multipers 2.3.3b6__cp310-cp310-macosx_11_0_arm64.whl

multipers/mma_structures.pyx.tp

@@ -0,0 +1,1094 @@
+{{py:
+
+"""
+MMA core structures.
+"""
+
+## Value types : CTYPE, PYTHON_TYPE, short
+# value_types = [
+#   ("int32_t", "np.int32", "i32"),
+#   ("int64_t", "np.int64", "i64"),
+#   ("float", "np.float32", "f32"),
+#   ("double", "np.float64", "f64"),
+# ]
+
+import pickle as pkl
+import numpy as np
+
+D = pkl.load(open("build/tmp/_slicer_names.pkl", "rb"))
+value_types = np.unique([tuple((x['C_VALUE_TYPE'],x['PY_VALUE_TYPE'], x['SHORT_VALUE_TYPE'])) for x in D], axis=0)
+
+}}
+
+
+
+
+
+"""!
+@package mma
+@brief Files containing the C++ cythonized functions.
+@author David Loiseaux
+@copyright Copyright (c) 2022 Inria.
+"""
+
+# distutils: language = c++
+
+###########################################################################
+## PYTHON LIBRARIES
+import gudhi as gd
+import numpy as np
+from typing import Union, Literal
+from collections.abc import Callable, Iterable, Sequence
+import pickle
+import multipers.grids as mpg
+
+###########################################################################
+## CPP CLASSES
+from libc.stdint cimport intptr_t
+from libc.stdint cimport uintptr_t
+
+###########################################################################
+## CYTHON TYPES
+from libcpp.vector cimport vector
+from libcpp.utility cimport pair
+#from libcpp.list cimport list as clist
+from libcpp cimport bool
+from libcpp cimport int
+from cython.operator cimport dereference
+from libcpp.utility cimport move
+cimport cython
+#########################################################################
+## Multipersistence Module Approximation Classes
+from multipers.mma_structures cimport *
+from multipers.filtration_conversions cimport *
+cimport numpy as cnp
+
+
+#########################################################################
+## Small hack for typing
+from gudhi import SimplexTree
+from multipers.simplex_tree_multi import SimplexTreeMulti
+from joblib import Parallel, delayed
+
+available_pymodules = [
+    {{for CTYPE, PYTYPE, SHORT in value_types}}
+    PyModule_{{SHORT}},
+    {{endfor}}
+]
+
+PyModule_type = Union[
+    {{for CTYPE, PYTYPE, SHORT in value_types}}
+    PyModule_{{SHORT}},
+    {{endfor}}
+]
+
+{{for CTYPE, PYTYPE, SHORT in value_types}}
+cdef class PySummand_{{SHORT}}:
+    r"""
+    Stores a Summand of a PyModule
+    """
+    cdef Summand[{{CTYPE}}] sum
+
+    def get_birth_list(self): 
+        cdef vector[One_critical_filtration[{{CTYPE}}]] v = self.sum.get_birth_list()
+        return _vff21cview_{{SHORT}}(v, copy = True, duplicate = self.num_parameters())
+
+    def get_death_list(self):
+        cdef vector[One_critical_filtration[{{CTYPE}}]] v = self.sum.get_death_list()
+        return _vff21cview_{{SHORT}}(v, copy = True, duplicate = self.num_parameters())
+    @property
+    def degree(self)->int:
+        return self.sum.get_dimension()
+
+    cdef set(self, Summand[{{CTYPE}}]& summand):
+        self.sum = summand
+        return self
+    def get_bounds(self):
+        cdef pair[One_critical_filtration[{{CTYPE}}],One_critical_filtration[{{CTYPE}}]] cbounds
+        with nogil:
+            cbounds = self.sum.get_bounds().get_bounding_corners()
+        return _ff21cview_{{SHORT}}(&cbounds.first).copy(), _ff21cview_{{SHORT}}(&cbounds.second).copy()
+    @property
+    def dtype(self):
+      return {{PYTYPE}}
+
+    def num_parameters(self):
+        cdef vector[One_critical_filtration[{{CTYPE}}]] v = self.sum.get_birth_list()
+        if v[0].is_finite():
+            return v[0].num_parameters()
+        v = self.sum.get_death_list()
+        return v[0].num_parameters()
+    def __eq__(self, PySummand_{{SHORT}} other):
+        return self.sum == other.sum 
+
+
+
+
+cdef inline get_summand_filtration_values_{{SHORT}}(Summand[{{CTYPE}}] summand):
+    r"""
+    Returns a list (over parameter) of the filtrations values of this parameter.
+    """
+    cdef vector[One_critical_filtration[{{CTYPE}}]] vb = summand.get_birth_list()
+    cdef vector[One_critical_filtration[{{CTYPE}}]] vd = summand.get_death_list()
+
+    if vb[0].is_finite():
+        if vd[0].is_finite():
+            pts = np.concatenate([_vff21cview_{{SHORT}}(vb, copy=True),
+                                  _vff21cview_{{SHORT}}(vd, copy=True)],axis=0)
+        else:
+            pts = np.array(_vff21cview_{{SHORT}}(vb, copy=True))
+    else:
+        if vd[0].is_finite():
+            pts = np.array(_vff21cview_{{SHORT}}(vd, copy=True))
+        else:
+            return []
+    
+    num_parameters = pts.shape[1]
+    out = [np.unique(pts[:,parameter]) for parameter in range(num_parameters)]
+    out = [f[:-1] if f[-1] == np.inf else f for f in out]
+    out = [f[1:]  if f[0] == -np.inf else f for f in out]
+    return out
+
+cdef class PyBox_{{SHORT}}:
+    cdef Box[{{CTYPE}}] box
+    def __cinit__(self, vector[{{CTYPE}}]& bottomCorner, vector[{{CTYPE}}]& topCorner):
+        self.box = Box[{{CTYPE}}](bottomCorner, topCorner)
+    @property
+    def num_parameters(self):
+        cdef size_t dim = self.box.get_lower_corner().num_parameters()
+        if dim == self.box.get_upper_corner().num_parameters():	return dim
+        else:	print("Bad box definition.")
