Only update authorisation after all data is entered.

Pull the auth update function out of the database connection
contextmanager to prevent any problems with an auth update from
causing a rollback of all the data.

We can always manually update the authorisation later, therefore, we
do not want a failure in that causing the (potentially) time-consuming
data entry process from being rolled back.

1 files changed, 20 insertions, 9 deletions

diff --git a/scripts/load_phenotypes_to_db.py b/scripts/load_phenotypes_to_db.py
index 0b1f738..d5d2d16 100644
--- a/scripts/load_phenotypes_to_db.py
+++ b/scripts/load_phenotypes_to_db.py
@@ -1,3 +1,4 @@
+import sys
 import uuid
 import json
 import logging
@@ -224,7 +225,7 @@ def cross_reference_phenotypes_publications_and_data(
     return tuple()
 
 
-def update_auth(token, species, population, dataset, phenos):
+def update_auth(authserver, token, species, population, dataset, xrefdata):
     """Grant the user access to their data."""
     raise NotImplemented("Please implement this!")
 
@@ -285,7 +286,7 @@ def load_data(conn: mysqldb.Connection, job: dict) -> int:
     logger.info("Saved %s new phenotype data rows.", _num_data_rows)
     # 4. Cross-reference Phenotype, Publication, and PublishData in PublishXRef
     logger.info("Cross-referencing new phenotypes to their data and publications.")
-    xrefs = cross_reference_phenotypes_publications_and_data(
+    _xrefs = cross_reference_phenotypes_publications_and_data(
         conn, tuple(dataidmap.values()))
     # 5. If standard errors and N exist, save them too
     #    (use IDs returned in `3. b.` above).
@@ -306,10 +307,7 @@ def load_data(conn: mysqldb.Connection, job: dict) -> int:
                                     control_data=_control_data,
                                     filesdir=_outdir)
     logger.info("Saved %s new phenotype sample counts rows.", num_n_rows)
-    # 6. If entirely new data, update authorisations (break this down)
-    logger.info("Updating authorisation.")
-    update_auth(_token, _species, _population, _dataset, _phenos)
-    return 0
+    return (_species, _population, _dataset, _xrefs)
 
 
 if __name__ == "__main__":
@@ -344,6 +342,8 @@ if __name__ == "__main__":
         with (mysqldb.database_connection(args.db_uri) as conn,
               conn.cursor(cursorclass=DictCursor) as cursor,
               sqlite3.connection(args.jobs_db_path) as jobs_conn):
+            job = jobs.job(jobs_conn, args.job_id)
+
             # Lock the PublishXRef/PublishData/PublishSE/NStrain here: Why?
             #     The `DataId` values are sequential, but not auto-increment
             #     Can't convert `PublishXRef`.`DataId` to AUTO_INCREMENT.
@@ -371,12 +371,23 @@ if __name__ == "__main__":
             cursor.execute(# Lock the tables to avoid race conditions
                 "LOCK TABLES " + ", ".join(
                     f"{_table} WRITE" for _table in _db_tables_))
-            return load_data(conn, jobs.job(jobs_conn, args.job_id))
+
+            db_results = load_data(conn, job)
 
             logger.debug("Unlocking all database tables.")
             cursor.execute("UNLOCK TABLES")
 
-        return 1
+        # Update authorisations (break this down) — maybe loop until it works?
+        logger.info("Updating authorisation.")
+        _job_metadata = job["metadata"]
+        return update_auth(_job_metadata["authserver"],
+                           _job_metadata["token"],
+                           *db_results)
 
 
-    main()
+    try:
+        sys.exit(main())
+    except:
+        logger.debug("Data loading failed… Halting!",
+                     exc_info=True)
+        sys.exit(1)