PDL bitfield type

etj has asked for the wisdom of the Perl Monks concerning the following question:

I'm toying with the idea of adding a bitfield type to PDL, so that along the zeroth dimension it's not a whole number of bytes, but a calculated bit read (when it's used as an rvalue) or write (probably needs a function to set that, since PDL can't parse the C code yet to spot an lvalue).

Feedback on whether that would be valuable enough welcome!

Comment on PDL bitfield type - wanted?

Replies are listed 'Best First'.
Re: PDL bitfield type - wanted? by swl (Prior) on May 13, 2024 at 00:46 UTC
Sounds good to me. I assume the plan would be to add many of the functions from modules like Algorithm::BitVector, Bit::Vector and so forth?	[reply]
Re^2: PDL bitfield type - wanted? by etj (Priest) on May 13, 2024 at 16:46 UTC
You assume correctly! If you are prepared to help out (if you are, obviously I'd be very pleased to give lots and lots of help). I think the way forward would be for me to add (and use, to shake out bugs) the type implementation, then these algorithms would be in a separate PDL::BitOps or similar distro. My immediate "itch" was to use a bitmap to implement `seenmap` in the fancy new `contour_polylines`; it annoyed me a bit that while I'd made that be a `byte` (i.e. 8-bit) ndarray, I was only using 2 of the bits in each of those. It's not a vast amount of memory being wasted, but that has an impact on cache usage.	[reply] [d/l] [select]
Re^3: PDL bitfield type - wanted? by swl (Prior) on May 14, 2024 at 09:54 UTC
I can help with testing and debugging. Any initial implementation is of course best done by yourself given familiarity with the innards of PDL.	[reply]
Re^4: PDL bitfield type - wanted? by etj (Priest) on May 14, 2024 at 12:56 UTC
Re^5: PDL bitfield type - wanted? by swl (Prior) on May 14, 2024 at 22:18 UTC