+    def contains(self, x):
+        return self.box.contains(x)
+    cdef set(self, Box[{{CTYPE}}]& b):
+        self.box = b
+        return self
+
+    def get(self):
+        return [<vector[{{CTYPE}}]>self.box.get_lower_corner(), <vector[{{CTYPE}}]>self.box.get_upper_corner()]
+    def to_multipers(self):
+        #assert (self.get_dimension() == 2) "Multipers only works in dimension  2 !"
+        return np.array(self.get()).flatten(order = 'F')
+    @property
+    def dtype(self):
+      return {{PYTYPE}}
+
+
+
+cdef class PyModule_{{SHORT}}:
+    r"""
+    Stores a representation of a n-persistence module.
+    """
+    cdef Module[{{CTYPE}}] cmod
+    
+    @property
+    def dtype(self):
+        return {{PYTYPE}}
+
+    cdef set(self, Module[{{CTYPE}}] m):
+        self.cmod = m
+
+    def __eq__(self, PyModule_{{SHORT}} other):
+        return self.cmod == other.cmod
+
+    def merge(self, PyModule_{{SHORT}} other, int dim=-1):
+        r"""
+        Merges two modules into one
+        """
+        cdef Module[{{CTYPE}}] c_other = other.cmod
+        with nogil:
+            for summand in c_other:
+                self.cmod.add_summand(summand, dim)
+        return self
+
+    def _set_from_ptr(self, intptr_t module_ptr):
+        r"""
+        Copy module from a memory pointer. Unsafe.
+        """
+        self.cmod = move(dereference(<Module[{{CTYPE}}]*>(module_ptr)))
+    def set_box(self, box):
+        assert len(box) == 2, "Box format is [low, hight]"
+        pybox = PyBox_{{SHORT}}(box[0], box[1])
+        cdef Box[{{CTYPE}}] cbox = pybox.box
+        with nogil:
+            self.cmod.set_box(cbox)
+        return self
+    def get_module_of_degree(self, int degree)->PyModule_{{SHORT}}: # TODO : in c++ ?
+        r"""
+        Returns a copy of a module of fixed degree.
+        """
+        pmodule = PyModule_{{SHORT}}()
+        cdef Box[{{CTYPE}}] c_box = self.cmod.get_box()
+        pmodule.cmod.set_box(c_box) 
+        with nogil:
+            for summand in self.cmod:
+                if summand.get_dimension() == degree:
+                    pmodule.cmod.add_summand(summand)
+        return pmodule
+    def get_module_of_degrees(self, degrees:Iterable[int])->PyModule_{{SHORT}}: # TODO : in c++ ?
+        r"""
+        Returns a copy of the summands of degrees in `degrees`
+        """
+        pmodule = PyModule_{{SHORT}}()
+        cdef Box[{{CTYPE}}] c_box = self.cmod.get_box()
+        pmodule.cmod.set_box(c_box)
+        cdef vector[int] cdegrees = degrees
+        with nogil:
+            for summand in self.cmod:
+                for d in cdegrees:
+                    if d == summand.get_dimension():
+                        pmodule.cmod.add_summand(summand)
+        return pmodule
+    def __len__(self)->int:
+        return self.cmod.size()
+    def get_bottom(self)->np.ndarray:
+        r"""
+        Bottom of the box of the module
+        """
+        return np.asarray(<vector[{{CTYPE}}]>(self.cmod.get_box().get_lower_corner()))
+    def get_top(self)->np.ndarray:
+        r"""
+        Top of the box of the module
+        """
+        return np.asarray(<vector[{{CTYPE}}]>(self.cmod.get_box().get_upper_corner()))
+    def get_box(self)->np.ndarray:
+        r"""
+        Returns the current bounding box of the module.
+        """
+        return np.asarray([self.get_bottom(), self.get_top()])
+    @property
+    def max_degree(self)->int:
+        r"""
+        Returns the maximum degree of the module.
+        """
+        return self.cmod.get_dimension()
+    @property
+    def num_parameters(self)->int:
+        cdef size_t dim = self.cmod.get_box().get_lower_corner().num_parameters()
+        assert dim == self.cmod.get_box().get_upper_corner().num_parameters(), "Bad box definition, cannot infer num_parameters."
+        return dim
+    def dump(self, path:str|None=None):
+        r"""
+        Dumps the module into a pickle-able format.
+
+        Parameters
+        ----------
+
+        path:str=None (optional) saves the pickled module in specified path
+
+        Returns
+        -------
+        
+        list of list, encoding the module, which can be retrieved with the function `from_dump`.
+        """
+        ## TODO : optimize, but not really used.
+        return dump_cmod_{{SHORT}}(self.cmod) 
+    def __getstate__(self):
+        return self.dump()
+    def __setstate__(self,dump):
+        cdef Module[{{CTYPE}}] cmod = cmod_from_dump_{{SHORT}}(dump)
+        self.cmod = cmod
+        return
+    def __getitem__(self, int i) -> PySummand_{{SHORT}}:
+        if i == slice(None):
+            return self
+        summand = PySummand_{{SHORT}}()
+        summand.set(self.cmod.at(i % self.cmod.size()))
+        return summand
+    def __iter__(self):
+        cdef int num_summands = self.cmod.size()
+        for i in range(num_summands):
+            summand = PySummand_{{SHORT}}()
+            summand.set(self.cmod.at(i)) ## hmm copy. maybe do summand from ptr
+            yield summand
+
+    def get_bounds(self):
+        r"""
+        Computes bounds from the summands' bounds.
