PlasmaCalcs.hookups.eppic.eppic_runtime_info.EppicRuntimeInfoLoader

class PlasmaCalcs.hookups.eppic.eppic_runtime_info.EppicRuntimeInfoLoader

Bases: QuantityLoader

runtime info for EppicCalculator.

__init__()

Methods

`__init__`()
`attach_extra_coords`(arr)
`clock_times`()
`cls_help`([qstr, only, tree, modules, ...])
`cls_var_tree`(var, *[, missing_ok])
`copy`()
`direct_overrides_dynamic`()
`get_behavior`([keys])
`get_calc_time`()
`get_clock_times`()
`get_guess_cpu_seconds`(var, *[, _match])
`get_guess_cpu_seconds_per_each`(var, *[, _match])
`get_guess_cpu_seconds_per_timestep_or_ct`(var, *)
`get_guess_node_hours`(var, *[, _match])
`get_guess_runtime_seconds`(var, *[, _match])
`get_init_seconds`()
`get_run_time`()
`get_runtimes`()
`get_seconds2timer`()
`get_set_or_cached`(var)
`get_steps_seconds`()
`get_time_frac`(var, *[, _match])
`get_timer2seconds`()
`get_timers`()
`get_timestep_cost_or_dt_cost`(var, *[, _match])
`get_total_seconds`()
`get_vars`(vars, *args[, return_type, ...])
`get_write_time`()
`has_var`(var)
`help`([qstr, only, tree, modules, signature, ...])
`help_call_options`([search])
`help_quants_str`([qstr, only, tree, modules, ...])
`help_str`([qstr, only])
`jobfiles`()
`kw_call_options`(*[, sorted])
`load_direct`(var, args, *kw)
`load_fromfile`(var, args, *kw)
`maintaining_attrs`(attrs, *attrs_as_flags)
`match_var`(var, *[, check])
`match_var_loading_dims`(var, **kw_loading_dims)
`match_var_result_dims`(var, **kw_result_dims)
`match_var_result_size`(var, *[, maindims])
`match_var_tree`([var])
`quant_tree`([var])
`set_var`(var, value[, behavior_attrs, ...])
`set_var_internal`(var, value, behavior_attrs)
`timers_dat`(*[, with_snaps, as_array])
`tree`([var])
`unset_var`(var[, behavior_attrs, missing_ok])
`unset_var_internal`(var, behavior_attrs[, ...])
`using_at_call_depth`(depth, **attrs_and_values)
`using_at_next_call_depth`(**attrs_and_values)
`using_attrs`([attrs_as_dict, _unset_sentinel])

Attributes

`CPU_SECONDS_PER_EACH_INFO`
`KNOWN_PATTERNS`
`KNOWN_SETTERS`
`KNOWN_VARS`
`assign_behavior_attrs`
`assign_behavior_attrs_max_call_depth`
`assign_behavior_attrs_skip_xr`
`behavior`
`behavior_attrs`
`call_depth`
`call_depth_manager`
`cls_behavior_attrs`
`direct_overrides`
`enable_fromfile`
`extra_coords`
`get`
`known_pattern`
`known_setter`
`known_var`
`logfiles`
`maintaining`
`nondim_behavior_attrs`
`set`
`setvar`
`setvars`
`toplevel_scale_coords`
`typevar_crash_if_nan`
`unset`
`using`

property assign_behavior_attrs: whether to assign self.behavior values as attrs of result when calling self.

False –> don’t use self.behavior code architecture to assign attrs.

True –> equivalent to ‘nondefault’

‘nondefault’ –> self.behavior.assign_nondefault_attrs(result)

(for brevity, it does not assign behavior attrs with “default” value.)

‘all’ –> self.behavior.assign_attrs(result).

[EFF] only assigns attrs at call_depth >= self.assign_behavior_attrs_max_call_depth.

(default: only assigns attrs at call_depth=1, i.e. at top level.

property assign_behavior_attrs_max_call_depth: max call_depth at which to assign_behavior_attrs to result,

if self.assign_behavior_attrs indicates to assign behavior attrs.

default 1, i.e. only assign if at top level.

Use None to indicate “no max detph”.

property assign_behavior_attrs_skip_xr: whether to use include_xr=False if self.assign_behavior_attrs,

during self.behavior.assign_nondefault_attrs.

Use this if you want to assign behavior attrs EXCEPT array-valued behavior attrs.

attach_extra_coords(arr): attach any self.extra_coords to array arr but only if it is an xarray.DataArray or xarray.Dataset

property behavior: dict of {attr: self.attr} for attr in self.behavior_attrs. Note dims are separate;

dims go in behavior.dims. E.g. Behavior({‘units’:’si’,…}, dims={‘snap’:0,…}).

property behavior_attrs: list of attrs in self which control behavior of self.

