raw-philippine-data / scripts /load_documents_to_db.py
napppy's picture
feat: add extracted document text
c1cf2fd
#!/usr/bin/env python3
"""
Load document data from text files into a DuckDB database.
This script scans the data/document directory for text files and loads them
into a local DuckDB database. Documents are organized by type (sb/hb), congress,
and document number.
Directory structure: data/document/{type}/{congress}/{range}/{TYPE}-{number}.txt
Example: data/document/sb/20/00001-01000/SB-00002.txt
- type: sb (senate bill) or hb (house bill)
- congress: 20 (20th congress)
- document_number: 2
"""
import sys
import argparse
from pathlib import Path
import duckdb
import logging
from datetime import datetime
import re
# Add parent directory to path to import schemas and config
sys.path.insert(0, str(Path(__file__).parent.parent))
from schemas.document import DOCUMENT_SCHEMA
from config import DATABASE_PATH, DOCUMENT_DATA_DIR
def parse_document_path(file_path: Path) -> dict:
"""
Parse document file path to extract metadata.
Args:
file_path: Path to document file
Returns:
Dict with document_type, congress, document_number
Example:
data/document/sb/20/00001-01000/SB-00002.txt ->
{
'document_type': 'sb',
'congress': 20,
'document_number': 2
}
"""
# Get parts from path
parts = file_path.parts
# Find the document type directory (sb, hb, etc.)
try:
# Path structure: data/document/{type}/{congress}/{range}/{TYPE}-{number}.txt
doc_index = parts.index('document')
document_type = parts[doc_index + 1].lower()
congress = int(parts[doc_index + 2])
# Parse filename: SB-00002.txt -> document_number = 2
filename = file_path.stem # Remove .txt extension
match = re.match(r'^[A-Z]+-(\d+)$', filename)
if match:
document_number = int(match.group(1))
else:
raise ValueError(f"Cannot parse document number from filename: {filename}")
return {
'document_type': document_type,
'congress': congress,
'document_number': document_number
}
except (ValueError, IndexError) as e:
raise ValueError(f"Cannot parse document path: {file_path}") from e
def get_document_files(data_dir: Path):
"""Recursively find all .txt files in the document data directory (generator)."""
return data_dir.glob('**/*.txt')
def create_documents_table(conn: duckdb.DuckDBPyConnection):
"""Create the documents table using the explicit schema."""
create_sql = DOCUMENT_SCHEMA.get_create_table_sql()
print(f"Creating table '{DOCUMENT_SCHEMA.table_name}' with {len(DOCUMENT_SCHEMA.schema)} columns:")
print(f" Fields: {', '.join(DOCUMENT_SCHEMA.field_order)}")
conn.execute(create_sql)
def load_documents_to_db(
data_dir: Path,
db_path: Path,
export_parquet: bool = False,
parquet_path: Path = None,
batch_size: int = 1000,
progress_interval: int = 100
):
"""Load all document text files into the DuckDB database with batch inserts."""
# Setup error logging
error_log_path = db_path.parent / 'load_documents_errors.log'
logging.basicConfig(
filename=str(error_log_path),
level=logging.ERROR,
format='%(asctime)s - %(message)s',
filemode='w' # Overwrite previous log
)
print(f"Connecting to database: {db_path}")
print(f"Error log: {error_log_path}")
print(f"Batch size: {batch_size}")
conn = duckdb.connect(str(db_path))
# Create table with explicit schema
create_documents_table(conn)
# Build INSERT statement using schema definition
insert_sql = DOCUMENT_SCHEMA.get_insert_sql()
# Get all text files
print("\nScanning for document files...")
doc_files = list(get_document_files(data_dir))
total_files = len(doc_files)
print(f"Found {total_files} files to process")
# Load data with batch inserts
print("\nLoading document data...")
loaded_count = 0
error_count = 0
processed_count = 0
# Batch for accumulating records
document_batch = []
# Track timing for ETA
start_time = datetime.now()
def flush_batch():
"""Helper to insert accumulated batch and commit."""
nonlocal loaded_count
if not document_batch:
return
batch_size_to_commit = len(document_batch)
print(f" Committing batch of {batch_size_to_commit} records...", end='', flush=True)
try:
conn.execute("BEGIN TRANSACTION")
conn.executemany(insert_sql, document_batch)
conn.execute("COMMIT")
loaded_count += len(document_batch)
print(f" done!")
except Exception as e:
conn.execute("ROLLBACK")
logging.error(f"Batch insert failed: {e}")
print(f"\n Warning: Batch insert failed, see error log")
document_batch.clear()
# Process files
try:
for doc_file in doc_files:
try:
# Parse metadata from path
metadata = parse_document_path(doc_file)
# Read file content
with open(doc_file, 'r', encoding='utf-8') as f:
content = f.read()
# Create document ID: {type}-{congress}-{number}
doc_id = f"{metadata['document_type']}-{metadata['congress']}-{metadata['document_number']}"
# Build values list
values = [
doc_id,
metadata['document_type'],
metadata['congress'],
metadata['document_number'],
str(doc_file),
content
]
document_batch.append(values)
except Exception as e:
# File processing failed
error_count += 1
logging.error(f"{doc_file}: {e}")
processed_count += 1
# Flush batch when reaching batch size
if len(document_batch) >= batch_size:
flush_batch()
# Progress indicator
if processed_count % progress_interval == 0 or processed_count == total_files:
pct = (processed_count / total_files * 100) if total_files > 0 else 0
pending = len(document_batch)
total_loaded = loaded_count + pending
print(f" [{pct:5.1f}%] {processed_count}/{total_files} files | "
f"{total_loaded} loaded ({loaded_count} committed, {pending} pending) | "
f"{error_count} errors")
except KeyboardInterrupt:
print("\n\n*** Interrupted by user (Ctrl+C) ***")
print("Flushing any pending records to database...")