+        Useful to change this' box.
+        """
+        cdef pair[One_critical_filtration[{{CTYPE}}],One_critical_filtration[{{CTYPE}}]] cbounds
+        with nogil:
+            cbounds = self.cmod.get_bounds().get_bounding_corners()
+        return _ff21cview_{{SHORT}}(&cbounds.first).copy(), _ff21cview_{{SHORT}}(&cbounds.second).copy()
+    def rescale(self,rescale_factors, int degree=-1):
+        r"""
+        Rescales the fitlration values of the summands by this rescaling vector.
+        """
+        cdef vector[{{CTYPE}}] crescale_factors = rescale_factors
+        with nogil:
+            self.cmod.rescale(crescale_factors,degree)
+    def translate(self,translation, int degree=-1):
+        r"""
+        Translates the module in the filtration space by this vector.
+        """
+        cdef vector[{{CTYPE}}] ctranslation = translation
+        with nogil:
+            self.cmod.translate(ctranslation,degree)
+
+    def get_filtration_values(self, bool unique=True):
+        r"""
+        Retrieves all filtration values of the summands of the module.
+
+        Output format 
+        -------------
+
+        list of filtration values for parameter.
+        """
+        if len(self) ==0:
+            return np.empty((self.num_parameters,0))
+        values = tuple(tuple(stuff) if len(stuff:=get_summand_filtration_values_{{SHORT}}(summand)) == self.num_parameters else list(stuff) + [[]]*(self.num_parameters - len(stuff)) for summand in self.cmod)
+        try:
+          values = tuple(np.concatenate([
+            f[parameter]
+            for f in values
+            ], axis=0) for parameter in range(self.num_parameters)
+          )
+        except:
+          return values
+        if unique:
+            return [np.unique(f) for f in values]
+        return values
+
+    def plot(self, int degree=-1,**kwargs)->None:
+        r"""Shows the module on a plot. Each color corresponds to an apprimation summand of the module, and its shape corresponds to its support.
+        Only works with 2-parameter modules.
+
+        Parameters
+        ----------
+        degree = -1 : integer
+            If positive returns only the image of dimension `dimension`.
+        box=None : of the form [[b_x,b_y], [d_x,d_y]] where b,d are the bottom and top corner of the rectangle.
+            If non-None, will plot the module on this specific rectangle.
+        min_persistence =0 : float
+            Only plots the summand with a persistence above this threshold.
+        separated=False : bool
+            If true, plot each summand in a different plot.
+        alpha=1 : float
+            Transparancy parameter
+        save = False : string
+            if nontrivial, will save the figure at this path
+
+
+        Returns
+        -------
+        The figure of the plot.
+        """
+        from multipers.plots import plot2d_PyModule
+        import matplotlib.pyplot as plt
+        box = kwargs.pop('box', self.get_box())
+        if (len(box[0]) != 2):
+            print("Filtration size :", len(box[0]), " != 2")
+            return
+        if(degree < 0):
+            dims = np.unique(self.get_dimensions())
+            separated = kwargs.pop("separated", False)
+            ndim = len(dims)
+            scale = kwargs.pop("scale", 4)
+            if separated:
+                fig = None
+                axes = None
+            elif ndim > 1:
+              fig, axes = plt.subplots(1,ndim, figsize=(ndim*scale,scale))
+            else:
+              fig = plt.gcf()
+              axes = [plt.gca()]
+            for dim_idx in range(ndim):
+                if not separated:
+                    plt.sca(axes[dim_idx])
+                self.plot(dims[dim_idx],box=box, separated = separated, **kwargs)
+            return
+        corners = self.cmod.get_corners_of_dimension(degree)
+        plot2d_PyModule(corners, box=box, dimension=degree, **kwargs)
+        return
+    def degree_splits(self):
+        return np.asarray(self.cmod.get_degree_splits(), dtype=np.int64)
+{{if SHORT[0] == 'f'}}
+    def _compute_pixels(self,coordinates:np.ndarray,
+                                         degrees=None, box=None, {{CTYPE}} delta=.1, 
+                                         {{CTYPE}} p=1., bool normalize=False, int n_jobs=0):
+        r"""
+        Computes the image of the module at the given coordinates
+        """
+        if degrees is not None: assert np.all(degrees[:-1] <= degrees[1:]), "Degrees have to be sorted"
+        cdef vector[int] cdegrees = np.arange(self.max_degree +1) if degrees is None else degrees
+        pybox = PyBox_{{SHORT}}(*self.get_box()) if box is None else PyBox_{{SHORT}}(*box)
+        cdef Box[{{CTYPE}}] cbox = pybox.box
+        cdef vector[vector[{{CTYPE}}]] ccoords = coordinates
+        cdef vector[vector[{{CTYPE}}]] out 
+        with nogil:
+            out = self.cmod.compute_pixels(ccoords, cdegrees, cbox, delta, p, normalize, n_jobs)
+        return np.asarray(out)
+    def barcode(self, basepoint, int degree = -1,*, bool threshold = False): # TODO direction vector interface
+        r"""Computes the barcode of module along a lines.
+
+        Parameters
+        ----------
+        
+        basepoint  : vector
+            basepoint of the lines on which to compute the barcodes, i.e. a point on the line
+        degree = -1 : integer
+            Homology degree on which to compute the bars. If negative, every dimension is computed
+        box (default) :
+            box on which to compute the barcodes if basepoints is not specified. Default is a linspace of lines crossing that box.
+        threshold = False : 
+            Thre
+
+        Warning
+        -------
+        
+        If the barcodes are not thresholded, essential barcodes will not be plot-able.
+
+        Returns
+        -------
+        
+        PyMultiDiagrams
+            Structure that holds the barcodes. Barcodes can be retrieved with a .get_points() or a .to_multipers() or a .plot().
