Outputs

Summary

PyFLASH output files fall into a few predictable families: figures, pipeline run folders, Excel workbooks, normality diagnostics, model-sweep artifacts, structured report records, and pickles. This section explains which functions create each family, where files are saved, what is inside them, and how to reuse them without re-running the full analysis.

Start with Where Results Go if you only need the basic folder map. Use these pages when you need output-specific file names, table meanings, or loading examples.

Created By

Output files are created by these main routes:

Route	Common outputs
Plot functions with save=True	SVG figures under batch.fig_path, plus plot-specific CSV or PNG diagnostics when the plot computes statistics.
Pipeline functions such as correlation, data_overview, and linear_model	Run folders with figures, CSV tables, manifest.json, _runs_index.csv, and often ! Overview Montage.png.
Batch.export_all_excel() and the Batch.export_*_excel() methods	.xlsx workbooks and matching *_RegexFilters.txt reports under batch.export_path or a supplied export folder.
iterative_model_sweep	Classifier sweep CSVs, PNG summaries, manifest.json, and a run README.
The PyFLASH runner around PyFLASH.report	.runtime/results_store/<run_id>.results.json, .results.md, and index.jsonl.
save_state, load_state, normalize_paths, and create_batch(..., pickle_path=...)	.pkl saved-state files and path-rebased objects.

Folder Layout

The default batch output roots are created lazily. A path can be recorded on the object before the corresponding folder exists on disk.

<batch-output>/
  Exports/
  Results/
    Python Figures/
    Data and Stats/
    Separate CSVs/
    Representative Images/
    Legends/

Current pipeline outputs are centered under Results/Python Figures, not Results/Data and Stats:

<batch-output>/Results/Python Figures/<Pipeline Name>/<run_label>/
  manifest.json
  *.csv
  *.svg
  ! Overview Montage.png

File Contents

Page	File family
Figure folders	Saved SVG plots, PNG montages, optional PNG diagnostics, and plot-side CSV statistics.
Pipeline run folders	Pipeline tables, figure subfolders, manifests, run indexes, and reuse policies.
Excel workbooks	IF summary, IF extended, behavior, and extra-summary workbooks plus regex audit reports.
Normality outputs	Plot-level Q-Q PNGs and data_overview normality tables/summary figures.
Model sweep outputs	Classifier sweep score tables, prediction tables, permutation tests, plots, README, and manifest.
Report records	Structured JSON result records, deterministic Markdown digests, index ledger, and lab notebook entries.
Pickle files	Saved Batch or Experiment state, legacy migration, path normalization, and cache reuse.

How To Reuse

Inspect a run folder with standard Python tools:

from pathlib import Path
import pandas as pd

run_dir = Path("analysis-output/Results/Python Figures/Correlation Pipeline/scn_run")
manifest = pd.read_json(run_dir / "manifest.json", typ="series").to_dict()
selected = pd.read_csv(run_dir / "selected_pairs.csv")

print(manifest["run_label"])
print(selected.head())

Reopen a saved PyFLASH object when you need to make more plots from the same processed state:

from PyFLASH import load_state

batch = load_state("analysis-output/pickles/scn_batch.pkl")
print(batch.fig_path)

Notes

• save=False means plot and pipeline functions should return Python results without writing new output files.
• Config.SAVE_MODE=False disables figure writes at the shared save_fig(...) layer.
• Pipeline if_exists policies control existing run folders: overwrite, version, error, and skip.
• Do not use local analysis folders as reusable documentation examples. Use placeholder paths and replace them in your own scripts.