Python numpy.__dict__() Examples

def get_fake_data():

    # try to evaluate the source script
    try:
        # first get all the extra numpy globals
        g = _n.__dict__

        # update these globals with extra stuff needed for evaluation
        g.update(dict(t=d1[0]+s["settings/simulated_input/noise"]*_n.random.rand()))
        y = eval(s["settings/simulated_input/source"], g)+_n.random.random(len(d1['t']))

    # default to zeros
    except:
        print("ERROR: Invalid source script.")
        y = d1[0]*0.0

    # pretend this acquisition actually took time (avoids black holes)
    _t.sleep(s["settings/simulated_input/duration"])

    return y


# define a function to be called whenever the acquire button is pressed 

def _globals(self):
        """
        Returns the globals needed for eval() statements.
        """

        # start with numpy
        globbies = dict(_n.__dict__)
        globbies.update(_special.__dict__)

        # update with required stuff
        globbies.update({'h':self.h, 'c':self.c, 'd':self, 'self':self})

        # update with user stuff
        globbies.update(self.extra_globals)

        return globbies 

def _rescale(arr, ndv, dst_dtype, out_alpha=True):
    """Convert an array from output dtype, scaling up linearly
    """
    if dst_dtype == np.__dict__['uint16']:
        scale = 1
    else:
        # convert to 8bit value range in place
        scale = float(np.iinfo(np.uint16).max) / float(np.iinfo(np.uint8).max)

    res = (arr / scale).astype(dst_dtype)

    if out_alpha:
        mask = _simple_mask(
            arr.astype(dst_dtype),
            (ndv, ndv, ndv)).reshape(
                1, arr.shape[1], arr.shape[2])
        return np.concatenate([res, mask])
    else:
        return res 

def test_pansharp_data():
    b8_path = 'tests/fixtures/tiny_20_tiffs/LC81070352015122LGN00/'\
              'LC81070352015122LGN00_B8.tif'
    b4_path = 'tests/fixtures/tiny_20_tiffs/LC81070352015122LGN00/'\
              'LC81070352015122LGN00_B4.tif'
    b3_path = 'tests/fixtures/tiny_20_tiffs/LC81070352015122LGN00/'\
              'LC81070352015122LGN00_B3.tif'
    b2_path = 'tests/fixtures/tiny_20_tiffs/LC81070352015122LGN00/'\
              'LC81070352015122LGN00_B2.tif'
    band_paths = [b8_path, b4_path, b3_path, b2_path]
    pan_window = ((1536, 1792), (1280, 1536))
    g_args = {'half_window': False,
              'dst_aff': Affine(75.00483870967741, 0.0, 300892.5,
                                0.0, -75.00475285171103, 4107007.5),
              'verb': False, 'weight': 0.2,
              'dst_crs': {'init': u'epsg:32654'},
              'r_crs': {'init': u'epsg:32654'},
              'dst_dtype': np.__dict__['uint16'],
              'r_aff': Affine(150.0193548387097, 0.0, 300885.0,
                              0.0, -150.0190114068441, 4107015.0),
              'src_nodata': 0}

    return [rasterio.open(f) for f in band_paths],\
        pan_window, (6, 5), g_args 

def nbytes(self):
        """ number of bytes of the object """
        n = sum(k.nbytes if hasattr(k, 'nbytes') else sys.getsizeof(k)
                for k in self.__dict__.values())
        return n 

def evalexpr(self, expr, exprvars=None, dtype=float):
        """ evaluate expression based on the data and external variables
            all np function can be used (log, exp, pi...)