+        """
+        out = PyMultiDiagram_{{SHORT}}()
+        out.set(self.cmod.get_barcode(Line[{{CTYPE}}](_py21c_{{SHORT}}(np.asarray(basepoint, dtype={{PYTYPE}}))), degree, threshold))
+        return out
+    @staticmethod
+    cdef _threshold_bc(bc): 
+        return tuple(np.fromiter((a for a in stuff if a[0] < np.inf),  dtype=np.dtype(({{PYTYPE}},2)) ) for stuff in bc)
+    @staticmethod
+    def _bc_to_full(bcs, basepoint, direction=None):
+        # i, (b sv d), coords 
+        basepoint = np.asarray(basepoint)[None,None,:]
+        direction = 1 if direction is None else np.asarray(direction)[None,None,:]
+        return tuple(bc[:,:,None]*direction + basepoint for bc in bcs)
+    def barcode2(self, basepoint, direction=None, int degree = -1,*, bool threshold = False, bool keep_inf = True, bool full = False): # TODO direction vector interface
+        r"""
+        Compute the 1d-barcode a diagonal line based on basepoint, with some direction.
+        
+        Parameters
+        ----------
+
+         - basepoint: 1d array
+         - directiont: 1d array or None, if None: diagonal
+         - degree: int the degree to compute (-1 means all)
+         - threshold: bool if True, threshold the barcode to the modules box
+         - keep_inf: bool if False, removes trivial bars
+                     Note that this removes the order w.r.t. the summands in that case
+         - full:bool if True, returns the coordinates of the barcode instead of the coordinate in the line.
+
+        The output is of the form
+
+        tuple[np.ndarray of shape (num_bars,2)] or tuple[array of shape (num_bar, 2, num_parameters)]
+        """
+        basepoint = np.asarray(basepoint, dtype={{PYTYPE}})
+        if direction is None:
+            bc = tuple(np.asarray(x).reshape(-1,2) for x in self.cmod.get_barcode2(Line[{{CTYPE}}](_py21c_{{SHORT}}(basepoint)), degree))
+        else:
+            direction = np.asarray(direction, dtype = {{PYTYPE}})
+            bc = tuple(np.asarray(x).reshape(-1,2) for x in self.cmod.get_barcode2(Line[{{CTYPE}}](_py21c_{{SHORT}}(basepoint), _py21c_{{SHORT}}(direction)), degree))
+        if not keep_inf:
+            bc = PyModule_{{SHORT}}._threshold_bc(bc)
+        if full:
+            bc = PyModule_{{SHORT}}._bc_to_full(bc, basepoint, direction)
+
+        return bc
+
+    def get_dimensions(self):
+        cdef int num_summands =  len(self)
+        out = np.empty(shape=num_summands, dtype=np.int32)
+        cdef int32_t[:] c_out = out
+        for i in range(num_summands):
+            c_out[i] = self.cmod.at(i).get_dimension()
+        return out
+
+    def barcodes(self, int degree, basepoints = None, num=100, box = None,threshold = False):
+        r"""Computes barcodes of module along a set of lines.
+
+        Parameters
+        ----------
+        
+        basepoints = None : list of vectors
+            basepoints of the lines on which to compute the barcodes.
+        degree = -1 : integer
+            Homology degree on which to compute the bars. If negative, every dimension is computed
+        box (default) :
+            box on which to compute the barcodes if basepoints is not specified. Default is a linspace of lines crossing that box.
+        num:int=100
+            if basepoints is not specified, defines the number of lines to consider.
+        threshold = False : threshold t
+            Resolution of the image(s).
+
+        Warning
+        -------
+        
+        If the barcodes are not thresholded, essential barcodes will not be plot-able.
+
+        Returns
+        -------
+        
+        PyMultiDiagrams
+            Structure that holds the barcodes. Barcodes can be retrieved with a .get_points() or a .to_multipers() or a .plot().
+        """
+        out = PyMultiDiagrams_{{SHORT}}()
+        if box is None:
+            box = [self.get_bottom(), self.get_top()]
+        if (len(box[0]) != 2) and (basepoints is None):
+            raise ValueError("Basepoints has to be specified for filtration dimension >= 3 !")
+        elif basepoints is None:
+            h = box[1][1] - box[0][1]
+            basepoints = np.linspace([box[0][0] - h,box[0][1]], [box[1][0],box[0][1]], num=num) 
+        else :
+            num=len(basepoints)
+
+        cdef {{CTYPE}}[:,:] basepoints_view = np.asarray(basepoints, dtype = {{PYTYPE}})
+        cdef vector[One_critical_filtration[{{CTYPE}}]] cbasepoints = _py2v1c_{{SHORT}}(basepoints_view)
+
+        out.set(self.cmod.get_barcodes(cbasepoints, degree, threshold))
+        return out
+
+    def barcodes2(self, int degree = -1, basepoints = None, int num=100, box = None, bool threshold = False):
+        r"""Computes barcodes of module along a set of lines.
+
+        Parameters
+        ----------
+
+        basepoints = None : list of vectors
+            basepoints of the lines on which to compute the barcodes.
+        degree = -1 : integer
+            Homology degree on which to compute the bars. If negative, every dimension is computed
+        box (default) :
+            box on which to compute the barcodes if basepoints is not specified. Default is a linspace of lines crossing that box.
+        num:int=100
+            if basepoints is not specified, defines the number of lines to consider.
+        threshold = False : threshold t
+            Resolution of the image(s).
+
+        Warning
+        -------
+        
+        If the barcodes are not thresholded, essential barcodes will not be plot-able.
