import anndata as ad
import pandas as pd
import os
import glob

h5ad_dir = "C:/Users/Swamy/OneDrive - Zafrens/Desktop/Z-Screen Deepdive/Shared_Source_Data_June2026/external_reference/scperturb_downloads"
files = glob.glob(os.path.join(h5ad_dir, "*.h5ad"))

metadata_records = []

for f in files:
    try:
        adata = ad.read_h5ad(f, backed='r')
        dataset_name = os.path.basename(f)
        
        # extract metadata safely
        cell_lines = adata.obs['cell_line'].unique().tolist() if 'cell_line' in adata.obs else []
        perturbations = adata.obs['perturbation'].unique().tolist() if 'perturbation' in adata.obs else []
        n_cells = adata.n_obs
        n_vars = adata.n_vars
        
        metadata_records.append({
            "dataset_id": dataset_name.replace(".h5ad", ""),
            "file": dataset_name,
            "cell_lines": ", ".join([str(x) for x in cell_lines]),
            "num_perturbations": len(perturbations),
            "num_cells": n_cells,
            "num_genes": n_vars
        })
        print(f"Processed {dataset_name}")
    except Exception as e:
        print(f"Error reading {f}: {e}")

if metadata_records:
    df = pd.DataFrame(metadata_records)
    out_csv = os.path.join(h5ad_dir, "scperturb_summary_metadata.csv")
    df.to_csv(out_csv, index=False)
    print(f"\nMetadata successfully extracted to {out_csv}")
    print(df)
else:
    print("No complete metadata could be extracted yet.")