grabmonkey

Dot-path access, mutation, and flattening for deeply nested dict/JSON data — with error messages that tell you exactly which level failed.

Stop writing this:

data.get("response", {}).get("results", [{}])[0].get("metadata", {}).get("created_at", "")

Write this:

from grabmonkey import grab

grab(data, "response.results[0].metadata.created_at", default="")

Install

pip install grabmonkey

Requires Python 3.11+. No runtime dependencies.

Quick start

from grabmonkey import grab, grab_many, put, delete, flatten, unflatten

data = {"response": {"results": [{"metadata": {"id": 1, "created_at": "2026"}}]}}

# Read, any depth
grab(data, "response.results[0].metadata.created_at")   # "2026"

# Safe default instead of an exception
grab(data, "response.results[0].missing", default="")   # ""

# Type coercion on access
grab({"count": "42"}, "count", as_type=int)             # 42

# Wildcards over lists
grab({"items": [{"n": 1}, {"n": 2}]}, "items[*].n")     # [1, 2]

# Batch access
grab_many(data, {"id": "response.results[0].metadata.id"})   # {"id": 1}

# Set and delete (mutates in place, creates intermediate containers)
d = {}
put(d, "a.b[2].c", 9)              # {"a": {"b": [None, None, {"c": 9}]}}
delete(d, "a.b[2].c")              # {"a": {"b": [None, None, {}]}}
delete(d, "a.b[2]", prune=True)    # {"a": {"b": [None, None]}}  (prune drops empties)

# Flatten / unflatten
flatten({"a": {"b": [1, 2]}})           # {"a.b[0]": 1, "a.b[1]": 2}
unflatten({"a.b[0]": 1, "a.b[1]": 2})   # {"a": {"b": [1, 2]}}

Clear errors

grab(data, "response.results[0].nope")
# grabmonkey.errors.PathError:
# Path 'response.results[0].nope' failed at 'nope':
# key not found in dict with keys ['metadata']

PathError (a LookupError) means a structural miss — the case default= absorbs. PathSyntaxError means a malformed path string. CoercionError means the value was found but as_type could not convert it. The last two are not absorbed by default=, on purpose.

Path syntax

Works on dicts, lists, dataclasses, named tuples, and any object with attribute access. Strings and bytes are treated as leaf values, not navigable containers (see LIMITATIONS.md).

CLI

Reads JSON from a file (--input) or stdin, prints JSON:

echo '{"user":{"name":"Alice"}}' | grabmonkey grab user.name   # "Alice"
echo '{"a":{"b":1}}'             | grabmonkey flatten          # {"a.b": 1}
grabmonkey grab metadata.id --input response.json

Using with AI assistants

See SKILL.md for an LLM-consumable reference (decision table, worked examples, anti-patterns). See LIMITATIONS.md for deliberate design tradeoffs.

Development & review

grabmonkey is tested and reviewed under a portable review kit:

• CONTRIBUTING.md — testing philosophy and the competitive multi-model review-panel process.
• LIMITATIONS.md — deliberate tradeoffs reviewers should not re-litigate (read before "fixing" surprising behaviour).
• RELEASE_READINESS.md — the release rubric; run python scripts/readiness.py for the score and convergence verdict.
• REVIEW_HISTORY.md — the panel-by-panel trajectory.
• review-kit/ — the pristine, project-agnostic templates the above were derived from (inert to the project's own checks).

pip install -e ".[dev]"
pytest -q && ruff check src tests && mypy
python scripts/readiness.py

License

MIT