Source code for pynbody.analysis.profile

"""

profile
=======

A set of classes and functions for making profiles of simulation
properties.

"""

import numpy as np
import pynbody
from .. import family, units, array, util, backcompat
import math
import logging
import warnings
logger = logging.getLogger('pynbody.analysis.profile')


[docs]class Profile: """ A basic profile class for arbitrary profiles. Stores information about bins etc. Made to work with the pynbody SimSnap instances. The constructor only generates the bins and figures out which particles belong in which bin. Profiles are generated lazy-loaded when a given property is requested. **Input**: *sim* : a simulation snapshot - this can be any subclass of SimSnap **Optional Keywords**: *ndim* (default = 2): specifies whether it's a 2D or 3D profile - in the 2D case, the bins are generated in the xy plane *type* (default = 'lin'): specifies whether bins should be spaced linearly ('lin'), logarithmically ('log') or contain equal numbers of particles ('equaln') *rmin* (default = min(x)): minimum radial value to consider *rmax* (default = max(x)): maximum radial value to consider *nbins* (default = 100): number of bins *bins* : array like - predefined bin edges in units of binning quantity. If this keyword is set, the values of the keywords *type*, *nbins*, *rmin* and *rmax* will be ignored *calc_x* (default = None): function to use to calculate the value for binning. If None it defaults to the radial distance from origin (in either 2 or 3 dimensions), ut you can specify this function to return any value you want for making profiles along arbitrary axes. Depening on your function, the units of certain profiles (such as density) might not make sense. *weight_by* (default = 'mass'): name of the array to use for weighting averages across particles in each bin **Output**: a Profile object. To find out which profiles are available, use keys(). **Implemented profile functions**: *density* : density *mass* : mass in each bin *mass_enc* : enclosed mass *fourier* : provides fourier coefficients, amplitude and phase for m=0 to m=6. To access the amplitude profile of m=2 mode, do ``p['fourier']['amp'][2,:]`` *dyntime* : dynamical time *g_spherical*: GM_enc/r^2 *rotation_curve_spherical*: rotation curve from vc = sqrt(GM/R) - can be very wrong! *j_circ* : angular momentum of particles on circular orbits *v_circ* : circular velocity, aka rotation curve - calculated from the midplane gravity, so this can be expensive *E_circ* : energy of particles on circular orbits in the midplane *omega* : circular orbital frequency *kappa* : radial orbital frequency *beta* : 3-D velocity anisotropy parameter *magnitudes* : magnitudes in each bin - default band = 'v' *sb* : surface brightness - default band = 'v' Additional functions should use the profile_property to yield the desired profile. **Lazy-loading arrays:** The Profile class will automatically compute a mass-weighted profile for any lazy-loadable array of its parent SimSnap object. **Dispersions:** To obtain a dispersion profile, attach a ``_disp`` after the desired quantity name. **RMS:** The root-mean-square of a quantity can be obtained by using a ``_rms`` suffix **Derivatives:** To compute a derivative of a profile, prepend a ``d_`` to the profile string, as in ``p['d_temp']`` to get a temperature gradient. **Saving and loading previously generated profiles:** Use the :func:`~pynbody.analysis.profile.Profile.write` function to write the current profiles with all the necessary information to a file. Initialize a profile with the load_from_file=True keyword to automatically load a previously saved profile. The filename is chosen automatically and corresponds to a hash generated from the positions of the particles used in the profile. This is to ensure that you are always looking at the same set of particles, centered in the same way. It also means you *must* use the same centering method if you want to reuse a saved profile. """ _profile_registry = {} def _calculate_x(self, sim): return ((sim['pos'][:, 0:self.ndim] ** 2).sum(axis=1)) ** (1, 2) def __init__(self, sim, load_from_file=False, ndim=2, type='lin', calc_x=None, weight_by='mass', **kwargs): generate_new = True if calc_x is None: calc_x = self._calculate_x self.sim = sim self.type = type self.ndim = ndim self._weight_by = weight_by self._x = calc_x(sim) x = self._x if load_from_file: import pickle filename = self._generate_hash_filename() try: data = pickle.load(open(filename, 'r')) self._properties = data['properties'] self.max = data['max'] self.min = data['min'] self.nbins = data['nbins'] self._profiles = data['profiles'] self.binind = data['binind'] logger.info("Loaded profile from %s" % filename) generate_new = False except IOError: logger.warning( "Existing profile not found -- generating one from scratch instead") if generate_new: self._properties = {} # The profile object is initialized given some array of values # and optional keyword parameters if 'max' in kwargs: kwargs['rmax'] = kwargs.pop('max') warnings.warn("Use of max as a keyword argument is deprecated. Use rmax instead.", DeprecationWarning) if 'min' in kwargs: kwargs['rmin'] = kwargs.pop('min') warnings.warn("Use of min as a keyword argument is deprecated. Use rmin instead.", DeprecationWarning) if 'rmax' in kwargs: if isinstance(kwargs['rmax'], str): self.max = units.Unit(kwargs['rmax']).ratio(x.units, **sim.conversion_context()) else: self.max = kwargs['rmax'] else: self.max = np.max(x) if 'bins' in kwargs: self.nbins = len(kwargs['bins']) - 1 elif 'nbins' in kwargs: self.nbins = kwargs['nbins'] else: self.nbins = 100 if 'rmin' in kwargs: if isinstance(kwargs['rmin'], str): self.min = units.Unit(kwargs['rmin']).ratio(x.units, **sim.conversion_context()) else: self.min = kwargs['rmin'] else: self.min = np.min(x[x > 0]) if 'bins' in kwargs: self._properties['bin_edges'] = kwargs['bins'] self.min = kwargs['bins'].min() self.max = kwargs['bins'].max() elif type == 'log': self._properties['bin_edges'] = np.logspace( np.log10(self.min), np.log10(self.max), num=self.nbins + 1) elif type == 'lin': self._properties['bin_edges'] = np.linspace( self.min, self.max, num=self.nbins + 1) elif type == 'equaln': self._properties['bin_edges'] = util.equipartition( x, self.nbins, self.min, self.max) else: raise RuntimeError("Bin type must be one of: lin, log, equaln") self['bin_edges'] = array.SimArray(self['bin_edges'], x.units) self['bin_edges'].sim = self.sim n, bins = np.histogram(self._x, self['bin_edges']) self._setup_bins() # set up the empty list of profiles self._profiles = {'n': n} def _setup_bins(self): # middle of the bins for convenience self._properties['rbins'] = 0.5 * (self['bin_edges'][:-1] + self['bin_edges'][1:]) # no idea why this next line was there... for conversion_context # self['rbins'].sim = self.sim # Width of the bins self._properties['dr'] = np.gradient( self['rbins']).view(array.SimArray) # be extra cautious carrying over stuff because sometimes fails self._properties['dr'].units = self['rbins'].units self._properties['dr'].sim = self.sim self.binind = [] if len(self._x) > 0: self.partbin = np.digitize(self._x, self['bin_edges']) else: self.partbin = np.array([]) assert self.ndim in [2, 3] if self.ndim == 2: self._binsize = np.pi * (self['bin_edges'][1:] ** 2 - self['bin_edges'][:-1] ** 2) else: self._binsize = 4. / 3. * np.pi * (self['bin_edges'][1:] ** 3 - self['bin_edges'][:-1] ** 3) # sort the partbin array from bisect import bisect sortind = self.partbin.argsort() sort_pind = self.partbin[sortind] # create the bin index arrays prev_index = bisect(sort_pind, 0) for i in range(self.nbins): new_index = bisect(sort_pind, i + 1) self.binind.append(np.sort(sortind[prev_index:new_index])) prev_index = new_index def __len__(self): """Returns the number of bins used in this profile object""" return self.nbins def _get_profile(self, name): """Return the profile of a given kind""" x = name.split(",") if name in self._profiles: return self._profiles[name] elif x[0] in Profile._profile_registry: args = x[1:] self._profiles[name] = Profile._profile_registry[x[0]](self, *args) try: self._profiles[name].sim = self.sim except AttributeError: pass return self._profiles[name] elif name in list(self.sim.keys()) or name in self.sim.all_keys(): self._profiles[name] = self._auto_profile(name) self._profiles[name].sim = self.sim return self._profiles[name] elif name[-5:] == "_disp" and (name[:-5] in list(self.sim.keys()) or name[:-5] in self.sim.all_keys()): logger.info("Auto-deriving %s" % name) self._profiles[name] = self._auto_profile( name[:-5], dispersion=True) self._profiles[name].sim = self.sim return self._profiles[name] elif name[-4:] == "_rms" and (name[:-4] in list(self.sim.keys()) or name[:-4] in self.sim.all_keys()): logger.info("Auto-deriving %s" % name) self._profiles[name] = self._auto_profile(name[:-4], rms=True) self._profiles[name].sim = self.sim return self._profiles[name] elif name[-4:] == "_med" and (name[:-4] in list(self.sim.keys()) or name[:-4] in self.sim.all_keys()): logger.info("Auto-deriving %s" % name) self._profiles[name] = self._auto_profile(name[:-4], median=True) self._profiles[name].sim = self.sim return self._profiles[name] elif name[0:2] == "d_" and (name[2:] in list(self.keys()) or name[2:] in self.derivable_keys() or name[2:] in self.sim.all_keys()): # if np.diff(self['dr']).all() < 1e-13 : logger.info("Auto-deriving %s/dR" % name) self._profiles[name] = np.gradient(self[name[2:]], self['dr'][0]) self._profiles[name] = self._profiles[name] / self['dr'].units return self._profiles[name] # else : # raise RuntimeError, "Derivatives only possible for profiles of fixed bin width." else: raise KeyError(name + " is not a valid profile") def _auto_profile(self, name, dispersion=False, rms=False, median=False): result = np.zeros(self.nbins) # force derivation of array if necessary: self.sim[name] for i in range(self.nbins): subs = self.sim[self.binind[i]] name_array = subs[name].view(np.ndarray) mass_array = subs[self._weight_by].view(np.ndarray) if dispersion: sq_mean = (name_array ** 2 * mass_array).sum() / \ self['weight_fn'][i] mean_sq = ( (name_array * mass_array).sum() / self['weight_fn'][i]) ** 2 try: result[i] = math.sqrt(sq_mean - mean_sq) except ValueError: # sq_mean<mean_sq occasionally from numerical roundoff result[i] = 0 elif rms: result[i] = np.sqrt( (name_array ** 2 * mass_array).sum() / self['weight_fn'][i]) elif median: if len(subs) == 0: result[i] = np.nan else: sorted_name = sorted(name_array) result[i] = sorted_name[int(np.floor(0.5 * len(subs)))] else: result[i] = (name_array * mass_array).sum() / self['weight_fn'][i] result = result.view(array.SimArray) result.units = self.sim[name].units result.sim = self.sim return result def __getitem__(self, name): """Return the profile of a given kind""" if name in self._properties: return self._properties[name] else: return self._get_profile(name) def __setitem__(self, name, item): """Set the profile or property by hand""" if name in self._properties: self._properties[name] = item elif name in self._profiles: self._profiles[name] = item else: raise KeyError(name + " is not a valid profile or property") def __delitem__(self, name): del self._profiles[name] def __repr__(self): return ("<Profile: " + str(self.families()) + " ; " + str(self.ndim) + "D ; " + self.type) + " ; " + str(list(self.keys())) + ">"
[docs] def keys(self): """Returns a listing of available profile types""" return list(self._profiles.keys())
[docs] def derivable_keys(self): """Returns a list of possible profiles""" return list(self._profile_registry.keys())
[docs] def families(self): """Returns the family of particles used""" return self.sim.families()
