GotPsi / scripts /process_users_full.py

instNoeticSciences

claude opus 4.6

Flatten history: re-sanitized (128-bit hash) GotPsi public dataset

9deebf2 5 months ago

8.52 kB

"""
Process full Users dataset and export to Parquet.
Processes all user survey files (users14.dat and questions.dat) and exports cleaned data.
"""

import sys
import argparse
from pathlib import Path
from datetime import datetime

sys.path.insert(0, str(Path(__file__).parent.parent))

from src.core.config import Config
from src.core.output_lock import seal_output, unlock_for_write
from src.processors.users_processor import UsersProcessor


def main():
    parser = argparse.ArgumentParser(description='Process Users dataset')
    parser.add_argument('--audit', action='store_true',
                       help='Include source_file and source_row_number columns in output')
    parser.add_argument('--data-dir',
                       help='Override raw data directory (default: data/)')
    args = parser.parse_args()
    print("="*70)
    print("Users Dataset Full Processing")
    print("="*70)
    print()

    start_time = datetime.now()

    # Load config
    config = Config('config/cleaning_config.yaml')

    # Override audit mode if --audit flag is provided
    if args.audit:
        config.set('processing.audit_mode', True)
        print("Audit mode: ENABLED (including source_file and source_row_number columns)")
        print()

    # Override data directory if --data-dir flag is provided
    if args.data_dir:
        config.set('directories.raw_data', args.data_dir)


    # Create output directory
    output_dir = config.output_dir / 'parquet'
    output_dir.mkdir(parents=True, exist_ok=True)

    output_file = output_dir / 'users.parquet'

    print(f"Output file: {output_file}")
    print()

    # Create processor
    processor = UsersProcessor(config)

    # Get file count
    try:
        all_files = processor.get_file_list()
        print(f"Total files to process: {len(all_files)}")
        for f in all_files:
            file_type = processor.detect_file_type(f)
            print(f"  - {f.name} ({file_type})")
        print()
    except Exception as e:
        print(f"✗ Error getting file list: {e}")
        print()
        print("Expected directory: data/user_data/")
        print("Make sure the raw data is in the correct location.")
        print("Files expected: users14.dat and/or questions.dat")
        sys.exit(1)

    # Process all files
    print("Processing all user files...")
    print()

    try:
        df = processor.process()

        elapsed = datetime.now() - start_time

        print()
        print("="*70)
        print("✓ Processing Complete!")
        print("="*70)
        print()

        # Show stats
        stats = processor.get_stats()
        print("Processing Statistics:")
        print(f"  Files processed: {stats['files_processed']:,}")
        print(f"  Files failed: {stats['files_failed']:,}")
        print(f"  Success rate: {stats['files_processed'] / len(all_files) * 100:.1f}%")
        print()
        print(f"  users14.dat rows: {stats['users14_rows']:,}")
        print(f"  questions.dat rows: {stats['questions_rows']:,}")
        print()
        print(f"  Total rows (pre-dedup): {stats['rows_total']:,}")
        print(f"  Duplicates removed: {stats['duplicates_removed']:,}")
        print(f"  Final rows: {len(df):,}")
        print(f"  NA usernames (kept): {stats['na_usernames_count']:,}")
        print()

        # Data quality metrics
        print("Data Quality:")
        completeness = df.notna().sum() / len(df)
        print(f"  Overall completeness (avg): {completeness.mean():.1%}")
        print()

        # Column completeness
        print("Column Completeness:")
        key_cols = ['username', 'timestamp', 'email', 'state',
                    'coordinates', 'country', 'na_count_psi_and_hemi']
        for col in key_cols:
            if col in df.columns:
                complete = df[col].notna().sum()
                pct = complete / len(df) * 100
                print(f"  {col:25s}: {complete:6,} / {len(df):6,} ({pct:5.1f}%)")
        print()

