Expand description
Parallel normalisation of raw records into fixed-width rows of canonical residues.
Every sequence byte is classified through the alphabet’s ByteClassTable: residues become
their upper-case canonical letter (T/U folded), every other column becomes 0 (it excludes the
k-mers that contain it), whitespace is skipped (D37), a NUL byte is a structural error (D38)
and unknown characters follow the validation mode (D3). Each record must have exactly as many
alignment columns as the first (D36). Records are processed in parallel chunks and the results
merged in input order, so statistics, examples and the first error are deterministic.
Structs§
- Normalised
Batch - The result of normalising one batch.
- Normaliser
- Normalises batches for one alphabet, alignment width and header format.
- Sequence
Stats - Character and whitespace statistics of sequence lines, mergeable in input order.
Functions§
- count_
columns - Number of alignment columns in a sequence: every byte except line feeds and whitespace.
- describe_
byte - Printable description of a byte for messages:
'J' (0x4A)or0xC3.