geone 1.3.0__py313-none-manylinux_2_35_x86_64.whl

geone/tools.py

@@ -0,0 +1,861 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+# -------------------------------------------------------------------------
+# Python module:  'tools.py'
+# author:         Julien Straubhaar
+# date:           nov-2023
+# -------------------------------------------------------------------------
+
+"""
+Module for miscellaneous tools.
+"""
+
+import sys
+import multiprocessing
+
+import numpy as np
+import matplotlib.pyplot as plt
+from matplotlib.backend_bases import MouseButton
+
+from geone import img
+
+# -----------------------------------------------------------------------------
+def add_path_by_drawing(
+        path_list,
+        close=False,
+        show_instructions=True,
+        last_point_marker='o',
+        last_point_color='red',
+        **kwargs):
+    """
+    Add paths in a list, by interatively drawing on a plot.
+
+    The first argument of the function is a list that is updated when the
+    function is called by appending one (or more) path(s) (see notes below).
+    A path is a 2D array of floats with two columns, each row is (the x and y
+    coordinates of) a point. A path is interactively determined on the plot on
+    the current axis (get with `matplotlib.pyplot.gca()`), with the following
+    rules:
+
+    - left click: add the next point (or first one)
+    - right click: remove the last point
+
+    When pressing a key:
+
+    - key n/N: terminate the current path, and start a new path
+    - key ENTER (or other): terminate the current path and exits
+
+    Parameters
+    ----------
+    path_list : list
+        list of paths, that will be updated by appending the path interactively
+        drawn in the current axis (one can start with `path_list = []`)
+
+    close : bool, default: False
+        if `True`: when a path is terminated, the first points of the path is
+        replicated at the end of the path to form a close line / path
+
+    show_instructions : bool, default: True
+        if `True`: instructions are printed in the standard output
+
+    last_point_marker : "matplotlib marker", default: 'o'
+        marker used for highlighting the last clicked point
+
+    last_point_color : "matplotlib color", default: 'red'
+        color used for highlighting the last clicked point
+
+    kwargs : dict
+        keyword arguments passed to `matplotlib.pyplot.plot` to plot the path(s)
+
+    Notes
+    -----
+    * The function does not return anything. The first argument, `path_list` is \
+    updated, with the path(s) drawn interactively, for example `path_list[-1]` \
+    is the last path, a 2D array, where `path_list[-1][i]` is the i-th point \
+    (1D array of two floats) of that path
+    * An interactive maplotlib backend must be used, so that this function works \
+    properly
+    """
+    # fname = 'add_path_by_drawing'
+
+    ax = plt.gca()
+    obj_drawn = []
+    x, y = [], []
+
+    set_default_color = False
+    if 'color' not in kwargs.keys() and 'c' not in kwargs.keys():
+        set_default_color = True
+        col = plt.rcParams['axes.prop_cycle'].by_key()['color']
+        ncol = len(col)
+        col_ind = [0]
+        kwargs['color'] = col[col_ind[0]] # default color for lines (segments)
+    # if 'color' not in kwargs.keys() and 'c' not in kwargs.keys():
+    #     kwargs['color'] = 'tab:blue' # default color for lines (segments)
+
+    if show_instructions:
+        instruct0 = '\n'.join(['   Left click: add next point', '   Right click: remove last point', '   Key n/N: select a new line', '   key ENTER (or other): finish (quit)'])
+        print('\n'.join(['Draw path', instruct0]))
+        sys.stdout.flush()
+
+    def on_click(event):
+        if not event.inaxes:
+            return
+        if event.button is MouseButton.LEFT:
+            if len(x):
+                # remove last point from obj_drawn
+                ax.lines[-1].remove()
+            # add clicked point
+            x.append(event.xdata)
+            y.append(event.ydata)
+            if len(x) > 1:
+                # add line (segment) to obj_drawn
+                obj_drawn.append(plt.plot(x[-2:], y[-2:], **kwargs))
+            # add point to obj_drawn
+            obj_drawn.append(plt.plot(x[-1], y[-1], marker=last_point_marker, color=last_point_color))
+        if event.button is MouseButton.RIGHT:
+            if len(x):
+                # remove last clicked point
+                del x[-1], y[-1]
+                # remove last point from obj_drawn
+                ax.lines[-1].remove()
+            if len(x):
+                # point(s) are still in the line
+                # remove last line (segment) from obj_drawn
+                ax.lines[-1].remove()
+                # add last point to obj_drawn
+                obj_drawn.append(plt.plot(x[-1], y[-1], marker=last_point_marker, color=last_point_color))
+
+    def on_key(event):
+        if len(x):
+            # remove last point from obj_drawn
+            ax.lines[-1].remove()
+            if close:
+                # close the line
+                if len(x):
+                    x.append(x[0])
+                    y.append(y[0])
+                    # add line (closing segment) to obj_drawn
+                    obj_drawn.append(plt.plot(x[-2:], y[-2:], **kwargs))
+        # add path to path_list
+        if len(x):
+            path_list.append(np.array((x, y)).T)
+        if event.key.lower() == 'n':
+            if show_instructions:
+                print('\n'.join(['Draw next path', instruct0]))
+                sys.stdout.flush()
+            x.clear() # Do not use: x = [], because x is global for this function!
