diff options
Diffstat (limited to 'tests/utils.py')
-rw-r--r-- | tests/utils.py | 32 |
1 files changed, 11 insertions, 21 deletions
diff --git a/tests/utils.py b/tests/utils.py index 8fa9af92..94f97d59 100644 --- a/tests/utils.py +++ b/tests/utils.py @@ -23,28 +23,19 @@ def load_vectors_from_file(filename, loader): def load_nist_vectors(vector_data): - section = None - count = None - data = {} + test_data = None + data = [] for line in vector_data: line = line.strip() - # Blank lines are ignored - if not line: - continue - - # Lines starting with # are comments - if line.startswith("#"): - continue - - # Look for section headers - if line.startswith("[") and line.endswith("]"): - section = line[1:-1] + # Blank lines, comments, and section headers are ignored + if not line or line.startswith("#") or (line.startswith("[") + and line.endswith("]")): continue if line.strip() == "FAIL": - data[section, count]["fail"] = True + test_data["fail"] = True continue # Build our data using a simple Key = Value format @@ -52,16 +43,15 @@ def load_nist_vectors(vector_data): # COUNT is a special token that indicates a new block of data if name.upper() == "COUNT": - count = value - data[section, count] = {} + test_data = {} + data.append(test_data) + continue # For all other tokens we simply want the name, value stored in # the dictionary else: - data[section, count][name.lower()] = value.encode("ascii") + test_data[name.lower()] = value.encode("ascii") - # We want to test only for a particular operation, we sort them for the - # benefit of the tests of this function. - return [v for k, v in sorted(data.items(), key=lambda kv: kv[0])] + return data def load_cryptrec_vectors(vector_data): |