#!/usr/bin/env bash
set -eau -o pipefail

NEEDED_COLUMNS=("IdentifiantImmeuble" "EtatImmeuble" "CoordonneeImmeubleX" "CoordonneeImmeubleY" "NumeroVoieImmeuble" "TypeVoieImmeuble" "NomVoieImmeuble" "CodePostalImmeuble" "CommuneImmeuble" "DateDebutAcceptationCmdAcces" "DateMiseEnServiceCommercialeImmeuble")

if [ "$#" -ne 2 ]; then
    echo "Usage: ingest path-to-directory-containing-IPE-CSVs path-to-generated-db"
    echo ""
    exit 1
fi
fullIpeDirPath=$(realpath "${1}")
fullDbPath=$(realpath "${2}")

tmpSql=$(mktemp)
clean_tmp () {
    rm "${tmpSql}"
}
trap clean_tmp EXIT

echo "[+] Ingesting IPE data."
echo "    The following files will be ingested."
echo ""
ipeFiles=$(find "${fullIpeDirPath}" -name '*.csv')
cat > "${tmpSql}" <<EOF
.separator ";"
EOF

firstFile=true
for ipeFile in ${ipeFiles}; do
  echo "    ${ipeFile}"
  head -n1 $ipeFile | grep -q IdentifiantImmeuble && header=true || header=false
  import_opt=""
  if $firstFile || $header; then
    if ! $header; then
      echo "ERROR: first file ${ipeFile} does not have a good header"
      exit 1
    fi
    if ! $firstFile; then
      import_opt="-skip 1"
    else
      header=$(head -n1 $ipeFile)
      OLD_IFS=$IFS
      export IFS=";"
      idx=1
      idx_to_keep=()
      for column in $header; do
        export IFS=$OLD_IFS
        if [[ " ${NEEDED_COLUMNS[*]} " =~ " ${column} " ]]; then
          idx_to_keep+=("$idx")
        fi
        idx=$((idx+1))
        export IFS=";"
      done
      export IFS=$OLD_IFS
      cut_idx_to_keep=$(IFS=',';echo "${idx_to_keep[*]}";IFS=$' \t\n')
      echo "  Column indexes that will be kept in csv files: $cut_idx_to_keep (matching columns ${NEEDED_COLUMNS[*]})"
    fi
    firstFile=false
  fi

  useIpeFile=${ipeFile}
  if [[ "$cut_idx_to_keep" != "" ]]; then
    cut -d';' -f$cut_idx_to_keep $ipeFile > ${ipeFile}.cut
    useIpeFile=${ipeFile}.cut
  fi
  cat >> "${tmpSql}" <<EOF
.import ${import_opt} ${useIpeFile} ipe
EOF
done

echo ""

rc=0
sqlite3 "${fullDbPath}" < "${tmpSql}" 2>&1 | grep -v "filling the rest with NULL" || rc="${PIPESTATUS[0]}"
if [[ $rc -ne 0 ]]; then
  echo "Error executing sqlite import"
  exit "$rc"
fi

echo "[+] Create separate table with id immeuble index and its state."
cat > "${tmpSql}" <<EOF
CREATE TABLE refimm (IdentifiantImmeuble text NOT NULL, EtatImmeuble text NOT NULL, DateDebutAcceptationCmdAcces text NOT NULL);
CREATE UNIQUE INDEX idx_IdentifiantImmeuble on refimm (IdentifiantImmeuble);
INSERT OR REPLACE INTO refimm SELECT IdentifiantImmeuble,EtatImmeuble, DateDebutAcceptationCmdAcces FROM ipe;
EOF
sqlite3 "${fullDbPath}" < "${tmpSql}"

echo "[+] Ingesting spatial data."

cat > "${tmpSql}" <<EOF
SELECT load_extension("mod_spatialite");
SELECT InitSpatialMetaData();

-- Despite the SELECT, we're actually creating a new ImmeubleGeoPoint
-- column here. The spatialite API is a bit weird...
SELECT AddGeometryColumn('ipe','ImmeubleGeoPoint',4326,'POINT');

-- The geodata is expressed in RGF93 (SRID 2154). We need to project
-- it to WSG84 to display it on a OSM map.
UPDATE ipe SET ImmeubleGeoPoint =
    Transform(
        MakePoint(
            CAST(CoordonneeImmeubleX as DOUBLE),
            CAST(CoordonneeImmeubleY as DOUBLE),
            2154)
        ,4326);
EOF
sqlite3 "${fullDbPath}" < "${tmpSql}"

echo "[+] Creating Rtree index (spatial index). We're almost done."
cat > "${tmpSql}" <<EOF
SELECT load_extension("mod_spatialite");
SELECT CreateSpatialIndex('ipe','ImmeubleGeoPoint');
EOF
sqlite3 "${fullDbPath}" < "${tmpSql}"

echo "[+] SQLite database generated at ${fullDbPath}"