[docs] def create_particle_array(self, profile_name, particle_name=None, out_sim=None): """Create a particle array with the results of the profile calculation. After calling this function, sim[particle_name][i] == profile[profile_name][bin_in_which_particle_i_sits] If particle_name is not specified, it defaults to the same as profile_name. """ import scipy import scipy.interpolate if particle_name is None: particle_name = profile_name if out_sim is None: out_sim = self.sim out_x = self._x else: out_x = self._calculate_x(out_sim) # nearest-neighbour version if interpolation unavailable #px = np.digitize(out_x, self.bins)-1 #ok = np.where((px>=0) * (px<len(self))) #out_sim[particle_name, ok[0]] = self[profile_name][px[ok]] in_y = self[profile_name] if in_y.min() > 0: use_log = True in_y = np.log(in_y) else: use_log = False interp = scipy.interpolate.interp1d(np.log10(self.r), in_y, 'linear', copy=False, bounds_error=False) rep = interp(np.log10(out_x)) if use_log: out_sim[particle_name] = np.exp(rep) else: out_sim[particle_name] = rep rep[np.where(out_x > math.log10(self.r.max()))] = self[ profile_name][-1] rep[np.where(out_x < math.log10(self.r.min()))] = self[profile_name][0] out_sim[particle_name].units = self[profile_name].units
def _generate_hash_filename(self): """Create a filename for the saved profile from a hash using the binning data""" import hashlib try: filename = self.sim.base.filename + '.profile.' + \ (hashlib.md5(self._x, usedforsecurity=False)).hexdigest() except AttributeError: filename = self.sim.filename + '.profile.' + \ (hashlib.md5(self._x, usedforsecurity=False)).hexdigest() return filename
[docs] def write(self): """ Writes all the vital information of the profile to a file. To recover the profile, initialize a profile with the load_from_file=True keyword to automatically load a previously saved profile. The filename is chosen automatically and corresponds to a hash generated from the positions of the particles used in the profile. This is to ensure that you are always looking at the same set of particles, centered in the same way. It also means you *must* use the same centering method if you want to reuse a saved profile. """ import pickle # record all the important data except for the snapshot itself # use the hash generated from the particle list for the file name # suffix filename = self._generate_hash_filename() logger.info("Writing profile to %s", filename) pickle.dump({'properties': self._properties, 'max': self.max, 'min': self.min, 'nbins': self.nbins, 'profiles': self._profiles, 'binind': self.binind}, open(filename, 'w'))
@staticmethod def profile_property(fn): Profile._profile_registry[fn.__name__] = fn return fn
[docs]@Profile.profile_property def weight_fn(self, weight_by=None): """ Calculate mass in each bin """ if weight_by is None: weight_by = self._weight_by mass = array.SimArray(np.zeros(self.nbins), self.sim[weight_by].units) with self.sim.immediate_mode: pmass = self.sim[weight_by].view(np.ndarray) for i in range(self.nbins): mass[i] = (pmass[self.binind[i]]).sum() mass.sim = self.sim mass.units = self.sim[weight_by].units return mass
@Profile.profile_property def mass(self): return weight_fn(self, 'mass')
[docs]@Profile.profile_property def density(self): """ Generate a radial density profile for the current type of profile """ return self['mass'] / self._binsize
[docs]@Profile.profile_property def fourier(self, delta_t = "0.1 Myr", phi_bins=100): """ Generate a profile of fourier coefficients, amplitudes and phases """ delta_t = pynbody.units.Unit(delta_t) f = {'c': np.zeros((7, self.nbins), dtype=complex), 'c_delta': np.zeros((7, self.nbins), dtype=complex), 'amp': np.zeros((7, self.nbins)), 'phi': np.zeros((7, self.nbins)), 'dphi_dt': np.zeros((7, self.nbins))} for i in range(self.nbins): if self._profiles['n'][i] > 100: phi = np.arctan2( self.sim['y'][self.binind[i]], self.sim['x'][self.binind[i]]) mass = self.sim['mass'][self.binind[i]] x1 = self.sim['x'][self.binind[i]] + self.sim['vx'][self.binind[i]] * delta_t y1 = self.sim['y'][self.binind[i]] + self.sim['vy'][self.binind[i]] * delta_t phi1 = np.arctan2(y1,x1) hist_range = (-np.pi, np.pi) hist, binphi = np.histogram(phi, weights=mass, bins=phi_bins, range=(hist_range)) hist1, _ = np.histogram(phi1, weights=mass, bins=phi_bins, range=(hist_range)) binphi = binphi[:-1] + .5*np.diff(binphi) for m in range(7): f['c'][m, i] = np.sum(hist * np.exp(-1j * m * binphi)) f['c_delta'][m, i] = np.sum(hist1 * np.exp(-1j * m * binphi)) f['c'][:, self['mass'] > 0] /= self['mass'][self['mass'] > 0] f['amp'] = np.sqrt(np.imag(f['c']) ** 2 + np.real(f['c']) ** 2) f['phi'] = np.arctan2(np.imag(f['c']), np.real(f['c'])) dphi = np.arctan2(np.imag(f['c_delta']), np.real(f['c_delta'])) - f['phi'] dphi[dphi>np.pi] = dphi[dphi>np.pi]-2*np.pi dphi[dphi<-np.pi] = dphi[dphi<-np.pi]+2*np.pi dphi = array.SimArray(dphi,"1") f['dphi_dt'] = (dphi / delta_t).in_units(self.sim['vx'].units/self.sim['x'].units) return f
