starfit package

Subpackages

Submodules

starfit.direct module

find best fot for a selection of stars

class starfit.direct.Direct(*args, stars=None, offsets=None, fixed_offsets=False, optimize=True, **kwargs)

Bases: StarFit

Find the best fit for stars you specify

when multiple DBs are specified, provide tuples of (DB, index) where DB is 1-based

starfit.fit module

Fitting stars

starfit.fit.gen_map(gen_start, gen_end, num, size=None, com=None)
starfit.fit.get_fitness(trimmed_db, eval_data, z_exclude_index, sol, ejecta=[], fixed_offsets=False, local_search=True, **kwargs)

Evaluate the fitness of a set of solutions. If abundance is directly given in the case of smart GA, use that.

starfit.fit.get_solution(gen_start, gen_end, exclude_index, trimmed_db, fixed_offsets=False, ejecta=None, return_size=None, sol_size=None, num=None, size=None, com=None, index=None, **kwargs)

Local search solve for a set of stars. Used for splitting into different processes. This needs to be a separate function that can be pickled.

starfit.ga module

Genetic algorithm

class starfit.ga.Ga(*args, gen=1000, time_limit=20, pop_size=200, sol_size=None, tour_size=2, frac_mating_pool=1, frac_elite=0.5, mut_rate_index=0.2, mut_rate_offset=0.1, mut_offset_magnitude=1, local_search=True, fixed_offsets=False, seed=None, max_pop=8192, spread=None, group=None, pin=None, n_top=20, interactive=True, **kwargs)

Bases: StarFit

Genetic Algorithm.

plot_fitness(gen=False, fig=None, ax=None, show_copyright=True, figsize=(10, 6), dpi=102)

plot fitness as a function of time

print_empty_table()
print_header()
print_update()

starfit.multi module

class starfit.multi.Multi(*args, n_top=10, fixed_offsets=False, threads=None, nice=19, save=False, webfile=None, path=None, block_size=131072, sol_size=None, group=None, **kwargs)

Bases: StarFit

Find the best fit for multiple stars (complete search). The search results are printed live.

group:

if number of data bases matches sol_size, take one from each DB

sol_size:
None:

if group is True, set sol_size to number of DBs

TODO - allow multiple contributions form one group

(sol_size becomes vector with length of number of groups)

init_futures(threads=None, nice=19)
print_empty_table()
print_header()
print_update()
save_file(file_name)
working_arrays()

starfit.single module

class starfit.single.Single(*args, **kwargs)

Bases: StarFit

Find the best fit for single stars (complete search)

Example:

import starfit

s = starfit.Single(
    filename = 'HE1327-2326.dat',
    db = 'znuc2012.S4.star.el.y.stardb.gz',
    combine = [[6, 7, 8]],
    z_max = 30,
    z_exclude = [3, 24, 30],
    z_lolim = [21, 29],
    upper_lim = True,
    cdf = True,
    constraints = 'energy <= 5',
    )
sol_size = 1

Single has sol_size hardcoded to 1. All other arguments are inherited from StarFit

starfit.star module

Reads elemental abundance file into an array File format:

Version 100200

Add tetection threshold. Otherwise identical to and backward compatible with Format 10100

10200 <— VERSION NUMBER SDSS12345-6789 <– STAR NAME ApJ 222, 333 (1999) <– SOURCE/REFERENCE data automatically generated by Anna Frebel <– SOME COMMENT 1 1 <— DATA FORMAT, MODE (DETECTION THRESHOLD PRESENT = 1)