Here, returns self.cls_behavior_attrs.

Subclasses could override if any behavior attrs are not known at the class-level,

e.g. if MySubclass’s list of behavior attrs varies between instances of MySubclass.

property call_depth: depth of the current call to self. depth = number of calls to self from within self.

E.g., call_depth while calculating gyrofrequency:

# call_depth == 0, for any code run here (outside any call to self).

self(‘gyrof’)

# call_depth == 1, for any code run here (inside ‘gyrof’ call but not inside deeper calls).

q = self(‘q’)

# call_depth == 2, for code inside ‘q’ call.

mod_B = self(‘mod_B’)

# call_depth == 2, for code inside ‘mod_B’ call.

self(‘B’)

# call_depth == 3, for code inside ‘B’ call.

m = self(‘m’)

# call_depth == 2, for code inside ‘m’ call.

result = q * mod_B / m

Cannot be set directly; can only be manipulated via self.call_depth_manager.

property call_depth_manager: stores the value of call_depth, and helps to manage attrs dependent on call_depth value.

clock_times(): return dict of clock times from this run. (From reading self.jobfiles.)

Result can have keys:

‘start’: datetime telling when the run started.

‘stepstart’: datetime telling when the iterations started.

‘end’: datetime telling when the run ended.

‘init_seconds’: (stepstart - start) [seconds]

‘steps_seconds’: (end - stepstart) [seconds]

‘total_seconds’: (end - start) [seconds]

If jobfile missing any info, relevant keys will not appear in result.

classmethod cls_help(qstr=None, only=None, *, tree=None, modules=False, signature=False, doc=True, dense=False, print=True, **kw)

prints str for help with quants. Fails for any quants which depend on present values of a cls instance.

qstr: None or str: None –> tells info about this class & how to use this function.

in particular, tells that quants are stored cls.KNOWN_VARS and cls.KNOWN_PATTERNS,

and describes behavior of calling help with a string.

str –> return str for help with all quants related to str.

use empty str to get help for all quants.
only: None or str: None –> get help with all quantities related to qstr.

‘VARS’ or ‘vars’ –> only get help with KNOWN_VARS.

‘PATTERNS’ or ‘patterns’ –> only get help with KNOWN_PATTERNS.

‘TREE’ or ‘tree’ –> only get help with quantities in cls.cls_var_tree(str).

if provided when qstr is None, treat qstr as ‘’ instead.
tree: None or bool: How much help to give for quantities in cls.cls_var_tree(qstr).

False –> don’t even check cls.cls_var_tree(qstr).

True –> help for all quantities in cls.cls_var_tree.

None –> help for quantities in cls.cls_var_tree(qstr).flat_branches_until_vars()

i.e. patterns & vars in tree but ignore any nodes with LoadableVar ancestors.

e.g. qstr=’mean_mod_beta’ –> help with ‘mean_(.+)’, ‘mod_(.+)’, and ‘beta’,

but no help with dependencies of ‘beta’ (‘q’, ‘mod_B’, ‘m’).
modules: bool: Whether to include modules in result.

If True, result will be grouped into sections with modules written at top.
signature: signature: bool: whether to include line with signature in help string.

e.g. “help_str(f, *, module=True, signature=True, indent=None)”
doc: doc: bool: whether to include lines with docstring in help string.

e.g. “return str for help(f).” … and all the other docs in here.
dense: bool: Whether to reduce whitespace in result.

E.g. True –> no newlines between functions. False –> one newline between functions.
print: bool: whether to print the result. If False, return the result instead of printing.

classmethod cls_var_tree(var, *, missing_ok=False)

return QuantTree of MatchedQuantity objects from matching var and all dependencies,

using self.KNOWN_VARS and self.KNOWN_PATTERNS when searching for matches.

missing_ok: bool: whether to be lenient sometimes when missing details that would allow to fully determine deps.

see help(MatchedQuantity.dep_vars) for more details.

copy(): returns a deep copy of self.

[TODO] implement something less hacky than using the pickle module?

property direct_overrides: dict of {var: override} for all overrides of self which don’t depend on behavior_attrs of self.

For example, if user wants to set an override (or if setvars sets an override?), it will be here.

See also: self.direct_overrides_dynamic().

direct_overrides_dynamic(): returns dict of {var: override} for all overrides of self which depend on behavior_attrs of self.

property enable_fromfile: bool: whether self.load_fromfile is enabled during self.load_direct.

If False, raise QuantCalcError if load_direct can’t get value without load_fromfile().

property extra_coords: dict of {coord_name: coord_value} to attach to outputs of self(var).

Useful if planning to join the output of self(var) with output from a different QuantityLoader.

