add checks of the format of the paths in the database

corrlib/integrity.py

@@ -7,7 +7,7 @@ from .tracker import get
 import pyerrors.input.json as pj
 import os
 from configparser import ConfigParser
-
+from .find import list_ensembles, list_projects
 from typing import Any
 
 
@@ -64,6 +64,31 @@ def are_keys_unique(db: Path, table: str, col: str) -> bool:
     return bool(results[0] == results[1])
 
 
+def check_path_format(result: pd.Series, ensembles: list[str], projects: list[str]) -> None:
+    """
+    Check whether the path of the given result has the right format.
+
+    Parameters
+    ----------
+    result: pd.Series
+        The result to be checked.
+    """
+    p = result['path']
+    if not p.startswith('archive'):
+        raise ValueError(f'The path {p} does not start correctly')
+
+    meas_key = p.split('::')[1]
+    ensemble = p.split('/')[1]
+    project = p.split('/')[2].split('::')[0]
+    if not len(meas_key) == 64:
+        raise ValueError(f'meas_key of {p} is scrambled')
+    if ensemble not in ensembles:
+        raise ValueError(f'meas_key of {p} points to an unknown ensemble')
+    if project not in projects:
+        raise ValueError(f'meas_key of {p} points to an unknown project id')
+
+
+
 def check_db_integrity(path: Path) -> None:
     """
     Check intergrity of the database by checking the uniqueness of the record keys used to load the records
@@ -82,10 +107,13 @@ def check_db_integrity(path: Path) -> None:
     search_expr = "SELECT * FROM 'backlogs'"
     conn = sqlite3.connect(path / db)
     results = pd.read_sql(search_expr, conn)
+    ensembles = list_ensembles(path)
+    projects = [p[0] for p in list_projects(path)]
 
     for _, result in results.iterrows():
         if not has_valid_times(result):
             raise ValueError(f"Result with id {result[id]} has wrong time signatures.")
+        check_path_format(result, ensembles, projects)
     return