.
  39        prange : list, optional
  40            List containing the first and last timeslice of the plateau
  41            region indentified for this correlator.
  42        """
  43
  44        if isinstance(data_input, np.ndarray):
  45
  46            # This only works, if the array fulfills the conditions below
  47            if not len(data_input.shape) == 2 and data_input.shape[0] == data_input.shape[1]:
  48                raise Exception("Incompatible array shape")
  49            if not all([isinstance(item, Corr) for item in data_input.flatten()]):
  50                raise Exception("If the input is an array, its elements must be of type pe.Corr")
  51            if not all([item.N == 1 for item in data_input.flatten()]):
  52                raise Exception("Can only construct matrix correlator from single valued correlators")
  53            if not len(set([item.T for item in data_input.flatten()])) == 1:
  54                raise Exception("All input Correlators must be defined over the same timeslices.")
  55
  56            T = data_input[0, 0].T
  57            N = data_input.shape[0]
  58            input_as_list = []
  59            for t in range(T):
  60                if any([(item.content[t] is None) for item in data_input.flatten()]):
  61                    if not all([(item.content[t] is None) for item in data_input.flatten()]):
  62                        warnings.warn("Input ill-defined at different timeslices. Conversion leads to data loss!", RuntimeWarning)
  63                    input_as_list.append(None)
  64                else:
  65                    array_at_timeslace = np.empty([N, N], dtype="object")
  66                    for i in range(N):
  67                        for j in range(N):
  68                            array_at_timeslace[i, j] = data_input[i, j][t]
  69                    input_as_list.append(array_at_timeslace)
  70            data_input = input_as_list
  71
  72        if isinstance(data_input, list):
  73
  74            if all([isinstance(item, (Obs, CObs)) or item is None for item in data_input]):
  75                _assert_equal_properties([o for o in data_input if o is not None])
  76                self.content = [np.asarray([item]) if item is not None else None for item in data_input]
  77                self.N = 1
  78
  79            elif all([isinstance(item, np.ndarray) or item is None for item in data_input]) and any([isinstance(item, np.ndarray) for item in data_input]):
  80                self.content = data_input
  81                noNull = [a for a in self.content if not (a is None)]  # To check if the matrices are correct for all undefined elements
  82                self.N = noNull[0].shape[0]
  83                if self.N > 1 and noNull[0].shape[0] != noNull[0].shape[1]:
  84                    raise Exception("Smearing matrices are not NxN")
  85                if (not all([item.shape == noNull[0].shape for item in noNull])):
  86                    raise Exception("Items in data_input are not of identical shape." + str(noNull))
  87            else:
  88                raise Exception("data_input contains item of wrong type")
  89        else:
  90            raise Exception("Data input was not given as list or correct array")
  91
  92        self.tag = None
  93
  94        # An undefined timeslice is represented by the None object
  95        self.content = [None] * padding[0] + self.content + [None] * padding[1]
  96        self.T = len(self.content)
  97        self.prange = prange
  98
  99    def __getitem__(self, idx):
 100        """Return the content of timeslice idx"""
 101        if self.content[idx] is None:
 102            return None
 103        elif len(self.content[idx]) == 1:
 104            return self.content[idx][0]
 105        else:
 106            return self.content[idx]
 107
 108    @property
 109    def reweighted(self):
 110        bool_array = np.array([list(map(lambda x: x.reweighted, o)) for o in [x for x in self.content if x is not None]])
 111        if np.all(bool_array == 1):
 112            return True
 113        elif np.all(bool_array == 0):
 114            return False
 115        else:
 116            raise Exception("Reweighting status of correlator corrupted.")
 117
 118    def gamma_method(self, **kwargs):
 119        """Apply the gamma method to the content of the Corr."""
 120        for item in self.content:
 121            if not (item is None):
 122                if self.N == 1:
 123                    item[0].gamma_method(**kwargs)
 124                else:
 125                    for i in range(self.N):
 126                        for j in range(self.N):
 127                            item[i, j].gamma_method(**kwargs)
 128
 129    gm = gamma_method
 130
 131    def projected(self, vector_l=None, vector_r=None, normalize=False):
 132        """We need to project the Correlator with a Vector to get a single value at each timeslice.
 133
 134        The method can use one or two vectors.
 135        If two are specified it returns v1@G@v2 (the order might be very important.)
 136        By default it will return the lowest source, which usually means unsmeared-unsmeared (0,0), but it does not have to
 137        """
 138        if self.N == 1:
 139            raise Exception("Trying to project a Corr, that already has N=1.")
 140
 141        if vector_l is None:
 142            vector_l, vector_r = np.asarray([1.] + (self.N - 1) * [0.]), np.asarray([1.] + (self.N - 1) * [0.])
 143        elif (vector_r is None):
 144            vector_r = vector_l
 145        if isinstance(vector_l, list) and not isinstance(vector_r, list):
 146            if len(vector_l) != self.T:
 147                raise Exception("Length of vector list must be equal to T")
 148            vector_r = [vector_r] * self.T
 149        if isinstance(vector_r, list) and not isinstance(vector_l, list):
 150            if len(vector_r) != self.T:
 151                raise Exception("Length of vector list must be equal to T")
 152            vector_l = [vector_l] * self.T
 153
 154        if not isinstance(vector_l, list):
 155            if not vector_l.shape == vector_r.shape == (self.N,):
 156                raise Exception("Vectors are of wrong shape!")
 157            if normalize:
 158                vector_l, vector_r = vector_l / np.sqrt((vector_l @ vector_l)), vector_r / np.sqrt(vector_r @ vector_r)
 159            newcontent = [None if _check_for_none(self, item) else np.asarray([vector_l.T @ item @ vector_r]) for item in self.content]
 160
 161        else:
 162            # There are no checks here yet. There are so many possible scenarios, where this can go wrong.
 163            if normalize:
 164                for t in range(self.T):
 165                    vector_l[t], vector_r[t] = vector_l[t] / np.sqrt((vector_l[t] @ vector_l[t])), vector_r[t] / np.sqrt(vector_r[t] @ vector_r[t])
 166
 167            newcontent = [None if (_check_for_none(self, self.content[t]) or vector_l[t] is None or vector_r[t] is None) else np.asarray([vector_l[t].T @ self.content[t] @ vector_r[t]]) for t in range(self.T)]
 168        return Corr(newcontent)
 169
 170    def item(self, i, j):
 171        """Picks the element [i,j] from every matrix and returns a correlator containing one Obs per timeslice.
 172
 173        Parameters
 174        ----------
 175        i : int
 176            First index to be picked.
 177        j : int
 178            Second index to be picked.
 179        """
 180        if self.N == 1:
 181            raise Exception("Trying to pick item from projected Corr")
 182        newcontent = [None if (item is None) else item[i, j] for item in self.content]
 183        return Corr(newcontent)
 184
 185    def plottable(self):
 186        """Outputs the correlator in a plotable format.
