Skip to content

Instantly share code, notes, and snippets.

@pkgw

pkgw/closures.py

Created August 5, 2016 14:16
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save pkgw/b58d319d799c2c90ac69d1db265b1b6f to your computer and use it in GitHub Desktop.
Save pkgw/b58d319d799c2c90ac69d1db265b1b6f to your computer and use it in GitHub Desktop.
Download ZIP
Raw
closures.py
#! /usr/bin/env python
# -*- mode: python; coding: utf-8 -*-
# Copyright 2016 Peter Williams <peter@newton.cx> and collaborators.
# Licensed under the MIT License.
# XXX remove hashbang
"""Compute phase closure diagnostics from a Measurement Set
For most results to make sense, the data should be observations of a bright
point source at phase center.
"""
from __future__ import absolute_import, division, print_function, unicode_literals
__all__ = str ('Config closures').split ()
import collections
import six, numpy as np
from six.moves import range
# XXX relativize e.g. "...cli"
from pwkit.cli import check_usage, die
from pwkit.kwargv import ParseKeywords, Custom
from pwkit.environments.casa import util
from pwkit.environments.casa.util import sanitize_unicode as b
closures_doc = \
"""
<closures> vis=MS [keywords...]
Phase closure triples.
vis=
Path of the MeasurementSet dataset to read. Required.
array=, baseline=, field=, observation=, polarization=, scan=,
scanintent=, spw=, taql=, time=, uvdist=
MeasurementSet selectors used to filter the input data.
Default polarization is 'RR,LL'. All polarizations are averaged
together, so mixing parallel- and cross-hand pols is almost
never what you want to do.
datacol=
Name of the column to use for visibility data. Defaults to 'data'.
You might want it to be 'corrected_data'.
"""
class Config (ParseKeywords):
vis = Custom (str, required=True)
datacol = 'data'
believeweights = False
# MeasurementSet filters
array = str
baseline = str
field = str
observation = str
polarization = 'RR,LL'
scan = str
scanintent = str
spw = str
taql = str
time = str
uvdist = str
loglevel = 'warn'
# ########################################################################
# Begin copying/emulating mirtask.util
FPOL_X = 0
FPOL_Y = 1
FPOL_R = 2
FPOL_L = 3
FPOL_I = 4
FPOL_Q = 5
FPOL_U = 6
FPOL_V = 7
fpol_names = 'XYRLIQUV'
# This table helps split a CASA pol code into a pair of fpol values. The pair
# is packed into 8 bits, the upper 3 being for the left pol and the lower 4
# being for the right.
_pol_to_fpol = np.array ([
0xFFFF, # ?, illegal
0x44, 0x55, 0x66, 0x77, # I Q U V
0x22, 0x23, 0x32, 0x33, # RR RL LR LL
0x00, 0x01, 0x10, 0x11, # XX XY YX YY
0x20, 0x21, 0x30, 0x31, # RX RY LX LY
0x02, 0x03, 0x12, 0x13, # XR XL YR YL
# not bothering with the rest because seriously
])
pol_is_intensity = np.array ([
False, # ?, illegal
True, True, True, True, # I Q U V
True, False, False, True, # RR RL LR LL
True, False, False, True, # XX XY YX YY
False, False, False, False, # RX RY LX LY
False, False, False, False, # XR XL YR YL
# not bothering with the rest because seriously
])
# This table performs the reverse mapping, with index being the two f-pol
# values packed into four bits each. A value of 0xFF indicates an illegal
# pairing. The values come from pwkit.environments.casa.util:pol_names
_fpol_to_pol = np.ndarray (128, dtype=np.int8)
_fpol_to_pol.fill (0xFF)
_fpol_to_pol[0x00] = 9
_fpol_to_pol[0x01] = 10
_fpol_to_pol[0x10] = 11
_fpol_to_pol[0x11] = 12
_fpol_to_pol[0x22] = 5
_fpol_to_pol[0x23] = 5
_fpol_to_pol[0x32] = 7
_fpol_to_pol[0x33] = 8
_fpol_to_pol[0x44] = 1
_fpol_to_pol[0x55] = 2
_fpol_to_pol[0x66] = 3
_fpol_to_pol[0x77] = 4
# A "antpol" (AP) is an integer identifying an antenna/fpol pair. It
# can be decoded without any external information. We used zero-based
# integer antenna numbers so that
#
# AP = M << 3 + FP
def ap_format (ap, getname=str):
return '%s%c' % (getname (ap >> 3), fpol_names[ap & 0x7])