+
+        Returns
+        -------
+        
+        tuple of 1d barcodes, based on basepoint, with direction (1,1)
+        """
+        if box is None:
+            box = [self.get_bottom(), self.get_top()]
+        if (len(box[0]) != 2) and (basepoints is None):
+            raise ValueError("Basepoints has to be specified for filtration dimension >= 3 !")
+        elif basepoints is None:
+            h = box[1][1] - box[0][1]
+            basepoints = np.linspace([box[0][0] - h,box[0][1]], [box[1][0],box[0][1]], num=num) 
+        else :
+            num=len(basepoints)
+
+        basepoints = np.asarray(basepoints, dtype={{PYTYPE}})
+        cdef vector[Line[{{CTYPE}}]] cbasepoints
+        for i in range(num):
+            cbasepoints.push_back(Line[{{CTYPE}}](_py21c_{{SHORT}}(basepoints[i])))
+        return tuple(np.asarray(bc) for bc in self.cmod.get_barcodes2(cbasepoints, degree))   
+
+    def landscape(self, degree:int, k:int=0,box:Sequence|np.ndarray|None=None, resolution:Sequence=[100,100], bool plot=False):
+        r"""Computes the multiparameter landscape from a PyModule. Python interface only bifiltrations.
+
+        Parameters
+        ----------
+
+        degree : integer
+            The homology degree of the landscape.
+        k = 0 : int
+            the k-th landscape
+        resolution = [50,50] : pair of integers
+            Resolution of the image.
+        box = None : in the format [[a,b], [c,d]]
+            If nontrivial, compute the landscape of this box. Default is the PyModule box.
+        plot = True : Boolean
+            If true, plots the images;
+        Returns
+        -------
+        
+        The landscape of the module.
+
+        """
+        import matplotlib.pyplot as plt
+        if box is None:
+            box = self.get_box()
+        cdef Box[{{CTYPE}}] c_box = Box[{{CTYPE}}](box)
+        out = np.array(self.cmod.get_landscape(degree, k, c_box, resolution))
+        if plot:
+            aspect = (box[1][0]-box[0][0]) / (box[1][1]-box[0][1])
+            extent = [box[0][0], box[1][0], box[0][1], box[1][1]]
+            plt.imshow(out.T, origin="lower", extent=extent, aspect=aspect)
+        return out
+
+    def landscapes(self, degree:int, ks:list|np.ndarray=[0],box=None, resolution:list|np.ndarray=[100,100], bool plot=False):
+        r"""Computes the multiparameter landscape from a PyModule. Python interface only bifiltrations.
+
+        Parameters
+        ----------
+        
+         - degree : integer
+            The homology degree of the landscape.
+         - ks = 0 : list of int
+            the k-th landscape
+         - resolution = [50,50] : pair of integers
+            Resolution of the image.
+         - box = None : in the format [[a,b], [c,d]]
+            If nontrivial, compute the landscape of this box. Default is the PyModule box.
+         - plot = True : bool
+            If true, plots the images;
+        Returns
+        -------
+        
+        The landscapes of the module with parameters ks.
+
+        """
+        import matplotlib.pyplot as plt
+        if box is None:
+            box = self.get_box()
+        out = np.array(self.cmod.get_landscapes(degree, ks, Box[{{CTYPE}}](box), resolution))
+        if plot:
+            to_plot = np.sum(out, axis=0)
+            aspect = (box[1][0]-box[0][0]) / (box[1][1]-box[0][1])
+            extent = [box[0][0], box[1][0], box[0][1], box[1][1]]
+            plt.imshow(to_plot.T, origin="lower", extent=extent, aspect=aspect)
+        return out
+
+
+    def representation(self, degrees=None, double bandwidth=0.1,
+            resolution:Sequence[int]|int=50, 
+            kernel:Literal["gaussian","linear","linear2","exponential"]|Callable = "gaussian",
+            bool signed=False,
+            bool normalize=False, bool plot=False, 
+            bool save=False, int dpi=200,double p=2., box=None, 
+            bool flatten=False, int n_jobs=0,
+            grid = None)->np.ndarray:
+        r"""Computes a representation of the module, using 
+
+        [A Framework for Fast and Stable Representations of Multiparameter Persistent Homology Decompositions, Neurips2023]
+
+        Parameters
+        ----------
+        
+         - degrees = None : integer list
+            If given returns only the image(s) of homology degrees `degrees`.
+         - bandwidth = 0.1 : float
+            Image parameter. 
+         - resolution = [100,100] : pair of integers
+            Resolution of the image(s).
+         - normalize = True : Boolean
+            Ensures that the image belongs to [0,1].
+         - plot = False : Boolean
+            If true, plots the images;
+         - flatten=False :
+            If True, reshapes the output to a flattened shape.
+         - kernel: Either linear, gaussian, or callable
+           The kernel to apply to the matrix $(d(x,I), x \in \mathrm{pixels}, I\in \mathrm{summands})$.
+           signature should be : (distance matrix, summand_weights, bandwidth, p) -> representation
+
+        Returns
+        -------
+        
+        The list of images, or the image of fixed dimension.
+        """
+        import matplotlib.pyplot as plt
+        # box = kwargs.get("box",[self.get_bottom(),self.get_top()])
+        if box is None:
+            box = self.get_box()
+        num_parameters = self.num_parameters
+        if degrees is None:
+            degrees = np.arange(self.max_degree +1)
+        num_degrees = len(degrees)
+        try:
+            int(resolution)
+            resolution = [resolution]*num_parameters
+        except:
+            pass
+
+        if grid is None:
+            grid = [np.linspace(*np.asarray(box)[:,parameter], num=res) for parameter, res in zip(range(num_parameters), resolution)]
+        else:
+            resolution = tuple(len(g) for g in grid)
+        coordinates = mpg.todense(grid)
+
+        if kernel == "linear":
+            assert not signed, "This kernel is not compatible with signed."
