Getting Started¶

Installation¶

pip install versionable

With HDF5 support (needed for saving large arrays to .h5 files — see why):

pip install "versionable[hdf5]"

The Hash¶

Why it exists¶

Serialized files outlive your code. If you rename a field, change a type, or add a required field, files saved by the old version of your class will silently load with wrong values — or worse, load without error while corrupting downstream logic.

The hash parameter is a 6-character fingerprint of your class’s field names and types. versionable recomputes it every time your module is imported and raises HashMismatchError immediately after you change your code if the class definition no longer matches what was declared. You find out at first startup, when you’re changing the code, not buried in a production bug.

Computing the hash for a new class¶

Define your class without a hash:

@dataclass
class MyConfig(Versionable, version=1):
    name: str
    value: float

Run your code — validation is skipped when hash is absent or empty.
Print the hash:
```
print(MyConfig.hash())
# e.g. "4b7866"
```

Add it to the class definition:

@dataclass
class MyConfig(Versionable, version=1, hash="4b7866"):
    name: str
    value: float

What happens when the schema changes¶

Suppose you saved a file with the original class:

# v1 — field is called "value"
@dataclass
class MyConfig(Versionable, version=1, hash="4b7866"):
    name: str
    value: float

The YAML file on disk looks like:

name: experiment-A
value: 9.81
__versionable__:
  __OBJECT__: MyConfig
  __VERSION__: 1
  __HASH__: 4b7866

Now you rename the field without updating the hash:

@dataclass
class MyConfig(Versionable, version=1, hash="4b7866"):  # ← hash still matches v1, but fields changed!
    name: str
    magnitude: float          # renamed from "value"

Without hash validation, versionable.load(MyConfig, "config.yaml") would silently return an object where magnitude is unset (or a default), and the value = 9.81 from the file would be silently discarded. versionable catches this before you ever call load:

HashMismatchError: MyConfig — declared hash '4b7866' does not match computed 'a70249'

That’s your signal to recompute the hash, bump the version, and add a migration so existing files can be upgraded. Here’s the correct v2:

from versionable import Versionable, Migration

@dataclass
class MyConfig(Versionable, version=2, hash="a70249"):
    name: str
    magnitude: float          # renamed from "value"

    class Migrate:
        v1 = Migration().rename("value", "magnitude")

Now versionable.load(MyConfig, "config.yaml") reads the old value: 9.81 from the file, renames it to magnitude on the fly, and returns a fully populated MyConfig instance.

See Migrations for the full range of migration operations.

Dev Mode¶

During rapid iteration you can suppress hash errors so mismatches are warnings instead:

from versionable import ignoreHashErrors
ignoreHashErrors(True)

Turn it off before committing. (or submit a PR for #2)