Today work was centered around reading through the parsing chapter in the Appel book on a compiler implementation in ML, and on further advancing work on the ctest extension.
I started off by tweaking a bit the operator expression in the Typst compiler, and starting to read through chapter 3 of the Modern compiler implementation in ML book. I wanted to understand some more LALR(1) grammars, and possibly get some tips on both solving reduce/reduce conflicts and designing a grammar that is primarily left-recursive.
I have almost finished the chapter, and have learnt about the exact opposite to achieving left-recursion, which is great because reversing those steps seems feasible. It doesn’t seem to provide much insight into solving conflicts, as those don’t seem to get a lot of treatment from what I’ve read.
I have reached the SLR parser, but have not yet read that subsection.
In terms of rust-lang/libc work, today I opened a draft PR with the work I’ve put into ctest for the last couple of days. I also CCed some maintainers and connoisseurs of ctest to get their stance on whether I should let the users handle item resolution conflicts or otherwise generate separate tests for each module.
I have also worked on fixing up yesterday’s implementation of filtering, as I had missed one major thing. I was completely missing the fact that filtering items in submodules requires keeping track of the parent path of modules to the item until reaching the item.
That needed me refactoring the code that parses that information from syn’s visitor interface, such that instead of only catching the item identifier (i.e. the last element of the punctuated path,) we build a path up from the current module (saved now as state on the visitor type) and any future modules until we reach an item (where we extract the path gathered thus far.)
None.
Today I did not get any work done with respect to the C side of tests. Either way, what I did today was a pre-requisite to that, so there’s that. Tomorrow, I expect to have finished working through the filtering logic, such that I can actually yield paths that are syntactically identical to those parsed into syn data structures. Currently, they’re losslessly convertible in a roundtrip, but they don’t preserve the look users would expect when calling into TestGenerator’s public API.
In terms of compiler work, I believe I will be done with the rest of the chapter by tomorrow, and will then continue working on the parser and lexer. There’s not much more to comment here, except that before making any progress on the former two, I will have to adapt some token sequences that I previously considered expressions.