+            concatenated_images = self._compute_pixels(coordinates, degrees=degrees, box=box, delta=bandwidth, p=p, normalize=normalize,n_jobs=n_jobs)
+        else:
+            if kernel == "linear2":
+                def todo(PyModule_{{SHORT}} mod_degree):
+                  x = mod_degree.distance_to(coordinates,signed=signed)
+                  w = mod_degree.get_interleavings()[None]**p
+                  s = np.where(x>=0,1,-1) if signed else 1 
+                  x = np.abs(x)
+                  return (s*np.where(x<bandwidth,(bandwidth-x)/bandwidth,0)*w).sum(1)
+            elif kernel == "gaussian":
+                def todo(PyModule_{{SHORT}} mod_degree):
+                  x = mod_degree.distance_to(coordinates,signed=signed)
+                  w = mod_degree.get_interleavings()[None]**p
+                  s = np.where(x>=0,1,-1) if signed else 1 
+                  return (s*np.exp( -.5*((x/bandwidth)**2))*w).sum(1)
+            elif kernel == "exponential":
+                def todo(PyModule_{{SHORT}} mod_degree):
+                  x = mod_degree.distance_to(coordinates,signed=signed)
+                  w = mod_degree.get_interleavings()[None]**p
+                  s = np.where(x>=0,1,-1) if signed else 1 
+                  x = np.abs(x)
+                  return (s*np.exp( -((np.abs(x)/bandwidth)))*w).sum(1)
+            else:
+                assert callable(kernel), r"""
+                    Kernel should be 
+                    gaussian, linear, linear2, exponential or callable,
+                    with signature
+                     (array[shape=(N, M)], array[shape=(M)])-> array(shape=(N)),
+                    the first argument being a distance matrix (pts) vs (summands of the module)
+                    and the second argument is a weight vector (weight(summand) for summand in module).
+                    Note that the distance can be signed.
+                    """
+                def todo(PyModule_{{SHORT}} mod_degree):
+                      x = mod_degree.distance_to(coordinates,signed=signed, n_jobs=n_jobs)
+                      w = mod_degree.get_interleavings()[None]**p
+                      return kernel(x/bandwidth,w)
+            concatenated_images = np.stack(
+                Parallel(n_jobs = n_jobs if n_jobs else -1, backend= "threading")(
+                  delayed(todo)(self.get_module_of_degree(degree))
+                  for degree in degrees
+                )
+              )
+
+        if flatten:
+            image_vector = concatenated_images.reshape((len(degrees),-1))
+            if plot:
+                raise ValueError("Unflatten to plot.")
+            return image_vector
+        else:
+            image_vector = concatenated_images.reshape((len(degrees),*resolution))
+        if plot:
+            assert num_parameters == 2, "Plot only available for 2-parameter modules"
+            import multipers.plots
+            i=0
+            n_plots = len(image_vector)
+            scale = 4
+            if n_plots >1:
+              fig, axs = plt.subplots(1,n_plots, figsize=(n_plots*scale,scale))
+            else:
+              fig = plt.gcf()
+              axs = [plt.gca()]
+            for image, degree, i in zip(image_vector, degrees, range(num_degrees)):
+                ax = axs[i]
+                temp = multipers.plots.plot_surface(grid, image.T, ax=ax)
+                plt.colorbar(temp, ax = ax)
+                if degree < 0 :
+                    ax.set_title(rf"$H_{i}$ $2$-persistence image")
+                if degree >= 0:
+                    ax.set_title(rf"$H_{degree}$ $2$-persistence image")
+        return image_vector
+
+    def euler_char(self, points:list|np.ndarray) -> np.ndarray:
+        r""" Computes the Euler Characteristic of the filtered complex at given (multiparameter) time
+
+        Parameters
+        ----------
+        
+        points: list[float] | list[list[float]] | np.ndarray
+            List of filtration values on which to compute the euler characteristic.
+            WARNING FIXME : the points have to have the same dimension as the simplextree.	
+
+        Returns
+        -------
+        
+        The list of euler characteristic values
+        """
+        if len(points) == 0:
+            return []
+        if type(points[0]) is float:
+            points = [points]
+        if type(points) is np.ndarray:
+            assert len(points.shape) in [1,2]
+            if len(points.shape) == 1:
+                points = [points]
+
+        cdef {{CTYPE}}[:,:] points_view = np.asarray(points, dtype = {{PYTYPE}})
+        cdef vector[One_critical_filtration[{{CTYPE}}]] c_points = _py2v1c_{{SHORT}}(points_view)
+        # cdef One_critical_filtration temp
+        # for point in points:
+        #     temp.clear()
+        #     for truc in point:
+        #         temp.push_back(<{{CTYPE}}>(truc))
+        #     c_points.push_back(temp)
+        cdef Module[{{CTYPE}}] c_mod = self.cmod
+        with nogil:
+            c_euler = c_mod.euler_curve(c_points)
+        euler = c_euler
+        return np.asarray(euler, dtype=int)
+    def to_idx(self,grid):
+        cdef vector[vector[{{CTYPE}}]] cgrid = grid
+        cdef vector[vector[pair[vector[vector[int]],vector[vector[int]]]]] out
+        with nogil:
+            out = self.cmod.to_idx(cgrid)
+        return tuple(tuple((np.asarray(I.first,dtype=np.int64), np.asarray(I.second, dtype=np.int64)) for I in Is_of_degree) for Is_of_degree in out)
+    
+    @cython.wraparound(False)
+    @cython.boundscheck(False)
+    def to_flat_idx(self,grid):
+        if len(self) == 0:
+            return np.empty((2,0), dtype = np.int32), np.empty((0, self.num_parameters), dtype = np.int32),  np.empty((0, self.num_parameters), dtype = np.int32)