ap_ant = lambda ap: ap >> 3
ap_fpol = lambda ap: ap & 0x7
antpol_to_ap = lambda antnum, fpol: (antnum << 3) + fpol
# A "basepol" is 2-tuple of antpols.
def bp_format (bp, getname=str):
ap1, ap2 = bp
if ap1 < 0:
raise ValueError ('first antpol %d is negative' % ap1)
if ap2 < 0:
raise ValueError ('second antpol %d is negative' % ap2)
return '%s%c-%s%c' % (getname (ap1 >> 3), fpol_names[ap1 & 0x7],
getname (ap2 >> 3), fpol_names[ap2 & 0x7])
def bp_to_aap (bp):
"""Converts a basepol into a tuple of (ant1, ant2, pol)."""
ap1, ap2 = bp
if ap1 < 0:
raise ValueError ('first antpol %d is negative' % ap1)
if ap2 < 0:
raise ValueError ('second antpol %d is negative' % ap2)
pol = _fpol_to_pol[((ap1 & 0x7) << 4) + (ap2 & 0x7)]
if pol == 0xFF:
raise ValueError ('no CASA polarization code for pairing '
'%c-%c' % (fpol_names[ap1 & 0x7],
fpol_names[ap2 & 0x7]))
return ap1 >> 3, ap2 >> 3, pol
def aap_to_bp (ant1, ant2, pol):
"""Create a basepol from antenna numbers and a CASA polarization code."""
if ant1 < 0:
raise ValueError ('first antenna is below 0: %s' % ant1)
if ant2 < ant1:
raise ValueError ('second antenna is below first: %s' % ant2)
if pol < 1 or pol > 12:
raise ValueError ('illegal polarization code %s' % pol)
fps = _pol_to_fpol[pol]
ap1 = (ant1 << 3) + ((fps >> 4) & 0x07)
ap2 = (ant2 << 3) + (fps & 0x07)
return ap1, ap2
# End copying/emulating mirtask.util
# ########################################################################
class StatsCollector (object):
def __init__ (self, chunk0size=64):
self.chunk0size = chunk0size
self._keymap = collections.defaultdict (lambda: len (self._keymap))
self._m0 = None # 0th moment
self._m1 = None #
self._m2 = None # 2nd moment
def accum (self, key, value, weight=1):
index = self._keymap[key]
if self._m0 is None:
self._m0 = np.zeros ((self.chunk0size,), dtype=np.result_type (weight))
self._m1 = np.zeros ((self.chunk0size,), dtype=np.result_type (value, weight))
self._m2 = np.zeros_like (self._m1)
elif index >= self._m0.size:
self._m0 = np.concatenate ((self._m0, np.zeros_like (self._m0)))
self._m1 = np.concatenate ((self._m1, np.zeros_like (self._m1)))
self._m2 = np.concatenate ((self._m2, np.zeros_like (self._m2)))
self._m0[index] += weight
q = weight * value
self._m1[index] += q
q *= value
self._m2[index] += q
return self
def finish (self, keyset, mask=True):
"""Returns (weights, means, variances), where:
weights
ndarray of number of samples per key
means
computed mean value for each key
variances
computed variance for each key
"""
n_us = len (self._keymap)