+            y.clear()
+            if set_default_color:
+                # Set color for next line
+                col_ind[0] = (col_ind[0]+1)%ncol
+                kwargs['color'] = col[col_ind[0]] # default color for lines (segments)
+            return
+
+        plt.disconnect(cid_click)
+        plt.disconnect(cid_key)
+        return
+        # if
+        # path_list.append(np.array((x, y)).T)
+        # cid_click = plt.connect('button_press_event', on_click)
+
+
+    cid_click = plt.connect('button_press_event', on_click)
+    cid_key = plt.connect('key_press_event', on_key)
+# -----------------------------------------------------------------------------
+
+# -----------------------------------------------------------------------------
+def is_in_polygon(x, vertices, wrap=None, return_sum_of_angles=False, **kwargs):
+    """
+    Checks if point(s) is (are) in a polygon given by its vertices forming a close line.
+
+    To check if a point is in the polygon, the method consists in computing
+    the vectors from the given point to the vertices of the polygon and
+    the sum of signed angles between two successives vectors (and the last
+    and first one). Then, the point is in the polygon if and only if this sum
+    is equal to +/- 2 pi.
+
+    Note that if the sum of angles is +2 pi (resp. -2 pi), then the vertices form
+    a close line counterclockwise (resp. clockwise); this can be checked by
+    specifying a point `x` in the polygon and `return_sum_of_angles=True`.
+
+    Parameters
+    ----------
+    x : 2D array-like or 1D array-like
+        point(s) coordinates, each row `x[i]` (if 2D array-like) (or `x` if
+        1D array-like) contains the two coordinates of one point
+
+    vertices : 2D array
+        vertices of a polygon in 2D, each row of `vertices` contains the two
+        coordinates of one vertex; the segments of the polygon are obtained by
+        linking two successive vertices (as well as the last one with the first
+        one, if `wrap=True` (see below)), so that the vertices form a close line
+        (clockwise or counterclockwise)
+
+    wrap : bool, optional
+        - if `True`: last and first vertices has to be linked to form a close line
+        - if `False`: last and first vertices should be the same ones (i.e. the \
+        vertices form a close line);
+
+        by default (`None`): `wrap` is automatically computed
+
+    return_sum_of_angles : bool, default: False
+        if `True`, the sum of angles (computed by the method) is returned
+
+    kwargs :
+        keyword arguments passed to function `numpy.isclose`
+
+    Returns
+    -------
+    out : 1D array of bools, or bool
+        indicates for each point in `x` if it is inside (True) or outside (False)
+        the polygon;
+        note: if `x` is of shape (m, 2), then `out` is a 1D array of shape (m, ),
+        and if `x` is of shape (2, ) (one point), `out` is bool
+
+    sum_of_angles : 1D array of floats, or float
+        returned if `return_sum_of_angles=True`; for each point in `x`, the sum
+        of angles computed by the method is returned (`nan` if the sum is not
+        computed);
+        note: `sum_of_angles` is an array of same shape as `out` or a float
+    """
+    # fname = 'is_in_polygon'
+
+    # Set wrap (and adjust vertices) if needed
+    if wrap is None:
+        wrap = ~np.isclose(np.sqrt(((vertices[-1] - vertices[0])**2).sum()), 0.0)
+    if not wrap:
+        # remove last vertice (should be equal to the first one)
+        vertices = np.delete(vertices, -1, axis=0)
+
+    # Array of shifted indices (on vertices)