        # Psi columns completeness
        psi_cols = [f'psi_{i:02d}' for i in range(1, 16)]
        existing_psi = [col for col in psi_cols if col in df.columns]
        if existing_psi:
            psi_complete = df[existing_psi].notna().sum().sum()
            psi_total = len(df) * len(existing_psi)
            print(f"Psi columns (1-15) completeness: {psi_complete:,} / {psi_total:,} ({psi_complete/psi_total*100:.1f}%)")
        print()

        # Hemi columns completeness
        hemi_cols = [f'hemi_{i:02d}' for i in range(1, 11)]
        existing_hemi = [col for col in hemi_cols if col in df.columns]
        if existing_hemi:
            hemi_complete = df[existing_hemi].notna().sum().sum()
            hemi_total = len(df) * len(existing_hemi)
            print(f"Hemi columns (1-10) completeness: {hemi_complete:,} / {hemi_total:,} ({hemi_complete/hemi_total*100:.1f}%)")
        print()

        # Date range
        if 'timestamp' in df.columns:
            valid_ts = df['timestamp'].dropna()
            if not valid_ts.empty:
                print(f"Date range: {valid_ts.min()} to {valid_ts.max()}")
                print(f"Span: {(valid_ts.max() - valid_ts.min()).days} days")
                print()

        # File type distribution
        if 'file_type' in df.columns:
            print("File Type Distribution:")
            for file_type in df['file_type'].unique():
                count = (df['file_type'] == file_type).sum()
                pct = count / len(df) * 100
                print(f"  {file_type}: {count:,} rows ({pct:.1f}%)")
            print()

        # Country distribution (top 10)
        if 'country' in df.columns:
            print("Top 10 Countries:")
            country_counts = df['country'].value_counts().head(10)
            for country, count in country_counts.items():
                pct = count / len(df) * 100
                print(f"  {country}: {count:,} ({pct:.1f}%)")
            print()

        # NA count distribution
        if 'na_count_psi_and_hemi' in df.columns:
            print("NA Count (Psi + Hemi) Distribution:")
            na_stats = df['na_count_psi_and_hemi'].describe()
            print(f"  Min: {na_stats['min']:.0f}")
            print(f"  Max: {na_stats['max']:.0f}")
            print(f"  Mean: {na_stats['mean']:.1f}")
            print(f"  Median: {na_stats['50%']:.0f}")
            print()

        # Performance stats
        print(f"Processing time: {elapsed}")
        if elapsed.total_seconds() > 0:
            print(f"Speed: {len(df) / elapsed.total_seconds():.0f} rows/second")
        print()

        # Drop demographic columns scrubbed at the .dat level (see sanitize_raw.py)
        df = df.drop(columns=['city', 'how_find'], errors='ignore')

        # Export to Parquet
        print("Exporting to Parquet...")
        unlock_for_write(output_file)
        df.to_parquet(
            output_file,
            engine='pyarrow',
            compression='snappy',
            index=False
        )
        seal_output(output_file)

        file_size_mb = output_file.stat().st_size / (1024 * 1024)
        print(f"✓ Exported to: {output_file}")
        print(f"  File size: {file_size_mb:.1f} MB")
        print()

        # Compression ratio
        raw_size_estimate = len(df) * 400  # Rough estimate of ~400 bytes/row in CSV
        compression_ratio = raw_size_estimate / output_file.stat().st_size
        print(f"  Compression ratio: {compression_ratio:.1f}x")
        print()

        # Show errata summary
        errata_summary = processor.errata_logger.get_summary()
        print("Error Summary:")
        print(f"  Total errors: {errata_summary['total_errors']:,}")
        print(f"  Files with errors: {errata_summary['files_with_errors']:,}")
        if errata_summary['error_types']:
            print(f"  Error types:")
            for error_type, count in sorted(errata_summary['error_types'].items(),
                                           key=lambda x: x[1], reverse=True)[:10]:
                print(f"    - {error_type}: {count:,}")
        print(f"  Errata log: {errata_summary['log_file']}")
        print()

        print("="*70)
        print("✓ Users Dataset Processing Complete!")
        print("="*70)

    except KeyboardInterrupt:
        print("\n\n✗ Processing interrupted by user")
        sys.exit(1)
    except Exception as e:
        print(f"\n✗ Processing failed: {e}")
        import traceback
        traceback.print_exc()
        sys.exit(1)


if __name__ == '__main__':
    main()

	"""
	Process full Users dataset and export to Parquet.

