use BitArray for boolean arrays

[JeffBezanson]

Once #987 is settled, we should add BitArray to Base and use it for bool arrays. The goal is to eliminate the copy&pasted code, and settle the contentious issue of BitArray's type parameter.

[ViralBShah]

The is certainly a good idea. Would be nice to have a performance comparison to ensure that there are no hidden surprises.

[StefanKarpinski]

My performance tests of BitArray NA masks for data frames indicate that BitArrays perform quite nicely and give the expected speedups. They seem quite solid.

[ViralBShah]

That is really awesome.

-viral

On 12-Aug-2012, at 10:07 PM, Stefan Karpinski wrote:

My performance tests of BitArray NA masks for data frames indicate that BitArrays perform quite nicely and give the expected speedups. They seem quite solid.

—
Reply to this email directly or view it on GitHub.

[JeffBezanson]

Sparse bitarrays would be cool too.

[ViralBShah]

Sparse bitarrays can be done by not having the nzval array in SparseMatrixCSC. The entries in rowval can indicate the bits that are set to true, and the rest are assumed to be false.

The entire SparseMatrixCSC implementation needs to be made aware of this, and it should be the most efficient way to do sparse bitarrays.

-viral

On 13-Aug-2012, at 2:56 AM, Jeff Bezanson wrote:

Sparse bitarrays would be cool too.

—
Reply to this email directly or view it on GitHub.

[StefanKarpinski]

This is done by fec3387. Now it's mostly a matter of fine-tuning the behavior and interfaces.