+    ind = np.hstack((np.arange(1, vertices.shape[0]), [0]))
+
+    # Initialization
+    xx = np.atleast_2d(x)
+    res = np.zeros(xx.shape[0], dtype='bool')
+    if return_sum_of_angles:
+        res_sum = np.full((xx.shape[0],), np.nan)
+
+    xmin, ymin = vertices.min(axis=0)
+    xmax, ymax = vertices.max(axis=0)
+
+    for j, xj in enumerate(xx):
+        if xj[0] < xmin or xj[0] > xmax or xj[1] < ymin or xj[1] > ymax:
+            continue
+
+        # Set a, b, c: edge of the triangles to compute angle between a and b
+        a = xj - vertices
+
+        # a_norm2: square norm of a
+        a_norm2 = (a**2).sum(axis=1)
+
+        b = a[ind]
+        b_norm2 = a_norm2[ind]
+        ab_norm = np.sqrt(a_norm2*b_norm2)
+
+        if np.any(np.isclose(ab_norm, 0, **kwargs)):
+            # xj on the border of the polygon
+            continue
+
+        # Compute the sum of angles using the theorem of cosine
+        c = b - a
+        c_norm2 = (c**2).sum(axis=1)
+        sign = 2*(a[:,0]*b[:,1] - a[:,1]*b[:,0] > 0) - 1
+        sum_angles = np.sum(sign*np.arccos(np.minimum(1.0, np.maximum(-1.0, (a_norm2 + b_norm2 - c_norm2)/(2.0*ab_norm)))))
+
+        res[j] = np.isclose(np.abs(sum_angles), 2*np.pi)
+        if return_sum_of_angles:
+            res_sum[j] = sum_angles
+
+    if np.asarray(x).ndim == 1:
+        res = res[0]
+        if return_sum_of_angles:
+            res_sum = res_sum[0]
+
+    if return_sum_of_angles:
+        return res, res_sum
+
+    return res
+# -----------------------------------------------------------------------------
+
+# -----------------------------------------------------------------------------
+def is_in_polygon_mp(x, vertices, wrap=None, return_sum_of_angles=False, nproc=-1, **kwargs):
+    """
+    Computes the same as the function :func:`is_in_polygon`, using multiprocessing.
+
+    All the parameters except `nproc` are the same as those of the function
+    :func:`is_in_polygon`.
+
+    The number of processes used (in parallel) is n, and determined by the
+    parameter `nproc` (int, optional) as follows:
+
+    - if `nproc > 0`: n = `nproc`,
+    - if `nproc <= 0`: n = max(nmax+`nproc`, 1), where nmax is the total \
+    number of cpu(s) of the system (retrieved by `multiprocessing.cpu_count()`), \
+    i.e. all cpus except `-nproc` is used (but at least one)
+
+    See function :func:`is_in_polygon`.
+    """
+    # fname = 'is_in_polygon_mp'
+
+    # Set wrap (and adjust vertices) if needed
+    if wrap is None:
+        wrap = ~np.isclose(np.sqrt(((vertices[-1] - vertices[0])**2).sum()), 0.0)
+    if not wrap:
+        # remove last vertice (should be equal to the first one)
+        vertices = np.delete(vertices, -1, axis=0)
+
+    # Set wrap key in keywords arguments
+    kwargs['wrap'] = True
+
+    # Set return_sum_of_angles key in keywords arguments
+    kwargs['return_sum_of_angles'] = return_sum_of_angles
+
+    # Initialization
+    xx = np.atleast_2d(x)
+
+    # Set number of processes (n)
+    if nproc > 0:
+        n = nproc
+    else:
+        n = min(multiprocessing.cpu_count()+nproc, 1)
+
+    # Set index for distributing tasks
+    q, r = np.divmod(xx.shape[0], n)
+    ids_proc = [i*q + min(i, r) for i in range(n+1)]
+
+    # Set pool of n workers
+    pool = multiprocessing.Pool(n)
+    out_pool = []
+    for i in range(n):
+        # Set i-th process
+        out_pool.append(pool.apply_async(is_in_polygon, args=(xx[ids_proc[i]:ids_proc[i+1]], vertices), kwds=kwargs))
+
+    # Properly end working process
+    pool.close() # Prevents any more tasks from being submitted to the pool,
+    pool.join()  # then, wait for the worker processes to exit.