 187
 188        Outputs three lists containing the timeslice index, the value on each
 189        timeslice and the error on each timeslice.
 190        """
 191        if self.N != 1:
 192            raise Exception("Can only make Corr[N=1] plottable")
 193        x_list = [x for x in range(self.T) if not self.content[x] is None]
 194        y_list = [y[0].value for y in self.content if y is not None]
 195        y_err_list = [y[0].dvalue for y in self.content if y is not None]
 196
 197        return x_list, y_list, y_err_list
 198
 199    def symmetric(self):
 200        """ Symmetrize the correlator around x0=0."""
 201        if self.N != 1:
 202            raise Exception('symmetric cannot be safely applied to multi-dimensional correlators.')
 203        if self.T % 2 != 0:
 204            raise Exception("Can not symmetrize odd T")
 205
 206        if self.content[0] is not None:
 207            if np.argmax(np.abs([o[0].value if o is not None else 0 for o in self.content])) != 0:
 208                warnings.warn("Correlator does not seem to be symmetric around x0=0.", RuntimeWarning)
 209
 210        newcontent = [self.content[0]]
 211        for t in range(1, self.T):
 212            if (self.content[t] is None) or (self.content[self.T - t] is None):
 213                newcontent.append(None)
 214            else:
 215                newcontent.append(0.5 * (self.content[t] + self.content[self.T - t]))
 216        if (all([x is None for x in newcontent])):
 217            raise Exception("Corr could not be symmetrized: No redundant values")
 218        return Corr(newcontent, prange=self.prange)
 219
 220    def anti_symmetric(self):
 221        """Anti-symmetrize the correlator around x0=0."""
 222        if self.N != 1:
 223            raise Exception('anti_symmetric cannot be safely applied to multi-dimensional correlators.')
 224        if self.T % 2 != 0:
 225            raise Exception("Can not symmetrize odd T")
 226
 227        test = 1 * self
 228        test.gamma_method()
 229        if not all([o.is_zero_within_error(3) for o in test.content[0]]):
 230            warnings.warn("Correlator does not seem to be anti-symmetric around x0=0.", RuntimeWarning)
 231
 232        newcontent = [self.content[0]]
 233        for t in range(1, self.T):
 234            if (self.content[t] is None) or (self.content[self.T - t] is None):
 235                newcontent.append(None)
 236            else:
 237                newcontent.append(0.5 * (self.content[t] - self.content[self.T - t]))
 238        if (all([x is None for x in newcontent])):
 239            raise Exception("Corr could not be symmetrized: No redundant values")
 240        return Corr(newcontent, prange=self.prange)
 241
 242    def is_matrix_symmetric(self):
 243        """Checks whether a correlator matrices is symmetric on every timeslice."""
 244        if self.N == 1:
 245            raise Exception("Only works for correlator matrices.")
 246        for t in range(self.T):
 247            if self[t] is None:
 248                continue
 249            for i in range(self.N):
 250                for j in range(i + 1, self.N):
 251                    if self[t][i, j] is self[t][j, i]:
 252                        continue
 253                    if hash(self[t][i, j]) != hash(self[t][j, i]):
 254                        return False
 255        return True
 256
 257    def matrix_symmetric(self):
 258        """Symmetrizes the correlator matrices on every timeslice."""
 259        if self.N == 1:
 260            raise Exception("Trying to symmetrize a correlator matrix, that already has N=1.")
 261        if self.is_matrix_symmetric():
 262            return 1.0 * self
 263        else:
 264            transposed = [None if _check_for_none(self, G) else G.T for G in self.content]
 265            return 0.5 * (Corr(transposed) + self)
 266
 267    def GEVP(self, t0, ts=None, sort="Eigenvalue", **kwargs):
 268        r'''Solve the generalized eigenvalue problem on the correlator matrix and returns the corresponding eigenvectors.
 269
 270        The eigenvectors are sorted according to the descending eigenvalues, the zeroth eigenvector(s) correspond to the
 271        largest eigenvalue(s). The eigenvector(s) for the individual states can be accessed via slicing
 272        ```python
 273        C.GEVP(t0=2)[0]  # Ground state vector(s)
 274        C.GEVP(t0=2)[:3]  # Vectors for the lowest three states
 275        ```
 276
 277        Parameters
 278        ----------
 279        t0 : int
 280            The time t0 for the right hand side of the GEVP according to $G(t)v_i=\lambda_i G(t_0)v_i$
 281        ts : int
 282            fixed time $G(t_s)v_i=\lambda_i G(t_0)v_i$ if sort=None.
 283            If sort="Eigenvector" it gives a reference point for the sorting method.
 284        sort : string
 285            If this argument is set, a list of self.T vectors per state is returned. If it is set to None, only one vector is returned.
 286            - "Eigenvalue": The eigenvector is chosen according to which eigenvalue it belongs individually on every timeslice.
 287            - "Eigenvector": Use the method described in arXiv:2004.10472 to find the set of v(t) belonging to the state.
 288              The reference state is identified by its eigenvalue at $t=t_s$.
 289
 290        Other Parameters
 291        ----------------
 292        state : int
 293           Returns only the vector(s) for a specified state. The lowest state is zero.
 294        '''
 295
 296        if self.N == 1:
 297            raise Exception("GEVP methods only works on correlator matrices and not single correlators.")
 298        if ts is not None:
 299            if (ts <= t0):
 300                raise Exception("ts has to be larger than t0.")
 301
 302        if "sorted_list" in kwargs:
 303            warnings.warn("Argument 'sorted_list' is deprecated, use 'sort' instead.", DeprecationWarning)
 304            sort = kwargs.get("sorted_list")
 305
 306        if self.is_matrix_symmetric():
 307            symmetric_corr = self
 308        else:
 309            symmetric_corr = self.matrix_symmetric()
 310
 311        G0 = np.vectorize(lambda x: x.value)(symmetric_corr[t0])
 312        np.linalg.cholesky(G0)  # Check if matrix G0 is positive-semidefinite.
 313
 314        if sort is None:
 315            if (ts is None):
 316                raise Exception("ts is required if sort=None.")
 317            if (self.content[t0] is None) or (self.content[ts] is None):
 318                raise Exception("Corr not defined at t0/ts.")
 319            Gt = np.vectorize(lambda x: x.value)(symmetric_corr[ts])
 320            reordered_vecs = _GEVP_solver(Gt, G0)
 321
 322        elif sort in ["Eigenvalue", "Eigenvector"]:
 323            if sort == "Eigenvalue" and ts is not None:
 324                warnings.warn("ts has no effect when sorting by eigenvalue is chosen.", RuntimeWarning)
 325            all_vecs = [None] * (t0 + 1)
 326            for t in range(t0 + 1, self.T):
 327                try:
 328                    Gt = np.vectorize(lambda x: x.value)(symmetric_corr[t])
 329                    all_vecs.append(_GEVP_solver(Gt, G0))
 330                except Exception:
 331                    all_vecs.append(None)
 332            if sort == "Eigenvector":
 333                if ts is None:
 334                    raise Exception("ts is required for the Eigenvector sorting method.")