# By definition (for now), wt >= 1 everywhere, so we don't need to
# worry about div-by-zero.
wt_us = self._m0[:n_us]
mean_us = self._m1[:n_us] / wt_us
var_us = self._m2[:n_us] / wt_us - mean_us**2
n_them = len (keyset)
wt = np.zeros (n_them, dtype=self._m0.dtype)
mean = np.empty (n_them, dtype=self._m1.dtype)
mean.fill (np.nan)
var = np.empty_like (mean)
var.fill (np.nan)
us_idx = []
them_idx = []
for them_i, key in enumerate (keyset):
us_i = self._keymap[key]
if us_i < n_us:
them_idx.append (them_i)
us_idx.append (us_i)
# otherwise, we must not have seen that key
wt[them_idx] = wt_us[us_idx]
mean[them_idx] = mean_us[us_idx]
var[them_idx] = var_us[us_idx]
if mask:
m = ~np.isfinite (mean)
mean = np.ma.MaskedArray (mean, m)
var = np.ma.MaskedArray (var, m)
self._m0 = self._m1 = self._m2 = None
self._keymap.clear ()
return wt, mean, var
class StatsCollector2D (object):
def __init__ (self, chunk0size=64):
self.chunk0size = chunk0size
self._key1map = collections.defaultdict (lambda: len (self._key1map))
self._key2map = collections.defaultdict (lambda: len (self._key2map))
self._m0 = None
self._m1 = None
self._m2 = None
def accum (self, key1, key2, value, weight=1):
index1 = self._key1map[key1]
index2 = self._key2map[key2]
if self._m0 is None:
self._m0 = np.zeros ((self.chunk0size, self.chunk0size), dtype=np.result_type (weight))
self._m1 = np.zeros ((self.chunk0size, self.chunk0size), dtype=np.result_type (value, weight))
self._m2 = np.zeros_like (self._m1)
if index1 >= self._m0.shape[0]:
self._m0 = np.concatenate ((self._m0, np.zeros_like (self._m0)), axis=0)
self._m1 = np.concatenate ((self._m1, np.zeros_like (self._m1)), axis=0)
self._m2 = np.concatenate ((self._m2, np.zeros_like (self._m2)), axis=0)
if index2 >= self._m0.shape[1]:
self._m0 = np.concatenate ((self._m0, np.zeros_like (self._m0)), axis=1)
self._m1 = np.concatenate ((self._m1, np.zeros_like (self._m1)), axis=1)
self._m2 = np.concatenate ((self._m2, np.zeros_like (self._m2)), axis=1)
self._m0[index1,index2] += weight
q = weight * value
self._m1[index1,index2] += q
q *= value
self._m2[index1,index2] += q
return self
def finish (self, key1set, key2set, mask=True):
"""Returns (weights, means, variances), where:
weights
ndarray of number of samples per key; shape (n1, n2), where n1 is the
size of key1set and n2 is the size of key2set.
means
computed mean value for each key
variances
computed variance for each key
"""
n1_us = len (self._key1map)
n2_us = len (self._key2map)
wt_us = self._m0[:n1_us,:n2_us]
badwt = (wt_us == 0) | ~np.isfinite (wt_us)
wt_us[badwt] = 1
mean_us = self._m1[:n1_us,:n2_us] / wt_us
var_us = self._m2[:n1_us,:n2_us] / wt_us - mean_us**2
wt_us[badwt] = 0
mean_us[badwt] = np.nan
var_us[badwt] = np.nan
n1_them = len (key1set)
n2_them = len (key2set)
wt = np.zeros ((n1_them, n2_them), dtype=self._m0.dtype)
mean = np.empty ((n1_them, n2_them), dtype=self._m1.dtype)
mean.fill (np.nan)
var = np.empty_like (mean)
var.fill (np.nan)
# You can't fancy-index on two axes simultaneously, so we do a manual
# loop on the first axis.
us_idx2 = []
them_idx2 = []
for them_i2, key2 in enumerate (key2set):
us_i2 = self._key2map[key2]
if us_i2 < n2_us:
them_idx2.append (them_i2)
us_idx2.append (us_i2)
# otherwise, we must not have seen that key
for them_i1, key1 in enumerate (key1set):
us_i1 = self._key1map[key1]
if us_i1 >= n1_us:
continue # don't have this key
wt[them_i1,them_idx2] = wt_us[us_i1,us_idx2]
mean[them_i1,them_idx2] = mean_us[us_i1,us_idx2]
var[them_i1,them_idx2] = var_us[us_i1,us_idx2]
if mask:
m = ~np.isfinite (mean)
mean = np.ma.MaskedArray (mean, m)
var = np.ma.MaskedArray (var, m)
self._m0 = self._m1 = self._m2 = None
self._key1map.clear ()
self._key2map.clear ()
return wt, mean, var
class StatsCollectorND (object):
"""This is vaguely like StatsCollector2D, but rather than having two discrete
keyed axes, we are passed one key and an N-dimensional vector of values.