+
+    # Get result from each process
+    if return_sum_of_angles:
+        res = []
+        res_sum = []
+        for w in out_pool:
+            r, s = w.get()
+            res.append(r)
+            res_sum.append(s)
+        res = np.hstack(res)
+        res_sum = np.hstack(res_sum)
+    else:
+        res = np.hstack([w.get() for w in out_pool])
+
+    if np.asarray(x).ndim == 1:
+        res = res[0]
+        if return_sum_of_angles:
+            res_sum = res_sum[0]
+
+    if return_sum_of_angles:
+        return res, res_sum
+
+    return res
+# -----------------------------------------------------------------------------
+
+# -----------------------------------------------------------------------------
+def rasterize_polygon_2d(
+        vertices,
+        nx=None, ny=None,
+        sx=None, sy=None,
+        ox=None, oy=None,
+        xmin_ext=0.0, xmax_ext=0.0,
+        ymin_ext=0.0, ymax_ext=0.0,
+        wrap=None,
+        logger=None,
+        **kwargs):
+    """
+    Rasterizes a polygon (close line) in a 2D grid.
+
+    This function returns an image with one variable indicating for each cell
+    if it is inside (1) or outside (0) the polygon defined by the given
+    vertices.
+
+    The grid geometry of the output image is set by the given parameters or
+    computed from the vertices, as in function :func:`geone.img.imageFromPoints`,
+    i.e. for the x axis (similar for y):
+
+    - `ox` (origin), `nx` (number of cells) and `sx` (resolution, cell size)
+    - or only `nx`: `ox` and `sx` automatically computed
+    - or only `sx`: `ox` and `nx` automatically computed
+
+    In the two last cases, the parameters `xmin_ext`, `xmax_ext`, are used and
+    the approximate limit of the grid along x axis is set to x0, x1, where
+
+    - x0: min x coordinate of the vertices minus `xmin_ext`
+    - x1: max x coordinate of the vertices plus `xmax_ext`
+
+    Parameters
+    ----------
+    vertices : 2D array
+        vertices of a polygon in 2D, each row of `vertices` contains the two
+        coordinates of one vertex; the segments of the polygon are obtained by
+        linking two successive vertices (as well as the last one with the first
+        one, if `wrap=True` (see below)), so that the vertices form a close line
+        (clockwise or counterclockwise)
+
+    nx : int, optional
+        number of grid cells along x axis; see above for possible inputs
+
+    ny : int, optional
+        number of grid cells along y axis; see above for possible inputs
+
+    sx : float, optional
+        cell size along x axis; see above for possible inputs
+
+    sy : float, optional
+        cell size along y axis; see above for possible inputs
+
+    ox : float, optional
+        origin of the grid along x axis (x coordinate of cell border);
+        see above for possible
+
+    oy : float, optional
+        origin of the grid along y axis (y coordinate of cell border);
+        see above for possible
+
+        Note: `(ox, oy)` is the "lower-left" corner of the grid
+
+    xmin_ext : float, default: 0.0
+        extension beyond the min x coordinate of the vertices (see above)
+
+    xmax_ext : float, default: 0.0
+        extension beyond the max x coordinate of the vertices (see above)
+
+    ymin_ext : float, default: 0.0
+        extension beyond the min y coordinate of the vertices (see above)
+
+    ymax_ext : float, default: 0.0
+        extension beyond the max y coordinate of the vertices (see above)
+
+    wrap : bool, optional
+        - if `True`: last and first vertices has to be linked to form a close line
+        - if `False`: last and first vertices should be the same ones (i.e. the \
+        vertices form a close line);
+
+        by default (`None`): `wrap` is automatically computed
+
+    logger : :class:`logging.Logger`, optional
+        logger (see package `logging`)
+        if specified, messages are written via `logger` (no print)
+
+    kwargs:
+        keyword arguments passed to function :func:`is_in_polygon`
+
+    Returns
+    -------
