License
BSD-3-Clause
Frame and query adapters over the Arrow IPC codec: the eager file readers, the frame writer with the approved temp-then-rename publication policy, the lazy scan source and the incremental sink. Everything frame-, plan- or query-shaped lives here; the codec tier itself (IpcBatchReader, IpcBatchWriter, buildIpcBytes over ColumnBatch) stays in ddn.df.ipc and never imports this module.
this(string path, ColumnSchema[] schema,
IpcCompression compression)ColumnSchema[] schema() @safevoid close() @trustedScanSource acquire() @safebool supportsProjection() @safeScanSource project(scope string[] columns) @safethis(string path)DataFrame readIpcFile(string path) @trustedReads an Arrow IPC file (`.arrows`, Feather V2) from disk.DataFrame readIpc(const(ubyte)[] bytes) @trustedReads an Arrow IPC file from bytes (`.arrows`, Feather V2): the footer's schema names the columns, the dictionary batches arrive before the record batches, and each record batch becomes one chunk p...ubyte[] buildIpcBytes(DataFrame df,
IpcCompression compression = IpcCompression.NONE) @trustedBuilds the file bytes of a frame (the codec tier owns the chunk-alignment policy; this adapter hands it the frame's columns).void writeIpcFile(string path, DataFrame df,
IpcCompression compression = IpcCompression.NONE) @trustedWrites a frame to an Arrow IPC file (`.arrows`, Feather V2) under the approved failure policy: the bytes stage into a sibling temp file that renames into place on success and is removed on any fail...LazyFrame scanIpc(string path) @trustedOpens an Arrow IPC file (`.arrows`, Feather V2) as a lazy scan source: the header and footer are read eagerly, then each record batch is sought and decoded on demand - one batch per pull, no whole-...void writeIpcFile(string path, LazyFrame df,
IpcCompression compression = IpcCompression.NONE) @trustedWrites a deferred frame to an Arrow IPC file incrementally: the schema derives from metadata alone, batches pull through the row-wise chain one at a time (the input never materialises), and each pu...