<— SYMBOL OF ELEMENT OR VALUE TO WHICH DATA IS NORMALIZED, IF ANY

17 <— NUMBER OF ELEMENTS C -1.23 +0.07 -1.07 +0.05 -0.05 N 1.23 +0.14 -2.14 -0.10 +0.10 O -1.23 +0.21 -3.21 -0.15 -0.15 <— MEASUREMNT, Fe 1.23 +0.28 -4.28 +0.20 +0.20 UNCORRELATED ERROR, Mg -1.23 +0.35 -5.35 +0.25 -0.25 DETECTION THRESHOLD, Si 1.23 +0.42 -6.42 +0.30 +0.30 COVARIANCE VECTORS Sc 1.23 +0.57 -7.57 -0.35 -0.35 Ca -1.23 +0.49 -8.49 Ti -1.23 +0.64 -9.64 V +1.23 +0.64 -8.64 <— MEASUREMNT, Cr 1.23 +0.10 -6.64 UNCORRELATED ERROR, Mn -1.23 +0.20 -6.64 DETECTION THRESHOLD Co -1.23 -0.30 Ni -1.23 -0.40 Cu 1.23 -0.50 Sr -1.23 -0.60 <— UPPER LIMIT, Zr 1.23 -0.70 (UNCORRELATED) UNCERTAINTIES Ba -1.23 -0.80 Lo09 <— SOLAR REFERENCE; USE “-” IF NOT PROVIDED

DATA FORMAT 0 Y (mol/g), error is absolute 1 log epsilon (PREFERRED DUE TO DIFFERENT SOLAR ABUNDANCES, give H mol/g as norm!) 2 [ ] 3 [X/Y] Y= norm element, provide [Y/H] in column for Y 4 log X/Si + 6 (by number), norm is assumeed log(Si) + 6 (mol/g) 5 log mole fraction (REALLY PREFERRED) 6 [X/H] - provide [H] as norm, otherwise BBN value is assumed 7 X/Si * 1e6 (by number), norm is Si/Si_sun, error is absolute

The first number is the data, the second is the uncorrelated (random)error in dex. Use negative random error values to indicate upper limits.

The 3rd column is the detection threshold, if present. May be “-” if missing.

The following columns give first correlated errors.

The covariances can have positive or negative signs, the uncorrelated (random) error always is inherently non-negative.

The square of the total error is the sum of the squares of all errors.

If no covariances are provide, the entire error is random.

As of the time of this writing, StarFit does not use correlated upper limits, but that may change in the future.

Version 10100

Adds covariant errors, format compatible with Version 10002

10100 <— VERSION NUMBER SDSS12345-6789 <– STAR NAME ApJ 222, 333 (1999) <– SOURCE/REFERENCE data automatically generated by Anna Frebel <– SOME COMMENT 1 <— DATA FORMAT (RECOMEND USE LOG EPSILON OR LOG MOL FRACTION)

<— NAME OF ELEMENT OR VALUE TO WHICH DATA IS NORMALIZED, IF ANY

17 <— NUMBER OF ELEMENTS C -1.23 +0.07 +0.05 -0.05 N 1.23 +0.14 -0.10 +0.10 <— MEASUREMNT, O -1.23 +0.21 -0.15 -0.15 UNCORRELATED ERROR, Fe 1.23 +0.28 +0.20 +0.20 COVARIANCE VECTORS Mg -1.23 +0.35 +0.25 -0.25 Si 1.23 +0.42 +0.30 +0.30 Ca -1.23 +0.49 -0.35 -0.35 Sc 1.23 +0.57 Ti -1.23 +0.64 Cr 1.23 +0.10 <— MEASUREMENT, Mn -1.23 +0.20 UNCORRELATED ERRORS Ni -1.23 -0.40 Cu 1.23 -0.50 Sr -1.23 -0.60 <— UPPERL LIMIT, Zr 1.23 -0.70 (UNCORRELATED) UNCERTAINTIES Ba -1.23 -0.80 Lo09 <— SOLAR REFERENCE; USE “-” IF NOT PROVIDED

DATA FORMAT 1 log epsilon (PREFERRED DUE TO DIFFERENT SOLAR ABUNDANCES, give H mol/g as norm!) 2 [ ] 3 [X/Y] Y= norm element, provide [Y/H] in column for Y 4 log X/Si + 6 (by number), norm is assumeed log(Si) + 6 (mol/g) 5 log mole fraction (REALLY PREFERRED) 6 [X/H] - provide [H] as norm, otherwise BBN value is assumed 7 X/Si * 1e6 (by number), norm is Si/Si_sun, error is absolute

The first number is the data, the second is the uncorrelated error in dex. Use negative random error values to indicate upper limits.

