We introduced the capability of repeating the shuffle filter (and only the shuffle) in C-Blosc2, but we recently discovered issues when using repeat values that are odd (i.e. not even). This PR removes this capability because:
1) It is buggy
2) It was not documented
3) In our tests, it does not provide any significant advantage in compression ratio
4) It complicates code
As it was not documented (besides than buggy), we don't expect this would break anybody's code.
Also, this PR completes documentation on internal filters (essentially, it explicitly says that filters_meta does not have any effect on shuffle, bitshuffle and delta filters).
We introduced the capability of repeating the shuffle filter (and only the shuffle) in C-Blosc2, but we recently discovered issues when using repeat values that are odd (i.e. not even). This PR removes this capability because:
1) It is buggy 2) It was not documented 3) In our tests, it does not provide any significant advantage in compression ratio 4) It complicates code
As it was not documented (besides than buggy), we don't expect this would break anybody's code.
Also, this PR completes documentation on internal filters (essentially, it explicitly says that
filters_metadoes not have any effect on shuffle, bitshuffle and delta filters).