curate formate-dates: Mask failed date strings

By default, completely mask date strings with "XXXX-XX-XX" for dates
that failed date formatting so that they are still in the proper
ISO 8601 dates format for downstream Augur commands. Users can turn off
the masking with the added `--no-mask-failure` option.

Note the `store_false` actions produce semi-misleading docs in the help
output, so the new option suppresses the default value in the help message
with `SKIP_AUTO_DEFAULT_IN_HELP` as suggested by @tsibley in review.

augur/argparse_.py

@@ -4,6 +4,20 @@
 from argparse import Action, ArgumentDefaultsHelpFormatter
 
 
+# Include this in an argument help string to suppress the automatic appending
+# of the default value by argparse.ArgumentDefaultsHelpFormatter.  This works
+# because the automatic appending is conditional on the presence of %(default),
+# so we include it but then format it as a zero-length string .0s.  🙃
+#
+# Another solution would be to add an extra attribute to the argument (the
+# argparse.Action instance) and then subclass ArgumentDefaultsHelpFormatter to
+# condition on that new attribute, but that seems more brittle.
+#
+# Copied from the Nextstrain CLI repo
+# https://github.com/nextstrain/cli/blob/017c53805e8317951327d24c04184615cc400b09/nextstrain/cli/argparse.py#L13-L21
+SKIP_AUTO_DEFAULT_IN_HELP = "%(default).0s"
+
+
 def add_default_command(parser):
     """
     Sets the default command to run when none is provided.

augur/curate/format_dates.py

@@ -5,6 +5,7 @@
 import re
 from datetime import datetime
 
+from augur.argparse_ import SKIP_AUTO_DEFAULT_IN_HELP
 from augur.errors import AugurError
 from augur.io.print import print_err
 from augur.types import DataErrorMethod
@@ -31,6 +32,10 @@ def register_parser(parent_subparsers):
         choices=list(DataErrorMethod),
         default=DataErrorMethod.ERROR_FIRST,
         help="How should failed date formatting be reported.")
+    optional.add_argument("--no-mask-failure", dest="mask_failure",
+        action="store_false",
+        help="Do not mask dates with 'XXXX-XX-XX' and return original date string if date formatting failed. " +
+             f"(default: False{SKIP_AUTO_DEFAULT_IN_HELP})")
 
     return parser
 
@@ -169,6 +174,10 @@ def run(args, records):
 
             formatted_date_string = format_date(date_string, args.expected_date_formats)
             if formatted_date_string is None:
+                # Mask failed date formatting before processing error methods
+                # to ensure failures are masked even when failures are "silent"
+                if args.mask_failure:
+                    record[field] = "XXXX-XX-XX"
 
                 if failure_reporting is DataErrorMethod.SILENT:
                     continue

tests/functional/curate/cram/format_dates.t

@@ -41,7 +41,7 @@ This is expected to fail with an error, so redirecting stdout since we don't car
   [2]
 
 Test output with unmatched expected date formats while warning on failures.
-This is expected to print warnings for failures and return the date strings in their original format.
+This is expected to print warnings for failures and return the masked date strings for failures.
 
   $ cat $TMP/records.ndjson \
   >   | ${AUGUR} curate format-dates \
@@ -53,16 +53,27 @@ This is expected to print warnings for failures and return the date strings in t
   WARNING: Unable to format dates for the following (record, field, date string):
   (0, 'collectionDate', '2020-01')
   (0, 'releaseDate', '2020-01')
-  {"record": 1, "date": "2020-XX-XX", "collectionDate": "2020-01", "releaseDate": "2020-01", "updateDate": "2020-07-18"}
+  {"record": 1, "date": "2020-XX-XX", "collectionDate": "XXXX-XX-XX", "releaseDate": "XXXX-XX-XX", "updateDate": "2020-07-18"}
 
 Test output with unmatched expected date formats while silencing failures.
-This is expected to return the date strings in their original format.
+This is expected to return the masked date strings for failures.
 
   $ cat $TMP/records.ndjson \
   >   | ${AUGUR} curate format-dates \
   >     --date-fields "date" "collectionDate" "releaseDate" "updateDate" \
   >     --expected-date-formats "%Y" "%Y-%m-%dT%H:%M:%SZ" \
   >     --failure-reporting "silent"
+  {"record": 1, "date": "2020-XX-XX", "collectionDate": "XXXX-XX-XX", "releaseDate": "XXXX-XX-XX", "updateDate": "2020-07-18"}
+
+Test output with unmatched expected date formats while silencing failures with `--no-mask-failure`.
+This is expected to return the date strings in their original format.
+
+  $ cat $TMP/records.ndjson \
+  >   | ${AUGUR} curate format-dates \
+  >     --date-fields "date" "collectionDate" "releaseDate" "updateDate" \
+  >     --expected-date-formats "%Y" "%Y-%m-%dT%H:%M:%SZ" \
+  >     --failure-reporting "silent" \
+  >     --no-mask-failure
   {"record": 1, "date": "2020-XX-XX", "collectionDate": "2020-01", "releaseDate": "2020-01", "updateDate": "2020-07-18"}
 
 Test output with multiple matching expected date formats.