+    im : :class:`geone.img.Img`
+        output image (see above);
+        note: the image grid is defined in 3D with `nz=1`, `sz=1.0`, `oz=-0.5`
+    """
+    # fname = 'rasterize_polygon_2d'
+
+    # Define grid geometry (image with no variable)
+    im = img.imageFromPoints(vertices,
+                             nx=nx, ny=ny, sx=sx, sy=sy, ox=ox, oy=oy,
+                             xmin_ext=xmin_ext, xmax_ext=xmax_ext,
+                             ymin_ext=ymin_ext, ymax_ext=ymax_ext, 
+                             logger=logger)
+
+    # Rasterize: for each cell, check if its center is within the grid
+    v = np.asarray(is_in_polygon(np.array((im.xx().reshape(-1), im.yy().reshape(-1))).T, vertices, wrap=wrap, **kwargs)).astype('float')
+    im.append_var(v, varname='in', logger=logger)
+
+    return im
+# -----------------------------------------------------------------------------
+
+# -----------------------------------------------------------------------------
+def rasterize_polygon_2d_mp(
+        vertices,
+        nx=None, ny=None,
+        sx=None, sy=None,
+        ox=None, oy=None,
+        xmin_ext=0.0, xmax_ext=0.0,
+        ymin_ext=0.0, ymax_ext=0.0,
+        wrap=None,
+        nproc=-1,
+        logger=None,
+        **kwargs):
+    """
+    Computes the same as the function :func:`rasterize_polygon_2d`, using multiprocessing.
+
+    All the parameters except `nproc` are the same as those of the function
+    :func:`rasterize_polygon_2d`.
+
+    The number of processes used (in parallel) is n, and determined by the
+    parameter `nproc` (int, optional) as follows:
+
+    - if `nproc > 0`: n = `nproc`,
+    - if `nproc <= 0`: n = max(nmax+`nproc`, 1), where nmax is the total \
+    number of cpu(s) of the system (retrieved by `multiprocessing.cpu_count()`), \
+    i.e. all cpus except `-nproc` is used (but at least one).
+
+    See function :func:`rasterize_polygon_2d`.
+    """
+    # fname = 'rasterize_polygon_2d_mp'
+
+    # Define grid geometry (image with no variable)
+    im = img.imageFromPoints(vertices,
+                             nx=nx, ny=ny, sx=sx, sy=sy, ox=ox, oy=oy,
+                             xmin_ext=xmin_ext, xmax_ext=xmax_ext,
+                             ymin_ext=ymin_ext, ymax_ext=ymax_ext,
+                             logger=logger)
+
+    # Rasterize: for each cell, check if its center is within the grid
+    v = np.asarray(is_in_polygon_mp(np.array((im.xx().reshape(-1), im.yy().reshape(-1))).T, vertices, wrap=wrap, nproc=nproc, **kwargs)).astype('float')
+    im.append_var(v, varname='in', logger=logger)
+
+    return im
+# -----------------------------------------------------------------------------
+
+# -----------------------------------------------------------------------------
+def curv_coord_2d_from_center_line(
+        x, cl_position, im_cl_dist,
+        cl_u=None,
+        gradx=None,
+        grady=None,
+        dg=None,
+        gradtol=1.e-5,
+        path_len_max=10000,
+        return_path=False,
+        verbose=1,
+        logger=None):
+    """
+    Computes curvilinear coordinates in 2D from a center line, for points given in standard coordinates.
+
+    This functions allows to change coordinates system in 2D. For a point in 2D,
+    let the coordinates
+
+    - u = (u1, u2) (in 2D) in curvilinear system
+    - x = (x1, x2) (in 2D) in standard system
+
+    The curvilinear coordinates system (u) is defined according to a center line
+    in a 2D grid as follows:
+
+    - considering the distance map (geone image `im_cl_dist`) of L2 distance \
+    to the center line (`cl_position`)
+    - the path from x to the point I on the center line is computed, descending \
+    the gradient (`gradx`, `grady`) of the distance map
+    - u = (u1, u2) is defined as:
+        - u1: the distance along the center line to the point I,
+        - u2: +/-the value of the distance map at x, with
+            * sign + for point "at left" of the center line and,
+            * sign - for point "at right" of the center line.
