Marivo 0.4.2

Marivo 0.4.2 unifies temporal semantics across semantic authoring, analysis selection, alignment, evidence, and forecasting. It adds certified custom periods, named occurrences, working-day schedules, and the typed contracts needed to use them reproducibly.

Highlights

Temporal semantics now have separate, reusable authorities: PeriodCalendar for fiscal and retail periods, TemporalSet for named intervals or dates, and WorkSchedule for certified working-day status. Grain, TimeScope, and AlignmentPolicy carry the exact semantic variant and provenance.
Semantic authoring and catalog loading can certify period calendars, calendar levels, temporal occurrences, and work schedules. Analysis reuses the exact project-local snapshot and digest instead of deriving boundaries or holidays independently.
Certified semantic grains support observation, roll-up, comparison, and custom-period forecasting over ordinal periods with exact keys and bounds. Incomplete, stale, or mismatched snapshots fail before an artifact is written.
Comparison adds typed period_progress, period_correspondence, occurrence_progress, and working_day_progress policies for equal progress inside periods, named period mappings, named occurrences, and working-day ordinals.

Breaking changes

Temporal authorities require exact typed semantic references or catalog entries. Analysis no longer infers fiscal periods, holidays, occurrences, or working-day status from names, locale, or local parameters.
Analysis artifacts use analysis-artifact/v8. Older artifact schemas and removed fields such as component_graph_ref are rejected fail-closed rather than migrated or dual-read; recreate the analysis session after upgrading.

Changed

Temporal bindings, period correspondence, occurrence scopes, work schedules, and alignment policies are persisted as typed evidence, including the authority ref, snapshot digest, boundary timezone, and exact half-open bounds.
Frame metadata now preserves the report timezone used for coverage and quality calculations. Timezone mixing is canonicalized before coverage ratios are computed.
Public errors and contracts expose more precise repair locations for missing temporal authority, invalid coverage, reserved columns, and unsupported metric arity. Live help and exported symbols remain deterministic with the runtime surface.
The documentation site adds a bilingual homepage and design blog, exposes search and indexing signals, and builds the Sphinx API reference as part of the repository check.

Fixed

Comparison and correlation evidence now remains bound to the exact selected period or lag used for the reported result.
Coverage without a coverage sidecar returns no coverage claim, while attribution extraction failures no longer hide rows that remain readable and reconciled.
Certified roll-up keys, temporal occurrence identity, and forecast period boundaries are preserved through persistence and reload.

Documentation

Added the unified temporal semantics specification covering period calendars, temporal sets, work schedules, grains, scopes, and alignment.
Updated English and Chinese analysis workflow guidance for certified temporal selection, alignment, working-day comparison, and semantic-grain forecasting.

Installation

pip install marivo

Install the backend extra for each datasource you plan to use. Marivo 0.4.2 requires Python 3.12 or newer.