 335                all_vecs = _sort_vectors(all_vecs, ts)
 336
 337            reordered_vecs = [[v[s] if v is not None else None for v in all_vecs] for s in range(self.N)]
 338        else:
 339            raise Exception("Unkown value for 'sort'.")
 340
 341        if "state" in kwargs:
 342            return reordered_vecs[kwargs.get("state")]
 343        else:
 344            return reordered_vecs
 345
 346    def Eigenvalue(self, t0, ts=None, state=0, sort="Eigenvalue"):
 347        """Determines the eigenvalue of the GEVP by solving and projecting the correlator
 348
 349        Parameters
 350        ----------
 351        state : int
 352            The state one is interested in ordered by energy. The lowest state is zero.
 353
 354        All other parameters are identical to the ones of Corr.GEVP.
 355        """
 356        vec = self.GEVP(t0, ts=ts, sort=sort)[state]
 357        return self.projected(vec)
 358
 359    def Hankel(self, N, periodic=False):
 360        """Constructs an NxN Hankel matrix
 361
 362        C(t) c(t+1) ... c(t+n-1)
 363        C(t+1) c(t+2) ... c(t+n)
 364        .................
 365        C(t+(n-1)) c(t+n) ... c(t+2(n-1))
 366
 367        Parameters
 368        ----------
 369        N : int
 370            Dimension of the Hankel matrix
 371        periodic : bool, optional
 372            determines whether the matrix is extended periodically
 373        """
 374
 375        if self.N != 1:
 376            raise Exception("Multi-operator Prony not implemented!")
 377
 378        array = np.empty([N, N], dtype="object")
 379        new_content = []
 380        for t in range(self.T):
 381            new_content.append(array.copy())
 382
 383        def wrap(i):
 384            while i >= self.T:
 385                i -= self.T
 386            return i
 387
 388        for t in range(self.T):
 389            for i in range(N):
 390                for j in range(N):
 391                    if periodic:
 392                        new_content[t][i, j] = self.content[wrap(t + i + j)][0]
 393                    elif (t + i + j) >= self.T:
 394                        new_content[t] = None
 395                    else:
 396                        new_content[t][i, j] = self.content[t + i + j][0]
 397
 398        return Corr(new_content)
 399
 400    def roll(self, dt):
 401        """Periodically shift the correlator by dt timeslices
 402
 403        Parameters
 404        ----------
 405        dt : int
 406            number of timeslices
 407        """
 408        return Corr(list(np.roll(np.array(self.content, dtype=object), dt)))
 409
 410    def reverse(self):
 411        """Reverse the time ordering of the Corr"""
 412        return Corr(self.content[:: -1])
 413
 414    def thin(self, spacing=2, offset=0):
 415        """Thin out a correlator to suppress correlations
 416
 417        Parameters
 418        ----------
 419        spacing : int
 420            Keep only every 'spacing'th entry of the correlator
 421        offset : int
 422            Offset the equal spacing
 423        """
 424        new_content = []
 425        for t in range(self.T):
 426            if (offset + t) % spacing != 0:
 427                new_content.append(None)
 428            else:
 429                new_content.append(self.content[t])
 430        return Corr(new_content)
 431
 432    def correlate(self, partner):
 433        """Correlate the correlator with another correlator or Obs
 434
 435        Parameters
 436        ----------
 437        partner : Obs or Corr
 438            partner to correlate the correlator with.
 439            Can either be an Obs which is correlated with all entries of the
 440            correlator or a Corr of same length.
 441        """
 442        if self.N != 1:
 443            raise Exception("Only one-dimensional correlators can be safely correlated.")
 444        new_content = []
 445        for x0, t_slice in enumerate(self.content):
 446            if _check_for_none(self, t_slice):
 447                new_content.append(None)
 448            else:
 449                if isinstance(partner, Corr):
 450                    if _check_for_none(partner, partner.content[x0]):
 451                        new_content.append(None)
 452                    else:
 453                        new_content.append(np.array([correlate(o, partner.content[x0][0]) for o in t_slice]))
 454                elif isinstance(partner, Obs):  # Should this include CObs?
 455                    new_content.append(np.array([correlate(o, partner) for o in t_slice]))
 456                else:
 457                    raise Exception("Can only correlate with an Obs or a Corr.")
 458
 459        return Corr(new_content)
 460
 461    def reweight(self, weight, **kwargs):
 462        """Reweight the correlator.
 463
 464        Parameters
 465        ----------
 466        weight : Obs
 467            Reweighting factor. An Observable that has to be defined on a superset of the
 468            configurations in obs[i].idl for all i.
 469        all_configs : bool
 470            if True, the reweighted observables are normalized by the average of
 471            the reweighting factor on all configurations in weight.idl and not
 472            on the configurations in obs[i].idl.
 473        """
 474        if self.N != 1:
 475            raise Exception("Reweighting only implemented for one-dimensional correlators.")
 476        new_content = []
 477        for t_slice in self.content:
 478            if _check_for_none(self, t_slice):
 479                new_content.append(None)
 480            else:
 481                new_content.append(np.array(reweight(weight, t_slice, **kwargs)))
 482        return Corr(new_content)
 483
 484    def T_symmetry(self, partner, parity=+1):
 485        """Return the time symmetry average of the correlator and its partner
 486
 487        Parameters
 488        ----------
 489        partner : Corr
 490            Time symmetry partner of the Corr
 491        partity : int
 492            Parity quantum number of the correlator, can be +1 or -1
 493        """
 494        if self.N != 1:
 495            raise Exception("T_symmetry only implemented for one-dimensional correlators.")
 496        if not isinstance(partner, Corr):
 497            raise Exception("T partner has to be a Corr object.")
 498        if parity not in [+1, -1]:
 499            raise Exception("Parity has to be +1 or -1.")
 500        T_partner = parity * partner.reverse()
 501
 502        t_slices = []
 503        test = (self - T_partner)
 504        test.gamma_method()
 505        for x0, t_slice in enumerate(test.content):
 506            if t_slice is not None:
 507                if not t_slice[0].is_zero_within_error(5):
 508                    t_slices.append(x0)
 509        if t_slices:
 510            warnings.warn("T symmetry partners do not agree within 5 sigma on time slices " + str(t_slices) + ".", RuntimeWarning)
 511
 512        return (self + T_partner) / 2
 513
 514    def deriv(self, variant="symmetric"):
 515        """Return the first derivative of the correlator with respect to x0.
 516
 517        Parameters
 518        ----------
 519        variant : str
 520            decides which definition of the finite differences derivative is used.
 521            Available choice: symmetric, forward, backward, improved, log, default: symmetric
 522        """
 523        if self.N != 1:
 524            raise Exception("deriv only implemented for one-dimensional correlators.")
