Introduce typed holes; improve disconnect support in human-readable encoding #169

apoelstra · 2023-08-14T18:10:13Z

This introduces the notion of "typed holes" which can be used as placeholders in the text encoding of Simplicity. It uses these holes for disconnect, so that at commitment time users can set their right children to placeholders. The old behavior, which allowed putting a complete expression into disconnect that wouldn't be committed to, was dangerous.

simpcli/example_programs/fixpoint.simpl

uncomputable · 2023-08-15T12:38:09Z

src/human_encoding/mod.rs

+                let cmr_str = /*if let node::Inner::Witness(..) = node.inner() {
                    String::new()
-                } else {
-                    format!("-- cmr {:.8}...", node.cmr())
-                };
+                } else { */
+                    format!("-- cmr {}...", node.cmr())
+                    + &format!("  -- imr {}...", node.imr().as_ref().map(ToString::to_string).unwrap_or("???".to_string()))
+//                };
+;


04f8ffa: We probably want to remove the special case for witness entirely.

Oops, this code just shouldn't be there. Or we should step back and consider whether we want to add CMRs, IMRs, or both (and how). Right now it's an ad-hoc choice based on what I currently needed to generate blog posts.

I think we should print the CMR for all nodes since it exists. As for the IMR, the ??? string might be confusing, because it looks like a bug in our software. I think we should remove it. Alternatively we could print something like (malleable) for witness IMRs.

I think malleable is misleading. How about [undetermined]? Or even [undetermined at commitment time] since we've got a lot of room for characters here.

simpcli/example_programs/hidden.simpl

simpcli/example_programs/twowit.simpl

simpcli/example_programs/bip340.simpl

simpcli/example_programs/disconnect.simpl

uncomputable

04f8ffa: Looks good. Unfortunately most of the example programs don't yet compile. In particular bip340.simpl causes errors on the latest commit because the holes are not filled somehow.

apoelstra · 2023-08-20T14:34:57Z

Rebased and renamed last commit. No other changes. I will fix the example programs tomorrow.

Right now if you use holes anywhere you'll get a parse error complaining about it, but the parse error will give the type of the hole.

… right child

apoelstra · 2023-08-21T20:30:50Z

FIxed the example programs. Deleted the hidden and disconnect ones, since those test bugs that are no longer architecturally possible.

uncomputable

7494aab: The example programs work now. Let's resolve the printing of CMRs and IMRs, and then this PR is ready to be merged.

…add more spacing

apoelstra · 2023-08-23T21:56:59Z

Ok, added commit which tweaks the printing output.

sanket1729 · 2023-08-24T01:29:43Z

I don't know what exactly is the issue. But as of 5465612
the following program does not rountrip. It can be disassembled. But cannot be assembled again.

sanket1729:~/rust-simplicity$ ./target/debug/simpcli disassemble 4MmyAAAACgRswIQKEGMTtoNQNpVbfx0Z7yNAFvcHOnVjPKpYTxVfJS0O0xfn5YYSBrFWbcQRgECZn3AgG8A= > ./simpcli/example_programs/test.simpl
sanket1729:~/rust-simplicity$ ./target/debug/simpcli assemble ./simpcli/example_programs/test.simpl 
Expression `main` not found.

uncomputable

5465612: The formatting looks good.

I could reproduce @sanket1729's error on my end. The disassembly output has a main, but the assembler doesn't find it.

uncomputable

5465612 ACK. Sanket's bug is already present on master.

uncomputable · 2023-08-28T17:45:33Z

I opened #173