E.g. self.extra_coords={‘run’: ‘run 0’} and other.extra_coords={‘run’: ‘run 1’},

then xr.concat([self(‘n’), other(‘n’)], ‘run’) gives ‘n’ from self AND other.

(this is nice if self and other have same values for dims. Otherwise, might struggle.)

property get: alias to __call__

get_behavior(keys=None)

return value of self.behavior.

keys: None or iterable: if provided, only include these attrs.

from nondim_behavior_attrs, or dims.

get_calc_time(): Wall clock runtime, ignoring time spent writing output files. Same units as timers_dat info.

get_clock_times(): dataset of clock times from this run. (From reading self.jobfiles.)

Result can have keys:

‘start’: datetime telling when the run started.

‘stepstart’: datetime telling when the iterations started.

‘end’: datetime telling when the run ended.

‘init_seconds’: (stepstart - start) [seconds]

‘steps_seconds’: (end - stepstart) [seconds]

‘total_seconds’: (end - start) [seconds]

If jobfile missing any info, relevant keys will not appear in result.

get_guess_cpu_seconds(var, *, _match=None): guess run cost, in cpu seconds. nt * guess_cpu_seconds_sum_per_timestep_safe

Uses ‘safe’ guess (unless ‘_unsafe’ included in name) but excludes initialization time.

Might want to add a safety factor (e.g. assume 20% larger than this time).

get_guess_cpu_seconds_per_each(var, *, _match=None): guess log10(cpu seconds per (particle, per) grid cell, per timestep), for the indicated timer.

if ‘safe’ included in var, uses mean + stddev instead of just mean, to get a conservative estimate.

Result excludes initialization time (before iterations begin).

timer options:

‘all’: result will be a Dataset of results from all timers, with timer names as data_vars.

(timer names are ‘vadv time’, ‘xadv time’, ‘charge’, ‘collect’, ‘efield’, ‘output’.)

‘vadv’, ‘xadv’, ‘charge’, ‘collect’, ‘efield’, ‘output’: result for corresponding timer.

if timer is ‘all’, return sum of the results from other timers.

These results are per particle per grid cell per timestep:

‘vadv time’, ‘xadv time’, ‘charge’, ‘output’.

These results are per particle per timestep:

‘collect’, ‘efield’.

The guess is informed by self.CPU_SECONDS_PER_EACH_INFO,

which contains numerical values derived from some simulations run on Frontera.

Assumes ‘output’ is not very expensive.

The “output” timer was not tested across a large variety of ‘nout’.

(trying to detrend with respect to nout didn’t affect stddev significantly.)

get_guess_cpu_seconds_per_timestep_or_ct(var, *, _match=None): guess cpu seconds required per (cell, per) timestep, for the indicated timer.

if ‘safe’ included in var, uses mean + stddev instead of just mean, to get a conservative estimate.

Use ‘ct’ to indicate ‘per cell per timestep’.

(result using ‘timestep’) == ncells_sim * (result using ‘ct’).

Result excludes initialization time (before iterations begin).

timer options:

‘all’: result will be a Dataset of results from all timers, with timer names as data_vars.

(timer names are ‘vadv time’, ‘xadv time’, ‘charge’, ‘collect’, ‘efield’, ‘output’.)

‘sum’: result will be the sum of the results from other timers.

‘vadv’, ‘xadv’, ‘charge’, ‘collect’, ‘efield’, ‘output’: result for corresponding timer.

Assumes ‘output’ is not very expensive; it was not tested as much as the others.

The guess is informed by self.CPU_SECONDS_PER_EACH_INFO.

For more details see: guess_cpu_seconds_per_each.

get_guess_node_hours(var, *, _match=None): guess run cost, in node hours. guess_cpu_seconds * seconds2hours / tasks_per_node

(seconds2hours = 1/3600)

Uses ‘safe’ guess (unless ‘_unsafe’ included in name) but excludes initialization time.

Might want to add a safety factor (e.g. assume 20% larger than this time).

get_guess_runtime_seconds(var, *, _match=None): guess run cost, in wall clock time [seconds]. guess_cpu_seconds / n_processors.

Uses ‘safe’ guess (unless ‘_unsafe’ included in name) but excludes initialization time.

Might want to add a safety factor (e.g. assume 20% larger than this time).

Compare directly with self(‘steps_seconds’).

get_init_seconds(): time [in seconds] spent initializing the run. (between start and when steps start.)

get_run_time(): Wall clock runtime for each snap. Same units as timers_dat info.

get_runtimes(): timers_dat info as an xarray.DataArray, at snaps in self.snap. see also: ‘timers’.

get_seconds2timer(): conversion factor from seconds to timers.dat units.

