FCCS/fccs_reorganize.py

"""
Step 3: Reorganize flat exported files into a proper folder structure.

Parses FCCS export filenames using known folder templates and date anchors
to reconstruct the original drawer/folder hierarchy.

Filename format:
  {drawer_id}_{client_name}_{folder_name}_{creation_date MM-DD-YYYY}_{document_name}.ext

Examples:
  01069_ABRAHAM, REBEKAH L._2025 Tax Documents_03-03-2026_030126 E-mail re Tax Info.pdf
  01069_ABRAHAM, REBEKAH L._Permanent File_02-24-2018_Driver's License.pdf
  01069_ABRAHAM, REBEKAH L._UltraTax CS 12-31-2008_02-12-2009_2008 Form 1040 Filing Instructions.doc

Output structure:
  output/{drawer_id}_{client_name}/{folder_name}/{original_filename}

Files that cannot be parsed go to output/_unparsed/ for manual review.
"""

import os
import re
import shutil
import sys

from fccs_config import (
    parse_args, load_config, make_logger,
    load_folder_list, build_folder_patterns,
)

# Regex for the creation date field (MM-DD-YYYY) bounded by underscores
_DATE_RE = re.compile(r"_(\d{2}-\d{2}-\d{4})_")


def parse_filename(filename, folder_patterns):
    """
    Parse an FCCS export filename into (drawer_id, client_name, folder_name,
    date, doc_name_with_ext) or return None if it cannot be parsed.

    folder_patterns is a list of (compiled_regex, template_name) from
    build_folder_patterns(), sorted longest-first.

    Strategy:
      1. Drawer ID: first token before first underscore
      2. Creation date: find MM-DD-YYYY pattern as anchor
      3. Folder name: regex-match a known folder template at the end of
         the text between client_name and creation_date
      4. Client name: whatever is left between drawer_id and folder_name
    """
    stem, ext = os.path.splitext(filename)

    # 1. Drawer ID — first underscore-delimited token
    sep = stem.find("_")
    if sep == -1:
        return None
    drawer_id = stem[:sep]
    rest = stem[sep + 1:]

    # 2. Find all creation date candidates (MM-DD-YYYY bounded by underscores)
    # Prepend underscore so a date right at the start of `rest` is also found
    search_str = "_" + rest
    date_matches = list(_DATE_RE.finditer(search_str))
    if not date_matches:
        return None

    # 3. Try each date match; for each, try to match a known folder template
    for m in date_matches:
        date_str = m.group(1)
        # Calculate positions relative to `rest` (adjust for prepended _)
        # m.start() is the position of the leading _ in search_str
        # In `rest`, the content before the date ends at (m.start() - 1)
        before_date = rest[: m.start() - 1]
        after_date_pos = m.start() - 1 + len("_") + len(date_str) + len("_")
        doc_name = rest[after_date_pos:]

        # Try each folder pattern against the end of before_date
        for pattern, template_name in folder_patterns:
            match = pattern.search(before_date)
            if match and match.end() == len(before_date):
                # Folder matched at the end — check for underscore separator before it
                folder_start = match.start()
                if folder_start > 0 and before_date[folder_start - 1] == "_":
                    client_name = before_date[: folder_start - 1]
                    folder_name = match.group()  # the actual expanded name
                    return (drawer_id, client_name, folder_name, date_str, doc_name + ext)

    return None


def main():
    args = parse_args()
    cfg = load_config(args.config)
    log = make_logger(cfg.get("paths", "log_file"))

    export_dir = cfg.get("paths", "export_dir")
    output_dir = cfg.get("paths", "output_dir")
    folder_list_path = cfg.get("paths", "folder_list")

    if not os.path.isdir(export_dir):
        log(f"ERROR: export directory not found: {export_dir}")
        sys.exit(1)

    folder_templates = load_folder_list(folder_list_path)
    if not folder_templates:
        log("WARNING: folder list is empty. Only built-in patterns (UltraTax CS) will match.")
    folder_patterns = build_folder_patterns(folder_templates)
    log(f"Loaded {len(folder_patterns)} folder patterns")

    files = [
        f for f in os.listdir(export_dir)
        if os.path.isfile(os.path.join(export_dir, f))
    ]

    log(f"Found {len(files)} files to reorganize")

    success = 0
    failed = 0

    for filename in files:
        result = parse_filename(filename, folder_patterns)

        if result is None:
            log(f"  UNPARSED: {filename}")
            dest_dir = os.path.join(output_dir, "_unparsed")
            failed += 1
        else:
            drawer_id, client_name, folder_name, date, doc_name = result
            dest_dir = os.path.join(
                output_dir,
                f"{drawer_id}_{client_name}",
                folder_name,
            )
            success += 1

        os.makedirs(dest_dir, exist_ok=True)
        src = os.path.join(export_dir, filename)
        dst = os.path.join(dest_dir, filename)

        # Handle duplicate filenames
        if os.path.exists(dst):
            base, fext = os.path.splitext(filename)
            counter = 1
            while os.path.exists(dst):
                dst = os.path.join(dest_dir, f"{base}_{counter}{fext}")
                counter += 1
            log(f"  DUPLICATE renamed: {filename} -> {os.path.basename(dst)}")

        shutil.copy2(src, dst)

    log("=" * 60)
    log(f"Reorganization complete: {success} organized, {failed} unparsed")
    if failed:
        log(f"Review unparsed files in: {os.path.join(output_dir, '_unparsed')}")
    log("=" * 60)


if __name__ == "__main__":
    main()