"""
def __init__ (self, chunk0size=64):
self.chunk0size = chunk0size
self._keymap = collections.defaultdict (lambda: len (self._keymap))
self._m0 = None
self._m1 = None
self._m2 = None
def accum (self, key, values, weights=1):
index = self._keymap[key]
values = np.asarray (values)
weights = np.broadcast_to (weights, values.shape)
if self._m0 is None:
self._m0 = np.zeros ((self.chunk0size,) + values.shape, dtype=weights.dtype)
self._m1 = np.zeros ((self.chunk0size,) + values.shape, dtype=np.result_type (weights, values))
self._m2 = np.zeros_like (self._m1)
elif index >= self._m0.size:
self._m0 = np.concatenate ((self._m0, np.zeros_like (self._m0)))
self._m1 = np.concatenate ((self._m1, np.zeros_like (self._m1)))
self._m2 = np.concatenate ((self._m2, np.zeros_like (self._m2)))
if values.shape != self._m0.shape[1:]:
raise ValueError ('inconsistent `values` shapes: had %r, got %r' % (
self._m0.shape[1:], values.shape))
self._m0[index] += weights
q = weights * values
self._m1[index] += q
q *= values
self._m2[index] += q
return self
def finish (self, keyset, mask=True):
"""Returns (weights, means, variances), where:
weights
total weights per key
means
computed mean value for each key
variances
computed variance for each key
The arrays all have a shape of `(nkeys,)+shape(values)`.
"""
n_us = len (self._keymap)
wt_us = self._m0[:n_us]
badwt = (wt_us == 0) | ~np.isfinite (wt_us)
wt_us[badwt] = 1.
mean_us = self._m1[:n_us] / wt_us
var_us = self._m2[:n_us] / wt_us - mean_us**2
wt_us[badwt] = 0
mean_us[badwt] = np.nan
var_us[badwt] = np.nan
n_them = len (keyset)
wt = np.zeros ((n_them,) + mean_us.shape[1:], dtype=self._m0.dtype)
mean = np.empty ((n_them,) + mean_us.shape[1:], dtype=self._m1.dtype)
mean.fill (np.nan)
var = np.empty_like (mean)
var.fill (np.nan)
us_idx = []
them_idx = []
for them_i, key in enumerate (keyset):
us_i = self._keymap[key]
if us_i < n_us:
them_idx.append (them_i)
us_idx.append (us_i)
# otherwise, we must not have seen that key
wt[them_idx] = wt_us[us_idx]
mean[them_idx] = mean_us[us_idx]
var[them_idx] = var_us[us_idx]
if mask:
m = ~np.isfinite (mean)
mean = np.ma.MaskedArray (mean, m)
var = np.ma.MaskedArray (var, m)
self._m0 = self._m1 = self._m2 = None
self._keymap.clear ()
return wt, mean, var
def postproc (stats_result):
"""Simple helper to postprocess angular outputs from StatsCollectors in the
way we want.
"""
n, mean, scat = stats_result
mean *= 180 / np.pi # rad => deg
scat /= n # variance-of-samples => variance-of-mean
scat **= 0.5 # variance => stddev
scat *= 180 / np.pi # rad => deg
return mean, scat
def postproc_mask (stats_result):
"""Simple helper to postprocess angular outputs from StatsCollectors in the
way we want.
"""
n, mean, scat = stats_result
ok = np.isfinite (mean)
n = n[ok]
mean = mean[ok]
scat = scat[ok]
mean *= 180 / np.pi # rad => deg
scat /= n # variance-of-samples => variance-of-mean
scat **= 0.5 # variance => stddev
scat *= 180 / np.pi # rad => deg
return ok, mean, scat
def grid_bp_data (bps, items, mask=True):
"""Given a bunch of scalars associated with intensity-type basepols, place
them onto a grid. There should be only two polarizations represented (e.g.
RR, LL); baselines for one of them will be put into the upper triangle of
the grid, while baselines for the other will be put into the lower triangle.
