PLT-8171: Combine PIR, PLC, and UPLC into a single executable

[bezirg]

DON'T LOOK HERE YET

Pre-submit checklist:

• Branch
  • Tests are provided (if possible)
  • Commit sequence broadly makes sense
  • Key commits have useful messages
  • Changelog fragments have been written (if appropriate)
  • Relevant tickets are mentioned in commit messages
  • Formatting, PNG optimization, etc. are updated
• PR
  • (For external contributions) Corresponding issue exists and is linked in the description
  • Targeting master unless this is a cherry-pick backport
  • Self-reviewed the diff
  • Useful pull request description
  • Reviewer requested

[zliu41]

This singleton-based approach looks quite sensible. With a small amount of work this should be mergeable.

• Replace undefined with either something that works, or at least error "this is not yet implemented".
• Fix the incomplete pattern matches.
• Add a page to the ReadTheDocs site that explains how to use it, with a few examples.

[zliu41]

There should be the ability to specify a list of particular optimizations to run, and the order to run them. Can be done separately.

[kwxm]

There should be the ability to specify a list of particular optimizations to run

See PLC-1046 and PLC-1047. I did look at these a while ago but it was a bit tricky because (I think) not all of the optimisations worked on exactly the same type of AST. Some work may be required on the way the optimisations are expressed internally to make it easier to rearrange them before we can expose a way for users to configure the pipeline.

[zliu41]

It should be straightforward to allow output ann type to be the same as the input ann type, isn't it?

[bezirg]

The Provenance gets in the way for this case. We can definitely find a way to make this work, but I left it out until Provenance is fixed (or removed)

[zliu41]

Looks reasonable overall, except for undefined and the compile errors.

[zliu41]

Suggested change

	- An experimental "plutus" tool that combines PIR,PLC, and UPLC compiling&manipulation into one executable file
	- An experimental "plutus" tool that combines PIR, PLC, and UPLC compiling, manipulation and evaluation into one executable file

[effectfully]

Didn't review some of the details.

Looks OK to me, you could reduce the amount of boilerplate, but it doesn't really matter.

[effectfully]

An error. You can make it a TypeError instead of you want.

[bezirg]

Yes, there is plan to move away from error and fail, and go completely to pure errors. But i leave it for another PR

[effectfully]

Normally with singletons stuff those two would be constraints, 'cause the compiler can figure them out for you, given the right instances.

[effectfully]

On a second thought, I'd just make this whole function a type class, so that all the various combinations are instances. Not important though.

[effectfully]

Do we need this comment for something?

[bezirg]

Not really, it was for you, so you don't have to review the following code :D

[kwxm]

I don't fully understand this because of the singletons etc. I have a few general comments though.

• It looks as if a lot of the complexity comes from the fact that apply takes multiple arguments. Maybe we could just get rid of it and then (I think) every "command" would have at most one input and one output: that might let us go back to applicative parsing. If we still want the apply functionality it might be easier to have a separate executable for it. I did need this in the past when I had unapplied validators and I wanted to supply them with data arguments, but I haven't used it recently.

• I like the idea of the smart suffixes, but they're almost not smart enough because we sometimes have to specify the name type and the annotation type. It would be unbearable to have to have files called fib.uplc.flat.NamedDeBruijn.unitAnnotations though. I had an idea of having some default format/name/annotation combination for each language and then letting the user override it with a sort of fake filename that says what those could be, like you have a file called fib.flat but the parameter you pass to plutus is something like fib.flat@NamedDeBruijn@unitAnnotations. Maybe that's no better than -xand friends though.

• I sometimes have a file and find myself trying to work out what's in it by running uplc with different combinations of options, which is pretty annoying. The best way to deal with this would probably have been to have some wrapper round the file containing metadata that says what's in it, but it's probably too late to do that. Maybe we could write a program that tries to read the input using different combinations until it succeeds, or even have a --auto option for plutus that does that for you. I'm not sure if that would work though: some failures may be untrappable.

• Something else that I've forgotten!

[kwxm]

Maybe this is incomplete, but I think it's ignoring the annotation type. I can't get it to read any of our validation scripts, for example.

[kwxm]

Oh wait. I can get it to work by typing

cabal run plutus -- auction_1-3.uplc.flat -a Unit --stdout

but not

cabal run plutus -- auction_1-3.uplc.flat -a Unit --stdout

which gives

TooMuchSpace (0x00000042005f9514,S {currPtr = 0x00000042005e5013, usedBits = 5})

presumably because I'm only telling it to use unit annotations after I've told it which file I want to read. But then what happens to the annotation type? Is that applied to the output?

Also, it looks as if the file really does have to have extension .uplc.flat. If I just call it auction_1-3.flat then I get

$ cabal run plutus -- -a Unit auction_1-3.flat  --stdout
Error from the PLC compiler:
test:1:1:
  |
1 | 33232"223"32"2232"32"3333""""3"332""33""32"3"3"3"32"3"3"32"3"3"3"3"222222223"222222222222222222P5PW"""03003PS503!5sS5s�!Pd$������������sUs��(�:L&`@��&ƮL�]     ��PUs�"n����b222222222222222222222233W4f��UΨ
  | ^
unexpected start of heading
expecting '('

so I think it's trying to read textual UPLC. The Read instance for FileType says that .flat is the same as .uplc.flat though, so I'm confused.

[bezirg]

Y the -x is supposed to be applied to everything after (including the output). If -o or --stdout is missing, then the last -x will be in effect

[bezirg]

What is the expected type of these auctions? Is it Flat (Program ... DeBruijn .... () ?

[bezirg]

I kind of hate annotations, mostly because of Provenance