Datasets:

---

mlfoundations-dev

/

github-issues

like 4

Follow

ML Foundations Development 133

ERROR: type should be string, got "https://github.com/joke2k/django-environ/blob/628ed38e540ec9a6f3aff622953948e54f14a76d/environ/environ.py#L633-L638\r\n\r\nHi, I was using the `Env.read_env()` function to load a file consisting of dictionary pattern:\r\n\r\n```\r\nMY_DICTIONARY=\"Tom\"=\"Jerry\"\r\nMY_OTHER_DICTIONARY='Tom'='Jerry'\r\n```\r\nAfter parsing the lines, what is pulled by the regex matching is:\r\n```\r\nTom\"=\"Jerry\r\nTom'='Jerry\r\n```\r\nwhich gives us:\r\n```\r\n>>> env.dict(MY_DICTIONARY) = {'Tom\"': '\"Jerry'} # notice the extra double quotes\r\n>>> env.dict(MY_OTHER_DICTIONARY) = {\"Tom'\": \"'Jerry\"} # notice the extra single quotes\r\n```\r\n\r\nI just found this out as i debugged for quite a while. It doesn't happen if the variables are already loaded into the shell environment via shell export. It only happens from reading the `.env` file (and overriding the environment variables).\r\n\r\nMaybe the intent is to parse shell string variables and not used that much for dictionary patterns, so this edge case came up.\r\n\r\nThe apostrophes might be needed, as some key or values itself may contain characters like `=`, `'`, `\"`, which are valid in a dictionary key or value.\r\n\r\nLooks like it'll be difficult to cover for dictionary use case with apostrophes via simple regex.\r\n\r\nWrapping the whole string with more quotes will still have a 50% chance to unknowingly hit the regex parsing of `m2` and `m3`."