`bps` is an iterable that yields a superset of all bps to be gridded.
`items` is an iterable that yields tuples of (bp, value). (`bps` is
therefore a bit redundant with it, but this structure makes it so that we
don't need to iterate of `items` twice, which can be convenient.)
Returns: (pol1, pol2, ants, data), where
pol1
The polarization gridded into the upper triangle
pol2
The polarization gridded into the lower triangle
ants
An array of the antenna numbers seen in `bps`
data
An n-by-n grid of the values, where n is the size of `ants`. Unsampled
basepols are filled with NaNs, or masked if `mask` is True.
The basepol (ant1, ant2, pol1) is gridded into `data[i1,i2]`, where `i1`
is the index of `ant1` in `ants`, etc. The basepol (ant1, ant2, pol2) is
gridded into `data[i2,i1]`.
"""
seen_ants = set ()
seen_pols = set ()
for ant1, ant2, pol in (bp_to_aap (bp) for bp in bps):
seen_ants.add (ant1)
seen_ants.add (ant2)
seen_pols.add (pol)
if len (seen_pols) != 2:
raise Exception ('can only work with 2 polarizations')
pol1, pol2 = sorted (seen_pols)
seen_ants = np.array (sorted (seen_ants))
ant_map = dict ((a, i) for (i, a) in enumerate (seen_ants))
data = None
n = len (seen_ants)
for bp, value in items:
if data is None:
data = np.empty ((n, n), dtype=np.result_type (value))
data.fill (np.nan)
ant1, ant2, pol = bp_to_aap (bp)
i1 = ant_map[ant1]
i2 = ant_map[ant2]
if pol == pol1:
data[i1,i2] = value
else:
data[i2,i1] = value
if mask:
data = np.ma.MaskedArray (data, ~np.isfinite (data))
return pol1, pol2, seen_ants, data
class ClosureCalculator (object):
def process (self, cfg):
# Initialize whole-run buffers.
self.all_aps = set ()
self.all_bps = set ()
self.all_times = set ()
self.global_stats_by_time = StatsCollector ()
self.ap_stats_by_ddid = collections.defaultdict (StatsCollector)
self.bp_stats_by_ddid = collections.defaultdict (StatsCollector)
self.ap_spec_stats_by_ddid = collections.defaultdict (StatsCollectorND)
self.ap_time_stats_by_ddid = collections.defaultdict (StatsCollector2D)
self._process_ms (cfg)
return self
def _process_ms (self, cfg):
tb = util.tools.table ()
ms = util.tools.ms ()
me = util.tools.measures ()
# Prep.
ms.open (b(cfg.vis))
ms.msselect (b(dict ((n, cfg.get (n)) for n in util.msselect_keys
if cfg.get (n) is not None)))
rangeinfo = ms.range (b'data_desc_id field_id'.split ())
ddids = rangeinfo['data_desc_id']
fields = rangeinfo['field_id']
tb.open (b(cfg.vis + '/DATA_DESCRIPTION'), nomodify=True)
ddid_to_polid = tb.getcol (b'POLARIZATION_ID')
ddid_to_spwid = tb.getcol (b'SPECTRAL_WINDOW_ID')
tb.close ()
tb.open (b(cfg.vis + '/POLARIZATION'), nomodify=True)
polid_to_polns = tb.getcol (b'CORR_TYPE')
tb.close ()
tb.open (b(cfg.vis + '/ANTENNA'), nomodify=True)
self.ant_names = tb.getcol (b'NAME')
self.ant_stations = tb.getcol (b'STATION')
tb.close ()
# Read stuff in. We can't expect that weight values have their
# absolute scale set correctly, but we can still use them to set the
# relative weighting of the data points.
#
# datacol is (ncorr, nchan, nchunk)
# flag is (ncorr, nchan, nchunk); zero means OK data
# weight is (ncorr, nchunk) [XXX: WEIGHT_SPECTRUM?]
# uvw is (3, nchunk)
# time is (nchunk)
#
# Iteration order, from slowest varying to fastest varying:
#
# spw, time, baseline, poln, frequency
#
# We are encouraged to use the new iteration interface
# (ms.niterinit(), etc), but as of 4.6.0 it is fundamentally broken
# (asking for ANTENNA2 gets you ANTENNA1) so never mind.
colnames = [cfg.datacol] + 'flag weight time antenna1 antenna2'.split ()
colnames = b(colnames)
for ddid in ddids:
ms.selectinit (ddid)
ms.iterinit (maxrows=4096)
ms.iterorigin ()
self.cur_ddid = ddid
self.cur_time = -1.
