json.load and json.dump — file versions

import json

with open("data.json") as f:
    data = json.load(f)         # parse JSON file → Python value

with open("out.json", "w") as f:
    json.dump(data, f)          # serialize Python value → JSON file

loads / dumps vs load / dump — the s is for string

Type mapping

Python ↔ JSON:

Note: a tuple goes IN as JSON array but comes OUT as a list. Sets and datetimes are not JSON-native — convert sets to lists, format dates as ISO strings.

Pretty-print with indent

json.dump(data, f, indent=2)
# Result file:
# {
#   "a": 1,
#   "b": [2, 3]
# }

Use indent=2 for human-readable files, no indent for compact wire-format.

sort_keys

json.dump(data, f, sort_keys=True, indent=2)

Useful for stable diffs in version control.

Common gotchas

• Trailing commas, comments, single quotes — none of those are valid JSON. Python's parser is strict. If your file has them, it isn't JSON.
• NaN / Infinity are non-standard. json allows them by default but other systems often reject. Use json.dump(..., allow_nan=False) if interoperability matters.
• Encoding. json.dumps produces ASCII by default, escaping non-ASCII characters. Pass ensure_ascii=False to keep emoji and accented characters readable.

When to use which form

• Working with files: json.load(open(...)) and json.dump(data, open(...)) — use them inside with blocks.
• Working with API responses: almost always json.loads(response.text) — the JSON arrives as a string.
• Logging or wire-format: json.dumps(data) — shape it for transmission.