Release notes

0.1.0

Added:

  • Add chunks parameter (False / True / int MB) to WDFReader

for lazy, memory-efficient reading of large map files backed by dask; chunks align to whole Y-rows and target ~128 MB each, capped at 20 chunks to avoid scheduler overhead. * Add pre-read RAM guard before the DATA block: raise MemoryError when the full array would exceed free RAM, or emit a UserWarning when it would exceed 75 % of free RAM, with instructions to re-open with chunks=True. * Add handler-based dispatch architecture: each WDF scan kind (single, series, points, line_xy, raster_rowmajor, raster_columnmajor, raster_snake, linefocus, volume) has its own module under wdf/handlers/. * Add wdfkit.read(path) module-level function returning an xr.DataArray directly. * Add wdfkit.classify(path) for fast header-only scan-type triage without loading spectrum data. * Add wdf/parsed.py with the ParsedWDF immutable dataclass as the intermediate representation between block parsing and handler assembly. * Add wdf/types.py with MeasurementType, ScanType, DataType, UnitType enums and MapFlag bitmask constants. * Add wdf/dispatch.py with classify_kind and dispatch for routing a ParsedWDF to the correct handler. * Add 7 new test modules covering each WDF scan-kind handler. * Support 2-D data arrays in CosmicRayRemover (treated as line scans). * Add max_passes parameter (default 3) to CosmicRayRemover for iterative 1-D cosmic-ray detection so that large spikes no longer hide smaller ones. * Extract ExposureTime (CCD exposure in seconds) and LaserPower (ND-filter transmission in percent) from the WXDM and WXIS blocks and store them as DataArray attributes for single spectra, maps, and volume scans. * Added test data for the new API. * Add ensure_in_memory shared helper in wdf/utils.py to centralise the Dask warn-and-compute pattern across CosmicRayRemover, SpectraCleaner, and normalize. * Add _build_cr_meta_1d method to CosmicRayRemover to deduplicate repeated construction of the 1-D cosmic-ray correction metadata dict. * Add py.typed marker (PEP 561) so that type checkers treat wdfkit as a typed package. * Add type annotations to read_wdf_file and the private assembly helpers. * Add tests/test_spectra_cleaner.py with 18 tests covering SpectraCleaner and the underlying denoise_spectra_pca.

Changed:

methods ("l1", "l2", "max", "min_max", "area") process each chunk independently via xarray.apply_ufunc() and keep the result lazy; global methods ("robust_scale", "wave_number") load the full array and emit a UserWarning explaining the memory impact. * Make CosmicRayRemover and SpectraCleaner detect Dask-backed input, call .compute() before their NumPy/sklearn kernels, and emit a UserWarning with the estimated RAM cost. * Consolidate all WDF I/O into a single wdf/ subpackage: internal/, spectral/, top-level handlers/, and the root _types.py, _parsed.py, _dispatch.py files are all absorbed into wdf/. * Collapse spectral/axis.py to a flat spectral.py module; the public import path from wdfkit.spectral import SpectralAxisSpec is unchanged. * Replace wdf/assemble.py with handler dispatch in read_wdf_file; WDFReader, wdfkit.read, and the legacy read_wdf_file now all share the same code path. * Rename map xr.DataArray dimensions to lowercase x/y (were X/Y). * Return single-scan xr.DataArray as 1-D (spectral axis only) instead of 2-D (Time, spectral). * Sort spectral coordinate ascending (low → high) for all scan kinds. * Remove duplicate lowercase keys from DataArray.attrs; only the existing CamelCase parameter keys are kept. * Use linear interpolation from the original signal for 1-D spike repair in all cases (was: "median" method replaced spikes with the biased median-filter value). * Dilate the 1-D spike mask by 1 channel on each side before repair to cover sub-threshold spike edges. * Include the DataArray name and file identifier in CosmicRayRemover error messages for easier debugging. * Update documentation to reflect the new API. * Change the logo to a more modern one. * Fix WDFReader(chunks=True) map assembly by splitting sortby into two 1-D calls (da.sortby(sdim).sortby("Time")) to avoid an unsupported nd fancy-index in Dask. * Raise ValueError in normalize immediately when an unrecognised method is passed (was: silently returned an unchanged copy). * Replace assert statements for coordinate validation with explicit raise ValueError (asserts are stripped in optimised mode). * Replace except Exception: pass in print_coord_lengths_if_verbose with a UserWarning that surfaces the failure message. * Replace assert X or True pyflakes workaround in __init__.py with # noqa: F401 on each import line. * Correct memory_check.py docstring to accurately describe that the chunked path skips all RAM checks. * Remove unused max_depth parameter from find_in_pset. * Fix grammar in incomplete-recording warning: “was recorded” → “were recorded”. * Rename test_normalize_invalid_method_warns to test_normalize_invalid_method_raises to match the new ValueError behaviour.

Fixed:

  • Fix WDFReader(chunks=True) on map files raising

NotImplementedError: Don't yet support nd fancy indexing. * Fix wdfkit CLI entry point in pyproject.toml: wdfkit.app:main → wdfkit.wdfkit_app:main (caused ModuleNotFoundError when the installed command was invoked). * Fix origin.py dtype 16/17 branch to advance the file pointer without discarding the result of an unassigned np.array call. * Chain the original exception in io.py file-not-found error (raise IOError(...) from e instead of from None). * Move ImageFile.LOAD_TRUNCATED_IMAGES = True from module-level in io.py into parse_whtl to remove a hidden process-global PIL side effect on import. * Gate ylst.py debug print(y_values) behind ctx.verbose (was: printed unconditionally for every file with a YLST block).

Removed:

  • Remove wdf/assemble.py and its assemble_data_array helper.

  • Remove internal/ package (contents moved to wdf/).

  • Remove spectral/ package (replaced by spectral.py).

  • Remove top-level _types.py, _parsed.py, _dispatch.py, and

handlers/ (all moved into wdf/). * Remove dead user_filesystem fixture from tests/conftest.py.

0.0.1

Changed:

  • README.rst file updated.

  • Logo added to README.rst file.