E.g. seconds2timer = 100 <–> 1 timer unit = 0.01 seconds <–> 1 second = 100 timer units.

seconds2timer = sum(run_time) / steps_seconds

Might vary from machine to machine. Seems to be 100 on Frontera (plus small rounding errors).

get_set_or_cached(var): returns var if found in self.setvars or self.cache, with compatible behavior_attrs.

otherwise, raise CacheNotApplicableError.

if var is found in self.setvars and has relevant, but not matching behavior_attrs,

self.load_across_dims will be used to load the value.

get_steps_seconds(): time [in seconds] spent doing timesteps. (total duration minus init_seconds.)

get_time_frac(var, *, _match=None): fraction of runtime spent on writing or calculating.

var=’write_time_frac’ –> fraction of runtime spent writing output files.

var=’calc_time_frac’ –> fraction of runtime spent calculating, ignoring write_time.

get_timer2seconds(): conversion factor from timers.dat units to seconds.

E.g. timer2seconds = 0.01 <–> 1 timer unit = 0.01 seconds <–> 1 second = 100 timer units.

timer2seconds = steps_seconds / sum(run_time)

Might vary from machine to machine. Seems to be 0.01 on Frontera (plus small rounding errors).

get_timers(): timers_dat info as an xarray.Dataset, at snaps in self.snap. see also: ‘runtimes’.

get_timestep_cost_or_dt_cost(var, *, _match=None): total cpu time per simulated particle, per timestep (or per dt).

time_cost = (runtime / timestep_or_dt) * (n_processors / total number of particles)

total number of particles = n_processors * npart.

Note: n_processors cancels out; time_cost = (runtime / timestep_or_dt) / npart

timestep_or_dt = one timestep or one dt; see below.

npart = number of simulated particles, in one processor. Depends on {settings}; see below.

‘{clock}_{time}_cost{settings}’

E.g. ‘run_timestep_cost’, ‘write_dt_cost_f’, ‘calc_dt_cost_nosubf’

{clock} = ‘run’, ‘calc’, or ‘write’

tells which clock to use.

‘run’ –> ‘Wall clock’ | ‘calc’ –> ‘Wall Clock - output’ | ‘write’ –> ‘output’

{time} = ‘timestep’ or ‘dt’

‘timestep’ –> report result as cost per timestep, regardless of dt.

‘dt’ –> report result as cost per dt (converted to SI units).

{settings} = ‘’, ‘_f’, ‘_nosub’, ‘_fnosub’, or ‘_nosubf’

tells whether to return a separate value for each fluid, and whether to account for subcycling.

‘’ –> single value. account for subcycling. npart = self(‘npd/subcycle’).sum(‘fluid’)

‘_nosub’ –> single value. ignore subcycling. npart = self(‘npd’).sum(‘fluid’)

‘_f’ –> per-fluid values. account for subcycling. npart = self(‘npd/subcycle’)

‘_fnosub’ –> per-fluid values. ignore subcycling. npart = self(‘npd’)

‘_nosubf’ –> same as ‘_fnosub’; provided for convenience.

accounting for subcycling means dividing by the subcycling factor,

because less effort is spent on subcycled distributions.

get_total_seconds(): duration [in seconds] spent to run the entire run.

get_vars(vars, *args, return_type='dataset', missing_vars=UNSET, **kw)

returns values of vars from self.

result is probably an xarray.Dataset, but not guaranteed; also depends on return_type.

Equivalent to self(vars, *args, return_type=’dataset’, **kw).

(Actually, self(vars, …) will call self.get_vars(vars, …).)

vars: iterable of strs: Names of the vars to load. [‘n’, ‘u’] for number density & velocity.

if any of these vars returns a return_type object, expand its keys,

e.g. if ‘myDSvar’ returns dataset with ‘myvar1’, ‘myvar2’,

then [‘n’, ‘myDSvar’] gives dataset with ‘n’, ‘myvar1’, ‘myvar2’.
return_type: ‘dataset’ or ‘dict’: if ‘dataset’, return result as xarray.Dataset.

the data_var names will be the same as the var names.

if ‘dict’, return result as dict of {var: value}.
missing_vars: UNSET, ‘ignore’, ‘warn’, or ‘raise’: what to do if any vars cause FormulaMissingError.

UNSET –> use self.missing_vars if it exists, else ‘raise’.

‘ignore’ –> ignore missing vars, and don’t include them in the result.

‘warn’ –> ignore missing vars, but print a warning.

‘raise’ –> raise FormulaMissingError if any vars are missing.

additional args & kwargs are passed to self(…).

get_write_time(): time spent writing output files. Same units as timers_dat info.

has_var(var): return whether self can load var. True if self.match_var(var) is found, else False.

Subclasses might override, to include checks for whether var can be loaded from data.

