toksearch-signal-routing

name	toksearch-signal-routing
description	Read this FIRST when unsure which Signal class or data path a DIII-D quantity needs. Maps physics quantities (Ip, Wmhd, beta_N, q95, NBI power, density, Te, boundary, ...) to the correct Signal class and path, how to discover IMAS paths, and which skill to use for time-alignment/correlation. Avoids the common mistake of using PtDataSignal for a quantity that lives in MDSplus or IMAS.

Choosing the right signal class for a DIII-D quantity

Picking the wrong backend is the most common failure. PtData only serves raw PTDATA pointnames; asking it for an analysis quantity (e.g. wmhd, betan) fails with PtDataError: Shot.extension not found. Use this table.

Quantity	Class & path	Notes
Plasma current Ip	PtDataSignal('ip')	amps -> /1e6 for MA; times in ms
Ip from EFIT	MdsSignal(r'\ipmhd', 'efit01')	MdsSignal is imported from toksearch
Stored energy Wmhd	MdsSignal(r'\wmhd', 'efit01')	not a PTDATA pointname; Joules -> /1e6 for MJ
Normalized beta (beta_N)	ImasSignal('equilibrium.time_slice.global_quantities.beta_normal')	EFIT01-backed
q95	ImasSignal('equilibrium.time_slice.global_quantities.q_95')
Total NBI power	ImasSignal('nbi.unit.power_launched.data')	object array of per-unit W series (see recipe)
Core electron density	ImasSignal('core_profiles.profiles_1d.electrons.density_thermal')	2-D (time, rho)
Thomson Te (per channel)	ImasSignal('thomson_scattering.channel.t_e.data', split_by='channel')	dict keyed by channel
Plasma boundary outline	ImasSignal('equilibrium.time_slice.boundary.outline.r')	ragged object array

PtDataSignal/ImasSignal come from toksearch_d3d; MdsSignal from toksearch.

Discovering IMAS paths (don't guess)

from toksearch_d3d import list_imas_fields
list_imas_fields()          # all supported IDS names -> fields
list_imas_fields('nbi')     # fields under one IDS

Supported IDSs: core_profiles, ec_launchers, ece, equilibrium, gas_injection, magnetics, nbi, reflectometer_profile, tf, thomson_scattering, wall. There is no summary IDS; line-averaged density is not available via IMAS here.

Recipe: total injected NBI power (object array)

nbi.unit.power_launched.data returns a numpy object array of per-unit power series (watts). Sum over units, then convert:

import numpy as np
def peak_nbi_mw(rec):
    units = rec['pnbi']['data']                       # object array, ~8 units
    total = np.nansum(np.stack([np.asarray(u, float) for u in units]), axis=0)
    rec['peak_pnbi_MW'] = float(np.nanmax(total) / 1e6)   # 0.0 if no NBI

Aligning / correlating two signals

Do not hand-roll resampling. Use Pipeline.fetch_dataset + Pipeline.align (see the toksearch-datasets skill) so both signals share one grid:

pipe.fetch_dataset('ds', {'ip': PtDataSignal('ip'),
                          'wmhd': MdsSignal(r'\wmhd', 'efit01')})
pipe.align('ds', align_with=10.0, method='pad')   # uniform 10 ms grid, zero-order hold
@pipeline.map
def corr(rec):
    ds = rec['ds']; a = ds['ip'].values; b = ds['wmhd'].values
    m = np.isfinite(a) & np.isfinite(b)
    rec['corr'] = float(np.corrcoef(a[m], b[m])[0, 1])

SQL shot metadata

from toksearch_d3d.sql import connect_d3drdb        # sets TDSVER for you

Canonical shot-type classification is the shots_type table (shot_type='plasma'), joined on shots.shot = shots_type.shot — not the shot_type column on shots.

API reminders

• Times are milliseconds.
• Set record fields by item assignment in a def map: rec['k'] = v (Record has no .update(); lambdas can't assign).
• rec.get('k', default) requires the default argument.

See also

• toksearch-d3d-imas — full ImasSignal API (ragged data, split_by, dims).
• toksearch-datasets — fetch_dataset/align for multi-signal grids.
• toksearch-d3d-ptdata / toksearch-mds — per-backend details.