#!/bin/bash
# Batch-extract every document under a directory through process_doc.py,
# mirroring the input folder structure into the output base. The doocus twin of
# ctrl/batch.sh (which does the same for meeting videos via process_meeting.py).
#
# Recursive and space-safe (paths come off a Windows mount, so they often have
# spaces). Each document's run folder is auto-created by process_doc.py inside
# the mirrored output subfolder.
#
# Usage:
#   ctrl/batch_docs.sh -i <input-dir> -o <output-dir> [-e "pdf docx ..."] [-n] \
#                      [-- <process_doc.py flags>]
#
# Examples:
#   # Everything under a downloaded Drive folder, mirrored into docs-output
#   ctrl/batch_docs.sh -i "/mnt/win/drive" -o docs-output
#
#   # Dry run: just show which docs map to which output folders
#   ctrl/batch_docs.sh -i "/mnt/win/drive" -o docs-output -n
#
#   # Only pdfs/docx, render page thumbnails
#   ctrl/batch_docs.sh -i "./download" -o docs-output -e "pdf docx" -- --render
#
# A doc at  <input>/2026/team a/notes.docx  produces
#   <output>/2026/team a/<YYYYMMDD-NNN-notes>/...
# i.e. the run folder lands inside the mirrored subtree.
set -euo pipefail

PROJECT_DIR="$(cd "$(dirname "$0")/.." && pwd)"
cd "$PROJECT_DIR"

# python: honor $PYTHON, else prefer python3
PYTHON="${PYTHON:-}"
if [ -z "$PYTHON" ]; then
  if command -v python3 >/dev/null 2>&1; then PYTHON=python3; else PYTHON=python; fi
fi

usage() { sed -n '2,26p' "$0"; }

INPUT=""
OUTPUT=""
EXTS="csv docx html htm jpg jpeg json md txt pdf png pptx xlsx yaml yml mp4"
DRY=false
FORWARD=()

while [[ $# -gt 0 ]]; do
  case "$1" in
    -i|--input)   INPUT="$2"; shift 2 ;;
    -o|--output)  OUTPUT="$2"; shift 2 ;;
    -e|--ext)     EXTS="$2"; shift 2 ;;
    -n|--dry-run) DRY=true; shift ;;
    -h|--help)    usage; exit 0 ;;
    --)           shift; FORWARD=("$@"); break ;;
    *) echo "Unknown arg: $1" >&2; usage; exit 1 ;;
  esac
done

[ -n "$INPUT" ]  || { echo "ERROR: -i/--input is required" >&2; exit 1; }
[ -n "$OUTPUT" ] || { echo "ERROR: -o/--output is required" >&2; exit 1; }
[ -d "$INPUT" ]  || { echo "ERROR: input dir not found: $INPUT" >&2; exit 1; }

# Absolute paths so relative-path math is stable regardless of where we cd'd.
INPUT="$(realpath "$INPUT")"
mkdir -p "$OUTPUT"
OUTPUT="$(realpath "$OUTPUT")"

# Build the find extension filter: \( -iname '*.pdf' -o -iname '*.docx' ... \)
find_expr=()
for ext in $EXTS; do
  find_expr+=( -iname "*.${ext}" -o )
done
unset 'find_expr[${#find_expr[@]}-1]'   # drop the trailing -o

echo "Input : $INPUT"
echo "Output: $OUTPUT"
echo "Exts  : $EXTS"
[ "$DRY" = true ] && echo "(dry run — nothing will be processed)"
echo

total=0 ok=0 fail=0
# Process substitution (not a pipe) so counters survive into the summary.
while IFS= read -r -d '' doc; do
  total=$((total + 1))

  rel="${doc#"$INPUT"/}"            # path relative to the input root
  reldir="$(dirname "$rel")"
  if [ "$reldir" = "." ]; then
    outdir="$OUTPUT"                 # doc sat directly in the input root
  else
    outdir="$OUTPUT/$reldir"         # mirror the subfolder structure
  fi

  echo "[$total] $rel"
  echo "      -> $outdir"

  if [ "$DRY" = true ]; then
    continue
  fi

  mkdir -p "$outdir"
  # stdin from /dev/null so any child process can't eat the file list and stall
  # the batch (same guard as ctrl/batch.sh).
  if "$PYTHON" "$PROJECT_DIR/process_doc.py" "$doc" \
        --output-dir "$outdir" "${FORWARD[@]+"${FORWARD[@]}"}" </dev/null; then
    ok=$((ok + 1))
  else
    echo "      !! FAILED (continuing)" >&2
    fail=$((fail + 1))
  fi
  echo
done < <(find "$INPUT" -type f \( "${find_expr[@]}" \) -print0 | sort -z)

echo "----------------------------------------"
if [ "$DRY" = true ]; then
  echo "Found $total document(s)."
else
  echo "Done. $total document(s): $ok ok, $fail failed."
fi
[ "$fail" -eq 0 ]