[TODO] also check if var in self.cache or self.setvars.

help(qstr=None, only=None, *, tree=None, modules=False, signature=False, doc=True, dense=False, print=True)

prints str for help with quants.

qstr: None or str: None –> tells info about this class & how to use this function.

in particular, tells that quants are stored cls.KNOWN_VARS and cls.KNOWN_PATTERNS,

and describes behavior of calling help with a string.

str –> return str for help with all quants related to str.

use empty str to get help for all quants.
only: None or str: None –> get help with all quantities related to qstr.

‘VARS’ or ‘vars’ –> only get help with KNOWN_VARS.

‘PATTERNS’ or ‘patterns’ –> only get help with KNOWN_PATTERNS.

‘TREE’ or ‘tree’ –> only get help with quantities in cls.cls_var_tree(str).

if provided when qstr is None, treat qstr as ‘’ instead.
tree: None or bool: How much help to give for quantities in cls.cls_var_tree(qstr).

False –> don’t even check cls.cls_var_tree(qstr).

True –> help for all quantities in cls.cls_var_tree.

None –> help for quantities in cls.cls_var_tree(qstr).flat_branches_until_vars()

i.e. patterns & vars in tree but ignore any nodes with LoadableVar ancestors.

e.g. qstr=’mean_mod_beta’ –> help with ‘mean_(.+)’, ‘mod_(.+)’, and ‘beta’,

but no help with dependencies of ‘beta’ (‘q’, ‘mod_B’, ‘m’).
modules: bool: Whether to include modules in result.

If True, result will be grouped into sections with modules written at top.
signature: signature: bool: whether to include line with signature in help string.

e.g. “help_str(f, *, module=True, signature=True, indent=None)”
doc: doc: bool: whether to include lines with docstring in help string.

e.g. “return str for help(f).” … and all the other docs in here.
dense: bool: Whether to reduce whitespace in result.

E.g. True –> no newlines between functions. False –> one newline between functions.

help_call_options(search=None): prints help for kw_call_options.

if search is provided, only print help for keys containing search.

classmethod help_quants_str(qstr=None, only=None, *, tree=None, modules=True, signature=False, doc=True, dense=False, _instance=None)

returns str for help with quants.

qstr: None or str: None –> tells info about this class & how to use this function.

in particular, tells that quants are stored cls.KNOWN_VARS and cls.KNOWN_PATTERNS,

and describes behavior of calling help with a string.

str –> return str for help with all quants related to str.

use empty str to get help for all quants.
only: None or str: None –> get help with all quantities related to qstr.

‘VARS’ or ‘vars’ –> only get help with KNOWN_VARS.

‘PATTERNS’ or ‘patterns’ –> only get help with KNOWN_PATTERNS.

‘TREE’ or ‘tree’ –> only get help with quantities in cls.cls_var_tree(str).

if provided when qstr is None, treat qstr as ‘’ instead.
tree: None or bool: How much help to give for quantities in cls.cls_var_tree(qstr).

False –> don’t even check cls.cls_var_tree(qstr).

True –> help for all quantities in cls.cls_var_tree.

None –> help for quantities in cls.cls_var_tree(qstr).flat_branches_until_vars()

i.e. patterns & vars in tree but ignore any nodes with LoadableVar ancestors.

e.g. qstr=’mean_mod_beta’ –> help with ‘mean_(.+)’, ‘mod_(.+)’, and ‘beta’,

but no help with dependencies of ‘beta’ (‘q’, ‘mod_B’, ‘m’).
modules: bool: Whether to include modules in result.

If True, result will be grouped into sections with modules written at top.
signature: signature: bool: whether to include line with signature in help string.

e.g. “help_str(f, *, module=True, signature=True, indent=None)”
doc: doc: bool: whether to include lines with docstring in help string.

e.g. “return str for help(f).” … and all the other docs in here.
dense: bool: Whether to reduce whitespace in result.

E.g. True –> no newlines between functions. False –> one newline between functions.
_instance: None or QuantityLoader instance: if provided, use _instance.match_var_tree() instead of cls.cls_var_tree().

classmethod help_str(qstr=None, only=None, **kw): returns cls.help_quants_str(qstr=qstr, only=only, **kw).

cls.help() calls help_str.

subclasses might overwrite help_str, but probably won’t touch help_quants_str.

jobfiles(): return list of all jobfiles within self.dirname directory.

kw_call_options(*, sorted=True): returns list of kwarg names which can be used to set attrs self during self.__call__.

(see self.__call__ for more details).

Here, returns list(self.behavior_attrs) + list(self._extra_kw_for_quantity_loader_call)

load_direct(var, *args, **kw): load var “directly”, from some source which is not known by the main part of PlasmaCalcs.

Attempt the following, returning the first successful attempt:

- return self.direct_overrides[var](self, *args, **kw).

- return self.direct_overrides_dynamic()[var](self, *args, **kw).

