open-christian-data / releases /v1.0.0-migration.md
OpenChristianData's picture
Clarify published v1.0.0 migration guide
984f9bd verified
|
Raw History Blame Contribute Delete
3.89 kB

Updating to Open Christian Data v1.0.0

The dataset keeps its repository ID: OpenChristianDataOrg/open-christian-data. Version 1.0.0 rebuilds the public format from accepted copies. Update the configuration and record handling in your code before switching versions.

Choose the new download group

Earlier configuration v1.0.0 starting point
bible_text bible
catechism_qa catechisms
commentary commentary, with changed record shapes and boundaries
devotional devotionals; also check prose for devotional books
doctrinal_document doctrinal_documents
hymn_collection hymns
prayer prayers; liturgical collections can also be in liturgies
reference_entry reference_entries
sermon sermons; source-defined prose/homily material can also be in prose
structured_text prose or poetry, according to the selected work
topical_reference topical_indexes
church_fathers No equivalent quotation group in this release; Church Fathers works appear in other groups

This is a guide to finding the new groups, not a guarantee of identical coverage or a one-to-one row conversion. Check the selected work and edition you need.

Update field handling

Earlier pattern New handling
_source_id identifies a source collection Use work_id, textual_form_id, and edition_id to distinguish the work and copy
_source_title, _author, _contributors are expected on every row Use available provenance and source notices; IDs alone are not author/title fields or complete citations
_source_url Use available provenance.source_url and source notices; a URL is not guaranteed on every row
_license Use the record's license and the matching source notices
A Bible reference is always top-level osis Read the category/source-specific locator and any reference-status/versification information
Each book record is a fixed-size passage Respect the new source-defined boundaries; some records contain a complete book or division
text includes all notes and formatting Inspect structure, notes, references, or front matter where provided
Tradition, era, audience, summary, key quotations, or stored word/token counts These earlier enrichment/count fields are not carried by the v1 public row format

Do not obtain the new IDs by renaming a legacy field mechanically. Names, identities, ordering, coverage, and passage boundaries can all differ.

For example, an earlier Bible reader might use:

source = row["_source_id"]
location = row["osis"]

The new reader distinguishes the selected copy and handles its locator:

work = row["work_id"]
copy = row["edition_id"]
location = row["locator"]

Keep the complete locator object when a normalized reference is unavailable. See the data guide for actual Bible and catechism examples.

Load and pin a version

pip install datasets
from datasets import load_dataset

rows = load_dataset(
    "OpenChristianDataOrg/open-christian-data",
    name="prose",
    split="train",
    revision="v1.0.0",
    streaming=True,
)

The v1.0.0 release is published. The tested reader versions are Datasets 5.1.0 and PyArrow 26.0.0. JSONL remains available for direct readers; Hugging Face loading uses the typed Arrow companions.

To keep using the previous upload and its earlier configuration names, pin:

old_rows = load_dataset(
    "OpenChristianDataOrg/open-christian-data",
    name="structured_text",
    split="train",
    revision="2f23018e069fd3900ff1bff499924814d40ffabf",
    streaming=True,
)

Leaving out revision follows main, so it will pick up future uploads. Old IDs or row positions do not establish correspondence with the new release.