misc.py

# -*- coding: utf-8 -*-
"""
Various additional utilities for ObsPy.

:copyright:
    The ObsPy Development Team (devs@obspy.org)
:license:
    GNU Lesser General Public License, Version 3
    (https://www.gnu.org/copyleft/lesser.html)
"""
from __future__ import (absolute_import, division, print_function,
                        unicode_literals)
from future.utils import PY2

import contextlib
import inspect
import io
import itertools
import locale
import math
import os
import shutil
from subprocess import STDOUT, CalledProcessError, check_output
import sys
import tempfile
import warnings
from pkg_resources import load_entry_point

from future.builtins import *  # NOQA

import numpy as np


WIN32 = sys.platform.startswith('win32')

# The following dictionary maps the first character of the channel_id to the
# lowest sampling rate this so called Band Code should be used for according
# to: SEED MANUAL p.124
# We use this e.g. in seishub.client.getWaveform to request two samples more on
# both start and end to cut to the samples that really are nearest to requested
# start/end time afterwards.
BAND_CODE = {'F': 1000.0,
             'G': 1000.0,
             'D': 250.0,
             'C': 250.0,
             'E': 80.0,
             'S': 10.0,
             'H': 80.0,
             'B': 10.0,
             'M': 1.0,
             'L': 1.0,
             'V': 0.1,
             'U': 0.01,
             'R': 0.0001,
             'P': 0.000001,
             'T': 0.0000001,
             'Q': 0.00000001}

# Dict that stores results from load entry points
_ENTRY_POINT_CACHE = {}

# The kwargs used by load_entry_point function
_LOAD_ENTRY_POINT_KEYS = ('dist', 'group', 'name')


def guess_delta(channel):
    """
    Estimate time delta in seconds between each sample from given channel name.

    :type channel: str
    :param channel: Channel name, e.g. ``'BHZ'`` or ``'H'``
    :rtype: float
    :return: Returns ``0`` if band code is not given or unknown.

    .. rubric:: Example

    >>> print(guess_delta('BHZ'))
    0.1

    >>> print(guess_delta('H'))
    0.0125

    >>> print(guess_delta('XZY'))  # doctest: +SKIP
    0
    """
    try:
        return 1. / BAND_CODE[channel[0]]
    except Exception:
        msg = "No or unknown channel id provided. Specifying a channel id " + \
              "could lead to better selection of first/last samples of " + \
              "fetched traces."
        warnings.warn(msg)
    return 0


def score_at_percentile(values, per, limit=(), issorted=True):
    """
    Calculates the score at the given per percentile of the sequence a.

    For example, the score at ``per=50`` is the median.

    If the desired quantile lies between two data points, we interpolate
    between them.

    If the parameter ``limit`` is provided, it should be a tuple (lower,
    upper) of two values.  Values of ``a`` outside this (closed) interval
    will be ignored.

    .. rubric:: Examples

    >>> a = [1, 2, 3, 4]
    >>> score_at_percentile(a, 25)
    1.75
    >>> score_at_percentile(a, 50)
    2.5
    >>> score_at_percentile(a, 75)
    3.25

    >>> a = [6, 47, 49, 15, 42, 41, 7, 255, 39, 43, 40, 36, 500]
    >>> score_at_percentile(a, 25, limit=(0, 100))
    25.5
    >>> score_at_percentile(a, 50, limit=(0, 100))
    40
    >>> score_at_percentile(a, 75, limit=(0, 100))
    42.5

    This function is taken from :func:`scipy.stats.score_at_percentile`.

    Copyright (c) Gary Strangman
    """
    if limit:
        values = [v for v in values if limit[0] < v < limit[1]]

    if issorted:
        values = sorted(values)

    def _interpolate(a, b, fraction):
        return a + (b - a) * fraction

    idx = per / 100. * (len(values) - 1)
    if (idx % 1 == 0):
        return values[int(idx)]
    else:
        return _interpolate(values[int(idx)], values[int(idx) + 1], idx % 1)


def flat_not_masked_contiguous(a):
    """
    Find contiguous unmasked data in a masked array along the given axis.