 525        if variant == "symmetric":
 526            newcontent = []
 527            for t in range(1, self.T - 1):
 528                if (self.content[t - 1] is None) or (self.content[t + 1] is None):
 529                    newcontent.append(None)
 530                else:
 531                    newcontent.append(0.5 * (self.content[t + 1] - self.content[t - 1]))
 532            if (all([x is None for x in newcontent])):
 533                raise Exception('Derivative is undefined at all timeslices')
 534            return Corr(newcontent, padding=[1, 1])
 535        elif variant == "forward":
 536            newcontent = []
 537            for t in range(self.T - 1):
 538                if (self.content[t] is None) or (self.content[t + 1] is None):
 539                    newcontent.append(None)
 540                else:
 541                    newcontent.append(self.content[t + 1] - self.content[t])
 542            if (all([x is None for x in newcontent])):
 543                raise Exception("Derivative is undefined at all timeslices")
 544            return Corr(newcontent, padding=[0, 1])
 545        elif variant == "backward":
 546            newcontent = []
 547            for t in range(1, self.T):
 548                if (self.content[t - 1] is None) or (self.content[t] is None):
 549                    newcontent.append(None)
 550                else:
 551                    newcontent.append(self.content[t] - self.content[t - 1])
 552            if (all([x is None for x in newcontent])):
 553                raise Exception("Derivative is undefined at all timeslices")
 554            return Corr(newcontent, padding=[1, 0])
 555        elif variant == "improved":
 556            newcontent = []
 557            for t in range(2, self.T - 2):
 558                if (self.content[t - 2] is None) or (self.content[t - 1] is None) or (self.content[t + 1] is None) or (self.content[t + 2] is None):
 559                    newcontent.append(None)
 560                else:
 561                    newcontent.append((1 / 12) * (self.content[t - 2] - 8 * self.content[t - 1] + 8 * self.content[t + 1] - self.content[t + 2]))
 562            if (all([x is None for x in newcontent])):
 563                raise Exception('Derivative is undefined at all timeslices')
 564            return Corr(newcontent, padding=[2, 2])
 565        elif variant == 'log':
 566            newcontent = []
 567            for t in range(self.T):
 568                if (self.content[t] is None) or (self.content[t] <= 0):
 569                    newcontent.append(None)
 570                else:
 571                    newcontent.append(np.log(self.content[t]))
 572            if (all([x is None for x in newcontent])):
 573                raise Exception("Log is undefined at all timeslices")
 574            logcorr = Corr(newcontent)
 575            return self * logcorr.deriv('symmetric')
 576        else:
 577            raise Exception("Unknown variant.")
 578
 579    def second_deriv(self, variant="symmetric"):
 580        """Return the second derivative of the correlator with respect to x0.
 581
 582        Parameters
 583        ----------
 584        variant : str
 585            decides which definition of the finite differences derivative is used.
 586            Available choice: symmetric, improved, log, default: symmetric
 587        """
 588        if self.N != 1:
 589            raise Exception("second_deriv only implemented for one-dimensional correlators.")
 590        if variant == "symmetric":
 591            newcontent = []
 592            for t in range(1, self.T - 1):
 593                if (self.content[t - 1] is None) or (self.content[t + 1] is None):
 594                    newcontent.append(None)
 595                else:
 596                    newcontent.append((self.content[t + 1] - 2 * self.content[t] + self.content[t - 1]))
 597            if (all([x is None for x in newcontent])):
 598                raise Exception("Derivative is undefined at all timeslices")
 599            return Corr(newcontent, padding=[1, 1])
 600        elif variant == "improved":
 601            newcontent = []
 602            for t in range(2, self.T - 2):
 603                if (self.content[t - 2] is None) or (self.content[t - 1] is None) or (self.content[t] is None) or (self.content[t + 1] is None) or (self.content[t + 2] is None):
 604                    newcontent.append(None)
 605                else:
 606                    newcontent.append((1 / 12) * (-self.content[t + 2] + 16 * self.content[t + 1] - 30 * self.content[t] + 16 * self.content[t - 1] - self.content[t - 2]))
 607            if (all([x is None for x in newcontent])):
 608                raise Exception("Derivative is undefined at all timeslices")
 609            return Corr(newcontent, padding=[2, 2])
 610        elif variant == 'log':
 611            newcontent = []
 612            for t in range(self.T):
 613                if (self.content[t] is None) or (self.content[t] <= 0):
 614                    newcontent.append(None)
 615                else:
 616                    newcontent.append(np.log(self.content[t]))
 617            if (all([x is None for x in newcontent])):
 618                raise Exception("Log is undefined at all timeslices")
 619            logcorr = Corr(newcontent)
 620            return self * (logcorr.second_deriv('symmetric') + (logcorr.deriv('symmetric'))**2)
 621        else:
 622            raise Exception("Unknown variant.")
 623
 624    def m_eff(self, variant='log', guess=1.0):
 625        """Returns the effective mass of the correlator as correlator object
 626
 627        Parameters
 628        ----------
 629        variant : str
 630            log : uses the standard effective mass log(C(t) / C(t+1))
 631            cosh, periodic : Use periodicitiy of the correlator by solving C(t) / C(t+1) = cosh(m * (t - T/2)) / cosh(m * (t + 1 - T/2)) for m.
 632            sinh : Use anti-periodicitiy of the correlator by solving C(t) / C(t+1) = sinh(m * (t - T/2)) / sinh(m * (t + 1 - T/2)) for m.