- use self.load_fromfile.

return the result (probably a numpy array, but not guaranteed).

Examples:

load Bx directly from a file

load n for H+, using a different module which somehow gives nH+

(PlasmaCalcs doesn’t need to know where the value came from.)

if used an override, instead of loading from file,

set self._load_direct_used_override = var.

Otherwise, set it to None.

This might be used, e.g., to determine if the output came directly from a file or not.

load_fromfile(var, *args, **kw): load var directly from a file. Other methods should usually use load_direct, instead.

the implementation here just raises LoadingNotImplementedError;

subclasses should implement this method in order to load any values from files.

property logfiles: alias to jobfiles

property maintaining: alias to maintaining_attrs

maintaining_attrs(*attrs, **attrs_as_flags): returns context manager which restores attrs of self to their original values, upon exit.

E.g. maintaining_attrs(obj, ‘attr1’, ‘attr2’, attr3=True, attr4=False)

–> will restore upon exit, original values of obj.attr1, attr2, and attr3, but not attr4.

classmethod match_var(var, *, check=['KNOWN_VARS', 'KNOWN_PATTERNS'])

match var from cls.KNOWN_VARS or cls.KNOWN_PATTERNS, or raise FormulaMissingError.

returns result=MatchedQuantity(var, loadable, _match=_match) where:

loadable is the LoadableQuantity associated with this var,

_match is:

None, if var in cls.KNOWN_VARS;

re.fullmatch(pattern, var), if var matches any pattern in cls.KNOWN_PATTERNS.

if var matches multiple patterns, only the first matching pattern is used.

Uses MatchedVar if match from KNOWN_VARS, MatchedPattern if from KNOWN_PATTERNS.

(note that both MatchedVar and MatchedPattern subclass MatchedQuantity.)

check: str or list of str from [‘KNOWN_VARS’, ‘KNOWN_PATTERNS’]: where to check for matches. Default is to check KNOWN_VARS and KNOWN_PATTERNS.

E.g. to only check KNOWN_PATTERNS, use check=[‘KNOWN_PATTERNS’].

loadable and _match can be retrieved via result.loadable and result._match.

match_var_loading_dims(var, **kw_loading_dims): return dims for loading var across.

Result will probably vary across these dims (but not guaranteed, if any dependency uses reduces_dims.)

These are all Dimension dims, not maindims. (E.g. ‘fluid’ and ‘snap’, but not ‘x’, ‘y’, ‘z’).

Equivalent: self.match_var_tree(var).loading_dims(**kw_loading_dims)

match_var_result_dims(var, **kw_result_dims): return dims which result of cls(var) will vary across.

These are all Dimension dims, not maindims. (E.g. ‘fluid’ and ‘snap’, but not ‘x’, ‘y’, ‘z’).

Equivalent: cls.match_var_tree(var).result_dims(**kw_result_dims)

match_var_result_size(var, *, maindims=True, **kw_result_dims)

return size (number of elements) which self(var) will have.
(Efficient; doesn’t actually get self(var).)
Depends on current values of relevant dims. (E.g., self.fluid, not self.fluids)

maindims: bool: if True, include maindims_shape when calculating size.

match_var_tree(var=UNSET, **kw_quant_tree_from_quantity_loader): return QuantTree of MatchedQuantity objects from matching var and all dependencies,

using self.KNOWN_VARS and self.KNOWN_PATTERNS when searching for matches.

var must be provided; var=UNSET will raise an error (helpful if tried calling this as a classmethod).

See also: type(self).cls_var_tree, for the classmethod version of this function.

Most of the time it is possible to get tree without any details from self,

but sometimes not. e.g. when getting collision frequencies, self.fluid affects deps.

additional kwargs will be passed to QuantTree.from_quantity_loader(…),

which passes kwargs from self.kw_call_options() into self.using(**kw) while getting deps.

matched_pattern_cls: alias of MatchedPattern

matched_var_cls: alias of MatchedVar

property nondim_behavior_attrs: list of attrs in self which control behavior of self, but which are NOT in self.dimensions.

quant_tree(var=UNSET, **kw_quant_tree_from_quantity_loader): return QuantTree of MatchedQuantity objects from matching var and all dependencies,

using self.KNOWN_VARS and self.KNOWN_PATTERNS when searching for matches.

var must be provided; var=UNSET will raise an error (helpful if tried calling this as a classmethod).

See also: type(self).cls_var_tree, for the classmethod version of this function.

Most of the time it is possible to get tree without any details from self,

but sometimes not. e.g. when getting collision frequencies, self.fluid affects deps.

additional kwargs will be passed to QuantTree.from_quantity_loader(…),

which passes kwargs from self.kw_call_options() into self.using(**kw) while getting deps.

quant_tree_cls: alias of QuantTree

property set: alias to set_var

set_var(var, value, behavior_attrs=None, forall=[], *, ukey=None, forced=False, **kw_using)

set var in self. When later doing self(var) to get var, return the set value,

but only if self.behavior is compatible with the relevant parts of self.behavior when var was set.