	Processes all user survey files (users14.dat and questions.dat) and exports cleaned data.
	"""

	import sys
	import argparse
	from pathlib import Path
	from datetime import datetime

	sys.path.insert(0, str(Path(__file__).parent.parent))

	from src.core.config import Config
	from src.core.output_lock import seal_output, unlock_for_write
	from src.processors.users_processor import UsersProcessor


	def main():
	parser = argparse.ArgumentParser(description='Process Users dataset')
	parser.add_argument('--audit', action='store_true',
	help='Include source_file and source_row_number columns in output')
	parser.add_argument('--data-dir',
	help='Override raw data directory (default: data/)')
	args = parser.parse_args()
	print("="*70)
	print("Users Dataset Full Processing")
	print("="*70)
	print()

	start_time = datetime.now()

	# Load config
	config = Config('config/cleaning_config.yaml')

	# Override audit mode if --audit flag is provided
	if args.audit:
	config.set('processing.audit_mode', True)
	print("Audit mode: ENABLED (including source_file and source_row_number columns)")
	print()

	# Override data directory if --data-dir flag is provided
	if args.data_dir:
	config.set('directories.raw_data', args.data_dir)


	# Create output directory
	output_dir = config.output_dir / 'parquet'
	output_dir.mkdir(parents=True, exist_ok=True)

	output_file = output_dir / 'users.parquet'

	print(f"Output file: {output_file}")
	print()

	# Create processor
	processor = UsersProcessor(config)

	# Get file count
	try:
	all_files = processor.get_file_list()
	print(f"Total files to process: {len(all_files)}")
	for f in all_files:
	file_type = processor.detect_file_type(f)
	print(f" - {f.name} ({file_type})")
	print()
	except Exception as e:
	print(f"✗ Error getting file list: {e}")
	print()
	print("Expected directory: data/user_data/")
	print("Make sure the raw data is in the correct location.")
	print("Files expected: users14.dat and/or questions.dat")
	sys.exit(1)

	# Process all files
	print("Processing all user files...")
	print()

	try:
	df = processor.process()

	elapsed = datetime.now() - start_time

	print()
	print("="*70)
	print("✓ Processing Complete!")
	print("="*70)
	print()

	# Show stats
	stats = processor.get_stats()
	print("Processing Statistics:")
	print(f" Files processed: {stats['files_processed']:,}")
	print(f" Files failed: {stats['files_failed']:,}")
	print(f" Success rate: {stats['files_processed'] / len(all_files) * 100:.1f}%")
	print()
	print(f" users14.dat rows: {stats['users14_rows']:,}")
	print(f" questions.dat rows: {stats['questions_rows']:,}")
	print()
	print(f" Total rows (pre-dedup): {stats['rows_total']:,}")
	print(f" Duplicates removed: {stats['duplicates_removed']:,}")
	print(f" Final rows: {len(df):,}")
	print(f" NA usernames (kept): {stats['na_usernames_count']:,}")
	print()

	# Data quality metrics
	print("Data Quality:")
	completeness = df.notna().sum() / len(df)
	print(f" Overall completeness (avg): {completeness.mean():.1%}")
	print()

	# Column completeness
	print("Column Completeness:")
	key_cols = ['username', 'timestamp', 'email', 'state',
	'coordinates', 'country', 'na_count_psi_and_hemi']
	for col in key_cols:
	if col in df.columns:
	complete = df[col].notna().sum()
	pct = complete / len(df) * 100
	print(f" {col:25s}: {complete:6,} / {len(df):6,} ({pct:5.1f}%)")
	print()

	# Psi columns completeness
	psi_cols = [f'psi_{i:02d}' for i in range(1, 16)]
	existing_psi = [col for col in psi_cols if col in df.columns]
	if existing_psi:
	psi_complete = df[existing_psi].notna().sum().sum()
	psi_total = len(df) * len(existing_psi)
	print(f"Psi columns (1-15) completeness: {psi_complete:,} / {psi_total:,} ({psi_complete/psi_total*100:.1f}%)")
	print()