 633            See, e.g., arXiv:1205.5380
 634            arccosh : Uses the explicit form of the symmetrized correlator (not recommended)
 635            logsym: uses the symmetric effective mass log(C(t-1) / C(t+1))/2
 636        guess : float
 637            guess for the root finder, only relevant for the root variant
 638        """
 639        if self.N != 1:
 640            raise Exception('Correlator must be projected before getting m_eff')
 641        if variant == 'log':
 642            newcontent = []
 643            for t in range(self.T - 1):
 644                if ((self.content[t] is None) or (self.content[t + 1] is None)) or (self.content[t + 1][0].value == 0):
 645                    newcontent.append(None)
 646                elif self.content[t][0].value / self.content[t + 1][0].value < 0:
 647                    newcontent.append(None)
 648                else:
 649                    newcontent.append(self.content[t] / self.content[t + 1])
 650            if (all([x is None for x in newcontent])):
 651                raise Exception('m_eff is undefined at all timeslices')
 652
 653            return np.log(Corr(newcontent, padding=[0, 1]))
 654
 655        elif variant == 'logsym':
 656            newcontent = []
 657            for t in range(1, self.T - 1):
 658                if ((self.content[t - 1] is None) or (self.content[t + 1] is None)) or (self.content[t + 1][0].value == 0):
 659                    newcontent.append(None)
 660                elif self.content[t - 1][0].value / self.content[t + 1][0].value < 0:
 661                    newcontent.append(None)
 662                else:
 663                    newcontent.append(self.content[t - 1] / self.content[t + 1])
 664            if (all([x is None for x in newcontent])):
 665                raise Exception('m_eff is undefined at all timeslices')
 666
 667            return np.log(Corr(newcontent, padding=[1, 1])) / 2
 668
 669        elif variant in ['periodic', 'cosh', 'sinh']:
 670            if variant in ['periodic', 'cosh']:
 671                func = anp.cosh
 672            else:
 673                func = anp.sinh
 674
 675            def root_function(x, d):
 676                return func(x * (t - self.T / 2)) / func(x * (t + 1 - self.T / 2)) - d
 677
 678            newcontent = []
 679            for t in range(self.T - 1):
 680                if (self.content[t] is None) or (self.content[t + 1] is None) or (self.content[t + 1][0].value == 0):
 681                    newcontent.append(None)
 682                # Fill the two timeslices in the middle of the lattice with their predecessors
 683                elif variant == 'sinh' and t in [self.T / 2, self.T / 2 - 1]:
 684                    newcontent.append(newcontent[-1])
 685                elif self.content[t][0].value / self.content[t + 1][0].value < 0:
 686                    newcontent.append(None)
 687                else:
 688                    newcontent.append(np.abs(find_root(self.content[t][0] / self.content[t + 1][0], root_function, guess=guess)))
 689            if (all([x is None for x in newcontent])):
 690                raise Exception('m_eff is undefined at all timeslices')
 691
 692            return Corr(newcontent, padding=[0, 1])
 693
 694        elif variant == 'arccosh':
 695            newcontent = []
 696            for t in range(1, self.T - 1):
 697                if (self.content[t] is None) or (self.content[t + 1] is None) or (self.content[t - 1] is None) or (self.content[t][0].value == 0):
 698                    newcontent.append(None)
 699                else:
 700                    newcontent.append((self.content[t + 1] + self.content[t - 1]) / (2 * self.content[t]))
 701            if (all([x is None for x in newcontent])):
 702                raise Exception("m_eff is undefined at all timeslices")
 703            return np.arccosh(Corr(newcontent, padding=[1, 1]))
 704
 705        else:
 706            raise Exception('Unknown variant.')
 707
 708    def fit(self, function, fitrange=None, silent=False, **kwargs):
 709        r'''Fits function to the data
 710
 711        Parameters
 712        ----------
 713        function : obj
 714            function to fit to the data. See fits.least_squares for details.
 715        fitrange : list
 716            Two element list containing the timeslices on which the fit is supposed to start and stop.
 717            Caution: This range is inclusive as opposed to standard python indexing.
 718            `fitrange=[4, 6]` corresponds to the three entries 4, 5 and 6.
 719            If not specified, self.prange or all timeslices are used.
 720        silent : bool
 721            Decides whether output is printed to the standard output.
 722        '''
 723        if self.N != 1:
 724            raise Exception("Correlator must be projected before fitting")
 725
 726        if fitrange is None:
 727            if self.prange:
 728                fitrange = self.prange
 729            else:
 730                fitrange = [0, self.T - 1]
 731        else:
 732            if not isinstance(fitrange, list):
 733                raise Exception("fitrange has to be a list with two elements")
 734            if len(fitrange) != 2:
 735                raise Exception("fitrange has to have exactly two elements [fit_start, fit_stop]")
 736
 737        xs = np.array([x for x in range(fitrange[0], fitrange[1] + 1) if not self.content[x] is None])
 738        ys = np.array([self.content[x][0] for x in range(fitrange[0], fitrange[1] + 1) if not self.content[x] is None])
 739        result = least_squares(xs, ys, function, silent=silent, **kwargs)
 740        return result
 741
 742    def plateau(self, plateau_range=None, method="fit", auto_gamma=False):
 743        """ Extract a plateau value from a Corr object
 744
 745        Parameters
 746        ----------
 747        plateau_range : list
 748            list with two entries, indicating the first and the last timeslice
 749            of the plateau region.
 750        method : str
 751            method to extract the plateau.
 752                'fit' fits a constant to the plateau region
 753                'avg', 'average' or 'mean' just average over the given timeslices.
 754        auto_gamma : bool
 755            apply gamma_method with default parameters to the Corr. Defaults to None
 756        """
 757        if not plateau_range:
 758            if self.prange:
 759                plateau_range = self.prange
 760            else:
 761                raise Exception("no plateau range provided")
 762        if self.N != 1:
 763            raise Exception("Correlator must be projected before getting a plateau.")
 764        if (all([self.content[t] is None for t in range(plateau_range[0], plateau_range[1] + 1)])):
 765            raise Exception("plateau is undefined at all timeslices in plateaurange.")
 766        if auto_gamma:
 767            self.gamma_method()
 768        if method == "fit":
 769            def const_func(a, t):
 770                return a[0]
 771            return self.fit(const_func, plateau_range)[0]
 772        elif method in ["avg", "average", "mean"]:
 773            returnvalue = np.mean([item[0] for item in self.content[plateau_range[0]:plateau_range[1] + 1] if item is not None])
 774            return returnvalue
 775
 776        else:
 777            raise Exception("Unsupported plateau method: " + method)
 778
 779    def set_prange(self, prange):
 780        """Sets the attribute prange of the Corr object."""
 781        if not len(prange) == 2:
 782            raise Exception("prange must be a list or array with two values")
 783        if not ((isinstance(prange[0], int)) and (isinstance(prange[1], int))):
 784            raise Exception("Start and end point must be integers")
 785        if not (0 <= prange[0] <= self.T and 0 <= prange[1] <= self.T and prange[0] < prange[1]):
 786            raise Exception("Start and end point must define a range in the interval 0,T")
 787
 788        self.prange = prange
 789        return
 790
 791    def show(self, x_range=None, comp=None, y_range=None, logscale=False, plateau=None, fit_res=None, ylabel=None, save=None, auto_gamma=False, hide_sigma=None, references=None, title=None):
 792        """Plots the correlator using the tag of the correlator as label if available.
 793
 794        Parameters
 795        ----------
 796        x_range : list
 797            list of two values, determining the range of the x-axis e.g. [4, 8].
 798        comp : Corr or list of Corr
 799            Correlator or list of correlators which are plotted for comparison.
 800            The tags of these correlators are used as labels if available.
 801        logscale : bool
 802            Sets y-axis to logscale.
 803        plateau : Obs
 804            Plateau value to be visualized in the figure.
 805        fit_res : Fit_result
 806            Fit_result object to be visualized.
 807        ylabel : str
 808            Label for the y-axis.
 809        save : str
 810            path to file in which the figure should be saved.
 811        auto_gamma : bool
 812            Apply the gamma method with standard parameters to all correlators and plateau values before plotting.
 813        hide_sigma : float
 814            Hides data points from the first value on which is consistent with zero within 'hide_sigma' standard errors.