The following columns give first covariant errors.

The covariant errors can have positive or negative signs, the uncorrelated (random) error always is inherently non-negative.

The square of the total error is the sum of the squares of all errors.

If no covariances are provide, the entire error is random.

As of the time of this writing, StarFit does not use correlated upper limits, but that may change in the future.

Version 100002

10002 <— VERSION NUMBER (FILE FORMAT) CS30336-049 <– STAR NAME ApJ 222, 333 (1999) <– SOURCE/REFERENCE data automatically generated by Anna Frebel <– SOME COMMENT 1 <— DATA FORMAT (USE LOG EPSILON OR LOG MOLE FRACTION)

<— NORM, IF ANY (ELEMENT OR ABUNDANCE OFFSET)

17 <— NUMBER OF ELEMENTS C -1.23 0.1 N 1.23 0.2 O -1.23 0.3 Fe 1.23 0.4 <— MEASUREMENT, Mg -1.23 0.5 UNCORRELATED ERRORS Si 1.23 0.6 Ca -1.23 0.7 Sc 1.23 0.8 Ti -1.23 0.9 Cr 1.23 -0.1 Mn -1.23 -0.2 Co 1.23 -0.3 Ni -1.23 -0.4 <— UPPERL LIMIT, Sr -1.23 -0.6 (UNCORRELATED) UNCERTAINTIES Zr 1.23 -0.7 Ba -1.23 -0.8 Lo09 <— SOLAR REFERENCE; USE “-” IF NOT PROVIDED

DATA FORMAT

1 log epsilon (PREFERRED DUE TO DIFFERENT SOLAR ABUNDANCES, give H mol/g as norm!) 2 [ ] 3 [X/Y] Y = norm element, provide [Y/H] in column for Y 4 log X/Si + 6 (by number), norm is assumeed log(Si) + 6 (mol/g) 5 log mole fraction (REALLY PREFERRED) 6 [X/H] - provide [H] as norm, otherwise BBN value is assumed 7 X/Si * 1e6 (by number), norm is Si/Si_sun, error is absolute

The first number is the data, the second is the error in dex. Use negative values to indicate upper limits; all need to be non-zero.

Version 100001

10001 <— VERSION NUMBER (FILE FORMAT) CS30336-049 <– STAR NAME ApJ 222, 333 (1999) <– SOURCE/REFERENCE data automatically generated by Anna Frebel <– SOME COMMENT 1 <— DATA FORMAT (USE LOG EPSILON OR LOG MOLE FRACTION)

<— NORM, IF ANY (ELEMENT OR ABUNDANCE OFFSET)

17 <— NUMBER OF ELEMENTS C -1.23 0.1 N 1.23 0.2 O -1.23 0.3 Fe 1.23 0.4 <— MEASUREMENT, Mg -1.23 0.5 UNCORRELATED ERRORS Si 1.23 0.6 Ca -1.23 0.7 Sc 1.23 0.8 Ti -1.23 0.9 Cr 1.23 -0.1 Mn -1.23 -0.2 Co 1.23 -0.3 Ni -1.23 -0.4 <— UPPERL LIMIT, Sr -1.23 -0.6 (UNCORRELATED) UNCERTAINTIES Zr 1.23 -0.7 Ba -1.23 -0.8

DATA FORMAT

1 log epsilon (PREFERRED DUE TO DIFFERENT SOLAR ABUNDANCES, give H mol/g as norm!) 2 [ ] 3 [X/Y] Y = norm element, provide [Y/H] in column for Y 4 log X/Si + 6 (by number), norm is assumeed log(Si) + 6 (mol/g) 5 log mole fraction (REALLY PREFERRED) 6 [X/H] - provide [H] as norm, otherwise BBN value is assumed 7 X/Si * 1e6 (by number), norm is Si/Si_sun, error is absolute

The first number is the data, the second is the error in dex. Use negative values to indicate upper limits; all need to be non-zero.