	# Hemi columns completeness
	hemi_cols = [f'hemi_{i:02d}' for i in range(1, 11)]
	existing_hemi = [col for col in hemi_cols if col in df.columns]
	if existing_hemi:
	hemi_complete = df[existing_hemi].notna().sum().sum()
	hemi_total = len(df) * len(existing_hemi)
	print(f"Hemi columns (1-10) completeness: {hemi_complete:,} / {hemi_total:,} ({hemi_complete/hemi_total*100:.1f}%)")
	print()

	# Date range
	if 'timestamp' in df.columns:
	valid_ts = df['timestamp'].dropna()
	if not valid_ts.empty:
	print(f"Date range: {valid_ts.min()} to {valid_ts.max()}")
	print(f"Span: {(valid_ts.max() - valid_ts.min()).days} days")
	print()

	# File type distribution
	if 'file_type' in df.columns:
	print("File Type Distribution:")
	for file_type in df['file_type'].unique():
	count = (df['file_type'] == file_type).sum()
	pct = count / len(df) * 100
	print(f" {file_type}: {count:,} rows ({pct:.1f}%)")
	print()

	# Country distribution (top 10)
	if 'country' in df.columns:
	print("Top 10 Countries:")
	country_counts = df['country'].value_counts().head(10)
	for country, count in country_counts.items():
	pct = count / len(df) * 100
	print(f" {country}: {count:,} ({pct:.1f}%)")
	print()

	# NA count distribution
	if 'na_count_psi_and_hemi' in df.columns:
	print("NA Count (Psi + Hemi) Distribution:")
	na_stats = df['na_count_psi_and_hemi'].describe()
	print(f" Min: {na_stats['min']:.0f}")
	print(f" Max: {na_stats['max']:.0f}")
	print(f" Mean: {na_stats['mean']:.1f}")
	print(f" Median: {na_stats['50%']:.0f}")
	print()

	# Performance stats
	print(f"Processing time: {elapsed}")
	if elapsed.total_seconds() > 0:
	print(f"Speed: {len(df) / elapsed.total_seconds():.0f} rows/second")
	print()

	# Drop demographic columns scrubbed at the .dat level (see sanitize_raw.py)
	df = df.drop(columns=['city', 'how_find'], errors='ignore')

	# Export to Parquet
	print("Exporting to Parquet...")
	unlock_for_write(output_file)
	df.to_parquet(
	output_file,
	engine='pyarrow',
	compression='snappy',
	index=False
	)
	seal_output(output_file)

	file_size_mb = output_file.stat().st_size / (1024 * 1024)
	print(f"✓ Exported to: {output_file}")
	print(f" File size: {file_size_mb:.1f} MB")
	print()

	# Compression ratio
	raw_size_estimate = len(df) * 400 # Rough estimate of ~400 bytes/row in CSV
	compression_ratio = raw_size_estimate / output_file.stat().st_size
	print(f" Compression ratio: {compression_ratio:.1f}x")
	print()

	# Show errata summary
	errata_summary = processor.errata_logger.get_summary()
	print("Error Summary:")
	print(f" Total errors: {errata_summary['total_errors']:,}")
	print(f" Files with errors: {errata_summary['files_with_errors']:,}")
	if errata_summary['error_types']:
	print(f" Error types:")
	for error_type, count in sorted(errata_summary['error_types'].items(),
	key=lambda x: x[1], reverse=True)[:10]:
	print(f" - {error_type}: {count:,}")
	print(f" Errata log: {errata_summary['log_file']}")
	print()

	print("="*70)
	print("✓ Users Dataset Processing Complete!")
	print("="*70)

	except KeyboardInterrupt:
	print("\n\n✗ Processing interrupted by user")
	sys.exit(1)
	except Exception as e:
	print(f"\n✗ Processing failed: {e}")
	import traceback
	traceback.print_exc()
	sys.exit(1)


	if __name__ == '__main__':
	main()