Add warning if xfel-calibrate may not be reproducible

Description

This adds a warning to xfel-calibrate if the chosen configuration may cause the result not be reproducible by xfel_calibrate.repeat, currently based on the presence of argument line flags. The user is forced to confirm this by pressing ENTER except when the additional --not-reproducible flag is passed. In the future, we may also use the latter flag to disable other checks or work only relevant for reproducibility.

Also, there's a new flag --skip-env-freeze to skip the pip freeze stage, which obviously breaks reproducibility.

Most important question and the reason I tag @calibration: Do you think the wording is good enough?

How Has This Been Tested?

Extensive testing of the HED fast path.

Types of changes

• New feature (non-breaking change which adds functionality)

Checklist:

Reviewers

@calibration

src/xfel_calibrate/calibrate.py

@@ -83,9 +83,18 @@ def make_initial_parser(**kwargs):
                         help='Filename (and optionally path) for output'
                              ' report')
 
+    parser.add_argument('--not-reproducible', action='store_true',
+                        help='Disable checks to allow the processing result '
+                             'to not be reproducible based on its metadata.')
+
     parser.add_argument('--skip-report', action='store_true',
                         help='Skip report generation in finalize step.')
 
+    parser.add_argument('--skip-env-freeze', action='store_true',
+                        help='Skip recording the Python environment for '
+                             'reproducibility purposes, requires '
+                             '--not-reproducible to run.')
+
     parser.add_argument('--concurrency-par', type=str,
                         help='Name of concurrency parameter.'
                               'If not given, it is taken from configuration.')
@@ -981,6 +990,35 @@ def run():
     caltype = args["type"].upper()
     sequential = args["no_cluster_job"]
 
+    # Pick out any arguments that may prevent reproducibility from
+    # working, sorted alphabetically and converted back to their
+    # canonical representation.
+    not_reproducible_args = sorted(
+        ('--' + x.replace('_', '-')
+         for x in ['skip_env_freeze']
+         if args[x]))
+
+    # If any of these arguments are set, present a warning.
+    if not_reproducible_args:
+        print('WARNING: One or more command line arguments ({}) may prevent '
+              'this specific correction result from being reproducible based '
+              'on its metadata. It may not be possible to restore identical '
+              'output data files when they have been deleted or lost. Please '
+              'ensure that the data retention policy of the chosen storage '
+              'location is sufficient for your '
+              'needs.'.format(', '.join(not_reproducible_args)))
+
+        if not args['not_reproducible']:
+            # If not explicitly specified that reproducibility may be
+            # broken, remind the user and exit.
+            print('To proceed, you can explicitly allow reproducibility to '
+                  'be broken by adding --not-reproducible')
+            sys.exit(1)
+
+        reproducible = False
+    else:
+        reproducible = True
+
     try:
         nb_info = notebooks[detector][caltype]
     except KeyError:
@@ -1096,6 +1134,7 @@ def run():
     metadata["pycalibration-version"] = version
     metadata["report-path"] = f"{report_to}.pdf" if report_to \
         else '# REPORT SKIPPED #'
+    metadata['reproducible'] = reproducible
     metadata["concurrency"] = {
         'parameter': concurrency_par,
         'default': concurrency_defval,
@@ -1116,8 +1155,9 @@ def run():
     metadata.save()
 
     # Record installed Python packages for reproducing the environment
-    with open(os.path.join(run_tmp_path, 'requirements.txt'), 'wb') as f:
-        check_call([python_exe, '-m', 'pip', 'freeze'], stdout=f)
+    if not args['skip_env_freeze']:
+        with open(os.path.join(run_tmp_path, 'requirements.txt'), 'wb') as f:
+            check_call([python_exe, '-m', 'pip', 'freeze'], stdout=f)
 
     folder = get_par_attr(parms, 'in_folder', 'value', '')