Version 100000

10001 <— VERSION NUMBER (FILE FORMAT) CS30336-049, data from a friend <– STAR NAME AND COMMENT 1 <— DATA FORMAT (USE LOG EPSILON OR LOG MOLE FRACTION)

<— NORM, IF ANY (ELEMENT OR ABUNDANCE OFFSET)

17 <— NUMBER OF ELEMENTS C -1.23 0.1 N 1.23 0.2 O -1.23 0.3 Fe 1.23 0.4 <— MEASUREMENT, Mg -1.23 0.5 UNCORRELATED ERRORS Si 1.23 0.6 Ca -1.23 0.7 Sc 1.23 0.8 Ti -1.23 0.9 Cr 1.23 -0.1 Mn -1.23 -0.2 Co 1.23 -0.3 Ni -1.23 -0.4 <— UPPERL LIMIT, Sr -1.23 -0.6 (UNCORRELATED) UNCERTAINTIES Zr 1.23 -0.7 Ba -1.23 -0.8

DATA FORMAT

1 log epsilon (PREFERRED DUE TO DIFFERENT SOLAR ABUNDANCES, give H mol/g as norm!) 2 [ ] 3 [X/Y] Y = norm element, provide [Y/H] in column for Y 4 log X/Si + 6 (by number), norm is assumeed log(Si) + 6 (mol/g) 5 log mole fraction (REALLY PREFERRED) 6 [X/H] - provide [H] as norm, otherwise BBN value is assumed 7 X/Si * 1e6 (by number), norm is Si/Si_sun, error is absolute

The first number is the data, the second is the error in dex. Use negative values to indicate upper limits; all need to be non-zero.

class starfit.star.Star(filename, silent=False)

Bases: Logged

All the data read from each file

get_covariances()
get_detection_thresholds()
get_elements()

return all elements in dataset

get_input_data_format()
get_measured()
get_n_covariances()
get_norm()
get_upper_limits()
list_to_array(data_list)
exception starfit.star.StarFileError

Bases: Exception

starfit.starfit module

Results objects from the various algorithms

exception starfit.starfit.ConstraintsParseError

Bases: Exception

class starfit.starfit.Convert_from_5(starfit, yscale=2, ynorm=None, ylabel=None)

Bases: object

err(err)
property ylabel
class starfit.starfit.StarFit(*args, silent=False, **kwargs)

Bases: Logged

Base class for running the various fitting algorithms.

Parameters:

  • filename (str) – Filename of star. Can be absolute or relative path. The files will also be searched for in the distribution files and in the search path specified by environment variable STARFIT_DATA in subdirectory stars.

  • db (str or pathlib.Path, optional) – database file or tuple of data base files. String or Path object. Can be absolute or relative path. Files will also be searched in the distribution files and in the search path specified by environment variable STARFIT_DATA in subdirectory db. You may also use the wildcard ("*") in the data base name. The code will then try to resolve all matching data bases in the first source directory that contains any matching file. The plain * argument will include all data bases in the first source that contains any data base; the matching is done against the pattern *.stardb.*. The Ellipis (... Python object, not in quotation marks) will do the same as the plain * argument, but will continue searching through all data souces. This allows for an easy way to search across all model data bases available. combine (list, optional): A list of lists of element charge numbers to treat as combined abundances (e.g. combine the CNO elements using [[6,7,8]]).

  • z_min (int, optional) – Lowest element charge number to fit.

  • z_max (int, optional) – Highest element charge number to fit.

  • z_exclude (list, optional) – Element charge numbers to exclude from fit.

  • lim_exclude (bool, optional) – Treat z_min and z_max limits as exclusions (default: True). Otherwise databases are “trimmed” to save memory and data cannot be plotted in interactive mode.

  • z_lolim (list, optional) – Elements that are model lower limits (effectively the same as observational upper limits).

  • upper_lim (bool, optional) – Include observational upper limits in data fitting.

  • cdf (bool, optional) – Use the uncertainty of upper limits to calculate a cumulative distribution function when calculating error contribution (otherwise treat the upper limit as a simple one-sided \({\chi^2}\) error).

