Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

tiferet-h5

HDF5 infrastructure extension for the Tiferet framework via PyTables.

tiferet-h5 mirrors the layered architecture of the Tiferet core — domain, interfaces, mappers, utils, repos — adapted to the hierarchical, columnar, and typed nature of HDF5 files. It introduces two mapper base classes that are the HDF5-native analogue of Tiferet's TransferObject: TableObject for row-oriented table storage and NodeObject for attribute-oriented node storage.

Requirements

  • Python ≥ 3.10
  • tiferet >= 2.0.3, < 2.1
  • tables >= 3.10.0 (PyTables)

Installation

pip install tiferet-h5

Or in development mode from the repository root:

pip install -e .

Quick Start

This tutorial stores a simple feature catalog — groups with scalar metadata and child step tables — to demonstrate the full stack: domain objects, mapper classes, H5Client, and H5Repository.

1. Define Your Mappers

import tables
from typing import ClassVar, Dict, Any, List
from pydantic import Field, AliasChoices
from tiferet_h5 import TableObject, NodeObject

# ── Group-level metadata → node attributes ────────────────────────────────
# One instance per feature group, stored on the HDF5 group node.
class FeatureGroupObject(NodeObject):
    name: str = Field(default='', description='Feature name.')
    description: str = Field(
        default='',
        serialization_alias='desc',
        validation_alias=AliasChoices('desc', 'description'),
        description='Stored as "desc" in HDF5 to keep attribute keys short.',
    )
    _ROLES: ClassVar[Dict[str, Dict[str, Any]]] = {
        'to_h5.attrs': {'by_alias': True, 'exclude_none': True},
    }

# ── Child collection → table rows ─────────────────────────────────────────
# One row per step, stored in a table nested inside the feature group.
class FeatureStepObject(TableObject):
    name: str = Field(default='', description='Step name.')
    service_id: str = Field(
        default='',
        serialization_alias='svc',
        validation_alias=AliasChoices('svc', 'service_id'),
        description='Service ID; stored as "svc" column in HDF5.',
    )
    pass_on_error: bool = Field(
        default=False,
        serialization_alias='pass_err',
        validation_alias=AliasChoices('pass_err', 'pass_on_error'),
    )
    _H5_TYPES: ClassVar[Dict[str, Any]] = {
        'name':     tables.StringCol(256),
        'svc':      tables.StringCol(256),
        'pass_err': tables.BoolCol(),
    }

2. Write to HDF5

from tiferet_h5 import H5Client

# HDF5 layout:
#   /features/                ← group, attr: schema_ver
#       calc/                 ← group, attrs: name, desc
#           steps             ← table, cols: name, svc, pass_err

with H5Client('catalog.h5', mode='w') as h5:
    # Catalog root
    h5.create_group('/features')
    h5.set_node_attr('/features', 'schema_ver', '1.0')

    # Feature group
    h5.create_group('/features/calc')
    group = FeatureGroupObject(
        name='Calculator Features',
        description='Basic arithmetic operations',
    )
    for k, v in group.to_attrs().items():
        h5.set_node_attr('/features/calc', k, v)

    # Child steps table
    t = h5.create_table(
        '/features/calc/steps',
        FeatureStepObject.get_description(),
        title='Feature Steps',
    )
    FeatureStepObject(name='Add numbers', service_id='add_event').to_row(t)
    FeatureStepObject(name='Validate',    service_id='validate_event', pass_on_error=True).to_row(t)
    t.flush()

3. Read Back

with H5Client('catalog.h5', mode='r') as h5:
    schema_ver = h5.get_node_attr('/features', 'schema_ver')
    print('Schema version:', schema_ver)

    group = FeatureGroupObject.from_attrs(h5.get_node_attrs('/features/calc'))
    print('Feature:', group.name, '—', group.description)

    steps: List[FeatureStepObject] = [
        FeatureStepObject.from_row(r)
        for r in h5.read_rows('/features/calc/steps')
    ]
    for s in steps:
        print(f'  step  {s.name!r}  service_id={s.service_id!r}  pass_on_error={s.pass_on_error}')

Output:

Schema version: 1.0
Feature: Calculator Features — Basic arithmetic operations
  step  'Add numbers'  service_id='add_event'  pass_on_error=False
  step  'Validate'     service_id='validate_event'  pass_on_error=True

4. Use a Repository

TableRepository and NodeRepository (tiferet_h5.repos.core) remove the hand-rolled get_or_create_table()/to_attrs() orchestration a repository would otherwise write itself -- compose one of each with H5Repository for the group's own metadata and its child table respectively:

from tiferet_h5 import H5Repository, NodeRepository, TableRepository

class FeatureMetaRepository(NodeRepository, H5Repository):
    node_cls = FeatureGroupObject
    node_path = '/features/{key}'

class FeatureStepsRepository(TableRepository, H5Repository):
    table_cls = FeatureStepObject
    table_path = '/features/{key}/steps'

meta_repo  = FeatureMetaRepository('catalog.h5')
steps_repo = FeatureStepsRepository('catalog.h5')

# Write
meta_repo.save(FeatureGroupObject(name='Calculator', description='Arithmetic ops'), key='calc')
steps_repo.save(FeatureStepObject(name='Add numbers', service_id='add_event'), key='calc')

# Read
group = meta_repo.get(key='calc')
steps = steps_repo.list(key='calc')

table_path/node_path are str.format() templates -- any **kwargs passed to a method (here, key='calc') interpolate into the path, so one repository instance serves every feature group in the file. See docs/guides/repos.md for the full method reference, the repository-level compression default, and why TableRepository/NodeRepository should never be multiply inherited into a single class.

Package Layout

tiferet_h5/
├── __init__.py          Public exports and aliases
├── domain/
│   └── h5.py            H5Column, H5TableSchema, H5Node
├── interfaces/
│   └── h5.py            H5Service abstract interface
├── mappers/
│   └── settings.py      TableObject, NodeObject base classes
├── utils/
│   └── h5.py            H5Client (alias: H5); also hosts H5 error code string constants
└── repos/
    ├── core.py          TableRepository, NodeRepository CRUD mixins
    └── h5.py            H5Repository base

Documentation

  • Domain ObjectsH5Column, H5TableSchema, H5Node
  • MappersTableObject, NodeObject, aliasing, nested modeling
  • H5Client — full method reference and error codes
  • ReposH5Repository, TableRepository, NodeRepository CRUD mixins

License

MIT

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

tiferet_h5-1.0.0a6.tar.gz (43.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

tiferet_h5-1.0.0a6-py3-none-any.whl (47.5 kB view details)

Uploaded Python 3

File details

Details for the file tiferet_h5-1.0.0a6.tar.gz.

File metadata

  • Download URL: tiferet_h5-1.0.0a6.tar.gz
  • Upload date:
  • Size: 43.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for tiferet_h5-1.0.0a6.tar.gz
Algorithm Hash digest
SHA256 fdf1b9523e175a60e859a08b2645c3826eb0c18f43a1fdede1bce1b9353b8812
MD5 3e7048b079a731b2e475bf9b8708cc23
BLAKE2b-256 c41a5169a05a2ce4ef959db6459d0a6220b7c8a138074dc9239a3749ce53cc00

See more details on using hashes here.

Provenance

The following attestation bundles were made for tiferet_h5-1.0.0a6.tar.gz:

Publisher: python-publish.yml on greatstrength/tiferet-h5

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file tiferet_h5-1.0.0a6-py3-none-any.whl.

File metadata

  • Download URL: tiferet_h5-1.0.0a6-py3-none-any.whl
  • Upload date:
  • Size: 47.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for tiferet_h5-1.0.0a6-py3-none-any.whl
Algorithm Hash digest
SHA256 bc5d51e1fe885cea71a8ae3bcc5b653a60a5dfa6003681fdd43057fa7d8093a9
MD5 9e4f4244a2c53269eceae2c9706a3d97
BLAKE2b-256 9780c64bad6e2fae37998312a80fc03892f227d5b648c493faf39b23caab3cb4

See more details on using hashes here.

Provenance

The following attestation bundles were made for tiferet_h5-1.0.0a6-py3-none-any.whl:

Publisher: python-publish.yml on greatstrength/tiferet-h5

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page