Wandas 0.8.0¶
This describes the published 0.8.0 tag. For the subsequent conditional WDF 0.5 writer, see Wandas 0.8.1.
Wandas 0.8.0 accepts existing file lists and recording catalogs as lazy Datasets, so you can select observations by metadata before reading their waveforms. It also extends the bilingual Learning Path with catalog inputs and reusable custom processing examples.
Wandas 0.8.0では、既存のファイル一覧・CSV・DataFrameから遅延Datasetを作り、 metadataで観測を選んでから必要な波形だけを読めます。日英Learning Pathには 表の取込と再利用可能な独自処理の実例を追加しました。
Highlights / 主な変更¶
wd.from_files(paths, base_dir=...)retains explicit input order and duplicates.wd.from_table(table, path_column=..., metadata_columns=..., base_dir=...)accepts a local CSV catalog or pandas DataFrame; one row is one observation.- CSV metadata stays strings. DataFrame basic scalar metadata preserves its
types and normalizes missing values to
None. Relative CSV paths use the catalog parent; file lists/DataFrames use construction-time cwd unless an explicit base is supplied. - Construction and metadata selection do not read WAV sources. Item access
inspects the selected WAV header;
frame.datamaterializes its waveform. Signal CSV sources differ: item access parses the complete signal table, and materialization parses it again. - Subsets snapshot only selected metadata while preserving isolation and
existing caches.
select()still scans the catalog; no constant-time or constant-memory guarantee is introduced. - Learning Path chapters 05 and 08 cover custom operations/Recipe replay and metadata selection from file lists, CSV and DataFrame catalogs. Automated lesson tests and static export use a noninteractive Matplotlib backend.
- Component test contracts, coverage and descriptive test names were improved.
ファイル一覧の順序・重複と、表の行ごとの属性を保持します。CSVの文字列とDataFrameの 型・欠損値を区別し、相対パスの基準を指定できます。subset生成は選択項目だけをコピー しますが、条件照合は全件走査です。構築時に全波形を展開する機能ではありません。
See file lists and recording catalogs and Learning Path chapter08 for executable examples and input/source-lifetime contracts.
Changes / 含まれる変更¶
- #464: lazy file-list and catalog inputs, subset snapshot optimization, examples and platform-path regressions.
- #461, #460: custom processing and reusable Recipe extension lessons.
- #459, #458: component test contracts, coverage, names and descriptions.
Compatibility / 互換性¶
The new documented Dataset entry points are Stable additive APIs. This release contains no removal or incompatible semantic change to stable public APIs or supported WDF/Recipe schemas, so no migration or deprecation is required. Existing folder discovery, basename string lookup, integer access, metadata propagation and lazy failure/cache behavior remain available.
新しいDataset入口はStableな追加APIです。既存の安定APIの削除・非互換変更や WDF/Recipe schema変更はなく、移行やdeprecationは不要です。フォルダ探索と既存の basename検索は維持します。入力照合の曖昧さ監査・ブラウザhandle/権限bridge・Dataset 永続化はこのリリースに含みません。