  • det (bool, optional) – Use the detection limits when calculating error contribution (experimental).

  • cov (bool, optional) – Use the error covariances when calculating error contribution (experimental).

  • dst (bool, optional) – Use statistical error only for detection treshold (default: True; experimental).

  • limit_solver (bool, optional) – Solver/search will only allow solutions for each star that contribute no more than 100%.

  • limit_solution (bool, optional) – Solver/search will only allow solutions where the total adds up to no more than 100% contributions from all stars. Results from the search are renormalised accordingly.

  • y_floor (bool, optional) – Floor value for abundances to assume in models (default: 1e.0e-99). This is useful for elements not produced in a model, otherwise \({\chi^2}\); or \({-\infty}\); may result.

  • db_label (list, optional) – A list of labels for the data bases to be used in plots and tables. Will only be shown if there is more than one database specified. If present, needs to match the number of databases specified. If not present, databases will be numbered starting with 0, unless the StarDB has a label field that will be used instead. The maximum label length currently allowed is 8.

  • show (bool, optional) – Show list of loaded databases with label/number

  • name (and) –

  • quit. (then) –

  • constraints (str, optional) – String with list of conditions separated by comma (acts as “and”). Conditions for specific databases can be prefixed with a number (zero-based) if the index of the database in the list, followed by a colon(:). Entries for different databases are separated by semicolon (;). The fieldname has to be given first, then the operator, and finally the comparison value. Allowed operators are <, <=, ==, >=, >, and !=.

  • constraints_error (str, optional) – one of warn (default), raise, ignore. How StarDB deal with errors in constraints.

W()
format(*args, **kwargs)
format_comments(npad=72, dbx=None)
format_db(ind=0, pad='', **kwargs)
info(i=0, **kwargs)
plot(num=0, fig=None, ax=None, fontsize=None, annosize=None, figsize=(10, 6), dpi=102, show_copyright=True, dist=None, xlim=None, ylim=None, data_size=3, save=None, save_format=None, return_plot_data=False, yscale=2, ynorm='Fe', range_det=False, range_lim=True, range_zmin=3, pad_abu=0.1, pad_det=0.05, multi=0, xlabel=None, ylabel=None)

Create a plot of the solution.

Note

The legend as well as the star name and copyright string can be moved (dragged).

Parameters:

  • num (int) – Number of solution, from the top (default: 0).

  • yscale (int) – select the y-scale of the plot. Numerical value identical to those used for the star data formats.

  • ynorm (str) – elements to use a norm for [X/Y] plots (yscale=3).

  • multi (int) – plot this many best solutions as grey lines (default: 0). For multi=-1 lines will be shaded according to relative data point probability based on \({\chi^2}\); and assuming multi-dimensional Gaussian error.

  • save (str) – filename to save plot.

  • range_det (bool) – adjust range to include detection thresholds (default: False)

  • range_lim – (bool) adjust range to include detection limits (default: True)

  • range_zmin (int) – minimum Z to consider for determining y range (default: 3)

  • pad_abu (float) – fraction of plot range to use at bottom/top boundary (default: 0.1).

  • pad_det (float) – fraction of plot range to pad detection thresholds (default: 0.05).

  • figsize (tuple) – dimensions of figure in inches (default: (10, 6)).

  • dpi (int) – resolution of image (default: 102).

  • xlim (tuple) – overwrite x range (low, high).

  • ylim (tuple) – overwrite y range (low, high).

  • data_size (int) – Size of data lines and symbols (default: 3).

  • fontsize (int) – size used for axis labels (default: 12).

  • annosize (str) – size used for element symbols (default: small).

  • dist (float) – distance of labels from data points.

  • fig (:class:matplotlib.figure.Figure) – figure object to use as canvas, otherwise as new figure is created.

  • ax (:class:matplotlib.axis.Axis) – axis objects to use for drawing, otherwise axis and parent figure are created as needed.

  • xlabel (str) – overwrite label for x-axis.

