gnott
commented
5 years ago May also consider to use OrderedDict as a replacement for all remaining plain old {} dict in the parser. We use OrderedDict regularly when generating some output, and it will probably have little negative speed or usability impacts. @giorgiosironi maybe you have some comments about switching to OrderedDict for older function calls?
There are JSON file test fixtures in the project, currently in the
elifetools/tests/JSON/folder. There's a scriptelifetools/tests/generate_expected_json.pywhich can be used to help re-generate the fixtures.Running the generate script (after some modifications) does still generate
.jsonfile output, except since the contents of the JSON files are mostly simpledictdata, the inconsistent order of the dict keys causes the git diff of the files to vary more than they should to be useful test fixtures. At the very least, the keys of the dict can be sorted alphabetically before saving them as.jsonfiles. Alternatively, output as anOrderedDictmay help with comparisons.Comparisons with the file fixtures can compare sorted dicts, or use the
assert equal dict-style assertions.Another couple potential improvements are to move the
generate_test_fixtures.pyto the root folder of the project, and move theelifetools/testsdirectory totestsinstead.In summary, some proposed tests refactoring:
elifetools/testsdirectory totestsgenerate_test_fixtures.pyto ~the root folder of the project~ thetests/foldergenerate_test_fixtures.pyto make it Python 3 compatiblegenerate_test_fixtures.pyto output json with alpha sorted keys, and/or employOrderedDictin some wayOrderedDictmore extensively in the parser itself (so far tests do still pass regardless of the json file dict key order)