ENH: add metadata necessary to download from previous idc-versions

CMakeLists.txt

@@ -19,6 +19,7 @@ add_custom_command(
   OUTPUT
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_CSV_ARCHIVE}>:${download_dir}/idc_index.csv.zip>
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_PARQUET}>:${download_dir}/idc_index.parquet>
+    $<$<BOOL:${IDC_INDEX_DATA_GENERATE_PARQUET}>:${download_dir}/prior_versions_index.parquet>
   COMMAND python ${CMAKE_CURRENT_SOURCE_DIR}/scripts/python/idc_index_data_manager.py
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_CSV_ARCHIVE}>:--generate-csv-archive>
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_PARQUET}>:--generate-parquet>
@@ -28,10 +29,12 @@ add_custom_target(run_idc_index_data_manager ALL
   DEPENDS
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_CSV_ARCHIVE}>:${download_dir}/idc_index.csv.zip>
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_PARQUET}>:${download_dir}/idc_index.parquet>
+    $<$<BOOL:${IDC_INDEX_DATA_GENERATE_PARQUET}>:${download_dir}/prior_versions_index.parquet>
 )
 
 install(
   FILES
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_CSV_ARCHIVE}>:${download_dir}/idc_index.csv.zip>
     $<$<BOOL:${IDC_INDEX_DATA_GENERATE_PARQUET}>:${download_dir}/idc_index.parquet>
+    $<$<BOOL:${IDC_INDEX_DATA_GENERATE_PARQUET}>:${download_dir}/prior_versions_index.parquet>
   DESTINATION "idc_index_data")

scripts/sql/prior_versions_index.sql

@@ -0,0 +1,89 @@
+-- For details on the syntax, see
+-- https://cloud.google.com/bigquery/docs/reference/standard-sql/procedural-language
+--
+-- Step 1: Declare variables
+DECLARE idc_versions ARRAY<INT64>;
+DECLARE latest_idc_version INT64 DEFAULT 18;
+DECLARE union_all_query STRING;
+
+--Step 2
+--SET latest_idc_version = (
+--SELECT max(idc_version)
+--FROM
+--bigquery-public-data.idc_current.version_metadata
+--);
+
+-- Step 3: Get all idc_versions
+SET idc_versions = (
+  SELECT GENERATE_ARRAY(1, latest_idc_version)
+  -- SELECT [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18]
+  --SELECT ARRAY_AGG(idc_version)
+  --FROM
+  --`bigquery-public-data.idc_current.version_metadata`
+);
+
+-- Step 4: Generate the UNION ALL query dynamically
+SET union_all_query = (
+  SELECT STRING_AGG(
+    FORMAT("""
+    SELECT
+    %d AS idc_version,
+    collection_id,
+    PatientID,
+    SeriesInstanceUID,
+    StudyInstanceUID,
+    Modality,
+    regexp_extract(gcs_url, 'gs://([^/]+)/') as gcs_bucket,
+    crdc_series_uuid,
+    ROUND(SUM(SAFE_CAST(instance_size AS float64))/1000000, 2) AS series_size_MB,
+  FROM
+  `bigquery-public-data.idc_v%d.dicom_all` AS dicom_all
+  where crdc_series_uuid not in (select distinct crdc_series_uuid from `bigquery-public-data.idc_v%d.dicom_all`)
+  GROUP BY
+  1,2,3,4,5,6,7,8
+
+  """,
+  version, version, latest_idc_version),
+    " UNION ALL "
+  )
+  FROM UNNEST(idc_versions) AS version
+);
+
+-- Step 5: Execute the complete query
+EXECUTE IMMEDIATE FORMAT("""
+WITH all_versions AS (
+  %s
+)
+SELECT
+  collection_id,
+  PatientID,
+  SeriesInstanceUID,
+  StudyInstanceUID,
+  Modality,
+  gcs_bucket,
+  crdc_series_uuid,
+  series_size_MB,
+  CASE
+
+  # map GCS bucket to AWS bucket, since for idc-index we prefer AWS
+  # if new buckets are included in IDC, this will need to be updated!
+
+  WHEN gcs_bucket='public-datasets-idc' THEN CONCAT('s3://','idc-open-data/',crdc_series_uuid, '/*')
+  WHEN gcs_bucket='idc-open-idc1' THEN CONCAT('s3://','idc-open-data-two/',crdc_series_uuid, '/*')
+  WHEN gcs_bucket='idc-open-cr' THEN CONCAT('s3://','idc-open-data-cr/',crdc_series_uuid, '/*')
+    END AS series_aws_url,
+  MIN(idc_version) AS min_idc_version,
+  MAX(idc_version) AS max_idc_version
+FROM all_versions
+
+where gcs_bucket not in ('idc-open-idc')
+
+#per @bcli4d:idc-open-idc was our public bucket before we moved most data to the Google owned public-datasets-idc.
+#We decided at the time to not touch BQ. To deal with this and other cases where some metadata can change (Licences),
+#we include the mutable_metadata table which maps crdc_instance_uuid to current gcs_url, aws_url, license, doi.
+
+GROUP BY
+ 1,2,3,4,5,6,7,8
+  """,
+  union_all_query
+);

src/idc_index_data/__init__.py

@@ -15,6 +15,7 @@
     "__version__",
     "IDC_INDEX_CSV_ARCHIVE_FILEPATH",
     "IDC_INDEX_PARQUET_FILEPATH",
+    "PRIOR_VERSIONS_INDEX_PARQUET_FILEPATH",
 ]
 
 
@@ -36,3 +37,6 @@ def _lookup(path: str, optional: bool = False) -> Path | None:
     "idc_index_data/idc_index.csv.zip", optional=True
 )
 IDC_INDEX_PARQUET_FILEPATH: Path | None = _lookup("idc_index_data/idc_index.parquet")
+PRIOR_VERSIONS_INDEX_PARQUET_FILEPATH: Path | None = _lookup(
+    "idc_index_data/prior_versions_index.parquet"
+)

tests/test_package.py

@@ -38,3 +38,8 @@ def test_reading_index():
         assert m.IDC_INDEX_PARQUET_FILEPATH.is_file()
         df_parquet = pd.read_parquet(m.IDC_INDEX_PARQUET_FILEPATH)
         assert not df_parquet.empty
+
+    if m.PRIOR_VERSIONS_INDEX_PARQUET_FILEPATH is not None:
+        assert m.PRIOR_VERSIONS_INDEX_PARQUET_FILEPATH.is_file()
+        df_parquet = pd.read_parquet(m.PRIOR_VERSIONS_INDEX_PARQUET_FILEPATH)
+        assert not df_parquet.empty