This function will use, if it exists:

self.KNOWN_SETTERS[var](self, value, behavior_attrs, forall=forall)

Otherwise, calls:

self.set_var_internal(var, value, self.behavior_attrs, forall=forall)

var: str: the var to set in self.
value: number, xarray, iterable or 1D array, array with shape matching self.maindims_shape.: the value to set var to.

number –> set var to this number.

xarray –> set var to this xarray.

[TODO](not yet implemented) iterable or 1D array –> set var to these values along dim=’testing’.

[TODO](not yet implemented) array with shape matching self.maindims_shape –> set var to this array.
behavior_attrs: None or list: tells which attrs from self control behavior of the set var.

The set var will only be retrieved when behavior_attrs of self are compatible.

E.g. set_var(‘n’, [‘fluid’, ‘snap’]) –> saves ‘n’ in cache with current fluid & snap.

Will only load ‘n’ if self.fluid and self.snap == cached fluid and snap for ‘n’.

if var in self.KNOWN_SETTERS, cannot provide behavior_attrs here.

else, use self.behavior_attrs if None.
forall: list of strings: if provided, tells which attrs of self do NOT control the behavior of the set var.

E.g. forall=[‘snap’] –> ‘snap’ will NOT be included in behavior_attrs.

(anything in behavior_attrs AND forall will be removed from the final behavior_attrs)
ukey: None or str: if provided, tells string to give to UnitsManager when converting value’s units.

When ukey is known, setting value in any unit system will enable to read it in all unit systems.

E.g. set_var(‘n’, 1e10, …, ukey=’n’, units=’si’)

–> self(‘n’, units=’raw’) == self(‘n’, units=’si’) * self.u(‘u’, ‘raw’, convert_from=’si’)

if not provided, value will be associated with current unit system;

attempted to read value in any other unit system will not used the cached value set here.

E.g. set_var(‘u’, 1e10, …, units=’si’) # ukey not provided

–> self(‘u’, units=’raw’) –> uses self’s other logic for getting ‘u’, not from setvars.

note: if provided, ‘units’ will be added to behavior_attrs if not already in there.
forced: bool, default True: handles the case where self.KNOWN_SETTERS[var] doesn’t exist. In that case…

True –> set var in self, anyway.

False –> crash; raise FormulaMissingError

additional kwargs, if provided, go to self.using(**kw) during the operation.

returns list of set quantities.

set_var_internal(var, value, behavior_attrs, forall=[], *, ukey=None)

set var in self. KNOWN_SETTERS functions may wish to use this method.

(KNOWN_SETTERS functions should NOT use self.set_var, to avoid recursion issue.)

This function has the internal logic for self.set_var;

set_var calls set_var_internal when self.KNOWN_SETTERS[var] not provided.

var: str: the var to set in self.
value: number, xarray, iterable or 1D array, array with shape matching self.maindims_shape.: the value to set var to. See help(self.set_var) for more info.
behavior_attrs: list of strings: the behavior attrs relevant to setting this var;

getting var only gives value when current behavior attrs values are compatible with the cached ones.
forall: list of strings: if provided, tells which behavior attrs do NOT control the behavior of the set var.

e.g. behavior_attrs=[‘snap’, ‘fluid’], forall=[‘snap’] –> use [‘fluid’], only.
ukey: None or str: if provided, tells string to give to UnitsManager when converting value’s units;

when ukey is provided, can retrieve value in any unit system (probably ‘si’ or ‘raw’).

when ukey not provided, if ‘units’ in used behavior attrs, can only retrieve value in that unit system.

property setvar: alias to set_var

property setvars: VarCache of vars set via self.set_var(). Returns these values when appropriate,

i.e. whenever self.behavior is compatible with the behavior in the cache.

To empty the cache, use self.setvars.clear() to empty the cache.

timers_dat(*, with_snaps=False, as_array=False)

return timers.dat as an xarray.Dataset. (dimension will be named ‘it’)

result will have the same units as timers.dat file.

with_snaps: bool: if True, attach snap & t coords and promote ‘snap’ to main dim.

based on self.snaps (not self.snap)
as_array: bool: whether to use xarray.Dataset.to_array() to return a DataArray instead of Dataset.

if True, vars from Dataset will be concatenated along the new dimension named ‘timer’.

property toplevel_scale_coords: dict of {coord_name: coord_scaling} to apply to top-level outputs of self(var).

(Never applies to internal calls of self(var), only applies at self.call_depth==1.)

Useful if making plots and want to scale coords by some factor.