+
+    Parameters
+    ----------
+    x : 2D array-like or 1D array-like
+        points coordinates in standard system (should be in the grid of the image
+        `im_cl_dist`, see below), each row `x[i]` (if 2D array-like) (or `x` if
+        1D array-like) contains the two coordinates of one point
+
+    cl_position : 2D array of shape (n, 2)
+        position of the points of the center line (in standard system);
+        note: the distance between two successive points of the center line gives
+        the resolution of the u1 coordinate
+
+    im_cl_dist : :class:`geone.img.Img`
+        image of the distance to the center line:
+
+        - its grid is the "support" of standard coordinate system and it \
+        should contain all the points `x`
+        - the center line (`cl_position`) should "separate" the image grid \
+        in two (disconnected) regions
+        - this image can be computed using the function \
+        `geone.geosclassicinterface.imgDistanceImage`
+
+    cl_u : 1D array-like of length n, optional
+        distance along the center line (automatically computed if not given
+        (`None`)), used for determining u1 coordinate
+
+    gradx, grady : 2D array-like, optional
+        gradient of the distance to the centerline, array of shape
+        `(im_cl_dist.ny, im_cl_dist.nx)` (automatically computed if not given
+        (`None`));
+        `gradx[iy, ix], grady[iy, ix]`: gradient in grid cell of index `iy`, `ix`
+        along x and y axes respectively
+
+    dg : float, optional
+        step (length) used for descending the gradient map; by default (`None`):
+        `dg` is set to the minimal distance between two successive points in
+        `cl_position`, (i.e. minimal difference between two succesive values in
+        `cl_u`)
+
+    gradtol : float, default: 1.e-5
+        tolerance for the gradient magnitude (if the magintude of the gradient
+        is below `gradtol`, it is considered as zero vector)
+
+    path_len_max : int, default: 10000
+        maximal length of the path from the initial point(s) (`x`) to the center
+        line
+
+    return_path : bool, default: False
+        indicates if the path(s) from the initial point(s) (`x`) to the point(s)
+        I of the centerline (descending the gradient of the distance map) is
+        (are) retrieved
+
+    verbose : int, default: 1
+        verbose mode, integer >=0, higher implies more display
+
+    logger : :class:`logging.Logger`, optional
+        logger (see package `logging`)
+        if specified, messages are written via `logger` (no print)
+
+    Returns
+    -------
+    u : 2D array or 1D array (same shape as `x`)
+        coordinates in curvilinear system of the input point(s) `x` (see above)
+
+    x_path : list of 2D arrays (or 2D array), optional
+        path(s) from the initial point(s) to the point(s) I of the centerline
+        (descending the gradient of the distance map):
+
+        - `x_path[i]` : 2D array of floats with 2 columns
+            * `xpath[i][j]` is the coordinates (in standard system) of the \
+            j-th point of the path from `x[i]` to the center line
+
+        note: if `x` is reduced to one point and given as 1D array-like, then
+        `x_path` is a 2D array of floats with 2 columns containing the path from
+        `x` to the center line;
+        returned if `returned_path=True`
+    """
+    fname = 'curv_coord_2d_from_center_line'
+
+    if cl_u is None:
+        cl_u = np.insert(np.cumsum(np.sqrt(((cl_position[1:,:] - cl_position[:-1,:])**2).sum(axis=1))), 0, 0.0)
+
+    if gradx is None or grady is None:
+        # Gradient of distance map
+        grady, gradx = np.gradient(im_cl_dist.val[0,0])
+        gradx = gradx / im_cl_dist.sx
+        grady = grady / im_cl_dist.sy
+
+    if dg is None:
+        dg = np.diff(cl_u).min()
+
+    x = np.asarray(x)
+    x_ndim = x.ndim
+    x = np.atleast_2d(x)
+    u = np.zeros_like(x)
+
+    if return_path:
