Francesc Alted

Yep, the `wheretrue()` method is only valid for unidimensional arrays. This has been documented now (rev 5b949bf). Not closing as a remind for a possible implementation in the future.

Hi Anders. In case you want to give this a go, I'd start with adding a new query mode (similar to [`wheretrue_mode`](https://github.com/Blosc/bcolz/blob/5b949bfab9837137deec1657b80c4c9650de21c1/carray/carrayExtension.pyx#L1994)), and then implementing multidimensional queries in the [`__next__`](https://github.com/Blosc/bcolz/blob/5b949bfab9837137deec1657b80c4c9650de21c1/carray/carrayExtension.pyx#L2040)...

@ankravch In case you just need a multi-dimensional container (i.e not a table-like interface), you might want to try [zarr](https://github.com/alimanfoo/zarr). Zarr follows the same principles than bcolz (chunked, compressed containers),...

Nice suggestion.

Yes, that would be a great addition. And yes, using ctable.fromdataframe() would be an easy way to convert to a carray before appending. Converting the dataframe to a numpy recarray...

Yes, forking is not a tested scenario, so I am not surprised to see this crashing. Probably the fix would be using the _ctx functions: https://github.com/Blosc/c-blosc/blob/master/blosc/blosc.h#L184 https://github.com/Blosc/c-blosc/blob/master/blosc/blosc.h#L222 which do not...

Looks good to me. Any performance hints on how much this PR can accelerate things? Could you add some estimates to the `RELEASE_NOTES` files?

Ok, so 2% is not really significant, although your modifications are not too intrusive, so I am still open to accept this PR. Thanks!

Yes, that's intended. `a.nchunks` is the number of chunks that are complete, whereas `iterblocks` iterates over the `remainder` block.

Could you please send an estimate on the kind of speedup this buys for your use case? Thanks!