v1.2.0

Added

• Database.__init__ now detects reserved-suffix column collisions. A collision exists when both <base> and <base><suffix> are present as columns, with <suffix> in (_lim, _has, _any) — in that case the kwarg <base><suffix>=v in Database.__call__ is ambiguous between the suffix predicate (range / set / substring) and a literal-column equality. Configurable via the new keyword-only on_reserved_suffix parameter: "warn" (default — emits a ReservedSuffixCollisionWarning), "raise" (raises ValueError), "ignore", or "rename" (auto-rename the colliding column via _lim → _limits, _has → _contains, _any → _options; raises if the rename target already exists). The warning message in "warn" mode points users to both fixes (rename the source manually, or pass on_reserved_suffix='rename'). Resolves the open question deferred from 1.1.0.
• ReservedSuffixCollisionWarning (subclass of UserWarning) — exported so callers can warnings.filterwarnings(...) against it specifically.

Notes

• Non-breaking. Existing Database(df) calls keep working; the new check only fires when the DataFrame has both a <base> and a <base><suffix> column. DatabaseContainer does not re-check after its overlapping-column rename — those renames almost never produce new reserved-suffix collisions in practice.
• Rename mode mutates self._data (via DataFrame.rename(columns=...), which returns a new frame) — the user's original DataFrame is not mutated in place.

v1.1.0

Added

• DatabaseContainer — manage hierarchical relationships between Database instances (e.g. subject → trial →
  action) with key-derivation lambdas linking levels. Supports add, add_data_field (via the contained
  Database instances), the same _lim / _has / _any keyword-suffix query syntax as Database, records,
  help, and parent/child column casting (_cast_column_to_db).
• DatabaseContainer is now exported (from datanest import DatabaseContainer).
• cache_me_if_you_can / cache_me_if_you_can_incremental — dill-backed file-cache decorators, migrated from
  immersionToolbox/immersionlab/__init__.py. The append_str: str (eval'd at call time) parameter is
  replaced by suffix: Callable[..., str]; pn-projects callers updated in the same release.
• GitHub Actions CI (.github/workflows/test.yml) — pytest matrix across Python 3.8–3.12 on Ubuntu 22.04 /
  macOS latest / Windows latest, triggered on push and PRs.
• [project.optional-dependencies] test group (pip install -e ".[test]") — installs pytest and openpyxl
  (the latter exercises Database's Excel I/O path in the test suite without bloating the runtime dep list).
• README "Payload types — anything goes" section documenting the modality-agnostic design: payloads can be any
  Python object, and pysampled.Data is a use convention rather than a hard dependency.

Dependencies

• dill is now a required dependency (used by the new cache_me_if_you_can* decorators).
• Python floor raised from 3.7 to 3.8 (3.7 is EOL and unsupported by the new CI matrix).

Tests

• test_database_container.py — _make_synthetic_signals now returns plain np.ndarray payloads instead of
  pysampled.Data. The two affected tests exercise add_data_field's modality-agnostic contract (key→object
  dict), which the payload type does not influence; removing the pysampled test-only import aligns the suite
  with datanest's design philosophy and drops a test-only dep.

Notes

• DatabaseContainer: this is its new home — previously implemented in
  immersionToolbox/immersionlab/__init__.py. The legacy classes there become a re-export shim of the
  datanest originals (mirroring the immersionlab/delsys.py pattern); existing
  immersionlab.DatabaseContainer subclasses continue to work unchanged.
• cache_me_if_you_can*: no re-export shim is left in immersionlab (in contrast with
  DatabaseContainer). pn-projects callers were flipped to import directly from datanest in lockstep with the
  migration; immersionlab.cache_me_if_you_can no longer exists.

Deferred to 1.2.0

• break_signals_into_actions stays in immersionlab (couples to pysampled / event semantics — does not
  belong in datanest's modality-agnostic core). immersionlab.DatabaseContainer keeps the method via a thin
  subclass.
• Kwarg-suffix collision detection (_lim / _has / _any reservations vs. real column names) — TODO at
  datanest/__init__.py DatabaseContainer docstring.

v1.0.0

Version: 1.0.0
Date: 2024.02.04
Summary: First major release of datanest, A simple, intuitive, pandas-based database that is ideal for handling data such as time series, images, or any Python objects alongside their metadata.
Installation: pip install datanest
Dependencies: python>=3.7, numpy, pandas, and optionally openpyxl
Testing: Performed using pytest. All 7 tests passed.
Coverage: 100%
Contributors: Praneeth Namburi
License: MIT