LZMA custom filter error

[rabernat]

In zarr-developers/zarr-python#435, I think I uncovered a bug in the LZMA compressor.

Minimal, reproducible code sample, a copy-pastable example if possible

import numpy as np
import lzma
from numcodecs import LZMA
filters = [dict(id=lzma.FILTER_DELTA, dist=4),
           dict(id=lzma.FILTER_LZMA2, preset=1)]
compressor = LZMA(filters=filters)
data = np.zeros(4)
encoded = compressor.encode(data)
compressor.decode(encoded)

raises

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-129-0aabeb1b1948> in <module>()
      6 data = np.zeros(4)
      7 encoded = compressor.encode(data)
----> 8 compressor.decode(encoded)

~/.conda/envs/geo_scipy/lib/python3.6/site-packages/numcodecs/lzma.py in decode(self, buf, out)
     63 
     64             # do decompression
---> 65             dec = _lzma.decompress(buf, format=self.format, filters=self.filters)
     66 
     67             # handle destination

~/.conda/envs/geo_scipy/lib/python3.6/lzma.py in decompress(data, format, memlimit, filters)
    330     results = []
    331     while True:
--> 332         decomp = LZMADecompressor(format, memlimit, filters)
    333         try:
    334             res = decomp.decompress(data)

ValueError: Cannot specify filters except with FORMAT_RAW

Problem description

This seems like a bug. If the options I specified for the compressor are somehow invalid, we should raise an error when the compressor is created, rather than encoding data that can't be decoded.

Version and installation information

Please provide the following:

• Value of numcodecs.__version__: 0.6.3
• Version of Python interpreter: 3.6
• Operating system (Linux/Windows/Mac): Mac
• How NumCodecs was installed (e.g., "using pip into virtual environment", or "using conda"): conda-forge

Also, if you think it might be relevant, please provide the output from pip list or
conda list depending on which was used to install NumCodecs.

[rabernat]

Seems like it can be fixed by just adding format=lzma.FORMAT_RAW to the LZMA constructor.

[jakirkham]

Yeah, I think this limitation comes from Python itself.

We could do that. Alternatively we could construct LZMACompressor and LZMADecompressor objects in the constructor. That should raise these errors as well.