no `seek()` on `PipedGzipReader`

[cv3d]

It looks like the file-like object returned by xopen() for Gzip compressed files misses the seek() method.

AttributeError: 'PipedGzipReader' object has no attribute 'seek'

This would be very helpful to define def seek(self, offset, whence=0)

[marcelm]

Normal gzip files do not support jumping to a certain position in the decompressed stream, so I believe that gzip.open just decompresses the entire file from the beginning when you attempt to seek to somewhere before the current position.

PipedGzipReader runs gzip (or similar) in a subprocess to do the actual decompression. Seeking forward would be possible by skipping some of the decompressed data that it returns, but seeking to an earlier position would only be possible by re-opening the subprocess and letting it decompress the original file from the beginning. (Some weird effects would be possible if the input file had been replaced or modified in the meantime.) And even that won’t work when the compressed input comes from stdin because the data is already gone.

If we implement this, it would therefore not always work and be inefficient for some use cases. However, since Python’s gzip module does allow seeking, I’m not totally against implementing it.

A first implementation could support only forward seeking.

[marcelm]

For completeness, I want to mention here that I’ve added seek() and tell() in #83, but it’s only for Windows.