quantlab.dataset.raw#
The store build path: datasets converted from raw vendor files into a store they own.
quantlab.dataset.base.BaseDataset is the read interface every dataset
answers; it has no build path. The datasets that convert a vendor’s raw
files (the raw tier, the files as downloaded) into a Zarr store of their
own derive from RawDataset here, or from RawMarketDataset for price
bars, and inherit the whole build path:
from_raw_dataandsaveconvert the whole configured range at once and write it;from_raw_data_chunkedconverts one time window at a time on a pinned symbol axis (every symbol of the range, worked out once), appending each window and recording it in a resume ledger (<store>.chunks.json), so an interrupted run resumes; a symbol the store lacks (a new listing) is refused, widened in with NaN history, or rebuilt in from the raw files, with the original store set aside and put back if a rebuild fails;updateisfrom_raw_data_chunkedwith the new-listing strategy chosen from raw-tier evidence.
A subclass implements _raw_data_to_xr (and, for RawMarketDataset,
_raw_data_to_xr_window and _to_kunquant); hooks such as
_raw_axes_in_range, _added_symbols_with_raw_history,
_continue_store, _widen_fill_values and _clean let it read its raw
files more cheaply or continue what is stored. IndexConstituentDataset is
the raw dataset of a point-in-time index membership built from intervals.
A view (a roster cut, a merge, a mapping) or a panel held in memory is not a raw dataset and has none of these methods.
Classes
Summary of one chunked conversion from raw files to a Zarr store. |
|
Daily point-in-time index-membership panel built from intervals. |
|
A dataset converted from raw vendor files into a store it owns. |
|
Market price bars converted from raw vendor files into a store they own. |