Plutarch.Extra.List and Plutarch.Extra.Ord
#580
Conversation
|
Test will be added shortly. Added. |
SeungheonOh
force-pushed
the
seungheonoh/listextra
branch
from
20dfafd
to
ac3413e
Compare
| argument returns `PJust`, the value is used in the result. | ||
| -} | ||
| pmapMaybe :: | ||
| forall (ell :: PType -> PType) (b :: PType) (a :: PType) (s :: S). |
There was a problem hiding this comment.
Like the capital letter of the alphabet. Calling to mind 'List'.
| _ -> Nothing | ||
|
|
||
| instance PlutusType POrdering where | ||
| type PInner POrdering = PInteger |
There was a problem hiding this comment.
Should this use SocttPlut?
There was a problem hiding this comment.
Possibly. I did the first thing that could possibly work, which seemed to perform OK, and actually allowed some optimizations (which I have measurements for).
There was a problem hiding this comment.
This is less efficient to match on and not really cheaper to construct. Scott-encoding is probably better.
| -- | ||
| -- We perform the layer parallelism sequentially; it doesn't affect the | ||
| -- semantics, it just makes us sad because it's slow. | ||
| psort4 :: |
There was a problem hiding this comment.
What's up with all this duplication? Just make a psortN
There was a problem hiding this comment.
It's heavily optimized by @cjay using some sorting network. You wouldn't get that performance without these.
There was a problem hiding this comment.
There is a huge amount of code duplication still. It seems like you can encode the sorting network as a table of numbers since the code is mostly the same. You can err for N more than 4.
There was a problem hiding this comment.
If you mean a table of numbers on the Haskell level, yeah it could probably be done while generating the same code on the script level, but that seems like a pointless exercise to me, unless you want to work with much larger sorting networks.
If you mean a table on the Plutarch-level, that seems useless if you want to sort arbitrary element types. If you can serialize into bytestrings then yeah, you could work with actual indices. But then you'd have to build the intermediate bytestrings. Maybe it would save significant script size on larger sorting networks, but probably with much higher execution cost.
There was a problem hiding this comment.
The former, it's much more succinct than adding a new function every time.
There was a problem hiding this comment.
Also, you can have a generic fallback for n > 4.
|
Do you still want this? |
|
Yes It would be nice to have, but not urgent. |
This PR should allow edits by maintainers now. It seems like I have to use personal fork to allow this due to some permission issue.
@kozross @peter-mlabs