+        cdef vector[vector[{{CTYPE}}]] cgrid = grid
+        cdef vector[vector[vector[int]]] out
+        cdef int num_summands, num_births, num_deaths
+        cdef int num_parameters = self.num_parameters
+        with nogil:
+            out = self.cmod.to_flat_idx(cgrid)
+            num_summands = out[0][0].size()
+            num_births = out[1].size()
+            num_deaths = out[2].size()
+        idx = np.empty((2, num_summands),dtype=np.int32 )
+        births = np.empty((num_births,num_parameters),dtype=np.int32)
+        deaths = np.empty((num_deaths,num_parameters),dtype=np.int32)
+
+        cdef int32_t[:,:] idx_view = idx
+        cdef int32_t[:,:] births_view = births
+        cdef int32_t[:,:] deaths_view = deaths
+
+        with nogil:
+            for i in range(num_summands):
+                idx_view[0,i] = out[0][0][i]
+                idx_view[1,i] = out[0][1][i]
+            for i in range(num_births):
+                for j in range(num_parameters):
+                    births_view[i,j] = out[1][i][j]
+            for i in range(num_deaths):
+                for j in range(num_parameters):
+                    deaths_view[i,j] = out[2][i][j]
+
+        return idx, births,deaths
+
+    def distances_idx_to(self, pts, bool full=False, int n_jobs=1):
+        pts = np.asarray(pts)
+        if pts.ndim == 1:
+            pts = pts[None]
+        cdef vector[vector[{{CTYPE}}]] cpts = pts
+        cdef vector[vector[vector[int]]] out
+        with nogil:
+            out = self.cmod.compute_distances_idx_to(cpts,full, n_jobs)
+        return np.asarray(out, dtype=np.int32)
+
+    def distance_to(self, pts, bool signed=False, int n_jobs = 0)->np.ndarray:
+        r"""
+        Distance from a point to each summand's support. 
+        Signed distance is the distance to the boundary,
+        with negative values inside the summands.
+
+        pts of shape (num_pts, num_parameters)
+
+        output shape : (num_pts,num_summands)
+        """
+        pts = np.asarray(pts)
+        if pts.ndim == 1:
+            pts = pts[None]
+        assert pts.shape[-1] == self.num_parameters
+        cdef vector[vector[{{CTYPE}}]] cpts = pts
+        # cdef vector[vector[{{CTYPE}}]] out
+        to_fill = np.empty(shape=(cpts.size(),len(self)), dtype = {{PYTYPE}})
+        cdef {{CTYPE}}[:,:] c_to_fill = to_fill
+        cdef {{CTYPE}}* data_ptr = &c_to_fill[0,0]
+        with nogil:
+            self.cmod.compute_distances_to(data_ptr, cpts,signed, n_jobs)
+        return to_fill
+
+    def get_interleavings(self,box=None):
+        if box is None:
+            box = self.get_box()
+        cdef Box[{{CTYPE}}] cbox = Box[{{CTYPE}}](box)
+        return np.asarray(self.cmod.get_interleavings(cbox))
+
+cdef class PyMultiDiagramPoint_{{SHORT}}:
+    cdef MultiDiagram_point[One_critical_filtration[{{CTYPE}}]] point
+    cdef set(self, MultiDiagram_point[One_critical_filtration[{{CTYPE}}]] pt):
+        self.point = pt
+        return self
+
+    def get_degree(self):
+        return self.point.get_dimension()
+    def get_birth(self):
+        cdef One_critical_filtration[{{CTYPE}}] v = self.point.get_birth()
+        return _ff21cview_{{SHORT}}(&v, copy=True)
+    def get_death(self):
+        cdef One_critical_filtration[{{CTYPE}}] v = self.point.get_death()
+        return _ff21cview_{{SHORT}}(&v, copy=True)
+
+
+cdef class PyMultiDiagram_{{SHORT}}:
+    r"""
+    Stores the diagram of a PyModule on a line
+    """
+    cdef MultiDiagram[One_critical_filtration[{{CTYPE}}], {{CTYPE}}] multiDiagram
+    cdef set(self, MultiDiagram[One_critical_filtration[{{CTYPE}}], {{CTYPE}}] m):
+        self.multiDiagram = m
+        return self
+    def get_points(self, degree:int=-1) -> np.ndarray:
+        cdef vector[pair[vector[{{CTYPE}}],vector[{{CTYPE}}]]] out = self.multiDiagram.get_points(degree)
+        if len(out) == 0 and len(self) == 0:
+            return np.empty() # TODO Retrieve good number of parameters if there is no points in diagram
+        if len(out) == 0:
+            return np.empty((0,2,self.multiDiagram.at(0).get_dimension())) # gets the number of parameters
+        return np.array(out)
+    def to_multipers(self, dimension:int):
+        return self.multiDiagram.to_multipers(dimension)
+    def __len__(self) -> int:
+        return self.multiDiagram.size()
+    def __getitem__(self,i:int) -> PyMultiDiagramPoint_{{SHORT}}:
+        return PyMultiDiagramPoint_{{SHORT}}().set(self.multiDiagram.at(i % self.multiDiagram.size()))
+cdef class PyMultiDiagrams_{{SHORT}}:
+    """
+    Stores the barcodes of a PyModule on multiple lines
+    """
+    cdef MultiDiagrams[One_critical_filtration[{{CTYPE}}], {{CTYPE}}] multiDiagrams
+    cdef set(self,MultiDiagrams[One_critical_filtration[{{CTYPE}}], {{CTYPE}}] m):
+        self.multiDiagrams = m
+        return self
+    def to_multipers(self):
+        out = self.multiDiagrams.to_multipers()
+        return [np.asarray(summand) for summand in out]
+    def __getitem__(self,i:int):
+        if i >=0 :
+            return PyMultiDiagram_{{SHORT}}().set(self.multiDiagrams.at(i))
+        else:
+            return PyMultiDiagram_{{SHORT}}().set(self.multiDiagrams.at( self.multiDiagrams.size() - i))
+    def __len__(self):
+        return self.multiDiagrams.size()
+    def get_points(self, degree:int=-1):
+        return self.multiDiagrams.get_points()
+    cdef _get_plot_bars(self, dimension:int=-1, min_persistence:float=0):
+        return self.multiDiagrams._for_python_plot(dimension, min_persistence);
+    def plot(self, degree:int=-1, min_persistence:float=0):
+        """
+        Plots the barcodes.
