Deborah.Sarah

Deborah.Sarah — Module

module Sarah

Deborah.Sarah — Common utilities for configuration, logging, resampling, I/O, and naming.

Deborah.Sarah is the utility toolkit shared across the Deborah.jl ecosystem. It offers job-aware logging, TOML helpers, string/name encoding, deterministic seeding, dataset partitioning and meta extraction, data loading, bootstrap/jackknife resampling, and summary formatting/collection. These utilities keep the higher-level modules small, consistent, and reproducible.

Scope & Responsibilities

Logging & diagnostics: structured println/warn, timing, and job-scoped messages (Deborah.Sarah.JobLoggerTools).
TOML helpers: append well-formed sections to existing TOML files (TOMLLogger.append_section_to_toml).
Naming & encoding: derive stable names/paths (Deborah.Sarah.NameParser) and map long tokens $\Leftrightarrow$ abbreviations/IDs (Deborah.Sarah.StringTranscoder).
Seeding: centralized RNG seeding mainly used by bootstrap resampling for reproducible runs (Deborah.Sarah.SeedManager).
Dataset partitioning: compute LB/TR/BC/UL splits and index ranges (Deborah.Sarah.DatasetPartitioner).
XY metadata & extraction: read columns, reshape per-(conf,src), and derive configuration indices (XYInfoGenerator.gen_conf_from_Y, XYInfoGenerator.gen_X_info).
Data loading: robust delimited-file loaders with fallbacks (Deborah.Sarah.DataLoader).
Resampling: bootstrap and jackknife engines plus runners/utilities (Deborah.Sarah.Bootstrap, Deborah.Sarah.Jackknife, Deborah.Sarah.BootstrapRunner).
Summaries: pretty (avg, err) formatting and printing; accumulate per-observable summaries (Deborah.Sarah.SummaryFormatter, Deborah.Sarah.SummaryCollector).

Key Components (selected)

Deborah.Sarah.JobLoggerTools — job-scoped logging and convenience printers.
TOMLLogger.append_section_to_toml — append a new [section] to a TOML file.
Deborah.Sarah.StringTranscoder — encode/decode feature/target names, abbreviation dicts.
Deborah.Sarah.NameParser — build canonical overall_name, model suffixes, and path pieces.
Deborah.Sarah.SeedManager — set/get seeds for RNG workflows.
Deborah.Sarah.DatasetPartitioner — compute label/train/bias-correction/unlabeled indices.
XYInfoGenerator.gen_conf_from_Y — pick per-config IDs from a Y matrix (or 1:N_cnf).
XYInfoGenerator.gen_X_info — extract a single component column into (channels, N_cnf, N_src).
Deborah.Sarah.Bootstrap, Deborah.Sarah.Jackknife — resampling cores; Deborah.Sarah.BootstrapRunner orchestrates block bootstrap schemes.
SummaryFormatter.print_bootstrap_average_error — compute & print AVG(ERR).
SummaryFormatter.print_jackknife_average_error — from JK samples.
SummaryFormatter.print_jackknife_average_error_from_raw — from raw data + bin size.
Deborah.Sarah.SummaryCollector — gather per-module summaries into dicts for downstream printers.

Minimal Usage (REPL)

julia> using Deborah

# Append metadata to an existing TOML
julia> using Deborah.Sarah.TOMLLogger, OrderedCollections
julia> TOMLLogger.append_section_to_toml("config.toml", "metadata", OrderedCollections.OrderedDict("creator"=>"alice","notes"=>"test"))  # appends a [metadata] section

# Extract configuration IDs and X info from raw matrices
julia> using Deborah.Sarah.XYInfoGenerator
julia> conf = XYInfoGenerator.gen_conf_from_Y(Y, N_cnf, N_src, 3)  # read conf IDs from 3rd column
julia> X_info = XYInfoGenerator.gen_X_info(X, N_cnf, N_src, 1)      # take column 1 as the component

# Print bootstrap and jackknife summaries
julia> using Deborah.Sarah.SummaryFormatter
julia> SummaryFormatter.print_bootstrap_average_error(bs_arr, "Y_P2", "OBS")
julia> SummaryFormatter.print_jackknife_average_error_from_raw(raw_arr, 20, "cond", "OJK")

Notes

Column indices are $1$-based throughout (files and arrays).
Bootstrap/Jackknife routines return (mean, std); Deborah.Sarah.SummaryFormatter also prints a compact AVG(ERR) string and warns when std == 0.
TOMLLogger.append_section_to_toml writes keys using repr, preserving Julia-literal fidelity in the TOML.

See Also

source