Diet of WORMs #143
Diet of WORMs #143
Conversation
This renames all State* FSM types to Fn* - StateLoaded -> FnReady - StateBuilt -> FnLoad - StateValidated -> FnBuild - StateProvisioned -> FnVerify - StateGenerated -> FnCrating This describes the state machine in present tense. It previously used the past tense, which proved confusing.
This moves all the logic back to the start, as the lib.rs is generated in a single step so that it can be written only once.
This allows not losing git history via squashing the commit that moves files along with this.
|
Probably the only thing is I feel all the As an aside, I've never liked how some tests work. Comparing generated v/s expected rust code using syn/quote causes us to write that generated code multiple times. And it make reviewing a diff like this one incredibly difficult. Doesn't have to be part of this PR, but is there something we can do about this? |
|
Yes, I want to do the Yeah, I made the tests use a more parametric form (partially generated, checking only the expected outline) to minimize that. I feel there's a lot of value in checking the overall outline against the fixture, and not as much in checking that every line is exactly the same, so I'm hoping the new form will tend to resist breaking on small changes. |
Okay. I'll approve this now knowing that'll follow.
I'm not sure there is. These code checking tests only validate we typed the same sequence of characters in different parts of our code. They don't ensure the character sequences are correct as it relates to plrust's promises of "safety" and "trust". I don't have a better idea at the moment, so just take that for what it's worth right now. |
This allows PL/Rust to feed on a filesystem that has been configured to only allow "write-once read-many" or WORM. It fixes #139 as one might imagine.
The unusual path remapping hack is only for this PR, I intend to fix it in the next. For various reasons I don't want to chance losing the git history for these files (yet).