+
+        Parameters
+        ----------
+        
+         - degree:int=-1
+            Only plots the bars of specified homology degree. Useful when the multidiagrams contains multiple dimenions
+         - min_persistence:float=0
+            Only plot bars of length greater than this value. Useful to reduce the time to plot.
+
+        Warning
+        -------
+        
+        If the barcodes are not thresholded, essential barcodes will not be displayed !
+
+        """
+        from cycler import cycler
+        import matplotlib
+        import matplotlib.pyplot as plt
+        if len(self) == 0: return
+        _cmap = matplotlib.colormaps["Spectral"]
+        multibarcodes_, colors = self._get_plot_bars(degree, min_persistence)
+        n_summands = np.max(colors)+1 if len(colors)>0 else 1
+
+        plt.rc('axes', prop_cycle = cycler('color', [_cmap(i/n_summands) for i in colors]))
+        return plt.plot(*multibarcodes_)
+{{endif}}
+
+
+cdef dump_summand_{{SHORT}}(Summand[{{CTYPE}}]& summand):
+    cdef vector[One_critical_filtration[{{CTYPE}}]] births = summand.get_birth_list()
+    cdef vector[One_critical_filtration[{{CTYPE}}]] deaths = summand.get_death_list()
+    return (
+            np.array(_vff21cview_{{SHORT}}(births)),
+            np.array(_vff21cview_{{SHORT}}(deaths)),
+            summand.get_dimension(),
+            )
+
+cdef inline Summand[{{CTYPE}}] summand_from_dump_{{SHORT}}(summand_dump):
+    cdef vector[One_critical_filtration[{{CTYPE}}]] births = _py2v1c_{{SHORT}}(summand_dump[0])
+    cdef vector[One_critical_filtration[{{CTYPE}}]] deaths = _py2v1c_{{SHORT}}(summand_dump[1]) 
+    cdef int dim = summand_dump[2]
+    return Summand[{{CTYPE}}](births,deaths,dim)
+
+cdef dump_cmod_{{SHORT}}(Module[{{CTYPE}}]& mod):
+    cdef Box[{{CTYPE}}] cbox = mod.get_box()
+    cdef int dim = mod.get_dimension()
+    cdef cnp.ndarray[{{CTYPE}}, ndim=1] bottom_corner = _ff21cview_{{SHORT}}(&cbox.get_lower_corner())
+    cdef cnp.ndarray[{{CTYPE}}, ndim=1] top_corner = _ff21cview_{{SHORT}}(&cbox.get_upper_corner())
+    box = np.asarray([bottom_corner, top_corner])
+    summands = tuple(dump_summand_{{SHORT}}(summand) for summand in mod)
+    return box, summands
+
+cdef Module[{{CTYPE}}] cmod_from_dump_{{SHORT}}(module_dump):
+    box = module_dump[0]
+    summands = module_dump[1]
+    cdef Module[{{CTYPE}}] out_module = Module[{{CTYPE}}]()
+    out_module.set_box(Box[{{CTYPE}}](box))
+    for i in range(len(summands)):
+        out_module.add_summand(summand_from_dump_{{SHORT}}(summands[i]))
+    return out_module
+
+
+def from_dump_{{SHORT}}(dump)->PyModule_{{SHORT}}:
+    r"""Retrieves a PyModule from a previous dump.
+
+    Parameters
+    ----------
+
+    dump: either the output of the dump function, or a file containing the output of a dump.
+        The dumped module to retrieve
+
+    Returns
+    -------
+    
+    PyModule
+        The retrieved module.
+    """
+    # TODO : optimize...
+    mod = PyModule_{{SHORT}}()
+    if type(dump) is str:
+        dump = pickle.load(open(dump, "rb"))
+    cdef Module[{{CTYPE}}] cmod = cmod_from_dump_{{SHORT}}(dump)
+    mod.cmod = cmod
+    return mod
+
+
+{{endfor}}
+
+
+global PyModule_type, PySummand_type
+PyModule_type = Union[
+{{for CTYPE, PYTYPE, SHORT in value_types}}
+    PyModule_{{SHORT}},
+{{endfor}}
+]
+PySummand_type = Union[
+{{for CTYPE, PYTYPE, SHORT in value_types}}
+    PySummand_{{SHORT}},
+{{endfor}}
+]
+
+PyBox_type = Union[
+{{for CTYPE, PYTYPE, SHORT in value_types}}
+    PyBox_{{SHORT}},
+{{endfor}}
+]
+
+
+PyMultiDiagram_type = Union[
+{{for CTYPE, PYTYPE, SHORT in value_types}}
+{{if SHORT[0] == 'f'}}
+    PyMultiDiagram_{{SHORT}},
+{{endif}}
+{{endfor}}
+]
+
+
+PyMultiDiagrams_type = Union[
+{{for CTYPE, PYTYPE, SHORT in value_types}}
+{{if SHORT[0] == 'f'}}
+    PyMultiDiagrams_{{SHORT}},
+{{endif}}
+{{endfor}}
+]
+
+def is_mma(stuff):
+    return (False 
+{{for CTYPE, PYTYPE, SHORT in value_types}}
+        or isinstance(stuff,PyModule_{{SHORT}})
+{{endfor}}
+    )