  • ylabel (str) – overwrite label for y-axis.

plot_error_matrix(num=0, zoom=False, nlab=9)
plot_star_inverse(zoom=False, nlab=9, compress=True)
plot_star_matrix(zoom=False, nlab=9, compress=True)
print(*args, **kwargs)
print_comments(*args, **kwargs)
print_db(*args, **kwargs)
run(stars, offsets=None, optimize=True, **kwargs)

Solve for the specified stars

text_db(dbx=None, filename=False, fields=False)
text_result(n=20, *, n0=0, format='unicode', best=True, wide=12, show_index=True, _return_dbx=False)

Print data of best fit.

static textpad(s, n)

starfit.starplot module

Various plotting routines

class starfit.starplot.IntFormatter(*args, **kwargs)

Bases: FuncFormatter

starfit.starplot.leg_info(ax, lines)
starfit.starplot.leg_starname(ax, name)

starfit.utils module

starfit.utils.find_all(subdir, pattern, complete=False)

Search through all the data directories for the filename. Return the absolute path of all files - in the first directory one is found if complete is False - in all subdirectories if complete is True Search precedence is: 1) absolute path if provided 2) STARFIT_DATA env var (set in __init__.py) 3) Installation DATA_DIR (set in __init__.py)

starfit.utils.find_data(subdir, filename)

Search through all the data directories for the filename. Return the absolute path if found. Precedence is: 1) absolute path provided 2) STARFIT_DATA env var (set in __init__.py) 3) Installation DATA_DIR (set in __init__.py)

starfit.utils.getch()
starfit.utils.set_priority(value: int)

Module contents

class starfit.Direct(*args, stars=None, offsets=None, fixed_offsets=False, optimize=True, **kwargs)

Bases: StarFit

Find the best fit for stars you specify

when multiple DBs are specified, provide tuples of (DB, index) where DB is 1-based

class starfit.Ga(*args, gen=1000, time_limit=20, pop_size=200, sol_size=None, tour_size=2, frac_mating_pool=1, frac_elite=0.5, mut_rate_index=0.2, mut_rate_offset=0.1, mut_offset_magnitude=1, local_search=True, fixed_offsets=False, seed=None, max_pop=8192, spread=None, group=None, pin=None, n_top=20, interactive=True, **kwargs)

Bases: StarFit

Genetic Algorithm.

plot_fitness(gen=False, fig=None, ax=None, show_copyright=True, figsize=(10, 6), dpi=102)

plot fitness as a function of time

print_empty_table()
print_header()
print_update()
class starfit.Multi(*args, n_top=10, fixed_offsets=False, threads=None, nice=19, save=False, webfile=None, path=None, block_size=131072, sol_size=None, group=None, **kwargs)

Bases: StarFit

Find the best fit for multiple stars (complete search). The search results are printed live.

group:

if number of data bases matches sol_size, take one from each DB

sol_size:
None:

if group is True, set sol_size to number of DBs

TODO - allow multiple contributions form one group

(sol_size becomes vector with length of number of groups)

init_futures(threads=None, nice=19)
print_empty_table()
print_header()
print_update()
save_file(file_name)
working_arrays()
class starfit.Single(*args, **kwargs)

Bases: StarFit

Find the best fit for single stars (complete search)

Example:

import starfit

s = starfit.Single(
    filename = 'HE1327-2326.dat',
    db = 'znuc2012.S4.star.el.y.stardb.gz',
    combine = [[6, 7, 8]],
    z_max = 30,
    z_exclude = [3, 24, 30],
    z_lolim = [21, 29],
    upper_lim = True,
    cdf = True,
    constraints = 'energy <= 5',
    )
sol_size = 1

Single has sol_size hardcoded to 1. All other arguments are inherited from StarFit

class starfit.Star(filename, silent=False)

Bases: Logged

All the data read from each file

get_covariances()
get_detection_thresholds()
get_elements()

return all elements in dataset

get_input_data_format()
get_measured()
get_n_covariances()
get_norm()
get_upper_limits()
list_to_array(data_list)