+        x_path = []
+
+    for i in range(x.shape[0]):
+        # Treat the i-th point
+        x_cur = x[i]
+        if return_path:
+            xi_path = [x_cur]
+
+        u2 = 0.0
+        d_prev = np.inf
+
+        for j in range(path_len_max):
+            # index in the grid (and interpolation factor) of the "current" point
+            tx = max(0, min((x_cur[0] - im_cl_dist.ox)/im_cl_dist.sx - 0.5, im_cl_dist.nx - 1.00001))
+            ix = int(tx)
+            tx = tx - int(tx)
+
+            ty = max(0, min((x_cur[1] - im_cl_dist.oy)/im_cl_dist.sy - 0.5, im_cl_dist.ny - 1.00001))
+            iy = int(ty)
+            ty = ty - int(ty)
+
+            # "current" distance to the center line
+            d_cur = (1.-ty)*((1.-tx)*im_cl_dist.val[0, 0, iy, ix] + tx*im_cl_dist.val[0, 0, iy, ix+1]) + ty*((1.-tx)*im_cl_dist.val[0, 0, iy+1, ix] + tx*im_cl_dist.val[0, 0, iy+1, ix+1])
+
+            # if np.abs(d_cur) < dg or d_cur*d_prev < 0:
+            if d_cur < dg or d_cur > d_prev:
+                break
+
+            # "current" gradient
+            gradx_cur = (1.-ty)*((1.-tx)*gradx[iy, ix] + tx*gradx[iy, ix+1]) + ty*((1.-tx)*gradx[iy+1, ix] + tx*gradx[iy+1, ix+1])
+            grady_cur = (1.-ty)*((1.-tx)*grady[iy, ix] + tx*grady[iy, ix+1]) + ty*((1.-tx)*grady[iy+1, ix] + tx*grady[iy+1, ix+1])
+
+            gradl = np.sqrt(gradx_cur**2 + grady_cur**2)
+            if gradl < gradtol:
+                break
+
+            # compute next point descending the gradient
+            # x_cur = x_cur - np.sign(d_cur) * dg*np.array([gradx_cur, grady_cur])/gradl
+            x_cur = x_cur - dg*np.array([gradx_cur, grady_cur])/gradl
+            if return_path:
+                xi_path.append(x_cur)
+
+            u2 = u2 + dg
+            d_prev = d_cur
+
+        # Finalize the computation of coordinate (u1, u2)
+        d_cur = ((cl_position[:-1,:] - x_cur)**2).sum(axis=1)
+        try:
+            k = np.where(d_cur == d_cur.min())[0][0]
+        except:
+            # k = len(cl_position[-2])
+            k = len(cl_position) - 2
+            if verbose > 0:
+                if logger:
+                    logger.warning(f'{fname}: closest point on center line not found (last segment selected)')
+                else:
+                    print(f'{fname}: WARNING: closest point on center line not found (last segment selected)')
+        
+        u1 = cl_u[k]
+
+        s = np.sign(np.linalg.det(np.vstack((cl_position[k+1] - cl_position[k], x[i] - x_cur))))
+        u2 = s*u2
+
+        u[i] = np.array([u1, u2])
+
+        if return_path:
+            x_path.append(np.asarray(xi_path))
+
+    if return_path:
+        if x_ndim == 1:
+            u = u[0]
+            x_path = x_path[0]
+        return u, x_path
+    else:
+        if x_ndim == 1:
+            u = u[0]
+        return u
+# -----------------------------------------------------------------------------
+
+# -----------------------------------------------------------------------------
+def curv_coord_2d_from_center_line_mp(
+        x, cl_position, im_cl_dist,
+        cl_u=None,
+        gradx=None,
+        grady=None,
+        dg=None,
+        gradtol=1.e-5,
+        path_len_max=10000,
+        return_path=False,
+        nproc=-1,
+        logger=None):
+    """
+    Computes the same as the function :func:`curv_coord_2d_from_center_line`, using multiprocessing.
+
+    All the parameters except `nproc` are the same as those of the function
+    :func:`curv_coord_2d_from_center_line`.
+
+    The number of processes used (in parallel) is n, and determined by the
+    parameter `nproc` (int, optional) as follows:
+
+    - if `nproc > 0`: n = `nproc`,
+    - if `nproc <= 0`: n = max(nmax+`nproc`, 1), where nmax is the total \
+    number of cpu(s) of the system (retrieved by `multiprocessing.cpu_count()`), \
+    i.e. all cpus except `-nproc` is used (but at least one).
+
+    See function :func:`curv_coord_2d_from_center_line`.