    This function is taken from
    :func:`numpy.ma.flatnotmasked_contiguous`.

    Copyright (c) Pierre Gerard-Marchant
    """
    np.ma.extras.flatnotmasked_contiguous
    m = np.ma.getmask(a)
    if m is np.ma.nomask:
        return slice(0, a.size, None)
    i = 0
    result = []
    for (k, g) in itertools.groupby(m.ravel()):
        n = len(list(g))
        if not k:
            result.append(slice(i, i + n))
        i += n
    return result


def complexify_string(line):
    """
    Converts a string in the form "(real, imag)" into a complex type.

    :type line: str
    :param line: String in the form ``"(real, imag)"``.
    :rtype: complex
    :return: Complex number.

    .. rubric:: Example

    >>> complexify_string("(1,2)")
    (1+2j)

    >>> complexify_string(" ( 1 , 2 ) ")
    (1+2j)
    """
    temp = line.split(',')
    return complex(float(temp[0].strip()[1:]), float(temp[1].strip()[:-1]))


def to_int_or_zero(value):
    """
    Converts given value to an integer or returns 0 if it fails.

    :param value: Arbitrary data type.
    :rtype: int

    .. rubric:: Example

    >>> to_int_or_zero("12")
    12

    >>> to_int_or_zero("x")
    0
    """
    try:
        return int(value)
    except ValueError:
        return 0


# import numpy loadtxt and check if ndmin parameter is available
try:
    from numpy import loadtxt
    loadtxt(np.array([0]), ndmin=1)
except TypeError:
    # otherwise redefine loadtxt
    def loadtxt(*args, **kwargs):
        """
        Replacement for older numpy.loadtxt versions not supporting ndmin
        parameter.
        """
        if 'ndmin' not in kwargs:
            return np.loadtxt(*args, **kwargs)
        # ok we got a ndmin param
        if kwargs['ndmin'] != 1:
            # for now we support only one dimensional arrays
            raise NotImplementedError('Upgrade your NumPy version!')
        del kwargs['ndmin']
        dtype = kwargs.get('dtype', None)
        # lets get the data
        try:
            data = np.loadtxt(*args, **kwargs)
        except IOError as e:
            # raises in older versions if no data could be read
            if 'reached before encountering data' in str(e):
                # return empty array
                return np.array([], dtype=dtype)
            # otherwise just raise
            raise
        # ensures that an array is returned
        return np.atleast_1d(data)


def get_untracked_files_from_git():
    """
    Tries to return a list of files (absolute paths) that are untracked by git
    in the repository.

    Returns `None` if the system call to git fails.
    """
    dir_ = os.path.abspath(
        os.path.dirname(inspect.getfile(inspect.currentframe())))
    dir_ = os.path.dirname(os.path.dirname(os.path.dirname(dir_)))
    try:
        # Check that the git root directory is actually the ObsPy directory.
        p = check_output(['git', 'rev-parse', '--show-toplevel'],
                         cwd=dir_, stderr=STDOUT)
        git_root_dir = p.decode().strip()
        if git_root_dir:
            git_root_dir = os.path.abspath(git_root_dir)
        if git_root_dir != dir_:
            raise ValueError('Git root directory (%s) does not match expected '
                             'path (%s).' % (git_root_dir, dir_))
        p = check_output(['git', 'status', '-u', '--porcelain'],
                         cwd=dir_, stderr=STDOUT)
        stdout = p.decode().splitlines()
        files = [os.path.abspath(os.path.join(dir_, line.split()[1].strip()))
                 for line in stdout
                 if line.startswith("??")]
    except (OSError, CalledProcessError):
        return None
    return files


if PY2:
    from cStringIO import StringIO as CaptureIO
else:
    class CaptureIO(io.TextIOWrapper):
        def __init__(self):
            super(CaptureIO, self).__init__(io.BytesIO(), encoding='utf-8',
                                            newline='\n', write_through=True)

        def getvalue(self):
            return self.buffer.getvalue().decode('utf-8')


@contextlib.contextmanager
def CatchOutput():  # NOQA
    """
    A context manager that captures input to stdout/stderr. Python level only!

    Always use with "with" statement. Does nothing otherwise.

    >>> with CatchOutput() as out:  # doctest: +SKIP
    ...    sys.stdout.write("mystdout")
    ...    sys.stderr.write("mystderr")
    >>> print(out.stdout)  # doctest: +SKIP
    mystdout
    >>> print(out.stderr)  # doctest: +SKIP
    mystderr
    """
    # Dummy class to transport the output.
    class Output():
        stdout = ''
        stderr = ''
    out = Output()

    # set current stdout/stderr to in-memory text streams
    sys.stdout = stdout_result = CaptureIO()
    sys.stderr = stderr_result = CaptureIO()

    try:
        raised = False
        yield out
    except SystemExit:
        raised = True
    finally:
        out.stdout = stdout_result.getvalue()
        out.stderr = stderr_result.getvalue()

        # reset to original stdout/stderr
        sys.stdout = sys.__stdout__
        sys.stderr = sys.__stderr__

        # normalize line breaks
        if WIN32:
            out.stdout = out.stdout.replace('\r', '')
            out.stderr = out.stderr.replace('\r', '')
        # remove encoding for PY2 -> we always want PY2 unicode/PY3 str
        if PY2:
            if sys.stdout.isatty():
                out.stdout.decode(sys.stdout.encoding)
            else:
                out.stdout.decode(locale.getpreferredencoding())
            if sys.stderr.isatty():
                out.stderr.decode(sys.stderr.encoding)
            else:
                out.stderr.decode(locale.getpreferredencoding())

    if raised:
        raise SystemExit(out.stderr)


def _py36_windowsconsoleio_workaround():
    """
    This monkey patch prevents crashing Py3.6 under Windows while using
    the SuppressOutput context manager.

    Python 3.6 implemented unicode console handling for Windows. This works
    by reading/writing to the raw console handle using
    ``{Read,Write}ConsoleW``.
    The problem is that we are going to ``dup2`` over the stdio file
    descriptors when doing ``FDCapture`` and this will ``CloseHandle`` the
    handles used by Python to write to the console. Though there is still some
    weirdness and the console handle seems to only be closed randomly and not
    on the first call to ``CloseHandle``, or maybe it gets reopened with the
    same handle value when we suspend capturing.
    The workaround in this case will reopen stdio with a different fd which
    also means a different handle by replicating the logic in
    "Py_lifecycle.c:initstdio/create_stdio".
    See https://github.com/pytest-dev/py/issues/103

    See http://bugs.python.org/issue30555
    """
    if not WIN32 or sys.version_info[:2] < (3, 6):
        return
    if not hasattr(sys.stdout, 'buffer'):
        return
    buffered = hasattr(sys.stdout.buffer, 'raw')
    raw_stdout = sys.stdout.buffer.raw if buffered else sys.stdout.buffer

    if not isinstance(raw_stdout, io._WindowsConsoleIO):
        return

    def _reopen_stdio(f, mode):
        if not buffered and mode[0] == 'w':
            buffering = 0
        else:
            buffering = -1

        return io.TextIOWrapper(
            open(os.dup(f.fileno()), mode, buffering),
            f.encoding,
            f.errors,
            f.newlines,
            f.line_buffering)

    sys.__stdin__ = sys.stdin = _reopen_stdio(sys.stdin, 'rb')
    sys.__stdout__ = sys.stdout = _reopen_stdio(sys.stdout, 'wb')
    sys.__stderr__ = sys.stderr = _reopen_stdio(sys.stderr, 'wb')


_py36_windowsconsoleio_workaround()


@contextlib.contextmanager
def SuppressOutput():  # noqa
    """
    A context manager that suppresses output to stdout/stderr.
    Always use with "with" statement. Does nothing otherwise.
    >>> with SuppressOutput():  # doctest: +SKIP
    ...    os.system('echo "mystdout"')
    ...    os.system('echo "mystderr" >&2')

    Note: Does not work reliably for Windows Python 3.6 under Windows - see
    function definition of _py36_windowsconsoleio_workaround().
    """
    with os.fdopen(os.dup(1), 'wb', 0) as tmp_stdout:
        with os.fdopen(os.dup(2), 'wb', 0) as tmp_stderr:
            with open(os.devnull, 'wb') as to_file:
                sys.stdout.flush()
                sys.stderr.flush()
                os.dup2(to_file.fileno(), 1)
                os.dup2(to_file.fileno(), 2)
                try:
                    yield
                finally:
                    sys.stdout.flush()
                    sys.stderr.flush()
                    os.dup2(tmp_stdout.fileno(), 1)
                    os.dup2(tmp_stderr.fileno(), 2)
    # reset to original stdout/stderr
    sys.stdout = sys.__stdout__
    sys.stderr = sys.__stderr__


@contextlib.contextmanager
def TemporaryWorkingDirectory():  # noqa --> this name is IMHO ok for a CM
    """
    A context manager that changes to a temporary working directory.

    Always use with "with" statement. Does nothing useful otherwise.

    >>> with TemporaryWorkingDirectory():  # doctest: +SKIP
    ...    os.system('echo "$PWD"')
    """
    tempdir = tempfile.mkdtemp(prefix='obspy-')
    old_dir = os.getcwd()
    os.chdir(tempdir)
    try:
        yield
    finally:
        os.chdir(old_dir)
        try:
            shutil.rmtree(tempdir)
        # Windows on appveyor is having issues with removing these temporary
        # directories (e.g. https://ci.appveyor.com/project/obspy/obspy/builds
        # /20718605/job/0vuja1b95rv5a0s2).
        except Exception as e:
            warnings.warn(e.__repr__())


def factorize_int(x):
    """
    Calculate prime factorization of integer.

    Could be done faster but faster algorithm have much more lines of code and
    this is fast enough for our purposes.

    https://stackoverflow.com/q/14550794

    >>> factorize_int(1800004)
    [2, 2, 450001]
    >>> factorize_int(1800003)
    [3, 19, 23, 1373]
    """
    if x == 1:
        return [1]
    factors, limit, check, num = [], int(math.sqrt(x)) + 1, 2, x
    for check in range(2, limit):
        while num % check == 0:
            factors.append(check)
            num /= check
    if num > 1:
        factors.append(int(num))
    return factors


def get_window_times(starttime, endtime, window_length, step, offset,
                     include_partial_windows):
    """
    Function calculating a list of times making up equal length windows from
    within a given time interval.

    :param starttime: The start time of the whole time interval.
    :type starttime: :class:`~obspy.core.utcdatetime.UTCDateTime`
    :param endtime: The end time of the whole time interval.
    :type endtime: :class:`~obspy.core.utcdatetime.UTCDateTime`
    :param window_length: The length of each window in seconds.
    :type window_length: float
    :param step: The step between the start times of two successive
        windows in seconds. Can be negative if an offset is given.
    :type step: float
    :param offset: The offset of the first window in seconds relative to
        the start time of the whole interval.
    :type offset: float
    :param include_partial_windows: Determines if windows that are
        shorter then 99.9 % of the desired length are returned.
    :type include_partial_windows: bool
    """
    if step > 0:
        end = endtime.timestamp - 0.001 * step
    else:
        # This minus is correct here due to the way the actual window times
        # are calculate later on.
        end = starttime.timestamp - 0.001 * abs(step)
    # Left sides of each window.
    indices = np.arange(start=starttime.timestamp + offset,
                        stop=end, step=step, dtype=np.float64)
    if step > 0:
        # Generate all possible windows.
        windows = [(_i, min(_i + window_length, endtime.timestamp))
                   for _i in indices]
    else:
        # Generate all possible windows.
        windows = [(max(_i - window_length, starttime.timestamp), _i)
                   for _i in indices]

    # Potentially remove partial windows not fulfilling the window length
    # criterion.
    if not include_partial_windows:
        windows = [_i for _i in windows
                   if abs(_i[1] - _i[0]) > 0.999 * window_length]

    t = type(starttime)
    return [(t(_i[0]), t(_i[1])) for _i in windows]


class MatplotlibBackend(object):
    """
    A helper class for switching the matplotlib backend.

    Can be used as a context manager to temporarily switch the backend or by
    using the :meth:`~MatplotlibBackend.switch_backend` staticmethod.

    The context manager has no effect when setting ``backend=None``.

    :type backend: str
    :param backend: Name of matplotlib backend to switch to.
    :type sloppy: bool
    :param sloppy: If ``True``, uses :func:`matplotlib.pyplot.switch_backend`
        and no warning will be shown if the backend was not switched
        successfully. If ``False``, additionally tries to use
        :func:`matplotlib.use` first and also shows a warning if the backend
        was not switched successfully.
    """
    def __init__(self, backend, sloppy=True):
        self.temporary_backend = backend
        self.sloppy = sloppy
        import matplotlib
        self.previous_backend = matplotlib.get_backend()

    def __enter__(self):
        if self.temporary_backend is None:
            return
        self.switch_backend(backend=self.temporary_backend, sloppy=self.sloppy)

    def __exit__(self, exc_type, exc_val, exc_tb):  # @UnusedVariable
        if self.temporary_backend is None:
            return
        self.switch_backend(backend=self.previous_backend, sloppy=self.sloppy)

    @staticmethod
    def switch_backend(backend, sloppy=True):
        """
        Switch matplotlib backend.

        :type backend: str
        :param backend: Name of matplotlib backend to switch to.
        :type sloppy: bool
        :param sloppy: If ``True``, only uses
            :func:`matplotlib.pyplot.switch_backend` and no warning will be
            shown if the backend was not switched successfully. If ``False``,
            additionally tries to use :func:`matplotlib.use` first and also
            shows a warning if the backend was not switched successfully.
        """
        import matplotlib
        # sloppy. only do a `plt.switch_backend(..)`
        if sloppy:
            import matplotlib.pyplot as plt
            plt.switch_backend(backend)
        else:
            # check if `matplotlib.use(..)` is emitting a warning
            try:
                with warnings.catch_warnings(record=True):
                    warnings.simplefilter("error", UserWarning)
                    matplotlib.use(backend)
            # if that's the case, follow up with `plt.switch_backend(..)`
            except UserWarning:
                import matplotlib.pyplot as plt
                plt.switch_backend(backend)
            # finally check if the switch was successful,
            # show a warning if not
            if matplotlib.get_backend().upper() != backend.upper():
                msg = "Unable to change matplotlib backend to '%s'" % backend
                raise Exception(msg)


def limit_numpy_fft_cache(max_size_in_mb_per_cache=100):
    """
    NumPy's FFT implementation utilizes caches to speedup subsequent FFTs of
    the same size. This accumulates memory when run for various length FFTs
    as can readily happen in seismology.

    This utility function clears both, full and real-only caches if their
    size is above the given threshold.

    The default 100 MB is fairly generous but we still want to profit from
    the cache where applicable.
    """
    for cache in ["_fft_cache", "_real_fft_cache"]:
        # Guard against different numpy versions just to be safe.
        if not hasattr(np.fft.fftpack, cache):
            continue
        cache = getattr(np.fft.fftpack, cache)
        # Check type directly and don't use isinstance() as future numpy
        # versions might use some subclass or what not.
        if type(cache) is not dict:
            continue
        # Its a dictionary with list's of arrays as the values. Wrap in
        # try/except to guard against future numpy changes.
        try:
            total_size = sum([_j.nbytes for _i in cache.values() for _j in _i])
        except Exception:
            continue
        if total_size > max_size_in_mb_per_cache * 1024 * 1024:
            cache.clear()


def buffered_load_entry_point(dist, group, name):
    """
    Return `name` entry point of `group` for `dist` or raise ImportError
    :type dist: str
    :param dist: The name of the distribution containing the entry point.
    :type group: str
    :param group: The name of the group containing the entry point.
    :type name: str
    :param name: The name of the entry point.
    :return: The loaded entry point
    """
    hash_str = '/'.join([dist, group, name])
    if hash_str not in _ENTRY_POINT_CACHE:
        _ENTRY_POINT_CACHE[hash_str] = load_entry_point(dist, group, name)
    return _ENTRY_POINT_CACHE[hash_str]


if __name__ == '__main__':
    import doctest
    doctest.testmod(exclude_empty=True)