 815        references : list
 816            List of floating point values that are displayed as horizontal lines for reference.
 817        title : string
 818            Optional title of the figure.
 819        """
 820        if self.N != 1:
 821            raise Exception("Correlator must be projected before plotting")
 822
 823        if auto_gamma:
 824            self.gamma_method()
 825
 826        if x_range is None:
 827            x_range = [0, self.T - 1]
 828
 829        fig = plt.figure()
 830        ax1 = fig.add_subplot(111)
 831
 832        x, y, y_err = self.plottable()
 833        if hide_sigma:
 834            hide_from = np.argmax((hide_sigma * np.array(y_err[1:])) > np.abs(y[1:])) - 1
 835        else:
 836            hide_from = None
 837        ax1.errorbar(x[:hide_from], y[:hide_from], y_err[:hide_from], label=self.tag)
 838        if logscale:
 839            ax1.set_yscale('log')
 840        else:
 841            if y_range is None:
 842                try:
 843                    y_min = min([(x[0].value - x[0].dvalue) for x in self.content[x_range[0]: x_range[1] + 1] if (x is not None) and x[0].dvalue < 2 * np.abs(x[0].value)])
 844                    y_max = max([(x[0].value + x[0].dvalue) for x in self.content[x_range[0]: x_range[1] + 1] if (x is not None) and x[0].dvalue < 2 * np.abs(x[0].value)])
 845                    ax1.set_ylim([y_min - 0.1 * (y_max - y_min), y_max + 0.1 * (y_max - y_min)])
 846                except Exception:
 847                    pass
 848            else:
 849                ax1.set_ylim(y_range)
 850        if comp:
 851            if isinstance(comp, (Corr, list)):
 852                for corr in comp if isinstance(comp, list) else [comp]:
 853                    if auto_gamma:
 854                        corr.gamma_method()
 855                    x, y, y_err = corr.plottable()
 856                    if hide_sigma:
 857                        hide_from = np.argmax((hide_sigma * np.array(y_err[1:])) > np.abs(y[1:])) - 1
 858                    else:
 859                        hide_from = None
 860                    ax1.errorbar(x[:hide_from], y[:hide_from], y_err[:hide_from], label=corr.tag, mfc=plt.rcParams['axes.facecolor'])
 861            else:
 862                raise Exception("'comp' must be a correlator or a list of correlators.")
 863
 864        if plateau:
 865            if isinstance(plateau, Obs):
 866                if auto_gamma:
 867                    plateau.gamma_method()
 868                ax1.axhline(y=plateau.value, linewidth=2, color=plt.rcParams['text.color'], alpha=0.6, marker=',', ls='--', label=str(plateau))
 869                ax1.axhspan(plateau.value - plateau.dvalue, plateau.value + plateau.dvalue, alpha=0.25, color=plt.rcParams['text.color'], ls='-')
 870            else:
 871                raise Exception("'plateau' must be an Obs")
 872
 873        if references:
 874            if isinstance(references, list):
 875                for ref in references:
 876                    ax1.axhline(y=ref, linewidth=1, color=plt.rcParams['text.color'], alpha=0.6, marker=',', ls='--')
 877            else:
 878                raise Exception("'references' must be a list of floating pint values.")
 879
 880        if self.prange:
 881            ax1.axvline(self.prange[0], 0, 1, ls='-', marker=',')
 882            ax1.axvline(self.prange[1], 0, 1, ls='-', marker=',')
 883
 884        if fit_res:
 885            x_samples = np.arange(x_range[0], x_range[1] + 1, 0.05)
 886            ax1.plot(x_samples,
 887                     fit_res.fit_function([o.value for o in fit_res.fit_parameters], x_samples),
 888                     ls='-', marker=',', lw=2)
 889
 890        ax1.set_xlabel(r'$x_0 / a$')
 891        if ylabel:
 892            ax1.set_ylabel(ylabel)
 893        ax1.set_xlim([x_range[0] - 0.5, x_range[1] + 0.5])
 894
 895        handles, labels = ax1.get_legend_handles_labels()
 896        if labels:
 897            ax1.legend()
 898
 899        if title:
 900            plt.title(title)
 901
 902        plt.draw()
 903
 904        if save:
 905            if isinstance(save, str):
 906                fig.savefig(save, bbox_inches='tight')
 907            else:
 908                raise Exception("'save' has to be a string.")
 909
 910    def spaghetti_plot(self, logscale=True):
 911        """Produces a spaghetti plot of the correlator suited to monitor exceptional configurations.
 912
 913        Parameters
 914        ----------
 915        logscale : bool
 916            Determines whether the scale of the y-axis is logarithmic or standard.
 917        """
 918        if self.N != 1:
 919            raise Exception("Correlator needs to be projected first.")
 920
 921        mc_names = list(set([item for sublist in [sum(map(o[0].e_content.get, o[0].mc_names), []) for o in self.content if o is not None] for item in sublist]))
 922        x0_vals = [n for (n, o) in zip(np.arange(self.T), self.content) if o is not None]
 923
 924        for name in mc_names:
 925            data = np.array([o[0].deltas[name] + o[0].r_values[name] for o in self.content if o is not None]).T
 926
 927            fig = plt.figure()
 928            ax = fig.add_subplot(111)
 929            for dat in data:
 930                ax.plot(x0_vals, dat, ls='-', marker='')
 931
 932            if logscale is True:
 933                ax.set_yscale('log')
 934
 935            ax.set_xlabel(r'$x_0 / a$')
 936            plt.title(name)
 937            plt.draw()
 938
 939    def dump(self, filename, datatype="json.gz", **kwargs):
 940        """Dumps the Corr into a file of chosen type
 941        Parameters
 942        ----------
 943        filename : str
 944            Name of the file to be saved.
 945        datatype : str
 946            Format of the exported file. Supported formats include
 947            "json.gz" and "pickle"
 948        path : str
 949            specifies a custom path for the file (default '.')
 950        """
 951        if datatype == "json.gz":
 952            from .input.json import dump_to_json
 953            if 'path' in kwargs:
 954                file_name = kwargs.get('path') + '/' + filename
 955            else:
 956                file_name = filename
 957            dump_to_json(self, file_name)
 958        elif datatype == "pickle":
 959            dump_object(self, filename, **kwargs)
 960        else:
 961            raise Exception("Unknown datatype " + str(datatype))
 962
 963    def print(self, print_range=None):
 964        print(self.__repr__(print_range))
 965
 966    def __repr__(self, print_range=None):
 967        if print_range is None:
 968            print_range = [0, None]
 969
 970        content_string = ""
 971        content_string += "Corr T=" + str(self.T) + " N=" + str(self.N) + "\n"  # +" filled with"+ str(type(self.content[0][0])) there should be a good solution here
 972
 973        if self.tag is not None:
 974            content_string += "Description: " + self.tag + "\n"
 975        if self.N != 1:
 976            return content_string
 977        if isinstance(self[0], CObs):
 978            return content_string
 979
 980        if print_range[1]:
 981            print_range[1] += 1
 982        content_string += 'x0/a\tCorr(x0/a)\n------------------\n'
 983        for i, sub_corr in enumerate(self.content[print_range[0]:print_range[1]]):
 984            if sub_corr is None:
 985                content_string += str(i + print_range[0]) + '\n'
 986            else:
 987                content_string += str(i + print_range[0])
 988                for element in sub_corr:
 989                    content_string += '\t' + ' ' * int(element >= 0) + str(element)
 990                content_string += '\n'
 991        return content_string
 992
 993    def __str__(self):
 994        return self.__repr__()
 995
 996    # We define the basic operations, that can be performed with correlators.
