snapfix 0.1.0

Capture real Python objects, scrub sensitive fields, emit pytest fixtures.

snapfix

Capture real Python objects from staging, scrub sensitive fields, emit @pytest.fixture files automatically.

---

The problem

You have a bug that only reproduces with real data. You need a test. You don't want to hand-build a factory that misses the edge case, and you don't want to copy-paste a production payload and accidentally commit a customer's email address.

snapfix solves this with one decorator.

---

Install

pip install snapfix

Python 3.10+ required. No other required dependencies.

---

Quickstart

# In your service code (staging or development only)
from snapfix import capture

@capture("invoice_response", scrub=["billing_name"])
def fetch_invoice(invoice_id: str) -> dict:
    return external_api.get(f"/invoices/{invoice_id}")

Call the function once against staging. snapfix writes this to tests/fixtures/snapfix_invoice_response.py:

# Generated by snapfix -- do not edit manually
# Captured : 2026-03-23T14:22:01
# Scrubbed : customer.email, meta.token, billing_name
import pytest
from snapfix import reconstruct

@pytest.fixture
def invoice_response():
    return reconstruct({
        "id": "INV-8821",
        "customer": {
            "email": "***SCRUBBED***",
            "billing_name": "***SCRUBBED***",
            "plan": "pro",
        },
        "amount": {"__snapfix_type__": "decimal", "value": "149.99"},
        "issued_at": {"__snapfix_type__": "datetime", "value": "2026-03-01T09:00:00"},
        "meta": {
            "token": "***SCRUBBED***",
            "retry": 0,
        },
    })

Use it in any test:

def test_invoice_total(invoice_response):
    assert invoice_response["amount"].quantize(Decimal("0.01")) == Decimal("149.99")

That's it. No factory definition. No manual scrubbing. No pasted JSON.

---

PII scrubbing

⚠ Important limitation: snapfix scrubs fields by key name only. It does NOT detect PII in field values. An email address stored as response["tags"][0] or inside a list will not be scrubbed. Always review generated fixtures before committing them.

Default scrubbed field names

Any key whose name contains one of these strings (case-insensitive, substring match) is scrubbed:

email · password · passwd · token · secret · api_key · apikey · access_token · refresh_token · ssn · credit_card · card_number · cvv · phone · mobile · dob · date_of_birth · address · ip_address · authorization · auth · bearer

customer_email → scrubbed (substring match on email)
billing_phone_number → scrubbed (substring match on phone)
retry_count → not scrubbed

Adding custom fields

@capture("order", scrub=["customer_id", "tax_number"])
def fetch_order(order_id: str) -> dict: ...

Replacement values

Field value type	Replacement
str	"***SCRUBBED***"
int / float	-1
None	"***SCRUBBED***"

---

Supported types

snapfix serializes the following Python types and restores them correctly via reconstruct():

Type	Serialized as	Restored on reconstruct()
dict, list, str, int, float, bool, None	JSON native	Same
datetime.datetime	ISO 8601 string + marker	datetime
datetime.date	ISO 8601 string + marker	date
datetime.time	ISO 8601 string + marker	time
datetime.timedelta	total seconds + marker	timedelta
uuid.UUID	string + marker	UUID
decimal.Decimal	string + marker	Decimal
bytes	base64 + marker	bytes
bytearray	base64 + marker	bytearray
pathlib.Path	string + marker	Path
enum.Enum	.value + marker	value only (enum class not preserved)
tuple	list + marker	list (documented: tuples become lists)
set / frozenset	sorted list + marker	set / frozenset
dataclass	dataclasses.asdict() then recurse	dict
pydantic.BaseModel	.model_dump() then recurse	dict
Circular reference	{"__snapfix_circular__": true}	sentinel dict
Unserializable type	{"__snapfix_unserializable__": true, ...}	sentinel dict

Note: tuple → list on roundtrip is intentional. JSON has no tuple type. If your tests depend on the exact type, assert isinstance(result, list) rather than tuple.

---

Configuration

Environment variables

Variable	Default	Description
SNAPFIX_OUTPUT_DIR	tests/fixtures	Where fixture files are written
SNAPFIX_MAX_DEPTH	10	Maximum serialization depth
SNAPFIX_MAX_SIZE	500000	Maximum payload size in bytes
SNAPFIX_ENABLED	true	Set to false to disable all capture

snapfix.yaml (project root)

snapfix:
  output_dir: tests/fixtures
  max_depth: 10
  max_size_bytes: 500000
  enabled: true

Priority order (highest to lowest)

1. Decorator parameters: @capture(name, scrub=[...], max_depth=5)
2. Environment variables
3. snapfix.yaml
4. Built-in defaults

Disabling in production

Set SNAPFIX_ENABLED=false in your production environment. The decorator becomes a no-op — zero overhead, no files written, no exceptions swallowed.

---

CLI

snapfix list              # List all captured fixtures with metadata
snapfix show  <name>      # Print a fixture to stdout
snapfix clear <name>      # Delete one fixture (prompts for confirmation)
snapfix clear-all         # Delete all fixtures (prompts for confirmation)

All commands accept --dir <path> to target a non-default fixture directory.

---

Decorator reference

@capture(
    name,                # str — fixture name; becomes the function name in the output file
    scrub=None,          # list[str] | None — extra field names to scrub (merged with defaults)
    max_depth=None,      # int | None — override max serialization depth
    max_size_bytes=None, # int | None — override max payload size
    config=None,         # SnapfixConfig | None — full config override
)

Works on both sync and async functions. The decorator is transparent:

• The return value is always preserved unchanged.
• If the wrapped function raises, the exception propagates normally and no file is written.
• If serialization fails for any reason, a warnings.warn() is emitted and execution continues.

---

FAQ

Is it safe to use against production traffic? No. Use snapfix against staging or a development environment. Production traffic contains real customer data. Even with scrubbing enabled, value-level PII (email addresses in list fields, etc.) will not be caught.

Does it slow down my application? Serialization adds latency proportional to payload size. For typical API responses (< 50 KB), the overhead is negligible in staging. Do not enable SNAPFIX_ENABLED=true in a production critical path.

What happens if the object is too large? If the serialized payload exceeds max_size_bytes, the fixture is not written and a warning is emitted. Increase SNAPFIX_MAX_SIZE or add depth limiting with max_depth.

What happens if a field contains an unserializable type? It is replaced with a sentinel dict: {"__snapfix_unserializable__": true, "__snapfix_repr__": "...", "__snapfix_type_name__": "..."}. The rest of the object is still captured. reconstruct() returns the sentinel as-is.

Can I regenerate a fixture? Yes. Re-run the decorated function. The fixture file is overwritten in place.

Does it work with pytest parametrize? The generated file is a standard @pytest.fixture. It works with everything pytest supports.

---

Development

git clone https://github.com/yourname/snapfix
cd snapfix
pip install -e ".[dev]"
pytest
ruff check src/

---

License

MIT