+    """
+    # fname = 'curv_coord_2d_from_center_line_mp'
+
+    # Initialization
+    xx = np.atleast_2d(x)
+
+    # Set number of processes (n)
+    if nproc > 0:
+        n = nproc
+    else:
+        n = min(multiprocessing.cpu_count()+nproc, 1)
+
+    # Set index for distributing tasks
+    q, r = np.divmod(xx.shape[0], n)
+    ids_proc = [i*q + min(i, r) for i in range(n+1)]
+
+    kwargs = dict(cl_u=cl_u, gradx=gradx, grady=grady, dg=dg, gradtol=gradtol, path_len_max=path_len_max, return_path=return_path, verbose=0, logger=logger)
+    # Set pool of n workers
+    pool = multiprocessing.Pool(n)
+    out_pool = []
+    for i in range(n):
+        # Set i-th process
+        out_pool.append(pool.apply_async(curv_coord_2d_from_center_line, args=(xx[ids_proc[i]:ids_proc[i+1]], cl_position, im_cl_dist), kwds=kwargs))
+
+    # Properly end working process
+    pool.close() # Prevents any more tasks from being submitted to the pool,
+    pool.join()  # then, wait for the worker processes to exit.
+
+    # Get result from each process
+    if return_path:
+        u = []
+        x_path = []
+        for p in out_pool:
+            u_p, x_path_p = p.get()
+            u.extend(u_p)
+            x_path.extend(x_path_p)
+        if np.asarray(x).ndim == 1:
+            u = u[0]
+            x_path = x_path[0]
+        return u, x_path
+    else:
+        u = np.vstack([p.get() for p in out_pool])
+        if np.asarray(x).ndim == 1:
+            u = u[0]
+        return u
+# -----------------------------------------------------------------------------
+
+##### OLD BELOW #####
+# # -----------------------------------------------------------------------------
+# def sector_angle(x, line, **kwargs):
+#     """
+#     Checks if point(s) (`x`) is (are) in a polygon given by its vertices
+#     `vertices` forming a close line.
+#
+#     To check if a point is in the polygon, the method consists in computing
+#     the vectors from the given point to the vertices of the polygon and
+#     the sum of signed angles between two successives vectors (and the last
+#     and first one). The point is in the polygon if and only if this sum is
+#     equal to +/- 2 pi.
+#
+#     :param x:       (1d-array of 2 floats, or 2d-array of shape (m, 2))
+#                         one or several points in 2D
+#     :param vertices:(2d-array of shape (n,2)) vertices of a polygon in 2D,
+#                         the segments of the polygon are obtained by linking
+#                         two successive vertices (and the last one with the
+#                         first one, if `wrap` is True (see below) or automatically
+#                         checked if it is needed), so that the vertices form a
+#                         close line (clockwise or counterclockwise)
+#     :param wrap:    (bool)
+#                         - if True, last and first vertices has to be linked
+#                             to form a close line,
+#                         - if False, last and first vertices should be the same
+#                             ones (i.e. the vertices form a close line)
+#                         - if None, automatically computed
+#     :param kwargs:  keyword arguments passed to `numpy.isclose` (see below)
+#
+#     :return:        (bool or 1d-array of bool of shape (m,)) True / False
+#                         value(s) indicating if the point(s) `x` is (are)
+#                         inside / outside the polygon
+#     """
+#
+#     # Initialization
+#     xx = np.atleast_2d(x)
+#     res = np.zeros(xx.shape[0], dtype='bool')
+#
+#     for j, xj in enumerate(xx):
+#         # Set a, b, c: edge of the triangles to compute angle between a and b
+#         v = xj - vertices
+#
+#         # v_norm2: square norm of v
+#         v_norm2 = (a**2).sum(axis=1)
+#
+#         a = v[:-1]
+#         b = v[1:]
+# #        b_norm2 = v_norm2[1:]
+#         ab_norm = np.sqrt(v_norm2[:-1]*v_norm2[1:])
+#
+#         ind = np.isclose(ab_norm, 0, **kwargs)
+#         # Compute the sum of angles using the theorem of cosine
+#         c = b - a
+#         c_norm2 = (c**2).sum(axis=1)
+#         sign = 2*(a[:,0]*b[:,1] - a[:,1]*b[:,0] > 0) - 1
+#         sum_angles = np.sum(sign[ind]*np.arccos(np.minimum(1.0, np.maximum(-1.0, (a_norm2[ind] + b_norm2[ind] - c_norm2[ind])/(2.0*ab_norm[ind])))))
+#         res[i] = sum_angles > 0
+#
+#     return res
+# # -----------------------------------------------------------------------------