 997    # While */+- get defined here, they only work for Corr*Obs and not Obs*Corr.
 998    # This is because Obs*Corr checks Obs.__mul__ first and does not catch an exception.
 999    # One could try and tell Obs to check if the y in __mul__ is a Corr and
1000
1001    def __add__(self, y):
1002        if isinstance(y, Corr):
1003            if ((self.N != y.N) or (self.T != y.T)):
1004                raise Exception("Addition of Corrs with different shape")
1005            newcontent = []
1006            for t in range(self.T):
1007                if _check_for_none(self, self.content[t]) or _check_for_none(y, y.content[t]):
1008                    newcontent.append(None)
1009                else:
1010                    newcontent.append(self.content[t] + y.content[t])
1011            return Corr(newcontent)
1012
1013        elif isinstance(y, (Obs, int, float, CObs)):
1014            newcontent = []
1015            for t in range(self.T):
1016                if _check_for_none(self, self.content[t]):
1017                    newcontent.append(None)
1018                else:
1019                    newcontent.append(self.content[t] + y)
1020            return Corr(newcontent, prange=self.prange)
1021        elif isinstance(y, np.ndarray):
1022            if y.shape == (self.T,):
1023                return Corr(list((np.array(self.content).T + y).T))
1024            else:
1025                raise ValueError("operands could not be broadcast together")
1026        else:
1027            raise TypeError("Corr + wrong type")
1028
1029    def __mul__(self, y):
1030        if isinstance(y, Corr):
1031            if not ((self.N == 1 or y.N == 1 or self.N == y.N) and self.T == y.T):
1032                raise Exception("Multiplication of Corr object requires N=N or N=1 and T=T")
1033            newcontent = []
1034            for t in range(self.T):
1035                if _check_for_none(self, self.content[t]) or _check_for_none(y, y.content[t]):
1036                    newcontent.append(None)
1037                else:
1038                    newcontent.append(self.content[t] * y.content[t])
1039            return Corr(newcontent)
1040
1041        elif isinstance(y, (Obs, int, float, CObs)):
1042            newcontent = []
1043            for t in range(self.T):
1044                if _check_for_none(self, self.content[t]):
1045                    newcontent.append(None)
1046                else:
1047                    newcontent.append(self.content[t] * y)
1048            return Corr(newcontent, prange=self.prange)
1049        elif isinstance(y, np.ndarray):
1050            if y.shape == (self.T,):
1051                return Corr(list((np.array(self.content).T * y).T))
1052            else:
1053                raise ValueError("operands could not be broadcast together")
1054        else:
1055            raise TypeError("Corr * wrong type")
1056
1057    def __truediv__(self, y):
1058        if isinstance(y, Corr):
1059            if not ((self.N == 1 or y.N == 1 or self.N == y.N) and self.T == y.T):
1060                raise Exception("Multiplication of Corr object requires N=N or N=1 and T=T")
1061            newcontent = []
1062            for t in range(self.T):
1063                if _check_for_none(self, self.content[t]) or _check_for_none(y, y.content[t]):
1064                    newcontent.append(None)
1065                else:
1066                    newcontent.append(self.content[t] / y.content[t])
1067            for t in range(self.T):
1068                if _check_for_none(self, newcontent[t]):
1069                    continue
1070                if np.isnan(np.sum(newcontent[t]).value):
1071                    newcontent[t] = None
1072
1073            if all([item is None for item in newcontent]):
1074                raise Exception("Division returns completely undefined correlator")
1075            return Corr(newcontent)
1076
1077        elif isinstance(y, (Obs, CObs)):
1078            if isinstance(y, Obs):
1079                if y.value == 0:
1080                    raise Exception('Division by zero will return undefined correlator')
1081            if isinstance(y, CObs):
1082                if y.is_zero():
1083                    raise Exception('Division by zero will return undefined correlator')
1084
1085            newcontent = []
1086            for t in range(self.T):
1087                if _check_for_none(self, self.content[t]):
1088                    newcontent.append(None)
1089                else:
1090                    newcontent.append(self.content[t] / y)
1091            return Corr(newcontent, prange=self.prange)
1092
1093        elif isinstance(y, (int, float)):
1094            if y == 0:
1095                raise Exception('Division by zero will return undefined correlator')
1096            newcontent = []
1097            for t in range(self.T):
1098                if _check_for_none(self, self.content[t]):
1099                    newcontent.append(None)
1100                else:
1101                    newcontent.append(self.content[t] / y)
1102            return Corr(newcontent, prange=self.prange)
1103        elif isinstance(y, np.ndarray):
1104            if y.shape == (self.T,):
1105                return Corr(list((np.array(self.content).T / y).T))
1106            else:
1107                raise ValueError("operands could not be broadcast together")
1108        else:
1109            raise TypeError('Corr / wrong type')
1110
1111    def __neg__(self):
1112        newcontent = [None if _check_for_none(self, item) else -1. * item for item in self.content]
1113        return Corr(newcontent, prange=self.prange)
1114
1115    def __sub__(self, y):
1116        return self + (-y)
1117
1118    def __pow__(self, y):
1119        if isinstance(y, (Obs, int, float, CObs)):
1120            newcontent = [None if _check_for_none(self, item) else item**y for item in self.content]
1121            return Corr(newcontent, prange=self.prange)
1122        else:
1123            raise TypeError('Type of exponent not supported')
1124
1125    def __abs__(self):
1126        newcontent = [None if _check_for_none(self, item) else np.abs(item) for item in self.content]
1127        return Corr(newcontent, prange=self.prange)
1128
1129    # The numpy functions:
1130    def sqrt(self):
1131        return self ** 0.5
1132
1133    def log(self):
1134        newcontent = [None if _check_for_none(self, item) else np.log(item) for item in self.content]
1135        return Corr(newcontent, prange=self.prange)
1136
1137    def exp(self):
1138        newcontent = [None if _check_for_none(self, item) else np.exp(item) for item in self.content]
1139        return Corr(newcontent, prange=self.prange)
1140
1141    def _apply_func_to_corr(self, func):
1142        newcontent = [None if _check_for_none(self, item) else func(item) for item in self.content]
1143        for t in range(self.T):
1144            if _check_for_none(self, newcontent[t]):
1145                continue
1146            tmp_sum = np.sum(newcontent[t])
1147            if hasattr(tmp_sum, "value"):