E.g., self.toplevel_scale_coords = {‘t’: 1000} to convert s to ms.

CAUTION: coord units labels will remain unaffected.

tree(var=UNSET, **kw_quant_tree_from_quantity_loader): return QuantTree of MatchedQuantity objects from matching var and all dependencies,

using self.KNOWN_VARS and self.KNOWN_PATTERNS when searching for matches.

var must be provided; var=UNSET will raise an error (helpful if tried calling this as a classmethod).

See also: type(self).cls_var_tree, for the classmethod version of this function.

Most of the time it is possible to get tree without any details from self,

but sometimes not. e.g. when getting collision frequencies, self.fluid affects deps.

additional kwargs will be passed to QuantTree.from_quantity_loader(…),

which passes kwargs from self.kw_call_options() into self.using(**kw) while getting deps.

property typevar_crash_if_nan: bool. whether to crash methods if typevar output would be ‘nan’.

False –> return NaN when typevar gives ‘nan’, instead of crashing.

“typevar” here refers to any var used for checking which formula to use, from various options,

e.g. ‘ntype’ in MhdMultifluidLoader or ‘ionfrac_type’ in MhdIonizationLoader.

The relevant methods can check if self.typevar_crash_if_nan before returning a ‘nan’ result.

property unset: alias to unset_var

unset_var(var, behavior_attrs=[], *, missing_ok=True, **kw_using)

remove var from self.setvars (but only at values stored with relevant behavior).

[TODO] define rules for which vars unset which other vars…

e.g. for eppic right now, set_var(‘n’) sets ‘den’ but not ‘n’;

unset_var(‘n’) unsets nothing… but should probably alias to unset_var(‘den’).

behavior_attrs: list of strings: only remove cached values where self.behavior matches cached behavior for these attrs.

if empty, remove all cached values for var, regardless of associated behavior.
missing_ok: bool: whether it is okay for there to be zero matching cached values for var.

raise CacheNotApplicableError if missing_ok=False when there are no matching cached values.

additional kwargs, if provided, go to self.using(**kw) during the operation.

return list of CachedQuantity objects which were removed from self.setvars.

unset_var_internal(var, behavior_attrs, forall=[], *, ukey=None, missing_ok=True)

unset var from self.setvars.

KNOWN_SETTERS functions may wish to use this method, to unset dependent values.

E.g. if u depends on n, and n is changed, may wish to unset the value of u.

behavior_attrs: list of strings: the behavior attrs relevant to setting this var.
forall: list of strings: if provided, tells which behavior attrs to ignore when unsetting the var.
ukey: None or string: if provided, ignore ‘units’ behavior attr when unsetting the var

(due to assuming that ukey was provided when setting the var,

hence that the set var could be retrieved in any units system)
missing_ok: bool: whether it is okay for there to be zero matching cached values for var.

raise CacheNotApplicableError if missing_ok=False when there are no matching cached values.

return list of CachedQuantity objects which were removed from self.setvars.

property using: alias to using_attrs

using_at_call_depth(depth, **attrs_and_values): context manager for setting attrs_and_values but only while call_depth == depth.

E.g.:

with self.using_at_call_depth(3, verbose=3):

self(‘sgyrof’)

# while self.call_depth == 3 inside of this ‘with’ block, uses self.verbose=3.

# but everywhere else, uses original value of verbose.

# assuming originally verbose=False (or unset), this example will print:

| | (call_depth=2) get var=’q’

| | (call_depth=2) get var=’mod_B’

| | (call_depth=2) get var=’m’

# compare this to simply using self.verbose=3, which would print:

| (call_depth=1) get var=’sgyrof’

| | (call_depth=2) get var=’q’

| | (call_depth=2) get var=’mod_B’

| | | (call_depth=3) get var=’B_dot_B’

| | | | (call_depth=4) get var=’B_xyz’

| | | | | (call_depth=5) get var=’B’

| | (call_depth=2) get var=’m’

Equivalent to self.call_depth_manager.using_obj_attrs_at(depth, **attrs_and_values)

using_at_next_call_depth(**attrs_and_values): context manager for setting attrs_and_values but only while call_depth == self.call_depth + 1

Equivalent to self.using_at_call_depth(self.call_depth + 1, **attrs_and_values).

(Also equivalent to self.call_depth_manager.using_obj_attrs_at_next(**attrs_and_values).)

using_attrs(attrs_as_dict={}, _unset_sentinel=ATTR_UNSET, **attrs_and_values)

returns context manager which sets attrs of obj upon entry; restores original values upon exit.

_unset_sentinel: any value, default ATTR_UNSET: upon entry, delete any attrs with value _unset_sentinel (compared via ‘is’).

E.g. using_attrs(obj, _unset_sentinel=None, x=None) –> del obj.x upon entry.