[docs]@Profile.profile_property def pattern_frequency(pro): """Estimate the pattern speed from the m=2 Fourier mode""" return pro['fourier']['dphi_dt'][2,:]/2
[docs]@Profile.profile_property def mass_enc(self): """ Generate the enclosed mass profile """ return self['mass'].cumsum()
[docs]@Profile.profile_property def density_enc(self): """ Generate the mean enclosed density profile """ return self['mass_enc'] / ((4. * math.pi / 3) * self['rbins'] ** 3)
[docs]@Profile.profile_property def dyntime(self): """The dynamical time of the bin, sqrt(R^3/2GM).""" dyntime = (self['rbins'] ** 3 / (2 * units.G * self['mass_enc'])) ** (1, 2) return dyntime
[docs]@Profile.profile_property def g_spherical(self): """The naive gravitational acceleration assuming spherical symmetry = GM_enc/r^2""" return (units.G * self['mass_enc'] / self['rbins'] ** 2)
[docs]@Profile.profile_property def rotation_curve_spherical(self): """ The naive rotation curve assuming spherical symmetry: vc = sqrt(G M_enc/r) """ # .in_units('km s**-1') return ((units.G * self['mass_enc'] / self['rbins']) ** (1, 2))
[docs]@Profile.profile_property def j_circ(p): """Angular momentum of particles on circular orbits.""" return p['v_circ'] * p['rbins']
[docs]@Profile.profile_property def v_circ(p, grav_sim=None): """Circular velocity, i.e. rotation curve. Calculated by computing the gravity in the midplane - can be expensive""" import pynbody.gravity.calc as gravity from .. import config global config grav_sim = grav_sim or p.sim logger.warn( "Profile v_circ -- this routine assumes the disk is in the x-y plane") # If this is a cosmological run, go up to the halo level # if hasattr(grav_sim,'base') and grav_sim.base.properties.has_key("halo_id") : # while hasattr(grav_sim,'base') and grav_sim.base.properties.has_key("halo_id") : # grav_sim = grav_sim.base # elif hasattr(grav_sim,'base') : # grav_sim = grav_sim.base start = backcompat.clock() rc = gravity.midplane_rot_curve( grav_sim, p['rbins']).in_units(p.sim['vel'].units) end = backcompat.clock() logger.info("Rotation curve calculated in %5.3g s" % (end - start)) return rc
[docs]@Profile.profile_property def E_circ(p): """Energy of particles on circular orbits.""" return 0.5 * (p['v_circ'] ** 2) + p['pot']
[docs]@Profile.profile_property def pot(p): """Calculates the potential in the midplane - can be expensive""" #from . import gravity import pynbody.gravity.calc as gravity logger.warn( "Profile pot -- this routine assumes the disk is in the x-y plane") grav_sim = p.sim # Go up to the halo level while hasattr(grav_sim, 'base') and "halo_id" in grav_sim.base.properties: grav_sim = grav_sim.base start = backcompat.clock() pot = gravity.midplane_potential( grav_sim, p['rbins']).in_units(p.sim['vel'].units ** 2) end = backcompat.clock() logger.info("Potential calculated in %5.3g s" % (end - start)) return pot
[docs]@Profile.profile_property def omega(p): """Circular frequency Omega = v_circ/radius (see Binney & Tremaine Sect. 3.2)""" prof = p['v_circ'] / p['rbins'] prof.set_units_like('km s**-1 kpc**-1') return prof
[docs]@Profile.profile_property def kappa(p): """Radial frequency kappa = sqrt(R dOmega^2/dR + 4 Omega^2) (see Binney & Tremaine Sect. 3.2)""" dOmega2dR = np.gradient(p['omega'] ** 2)/np.gradient(p['rbins']) return np.sqrt(p['rbins'] * dOmega2dR + 4 * p['omega'] ** 2)
[docs]@Profile.profile_property def beta(p): """3D Anisotropy parameter as defined in Binney and Tremiane""" assert p.ndim == 3 return 1.5 - (p['vx_disp'] ** 2 + p['vy_disp'] ** 2 + p['vz_disp'] ** 2) / p['vr_disp'] ** 2 / 2.