1148                if np.isnan(tmp_sum.value):
1149                    newcontent[t] = None
1150        if all([item is None for item in newcontent]):
1151            raise Exception('Operation returns undefined correlator')
1152        return Corr(newcontent)
1153
1154    def sin(self):
1155        return self._apply_func_to_corr(np.sin)
1156
1157    def cos(self):
1158        return self._apply_func_to_corr(np.cos)
1159
1160    def tan(self):
1161        return self._apply_func_to_corr(np.tan)
1162
1163    def sinh(self):
1164        return self._apply_func_to_corr(np.sinh)
1165
1166    def cosh(self):
1167        return self._apply_func_to_corr(np.cosh)
1168
1169    def tanh(self):
1170        return self._apply_func_to_corr(np.tanh)
1171
1172    def arcsin(self):
1173        return self._apply_func_to_corr(np.arcsin)
1174
1175    def arccos(self):
1176        return self._apply_func_to_corr(np.arccos)
1177
1178    def arctan(self):
1179        return self._apply_func_to_corr(np.arctan)
1180
1181    def arcsinh(self):
1182        return self._apply_func_to_corr(np.arcsinh)
1183
1184    def arccosh(self):
1185        return self._apply_func_to_corr(np.arccosh)
1186
1187    def arctanh(self):
1188        return self._apply_func_to_corr(np.arctanh)
1189
1190    # Right hand side operations (require tweak in main module to work)
1191    def __radd__(self, y):
1192        return self + y
1193
1194    def __rsub__(self, y):
1195        return -self + y
1196
1197    def __rmul__(self, y):
1198        return self * y
1199
1200    def __rtruediv__(self, y):
1201        return (self / y) ** (-1)
1202
1203    @property
1204    def real(self):
1205        def return_real(obs_OR_cobs):
1206            if isinstance(obs_OR_cobs.flatten()[0], CObs):
1207                return np.vectorize(lambda x: x.real)(obs_OR_cobs)
1208            else:
1209                return obs_OR_cobs
1210
1211        return self._apply_func_to_corr(return_real)
1212
1213    @property
1214    def imag(self):
1215        def return_imag(obs_OR_cobs):
1216            if isinstance(obs_OR_cobs.flatten()[0], CObs):
1217                return np.vectorize(lambda x: x.imag)(obs_OR_cobs)
1218            else:
1219                return obs_OR_cobs * 0  # So it stays the right type
1220
1221        return self._apply_func_to_corr(return_imag)
1222
1223    def prune(self, Ntrunc, tproj=3, t0proj=2, basematrix=None):
1224        r''' Project large correlation matrix to lowest states
1225
1226        This method can be used to reduce the size of an (N x N) correlation matrix
1227        to (Ntrunc x Ntrunc) by solving a GEVP at very early times where the noise
1228        is still small.
1229
1230        Parameters
1231        ----------
1232        Ntrunc: int
1233            Rank of the target matrix.
1234        tproj: int
1235            Time where the eigenvectors are evaluated, corresponds to ts in the GEVP method.
1236            The default value is 3.
1237        t0proj: int
1238            Time where the correlation matrix is inverted. Choosing t0proj=1 is strongly
1239            discouraged for O(a) improved theories, since the correctness of the procedure
1240            cannot be granted in this case. The default value is 2.
1241        basematrix : Corr
1242            Correlation matrix that is used to determine the eigenvectors of the
1243            lowest states based on a GEVP. basematrix is taken to be the Corr itself if
1244            is is not specified.
1245
1246        Notes
1247        -----
1248        We have the basematrix $C(t)$ and the target matrix $G(t)$. We start by solving
1249        the GEVP $$C(t) v_n(t, t_0) = \lambda_n(t, t_0) C(t_0) v_n(t, t_0)$$ where $t \equiv t_\mathrm{proj}$
1250        and $t_0 \equiv t_{0, \mathrm{proj}}$. The target matrix is projected onto the subspace of the
1251        resulting eigenvectors $v_n, n=1,\dots,N_\mathrm{trunc}$ via
1252        $$G^\prime_{i, j}(t) = (v_i, G(t) v_j)$$. This allows to reduce the size of a large
1253        correlation matrix and to remove some noise that is added by irrelevant operators.
1254        This may allow to use the GEVP on $G(t)$ at late times such that the theoretically motivated
1255        bound $t_0 \leq t/2$ holds, since the condition number of $G(t)$ is decreased, compared to $C(t)$.
1256        '''
1257
1258        if self.N == 1:
1259            raise Exception('Method cannot be applied to one-dimensional correlators.')
1260        if basematrix is None:
1261            basematrix = self
1262        if Ntrunc >= basematrix.N:
1263            raise Exception('Cannot truncate using Ntrunc <= %d' % (basematrix.N))
1264        if basematrix.N != self.N:
1265            raise Exception('basematrix and targetmatrix have to be of the same size.')
1266
1267        evecs = basematrix.GEVP(t0proj, tproj, sort=None)[:Ntrunc]
1268
1269        tmpmat = np.empty((Ntrunc, Ntrunc), dtype=object)
1270        rmat = []
1271        for t in range(basematrix.T):
1272            for i in range(Ntrunc):
1273                for j in range(Ntrunc):
1274                    tmpmat[i][j] = evecs[i].T @ self[t] @ evecs[j]
1275            rmat.append(np.copy(tmpmat))
1276
1277        newcontent = [None if (self.content[t] is None) else rmat[t] for t in range(self.T)]
1278        return Corr(newcontent)
1279
1280
1281def _sort_vectors(vec_set, ts):
1282    """Helper function used to find a set of Eigenvectors consistent over all timeslices"""
1283    reference_sorting = np.array(vec_set[ts])
1284    N = reference_sorting.shape[0]
1285    sorted_vec_set = []
1286    for t in range(len(vec_set)):
1287        if vec_set[t] is None:
1288            sorted_vec_set.append(None)
1289        elif not t == ts:
1290            perms = [list(o) for o in permutations([i for i in range(N)], N)]
1291            best_score = 0
1292            for perm in perms:
1293                current_score = 1
1294                for k in range(N):
1295                    new_sorting = reference_sorting.copy()
1296                    new_sorting[perm[k], :] = vec_set[t][k]
1297                    current_score *= abs(np.linalg.det(new_sorting))
1298                if current_score > best_score:
1299                    best_score = current_score
1300                    best_perm = perm
1301            sorted_vec_set.append([vec_set[t][k] for k in best_perm])
1302        else:
1303            sorted_vec_set.append(vec_set[t])
1304
1305    return sorted_vec_set
1306
1307
1308def _check_for_none(corr, entry):
1309    """Checks if entry for correlator corr is None"""
1310    return len(list(filter(None, np.asarray(entry).flatten()))) < corr.N ** 2
1311
1312
1313def _GEVP_solver(Gt, G0):
1314    """Helper function for solving the GEVP and sorting the eigenvectors.
1315
1316    The helper function assumes that both provided matrices are symmetric and
1317    only processes the lower triangular part of both matrices. In case the matrices
1318    are not symmetric the upper triangular parts are effectively discarded."""
1319    return scipy.linalg.eigh(Gt, G0, lower=True)[1].T[::-1]