flush_batch()
# Calculate time up to interruption
end_time = datetime.now()
total_seconds = (end_time - start_time).total_seconds()
total_minutes = total_seconds / 60
print(f"\nPartial load completed:")
print(f" Files processed: {processed_count}/{total_files}")
print(f" Records loaded: {loaded_count}")
print(f" Errors: {error_count}")
print(f" Time elapsed: {total_minutes:.1f} minutes ({total_seconds:.0f} seconds)")
conn.close()
sys.exit(1)
# Flush any remaining records
flush_batch()
# Calculate total time
end_time = datetime.now()
total_seconds = (end_time - start_time).total_seconds()
total_minutes = total_seconds / 60
# Show summary
print(f"\n{'='*60}")
print(f"Load complete!")
print(f" Total files processed: {processed_count}")
print(f" Successfully loaded: {loaded_count}")
print(f" Errors: {error_count}")
print(f" Total time: {total_minutes:.1f} minutes ({total_seconds:.0f} seconds)")
if error_count > 0:
print(f" Error details logged to: {error_log_path}")
# Show database stats
documents_count = conn.execute("SELECT COUNT(*) as total FROM documents").fetchone()
print(f" Total documents in database: {documents_count[0]}")
print(f"{'='*60}")
# Show sample data
print("\nSample documents (first 5 rows):")
sample = conn.execute("""
SELECT id, document_type, congress, document_number, LENGTH(content) as content_length
FROM documents
ORDER BY document_type, congress, document_number
LIMIT 5
""").fetchall()
for row in sample:
print(f" {row[0]}: {row[1].upper()}-{row[3]} (Congress {row[2]}, {row[4]} chars)")
# Show statistics by type
print("\nDocument statistics by type:")
stats = conn.execute("""
SELECT
document_type,
COUNT(*) as count,
MIN(congress) as min_congress,
MAX(congress) as max_congress,
MIN(document_number) as min_doc_num,
MAX(document_number) as max_doc_num
FROM documents
GROUP BY document_type
ORDER BY document_type
""").fetchall()
for row in stats:
print(f" {row[0].upper()}: {row[1]} documents (Congress {row[2]}-{row[3]}, "
f"Doc# {row[4]}-{row[5]})")
# Export to Parquet if requested
if export_parquet:
print(f"\n{'='*60}")
print("Exporting to Parquet format...")
print(f" Output: {parquet_path}")
try:
conn.execute(f"COPY documents TO '{parquet_path}' (FORMAT PARQUET)")
if parquet_path.exists():
file_size = parquet_path.stat().st_size
file_size_mb = file_size / (1024 * 1024)
print(f" Successfully exported documents.parquet ({file_size_mb:.2f} MB)")
else:
print(" Warning: Export completed but file not found")
except Exception as e:
print(f" Error exporting: {e}")
print(f"{'='*60}")
conn.close()
print(f"\nDatabase saved to: {db_path}")
def main():
"""Main entry point."""
parser = argparse.ArgumentParser(
description='Load document data from text files into a DuckDB database',
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog="""
Examples:
# Use default database path (databases/data.duckdb)
python scripts/load_documents_to_db.py
# Specify custom database path
python scripts/load_documents_to_db.py --db-path /path/to/custom.duckdb
# Use a different data directory
python scripts/load_documents_to_db.py --data-dir /path/to/document/data
# Export to Parquet for Hugging Face dataset viewer
python scripts/load_documents_to_db.py --export-parquet
"""
)
parser.add_argument(
'--db-path',
type=Path,
default=DATABASE_PATH,
help=f'Path to the DuckDB database (default: {DATABASE_PATH})'
)
parser.add_argument(
'--data-dir',
type=Path,
default=DOCUMENT_DATA_DIR,
help=f'Path to the document data directory (default: {DOCUMENT_DATA_DIR})'
)
parser.add_argument(
'--export-parquet',
action='store_true',
help='Export the documents table to Parquet format after loading'
)
parser.add_argument(
'--parquet-path',
type=Path,
default=Path(__file__).parent.parent / 'databases' / 'documents.parquet',
help='Path for the exported Parquet file (default: databases/documents.parquet)'
)
parser.add_argument(
'--batch-size',
type=int,
default=1000,
help='Number of records to insert per batch/transaction (default: 1000)'
)
parser.add_argument(
'--progress-interval',
type=int,
default=100,
help='Show progress every N files (default: 100)'
)
args = parser.parse_args()
# Validate data directory exists
if not args.data_dir.exists():
print(f"Error: Data directory not found: {args.data_dir}")
sys.exit(1)
# Create databases directory if it doesn't exist
args.db_path.parent.mkdir(parents=True, exist_ok=True)
load_documents_to_db(
args.data_dir,
args.db_path,
args.export_parquet,
args.parquet_path,
args.batch_size,
args.progress_interval
)
if __name__ == '__main__':
main()