Source code for sunpy.spectra.sources.callisto

# -*- coding: utf-8 -*-
# Author: Florian Mayer <[email protected]>
from __future__ import absolute_import, print_function

import datetime

import numpy as np

from import fits

from collections import defaultdict

from bs4 import BeautifulSoup

from scipy.optimize import leastsq
from scipy.ndimage import gaussian_filter1d

from sunpy.time import parse_time
from sunpy.util import minimal_pairs, deprecated
from sunpy.util.cond_dispatch import ConditionalDispatch, run_cls
from import download_file

from sunpy.spectra.spectrogram import LinearTimeSpectrogram, REFERENCE
from sunpy.extern.six.moves import urllib
from sunpy.extern.six import next, itervalues

__all__ = ['CallistoSpectrogram']

TIME_STR = "%Y%m%d%H%M%S"
_DAY = datetime.timedelta(days=1)

DATA_SIZE = datetime.timedelta(seconds=15 * 60)

def parse_filename(href):
    name = href.split('.')[0]
        inst, date, time, no = name.rsplit('_')
        dstart = datetime.datetime.strptime(date + time, TIME_STR)
    except ValueError:
        # If the split fails, the file name does not match out
        # format,so we skip it and continue to the next
        # iteration of the loop.
        return None
    return inst, no, dstart

    # Everything starts with ""
    ("", parse_filename)

def search(start, end, instruments=None, url=DEFAULT_URL):
    """Get URLs for callisto data from instruments between start and end.

    start : `~sunpy.time.parse_time` compatible
    end : `~sunpy.time.parse_time` compatible
    instruments : sequence
        Sequence of instruments whose data is requested.
    url : str
        Base URL for the request.
    day = datetime.datetime(start.year, start.month,
    while day <= end:
        directory = url + day.strftime('%Y/%m/%d/')
        opn = urllib.request.urlopen(directory)
            soup = BeautifulSoup(opn, 'html.parser')
            for link in soup.find_all("a"):
                href = link.get("href")
                for prefix, parser in PARSERS:
                    if href.startswith(prefix):

                result = parser(href)
                if result is None:
                inst, no, dstart = result

                if (instruments is not None and
                    inst not in instruments and
                    (inst, int(no)) not in instruments):

                dend = dstart + DATA_SIZE
                if dend > start and dstart < end:
                    yield directory + href
        day += _DAY

@deprecated('0.8', alternative='')
def query(start, end, instruments=None, url=DEFAULT_URL):
    __doc__ = search.__doc__
    return search(start, end, instruments=instruments, url=url)

def download(urls, directory):
    """Download files from urls into directory.

    urls : list of str
        urls of the files to retrieve
    directory : str
        directory to save them in
    return [download_file(url, directory) for url in urls]

def _parse_header_time(date, time):
    """Returns `~datetime.datetime` object from date and time fields of
    header. """
    if time is not None:
        date = date + 'T' + time
    return parse_time(date)

[docs]class CallistoSpectrogram(LinearTimeSpectrogram): """ Class used for dynamic spectra coming from the Callisto network. Attributes ---------- header : `` main header of the FITS file axes_header : `` header for the axes table swapped : bool flag that specifies whether originally in the file the x-axis was frequency """ # XXX: Determine those from the data. SIGMA_SUM = 75 SIGMA_DELTA_SUM = 20 _create = ConditionalDispatch.from_existing(LinearTimeSpectrogram._create) create = classmethod(_create.wrapper()) # Contrary to what pylint may think, this is not an old-style class. # pylint: disable=E1002,W0142,R0902 # This needs to list all attributes that need to be # copied to maintain the object and how to handle them. COPY_PROPERTIES = LinearTimeSpectrogram.COPY_PROPERTIES + [ ('header', REFERENCE), ('swapped', REFERENCE), ('axes_header', REFERENCE) ] # List of instruments retrieved in July 2012 from # INSTRUMENTS = set([ 'ALASKA', 'ALMATY', 'BIR', 'DARO', 'HB9SCT', 'HUMAIN', 'HURBANOVO', 'KASI', 'KENYA', 'KRIM', 'MALAYSIA', 'MRT1', 'MRT2', 'OOTY', 'OSRA', 'SWMC', 'TRIEST', 'UNAM' ])
[docs] def save(self, filepath): """ Save modified spectrogram back to filepath. Parameters ---------- filepath : str path to save the spectrogram to """ main_header = self.get_header() data = fits.PrimaryHDU(self, header=main_header) ## XXX: Update axes header. freq_col = fits.Column( name="frequency", format="D8.3", array=self.freq_axis ) time_col = fits.Column( name="time", format="D8.3", array=self.time_axis ) cols = fits.ColDefs([freq_col, time_col]) table = fits.new_table(cols, header=self.axes_header) hdulist = fits.HDUList([data, table]) hdulist.writeto(filepath)
[docs] def get_header(self): """Returns the updated header.""" header = self.header.copy() if self.swapped: header['NAXIS2'] = self.shape[1] # pylint: disable=E1101 header['NAXIS1'] = self.shape[0] # pylint: disable=E1101 else: header['NAXIS1'] = self.shape[1] # pylint: disable=E1101 header['NAXIS2'] = self.shape[0] # pylint: disable=E1101 return header
[docs] @classmethod def read(cls, filename, **kwargs): """Reads in FITS file and return a new CallistoSpectrogram. Any unknown (i.e. any except filename) keyword arguments get passed to Parameters ---------- filename : str path of the file to read """ fl =, **kwargs) data = fl[0].data axes = fl[1] header = fl[0].header start = _parse_header_time( header['DATE-OBS'], header.get('TIME-OBS', header.get('TIME$_OBS')) ) end = _parse_header_time( header['DATE-END'], header.get('TIME-END', header.get('TIME$_END')) ) swapped = "time" not in header["CTYPE1"].lower() # Swap dimensions so x-axis is always time. if swapped: t_delt = header["CDELT2"] t_init = header["CRVAL2"] - t_delt * header["CRPIX2"] t_label = header["CTYPE2"] f_delt = header["CDELT1"] f_init = header["CRVAL1"] - t_delt * header["CRPIX1"] f_label = header["CTYPE1"] data = data.transpose() else: t_delt = header["CDELT1"] t_init = header["CRVAL1"] - t_delt * header["CRPIX1"] t_label = header["CTYPE1"] f_delt = header["CDELT2"] f_init = header["CRVAL2"] - t_delt * header["CRPIX2"] f_label = header["CTYPE2"] # Table may contain the axes data. If it does, the other way of doing # it might be very wrong. if axes is not None: try: # It's not my fault. Neither supports __contains__ nor .get tm =['time'] except KeyError: tm = None try: fq =['frequency'] except KeyError: fq = None if tm is not None: # Fix dimensions (whyever they are (1, x) in the first place) time_axis = np.squeeze(tm) else: # Otherwise, assume it's linear. time_axis = \ np.linspace(0, data.shape[1] - 1) * t_delt + t_init # pylint: disable=E1101 if fq is not None: freq_axis = np.squeeze(fq) else: freq_axis = \ np.linspace(0, data.shape[0] - 1) * f_delt + f_init # pylint: disable=E1101 content = header["CONTENT"] instruments = set([header["INSTRUME"]]) return cls( data, time_axis, freq_axis, start, end, t_init, t_delt, t_label, f_label, content, instruments, header, axes.header, swapped )
def __init__(self, data, time_axis, freq_axis, start, end, t_init=None, t_delt=None, t_label="Time", f_label="Frequency", content="", instruments=None, header=None, axes_header=None, swapped=False): # Because of how object creation works, there is no avoiding # unused arguments in this case. # pylint: disable=W0613 super(CallistoSpectrogram, self).__init__( data, time_axis, freq_axis, start, end, t_init, t_delt, t_label, f_label, content, instruments ) self.header = header self.axes_header = axes_header self.swapped = swapped
[docs] @classmethod def is_datasource_for(cls, header): """Check if class supports data from the given FITS file. Parameters ---------- header : `` main header of the FITS file """ return header.get('instrume', '').strip() in cls.INSTRUMENTS
[docs] def remove_border(self): """Remove duplicate entries on the borders.""" left = 0 while self.freq_axis[left] == self.freq_axis[0]: left += 1 right = self.shape[0] - 1 while self.freq_axis[right] == self.freq_axis[-1]: right -= 1 return self[left-1:right+2, :]
[docs] @classmethod def read_many(cls, filenames, sort_by=None): """Returns a list of CallistoSpectrogram objects read from filenames. Parameters ---------- filenames : list of str list of paths to read from sort_by : str optional attribute of the resulting objects to sort from, e.g. start to sort by starting time. """ objs = list(map(, filenames)) if sort_by is not None: objs.sort(key=lambda x: getattr(x, sort_by)) return objs
[docs] @classmethod def from_range(cls, instrument, start, end, **kwargs): """Automatically download data from instrument between start and end and join it together. Parameters ---------- instrument : str instrument to retrieve the data from start : `~sunpy.time.parse_time` compatible start of the measurement end : `~sunpy.time.parse_time` compatible end of the measurement """ kw = { 'maxgap': None, 'fill': cls.JOIN_REPEAT, } kw.update(kwargs) start = parse_time(start) end = parse_time(end) urls = query(start, end, [instrument]) data = list(map(cls.from_url, urls)) freq_buckets = defaultdict(list) for elem in data: freq_buckets[tuple(elem.freq_axis)].append(elem) try: return cls.combine_frequencies( [cls.join_many(elem, **kw) for elem in itervalues(freq_buckets)] ) except ValueError: raise ValueError("No data found.")
def _overlap(self, other): """ Find frequency and time overlap of two spectrograms. """ one, two = self.intersect_time([self, other]) ovl = one.freq_overlap(two) return one.clip_freq(*ovl), two.clip_freq(*ovl) @staticmethod def _to_minimize(a, b): """Function to be minimized for matching to frequency channels.""" def _fun(p): if p[0] <= 0.2 or abs(p[1]) >= a.max(): return float("inf") return a - (p[0] * b + p[1]) return _fun def _homogenize_params(self, other, maxdiff=1): """ Return triple with a tuple of indices (in self and other, respectively), factors and constants at these frequencies. Parameters ---------- other : `sunpy.spectra.CallistoSpectrogram` Spectrogram to be homogenized with the current one. maxdiff : float Threshold for which frequencies are considered equal. """ pairs_indices = [ (x, y) for x, y, d in minimal_pairs(self.freq_axis, other.freq_axis) if d <= maxdiff ] pairs_data = [ (self[n_one, :], other[n_two, :]) for n_one, n_two in pairs_indices ] # XXX: Maybe unnecessary. pairs_data_gaussian = [ (gaussian_filter1d(a, 15), gaussian_filter1d(b, 15)) for a, b in pairs_data ] # If we used integer arithmetic, we would accept more invalid # values. pairs_data_gaussian64 = np.float64(pairs_data_gaussian) least = [ leastsq(self._to_minimize(a,b), [1, 0])[0] for a, b in pairs_data_gaussian64 ] factors = [x for x, y in least] constants = [y for x, y in least] return pairs_indices, factors, constants
[docs] def homogenize(self, other, maxdiff=1): """ Return overlapping part of self and other as (self, other) tuple. Homogenize intensities so that the images can be used with combine_frequencies. Note that this works best when most of the picture is signal, so use :py:meth:`in_interval` to select the subset of your image before applying this method. Parameters ---------- other : `sunpy.spectra.CallistoSpectrogram` Spectrogram to be homogenized with the current one. maxdiff : float Threshold for which frequencies are considered equal. """ one, two = self._overlap(other) pairs_indices, factors, constants = one._homogenize_params( two, maxdiff ) # XXX: Maybe (xd.freq_axis[x] + yd.freq_axis[y]) / 2. pairs_freqs = [one.freq_axis[x] for x, y in pairs_indices] # XXX: Extrapolation does not work this way. # XXX: Improve. f1 = np.polyfit(pairs_freqs, factors, 3) f2 = np.polyfit(pairs_freqs, constants, 3) return (one, two * np.polyval(f1, two.freq_axis)[:, np.newaxis] + np.polyval(f2, two.freq_axis)[:, np.newaxis])
[docs] def extend(self, minutes=15, **kwargs): """Requests subsequent files from the server. If minutes is negative, retrieve preceding files. """ if len(self.instruments) != 1: raise ValueError instrument = next(iter(self.instruments)) if minutes > 0: data = CallistoSpectrogram.from_range( instrument, self.end, self.end + datetime.timedelta(minutes=minutes) ) else: data = CallistoSpectrogram.from_range( instrument, self.start - datetime.timedelta(minutes=-minutes), self.start ) data = data.clip_freq(self.freq_axis[-1], self.freq_axis[0]) return CallistoSpectrogram.join_many([self, data], **kwargs)
[docs] @classmethod def from_url(cls, url): """Returns CallistoSpectrogram read from URL. Parameters ---------- url : str URL to retrieve the data from Returns ------- newSpectrogram : `sunpy.spectra.CallistoSpectrogram` """ return
CallistoSpectrogram._create.add( run_cls('from_range'), lambda cls, instrument, start, end: True, check=False ) try: CallistoSpectrogram.create.im_func.__doc__ = ( """Create CallistoSpectrogram from given input dispatching to the appropriate from_* function. Possible signatures: """ + CallistoSpectrogram._create.generate_docs()) except AttributeError: CallistoSpectrogram.create.__func__.__doc__ = ( """Create CallistoSpectrogram from given input dispatching to the appropriate from_* function. Possible signatures: """ + CallistoSpectrogram._create.generate_docs())