Newer
Older
from binarycpython.utils.grid import Population
from binarycpython.utils.functions import (
get_help_all,
get_help,
create_hdf5,
output_lines,
)
from binarycpython.utils.custom_logging_functions import temp_dir
#########################################################
# This file serves as an example for running a population.
# The use of help(<function>) is a good way to inspect what parameters are there to use
#########################################################
def parse_function(self, output):
# extract info from the population instance
# Get some information from the
data_dir = self.custom_options["data_dir"]
base_filename = self.custom_options["base_filename"]
# Check directory, make if necessary
os.makedirs(data_dir, exist_ok=True)
seperator = " "
# Create filename
outfilename = os.path.join(data_dir, base_filename)
parameters = ["time", "mass", "zams_mass", "probability", "radius", "stellar_type"]
# Go over the output.
for el in output_lines(output):
headerline = el.split()[0]
# CHeck the header and act accordingly
if headerline == "MY_STELLAR_DATA":
values = el.split()[1:]
print("Amount of column names isnt equal to amount of columns")
raise ValueError
if not os.path.exists(outfilename):
with open(outfilename, "w") as f:
f.write(seperator.join(parameters) + "\n")
with open(outfilename, "a") as f:
f.write(seperator.join(values) + "\n")
# Create population object
example_pop = Population()
# If you want verbosity, set this before other things
example_pop.set(verbose=1)
# Setting values can be done via .set(<parameter_name>=<value>)
# Values that are known to be binary_c_parameters are loaded into bse_options.
# Those that are present in the default grid_options are set in grid_options
# All other values that you set are put in a custom_options dict
example_pop.set(
# binary_c physics options
M_1=10, # bse_options
separation=0, # bse_options
max_evolution_time=15000, # bse_options
eccentricity=0.02, # bse_options
M_2=0.08, # Since in the example we run a single system, we should set the companion mass here. If we donm't do this, the code will complain.
# grid_options
verbose=1, # verbosity. Not fully configured correctly yet but having it value of 1 prints alot of stuff
# Custom options # TODO: need to be set in grid_options probably
temp_dir(), "example_python_population_result"
), # custom_options
base_filename="example_pop.dat", # custom_options
)
# Creating a parsing function
example_pop.set(
parse_function=parse_function, # Setting the parse function thats used in the evolve_population
)
### Custom logging
## Below example requires changing the parse function
## very simple example of custom logging. Will work but need to change the parse function to handle that nicely.
# example_pop.set(
# C_auto_logging={
# "MY_HEADER_LINE": ["star[0].mass", "star[1].mass", "model.probability"]
# }
# )
# Log the moment when the star turns into neutron
C_logging_code="""
{
if (stardata->model.time < stardata->model.max_evolution_time)
{
Printf("MY_STELLAR_DATA %30.12e %g %g %g %g %d\\n",
//
stardata->model.time, // 1
stardata->star[0].mass, // 2
stardata->common.zero_age.mass[0], // 4
stardata->model.probability, // 5
stardata->star[0].radius, // 6
stardata->star[0].stellar_type // 7
);
};
/* Kill the simulation to save time */
stardata->model.max_evolution_time = stardata->model.time - stardata->model.dtm;
};
David Hendriks
committed
# Add grid variables
example_pop.add_grid_variable(
name="lnm1",
longname="Primary mass",
valuerange=[2, 150],
resolution="{}".format(resolution["M_1"]),
spacingfunc="const(math.log(2), math.log(150), {})".format(resolution["M_1"]),
precode="M_1=math.exp(lnm1)",
probdist="three_part_powerlaw(M_1, 0.1, 0.5, 1.0, 150, -1.3, -2.3, -2.3)*M_1",
dphasevol="dlnm1",
parameter_name="M_1",
condition="", # Impose a condition on this grid variable. Mostly for a check for yourself
# # Mass ratio
# test_pop.add_grid_variable(
# name="q",
# longname="Mass ratio",
# valuerange=["0.1/M_1", 1],
# resolution="{}".format(resolution['q']),
# spacingfunc="const(0.1/M_1, 1, {})".format(resolution['q']),
# probdist="flatsections(q, [{'min': 0.1/M_1, 'max': 1.0, 'height': 1}])",
# dphasevol="dq",
# precode="M_2 = q * M_1",
# parameter_name="M_2",
# condition="", # Impose a condition on this grid variable. Mostly for a check for yourself
# test_pop.add_grid_variable(
# name="log10per", # in days
# valuerange=[0.15, 5.5],
# resolution="{}".format(resolution["per"]),
# spacingfunc="const(0.15, 5.5, {})".format(resolution["per"]),
# precode="""orbital_period = 10** log10per
# sep = calc_sep_from_period(M_1, M_2, orbital_period)
# sep_min = calc_sep_from_period(M_1, M_2, 10**0.15)
# probdist="sana12(M_1, M_2, sep, orbital_period, sep_min, sep_max, math.log10(10**0.15), math.log10(10**5.5), -0.55)",
# parameter_name="orbital_period",
# dphasevol="dlog10per",
# )
# Exporting of all the settings can be done with .export_all_info()
# on default it exports everything, but can be supressed by turning it off:
# population settings (bse_options, grid_options, custom_options), turn off with include_population
# settings=False
# binary_c_defaults (all the commandline arguments that binary c accepts, and their defaults).
# turn off with include_binary_c_defaults=False
# include_binary_c_version_info (all the compilation info, and information about the compiled
# parameters), turn off with include_binary_c_version_info=False
# include_binary_c_help_all (all the help information for all the binary_c parameters),
# turn off with include_binary_c_help_all=Fase
# On default it will write this to the custom_options['data_dir'], but that can be overriden by
# setting use_datadir=False and providing an outfile=<>
## Executing a single system
## This uses the M_1 orbital period etc set with the set function
# output = example_pop.evolve_single()
# print(output)
## Executing a population
## This uses the values generated by the grid_variables
# Wrapping up the results to an hdf5 file can be done by using the create_hdf5
# (<directory containing data and settings>) This function takes the settings file
# (ending in _settings.json) and the data files (ending in .dat) from the data_dir
# and packing them into an hdf5 file, which is then written into the same data_dir directory
create_hdf5(data_dir=example_pop.custom_options["data_dir"], name="example_pop.hdf5")