[docs]@Profile.profile_property def magnitudes(self, band='v'): """ Calculate magnitudes in each bin """ from . import luminosity magnitudes = np.zeros(self.nbins) for i in range(self.nbins): magnitudes[i] = luminosity.halo_mag( self.sim[self.binind[i]], band=band) magnitudes = array.SimArray(magnitudes, units.Unit('1')) magnitudes.sim = self.sim return magnitudes
@Profile.profile_property def sb(self, band='v'): # At 10 pc (distance for absolute magnitudes), 1 arcsec is 10 AU=1/2.06e4 pc # In [5]: (np.tan(np.pi/180/3600)*10.0)**2 # Out[5]: 2.3504430539466191e-09 # 1 square arcsecond is thus 2.35e-9 pc^2 sqarcsec_in_bin = self._binsize.in_units('pc^2') / 2.3504430539466191e-09 bin_luminosity = 10.0 ** (-0.4 * self['magnitudes,' + band]) #import pdb; pdb.set_trace() surfb = -2.5 * np.log10(bin_luminosity / sqarcsec_in_bin) surfb = array.SimArray(surfb, units.Unit('1')) surfb.sim = self.sim return surfb
[docs]@Profile.profile_property def Q(self): """Toomre Q parameter""" return (self['vr_disp'] * self['kappa'] / (3.36 * self['density'] * units.G)).in_units("")
[docs]@Profile.profile_property def X(self): """X parameter defined as kappa^2*R/(2*pi*G*sigma*m) See Binney & Tremaine 2008, eq. 6.77""" lambda_crit = 4. * np.pi ** 2 * units.G * \ self['density'] / (self['kappa'] ** 2) kcrit = 2. * np.pi / lambda_crit return (kcrit * self['rbins'] / 2).in_units("")
[docs]@Profile.profile_property def jtot(self): """ Magnitude of the total angular momentum """ jtot = np.zeros(self.nbins) for i in range(self.nbins): subs = self.sim[self.binind[i]] jx = (subs['j'][:, 0] * subs['mass']).sum() / self['mass'][i] jy = (subs['j'][:, 1] * subs['mass']).sum() / self['mass'][i] jz = (subs['j'][:, 2] * subs['mass']).sum() / self['mass'][i] jtot[i] = np.sqrt(jx ** 2 + jy ** 2 + jz ** 2) return jtot
[docs]@Profile.profile_property def j_theta(self): """ Angle that the angular momentum vector of the bin makes with respect to the xy-plane. """ return np.arccos(self['jz'] / self['jtot'])
[docs]@Profile.profile_property def j_phi(self): """ Angle that the angular momentum vector of the bin makes with the x-axis in the xy plane. """ j_phi = np.zeros(self.nbins) for i in range(self.nbins): subs = self.sim[self.binind[i]] jx = (subs['j'][:, 0] * subs['mass']).sum() / self['mass'][i] jy = (subs['j'][:, 1] * subs['mass']).sum() / self['mass'][i] j_phi[i] = np.arctan(jy, jx) return j_phi
[docs]class InclinedProfile(Profile): """ Creates a profile object to be used with a snapshot inclined by some known angle to the xy plane. In addition to the SimSnap object, it also requires the angle to initialize. **Example:** >>> s = pynbody.load('sim') >>> pynbody.analysis.angmom.faceon(s) >>> s.rotate_x(60) >>> p = pynbody.profile.InclinedProfile(s, 60) """ def _calculate_x(self, sim): # calculate an ellipsoidal radius return (sim['x'] ** 2 + (sim['y'] / np.cos(np.radians(self.angle))) ** 2) ** (1, 2) def __init__(self, sim, angle, load_from_file=False, ndim=2, type='lin', **kwargs): self.angle = angle Profile.__init__( self, sim, load_from_file=load_from_file, ndim=ndim, type=type, **kwargs) # define the minor axis self._properties['rbins_min'] = np.cos( np.radians(self.angle)) * self['rbins']
[docs]class VerticalProfile(Profile): """ Creates a profile object that uses the absolute value of the z-coordinate for binning. **Input**: *sim*: snapshot to make a profile from *rmin*: minimum radius for particle selection in kpc *rmax*: maximum radius for particle selection in kpc *zmax*: maximum height to consider in kpc **Optional Keywords**: *ndim*: if ndim=2, an edge-on projected profile is produced, i.e. density is in units of mass/pc^2. If ndim=3 a volume profile is made, i.e. density is in units of mass/pc^3. """ def _calculate_x(self, sim): return array.SimArray(np.abs(sim['z']), sim['z'].units) def __init__(self, sim, rmin, rmax, zmax, load_from_file=False, ndim=3, type='lin', **kwargs): if isinstance(rmin, str): rmin = units.Unit(rmin) if isinstance(rmax, str): rmax = units.Unit(rmax) if isinstance(zmax, str): zmax = units.Unit(zmax) self.rmin = rmin self.rmax = rmax self.zmax = zmax # create a snapshot that only includes the section of disk we're # interested in assert ndim in [2, 3] if ndim == 3: sub_sim = sim[ pynbody.filt.Disc(rmax, zmax) & ~pynbody.filt.Disc(rmin, zmax)] else: sub_sim = sim[(pynbody.filt.BandPass('x', rmin, rmax) | pynbody.filt.BandPass('x', -rmax, -rmin)) & pynbody.filt.BandPass('z', -zmax, zmax)] Profile.__init__( self, sub_sim, load_from_file=load_from_file, ndim=ndim, type=type, **kwargs) def _setup_bins(self): Profile._setup_bins(self) dr = self.rmax - self.rmin if self.ndim == 2: self._binsize = ( self['bin_edges'][1:] - self['bin_edges'][:-1]) * dr else: area = array.SimArray( np.pi * (self.rmax ** 2 - self.rmin ** 2), "kpc^2") self._binsize = ( self['bin_edges'][1:] - self['bin_edges'][:-1]) * area
[docs]class QuantileProfile(Profile): """ Creates a profile object that returns the requested quantiles for a given array in a given bin. The quantiles may be mass weighted. **Input**: *sim*: snapshot to make a profile from *q (default: (0.16,0.5,0.84))*: The quantiles that will be returned. Default is median with 1-sigma on either side. q can be of arbitrary length allowing the user to select any quantiles they desire. *weights (default:None)*: What should be used to weight the quantile. You will usually want to use particle mass: sim['mass']. The default is to not weight by anything, weights=None. **Optional Keywords**: *ndim*: if ndim=2, an edge-on projected profile is produced, i.e. density is in units of mass/pc^2. If ndim=3 a volume profile is made, i.e. density is in units of mass/pc^3. """ def __init__(self, sim, q=(0.16, 0.50, 0.84), weights=None, load_from_file = False, ndim = 3, type = 'lin', **kwargs): # create a snapshot that only includes the section of disk we're # interested in self.quantiles = q self.qweights = weights Profile.__init__( self, sim, load_from_file=load_from_file, ndim=ndim, type=type, **kwargs) def _get_profile(self, name): """Return the profile of a given kind""" x = name.split(",") if name in self._profiles: return self._profiles[name] elif name in list(self.sim.keys()) or name in self.sim.all_keys(): self._profiles[name] = self._auto_profile(name) self._profiles[name].sim = self.sim return self._profiles[name] else: raise KeyError(name + " is not a valid QuantileProfile") def _auto_profile(self, name, dispersion=False, rms=False, median=False): result = np.zeros((self.nbins, len(self.quantiles))) for i in range(self.nbins): subs = self.sim[self.binind[i]] with self.sim.immediate_mode: name_array = subs[name].view(np.ndarray) sorted_array = sorted(name_array) topind = len(name_array) - 1 if self.qweights is not None: sorted_weights = self.qweights[np.argsort(name_array)] for iq, q in enumerate(self.quantiles): #import pdb; pdb.set_trace() if len(name_array) > 0: if self.qweights is None: ilow = int(np.floor(q * topind)) inc = q * topind - ilow lowval = sorted_array[ilow] hival = sorted_array[ilow + 1] result[i, iq] = lowval + inc * (hival - lowval) else: cumw = np.cumsum( sorted_weights) / np.sum(sorted_weights) imin = min( np.arange(len(sorted_array)), key=lambda x: abs(cumw[x] - q)) inc = q - cumw[imin] lowval = sorted_array[imin] if inc > 0: nextval = sorted_array[imin + 1] else: if imin == 0: nextval = lowval else: nextval = sorted_array[imin - 1] result[i, iq] = lowval + inc * (nextval - lowval) #if (result[i,iq] < 1e-6) : import pdb; pdb.set_trace() else: result[i, iq] = np.nan self['rbins'][i] = np.nan result = result.view(array.SimArray) result.units = self.sim[name].units result.sim = self.sim return result