self._start_timeslot ()
all_polns = polid_to_polns[:,ddid_to_polid[ddid]]
polinfo = [(i, p) for i, p in enumerate (all_polns) if pol_is_intensity[p]]
while True:
cols = ms.getdata (items=colnames)
for i in range (cols['time'].size): # all records
time = cols['time'][i]
if time != self.cur_time:
self._finish_timeslot ()
self.cur_time = time
self._start_timeslot ()
antenna1 = cols['antenna1'][i]
antenna2 = cols['antenna2'][i]
if antenna1 == antenna2:
continue # no autocorrelations
for j, poln in polinfo: # all polns
flags = cols['flag'][j,:,i]
if flags.all ():
continue # all flagged
data = cols[cfg.datacol][j,:,i]
# data and flags are now both shape (nchan,)
np.logical_not (flags, flags)
# flags=1 now indicates good data
self.all_times.add (time)
bp = aap_to_bp (antenna1, antenna2, poln)
self.all_bps.add (bp)
self.data_by_bp[bp] = (data, flags) # + weight spectrum?
# Here we exploit the fact that we're only considering
# intensity-type polarizations, so bp[0] and bp[1]
# always have the same fpol. Also that ap_by_fpol is a
# defaultdict.
for ap in bp:
self.all_aps.add (ap)
self.ap_by_fpol[ap & 0x7].add (ap)
if not ms.iternext ():
self._finish_timeslot ()
break
ms.close ()
return self
def _getname (self, antidx):
return '%s@%s:' % (self.ant_names[antidx], self.ant_stations[antidx])
def _start_timeslot (self):
self.data_by_bp = {}
self.ap_by_fpol = collections.defaultdict (set)
def _finish_timeslot (self):
"""We have loaded in all of the visibilities in one timeslot. We can now
compute the phase closure triples.
XXX: we should only process independent triples. Are we???
"""
for fpol, aps in self.ap_by_fpol.iteritems ():
aps = sorted (aps)
nap = len (aps)
for i1, ap1 in enumerate (aps):
for i2 in range (i1, nap):
ap2 = aps[i2]
bp1 = (ap1, ap2)
info = self.data_by_bp.get (bp1)
if info is None:
continue
data1, flags1 = info
for i3 in range (i2, nap):
ap3 = aps[i3]
bp2 = (ap2, ap3)
info = self.data_by_bp.get (bp2)
if info is None:
continue
data2, flags2 = info
bp3 = (ap1, aps[i3])
info = self.data_by_bp.get (bp3)
if info is None:
continue
data3, flags3 = info
# try to minimize allocations:
tflags = flags1 & flags2
np.logical_and (tflags, flags3, tflags)
if not tflags.any ():
continue
triple = data3.conj ()
np.multiply (triple, data1, triple)
np.multiply (triple, data2, triple)
self._process_sample (ap1, ap2, ap3, triple, tflags)
# Reset for next timeslot
self.cur_time = -1.
self.bp_by_ap = None
self.ap_by_fpol = None
def _process_sample (self, ap1, ap2, ap3, triple, tflags):
"""We have computed one independent phase closure triple in one timeslot.
"""
# Frequency-resolved:
np.divide (triple, np.abs (triple), triple)
phase = np.angle (triple)
self.ap_spec_stats_by_ddid[self.cur_ddid].accum (ap1, phase, tflags + 0.)
self.ap_spec_stats_by_ddid[self.cur_ddid].accum (ap2, phase, tflags + 0.)
self.ap_spec_stats_by_ddid[self.cur_ddid].accum (ap3, phase, tflags + 0.)