        Parameters
        ----------
        expr: str
            expression to evaluate on the table
            includes mathematical operations and attribute names

        exprvars: dictionary, optional
            A dictionary that replaces the local operands in current frame.

        dtype: dtype definition
            dtype of the output array

        Returns
        -------
        out : NumPy array
            array of the result
        """
        _globals = {}
        for k in ( list(self.colnames) + list(self._aliases.keys()) ):
            if k in expr:
                _globals[k] = self[k]

        if exprvars is not None:
            if (not (hasattr(exprvars, 'keys') & hasattr(exprvars, '__getitem__' ))):
                raise AttributeError("Expecting a dictionary-like as condvars")
            for k, v in ( exprvars.items() ):
                _globals[k] = v

        # evaluate expression, to obtain the final filter
        r    = np.empty( self.nrows, dtype=dtype)
        r[:] = eval(expr, _globals, np.__dict__)

        return r 

def test_initialization():
    """ Test structure initialization. """
    a = Structure()
    assert all(abs(a.cell - identity(3)) < 1e-8)
    assert abs(a.scale - 1e0 * angstrom) < 1e0
    assert len(a.__dict__) == 3

    a = Structure(identity(3) * 2.5, scale=5.45)
    assert all(abs(a.cell - identity(3) * 2.5) < 1e-8)
    assert abs(a.scale - 5.45 * angstrom) < 1e0
    assert len(a.__dict__) == 3

    a = Structure(identity(3) * 2.5, scale=0.545 * nanometer)
    assert all(abs(a.cell - identity(3) * 2.5) < 1e-8)
    assert abs(a.scale - 5.45 * angstrom) < 1e0
    assert len(a.__dict__) == 3

    a = Structure(2.5, 0, 0, 0, 2.5, 0, 0, 0, 2.5, scale=5.45)
    assert all(abs(a.cell - identity(3) * 2.5) < 1e-8)
    assert abs(a.scale - 5.45 * angstrom) < 1e0
    assert len(a.__dict__) == 3

    a = Structure([2.5, 0, 0], [0, 2.5, 0], [0, 0, 2.5], scale=5.45)
    assert all(abs(a.cell - identity(3) * 2.5) < 1e-8)
    assert abs(a.scale - 5.45 * angstrom) < 1e0
    assert len(a.__dict__) == 3

    a = Structure(cell=[[2.5, 0, 0], [0, 2.5, 0], [0, 0, 2.5]], scale=5.45)
    assert all(abs(a.cell - identity(3) * 2.5) < 1e-8)
    assert abs(a.scale - 5.45 * angstrom) < 1e0
    assert len(a.__dict__) == 3

    a = Structure(identity(3) * 2.5, scale=5.45, m=True)
    assert all(abs(a.cell - identity(3) * 2.5) < 1e-8)
    assert abs(a.scale - 5.45 * angstrom) < 1e0
    assert len(a.__dict__) == 4 and getattr(a, 'm', False) 

def test_representability():
    import quantities
    import numpy
    dictionary = {Structure.__name__: Structure}
    dictionary.update(numpy.__dict__)
    dictionary.update(quantities.__dict__)

    expected = Structure()
    actual = eval(repr(expected), dictionary)
    assert all(abs(expected.cell - actual.cell) < 1e-8)
    assert abs(expected.scale - actual.scale) < 1e-8
    assert len(expected) == len(actual)

    expected = Structure([1, 2, 0], [3, 4, 5], [6, 7, 8], m=True)
    actual = eval(repr(expected), dictionary)
    assert all(abs(expected.cell - actual.cell) < 1e-8)
    assert abs(expected.scale - actual.scale) < 1e-8
    assert len(expected) == len(actual)
    assert getattr(expected, 'm', False) == actual.m

    expected = Structure([1, 2, 0], [3, 4, 5], [6, 7, 8], m=True) \
        .add_atom(0, 1, 2, "Au", m=5) \
        .add_atom(0, -1, -2, "Pd")
    actual = eval(repr(expected), dictionary)
    assert all(abs(expected.cell - actual.cell) < 1e-8)
    assert abs(expected.scale - actual.scale) < 1e-8
    assert len(expected) == len(actual)
    assert all(abs(expected[0].pos - actual[0].pos) < 1e-8)
    assert getattr(expected[0], 'm', 0) == actual[0].m
    assert expected[0].type == actual[0].type
    assert all(abs(expected[1].pos - actual[1].pos) < 1e-8)
    assert expected[1].type == actual[1].type
    assert getattr(expected, 'm', False) == actual.m 

def substitute_with_eval(expression: sympy.Expr,
                         substitutions: Dict[str, Union[sympy.Expr, numpy.ndarray, str]]) -> sympy.Expr:
    """Substitutes only sympy.Symbols. Workaround for numpy like array behaviour. ~Factor 3 slower compared to subs"""
    substitutions = {k: v if isinstance(v, sympy.Expr) else sympify(v)
                     for k, v in substitutions.items()}

    for symbol in get_free_symbols(expression):
        symbol_name = str(symbol)
        if symbol_name not in substitutions:
            substitutions[symbol_name] = symbol

    string_representation = sympy.srepr(expression)
    return eval(string_representation, sympy.__dict__, {'Symbol': substitutions.__getitem__,
                                                        'Mul': numpy_compatible_mul}) 

def test_pansharpen_worker_uint8(test_pansharp_data):
    open_files, pan_window, _, g_args = test_pansharp_data
    g_args.update(dst_dtype=np.__dict__['uint8'])
    pan_output = _pansharpen_worker(open_files, pan_window, _, g_args)
    assert pan_output.dtype == np.uint8
    assert np.max(pan_output) <= 2**8 

def test_rescale(arr, ndv, dst_dtype):
    if dst_dtype == np.__dict__['uint16']:
        assert np.array_equal(
                _rescale(arr, ndv, dst_dtype),
                np.concatenate(
                    [
                        (arr).astype(dst_dtype),
                        _simple_mask(
                            arr.astype(dst_dtype),
                            (ndv, ndv, ndv)
                        ).reshape(1, arr.shape[1], arr.shape[2])
                    ]
                )
            )
    else:
        assert np.array_equal(
                _rescale(arr, ndv, dst_dtype),
                np.concatenate(
                    [
                        (arr / 257.0).astype(dst_dtype),
                        _simple_mask(
                            arr.astype(dst_dtype),
                            (ndv, ndv, ndv)
                        ).reshape(1, arr.shape[1], arr.shape[2])
                    ]
                )
            )


# Testing make_windows_block function's random element 

def nbytes(self):
        """ number of bytes of the object """
        n = sum(k.nbytes if hasattr(k, 'nbytes') else sys.getsizeof(k)
                for k in self.__dict__.values())
        return n 

def evalexpr(self, expr, exprvars=None, dtype=float):
        """ evaluate expression based on the data and external variables
            all np function can be used (log, exp, pi...)

        Parameters
        ----------
        expr: str
            expression to evaluate on the table
            includes mathematical operations and attribute names

        exprvars: dictionary, optional
            A dictionary that replaces the local operands in current frame.

        dtype: dtype definition
            dtype of the output array

        Returns
        -------
        out : NumPy array
            array of the result
        """
        _globals = {}
        for k in ( list(self.colnames) + list(self._aliases.keys()) ):
            if k in expr:
                _globals[k] = self[k]

        if exprvars is not None:
            if (not (hasattr(exprvars, 'keys') & hasattr(exprvars, '__getitem__' ))):
                raise AttributeError("Expecting a dictionary-like as condvars")
            for k, v in ( exprvars.items() ):
                _globals[k] = v

        # evaluate expression, to obtain the final filter
        r    = np.empty( self.nrows, dtype=dtype)
        r[:] = eval(expr, _globals, np.__dict__)

        return r 

def nbytes(self):
        """ number of bytes of the object """
        n = sum(k.nbytes if hasattr(k, 'nbytes') else sys.getsizeof(k)
                for k in self.__dict__.values())
        return n 

def evalexpr(self, expr, exprvars=None, dtype=float):
        """ evaluate expression based on the data and external variables
            all np function can be used (log, exp, pi...)

        Parameters
        ----------
        expr: str
            expression to evaluate on the table
            includes mathematical operations and attribute names

        exprvars: dictionary, optional
            A dictionary that replaces the local operands in current frame.

        dtype: dtype definition
            dtype of the output array

        Returns
        -------
        out : NumPy array
            array of the result
        """
        _globals = {}
        for k in ( list(self.colnames) + list(self._aliases.keys()) ):
            if k in expr:
                _globals[k] = self[k]

        if exprvars is not None:
            if (not (hasattr(exprvars, 'keys') & hasattr(exprvars, '__getitem__' ))):
                raise AttributeError("Expecting a dictionary-like as condvars")
            for k, v in ( exprvars.items() ):
                _globals[k] = v

        # evaluate expression, to obtain the final filter
        r    = np.empty( self.nrows, dtype=dtype)
        r[:] = eval(expr, _globals, np.__dict__)

        return r 

def select(self, fields, indices=None, **kwargs):
        """
        Select only a few fields in the table

        Parameters
        ----------
        fields: str or sequence
            fields to keep in the resulting table

        indices: sequence or slice
            extract only on these indices

        returns
        -------
        tab: SimpleTable instance
            resulting table
        """
        _fields = self.keys(fields)

        if fields == '*':
            if indices is None:
                return self
            else:
                tab = self.__class__(self[indices])
                for k in self.__dict__.keys():
                    if k not in ('data', ):
                        setattr(tab, k, deepcopy(self.__dict__[k]))
                return tab
        else:
            d = {}
            for k in _fields:
                _k = self.resolve_alias(k)
                if indices is not None:
                    d[k] = self[_k][indices]
                else:
                    d[k] = self[_k]
            d['header'] = deepcopy(self.header)
            tab = self.__class__(d)
            for k in self.__dict__.keys():
                if k not in ('data', ):
                    setattr(tab, k, deepcopy(self.__dict__[k]))
            return tab 

def load_values(dir_logs, metrics, nb_epochs=-1, best=None):
    json_files = {}
    values = {}

    # load argsup of best
    if best:
        if best['json'] not in json_files:
            with open(osp.join(dir_logs, f'{best["json"]}.json')) as f:
                json_files[best['json']] = json.load(f)

        jfile = json_files[best['json']]
        vals = jfile[best['name']]
        end = len(vals) if nb_epochs == -1 else nb_epochs
        argsup = np.__dict__[f'arg{best["order"]}'](vals[:end])

    # load logs
    for _key, metric in metrics.items():
        # open json_files
        if metric['json'] not in json_files:
            with open(osp.join(dir_logs, f'{metric["json"]}.json')) as f:
                json_files[metric['json']] = json.load(f)

        jfile = json_files[metric['json']]

        if 'train' in metric['name']:
            epoch_key = 'train_epoch.epoch'
        else:
            epoch_key = 'eval_epoch.epoch'

        if epoch_key in jfile:
            epochs = jfile[epoch_key]
        else:
            epochs = jfile['epoch']

        vals = jfile[metric['name']]
        if not best:
            end = len(vals) if nb_epochs == -1 else nb_epochs
            argsup = np.__dict__[f'arg{metric["order"]}'](vals[:end])

        try:
            values[metric['name']] = epochs[argsup], vals[argsup]
        except IndexError:
            values[metric['name']] = epochs[argsup - 1], vals[argsup - 1]
    return values 

def select(self, fields, indices=None, **kwargs):
        """
        Select only a few fields in the table

        Parameters
        ----------
        fields: str or sequence
            fields to keep in the resulting table

        indices: sequence or slice
            extract only on these indices

        returns
        -------
        tab: SimpleTable instance
            resulting table
        """
        _fields = self.keys(fields)

        if fields == '*':
            if indices is None:
                return self
            else:
                tab = self.__class__(self[indices])
                for k in self.__dict__.keys():
                    if k not in ('data', ):
                        setattr(tab, k, deepcopy(self.__dict__[k]))
                return tab
        else:
            d = {}
            for k in _fields:
                _k = self.resolve_alias(k)
                if indices is not None:
                    d[k] = self[_k][indices]
                else:
                    d[k] = self[_k]
            d['header'] = deepcopy(self.header)
            tab = self.__class__(d)
            for k in self.__dict__.keys():
                if k not in ('data', ):
                    setattr(tab, k, deepcopy(self.__dict__[k]))
            return tab