# Frequency-averaged:
triple = np.dot (triple, tflags) / tflags.sum ()
phase = np.angle (triple)
self.global_stats_by_time.accum (self.cur_time, phase)
self.ap_stats_by_ddid[self.cur_ddid].accum (ap1, phase)
self.ap_stats_by_ddid[self.cur_ddid].accum (ap2, phase)
self.ap_stats_by_ddid[self.cur_ddid].accum (ap3, phase)
self.bp_stats_by_ddid[self.cur_ddid].accum ((ap1, ap2), phase)
self.bp_stats_by_ddid[self.cur_ddid].accum ((ap1, ap3), phase)
self.bp_stats_by_ddid[self.cur_ddid].accum ((ap2, ap3), phase)
self.ap_time_stats_by_ddid[self.cur_ddid].accum (self.cur_time, ap1, phase)
self.ap_time_stats_by_ddid[self.cur_ddid].accum (self.cur_time, ap2, phase)
self.ap_time_stats_by_ddid[self.cur_ddid].accum (self.cur_time, ap3, phase)
def report (self, cfg):
import omega as om
import omega.gtk3
from pwkit import ndshow_gtk3
self.all_aps = np.sort (list (self.all_aps))
self.all_bps = sorted (self.all_bps)
self.all_times = np.sort (list (self.all_times))
# Antpols by DDID, time:
data = []
descs = []
for ddid, stats in self.ap_time_stats_by_ddid.iteritems ():
mean, scat = postproc (stats.finish (self.all_times, self.all_aps))
data.append (mean / scat)
descs.append ('DDID %d' % ddid)
ndshow_gtk3.cycle (data, descs, run_main=True)
# Antpols by DDID, freq:
data = []
descs = []
for ddid, stats in self.ap_spec_stats_by_ddid.iteritems ():
mean, scat = postproc (stats.finish (self.all_aps))
data.append (mean / scat)
descs.append ('DDID %d' % ddid)
ndshow_gtk3.cycle (data, descs, run_main=True)
# Antpols by DDID
p = om.RectPlot ()
for ddid, stats in self.ap_stats_by_ddid.iteritems ():
ok, mean, scat = postproc_mask (stats.finish (self.all_aps))
p.addXYErr (np.arange (len (self.all_aps))[ok], mean, scat, 'DDID %d' % ddid)
p.setBounds (-0.5, len (self.all_aps) - 0.5)
p.addHLine (0, keyText=None, zheight=-1)
p.show ()
# Basepols by DDID
data = []
descs = []
tostatuses = []
def bpgrid_status (pol1, pol2, ants, yx):
i, j = [int (_) for _ in np.floor (yx + 0.5)]
if i < 0 or j < 0 or i >= ants.size or j >= ants.size:
return ''
ni = self._getname (ants[i])
nj = self._getname (ants[j])
if i <= j:
return '%s-%s %s' % (ni, nj, util.pol_names[pol1])
return '%s-%s %s' % (nj, ni, util.pol_names[pol2])
for ddid, stats in self.bp_stats_by_ddid.iteritems ():
mean, scat = postproc (stats.finish (self.all_bps))
nmean = mean / scat
from itertools import izip
pol1, pol2, ants, grid = grid_bp_data (self.all_bps,
izip (self.all_bps, nmean))
data.append (grid)
descs.append ('DDID %d' % ddid)
tostatuses.append (lambda yx: bpgrid_status (pol1, pol2, ants, yx))
ndshow_gtk3.cycle (data, descs, tostatuses=tostatuses, run_main=True)
# Everything by time
ok, mean, scat = postproc_mask (self.global_stats_by_time.finish (self.all_times))
stimes = self.all_times[ok] / 86400
st0 = int (np.floor (stimes.min ()))
stimes -= st0
p = om.quickXYErr (stimes, mean, scat)
p.addHLine (0, keyText=None, zheight=-1)
p.setXLabel ('MJD - %d' % st0)
p.show ()
def closures_cli (argv):
check_usage (closures_doc, argv, usageifnoargs=True)
cfg = Config ().parse (argv[1:])
util.logger (cfg.loglevel)
ClosureCalculator ().process (cfg).report (cfg)
if __name__ == '__main__': # XXX TEMP
import sys
from pwkit import cli
cli.unicode_stdio ()
cli.propagate_sigint ()
closures_cli (sys.argv)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment