[wip] chore: use quote-based code generation#135
[wip] chore: use quote-based code generation#135nyurik wants to merge 11 commits intooxibus:mainfrom
Conversation
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
src/utils.rs
Outdated
| fn field_name(&self) -> Ident { | ||
| sanitize_name(self.get_name(), ToSnakeCase::to_snake_case).ident() | ||
| } | ||
| fn field_name2(&self, prefix: &str, suffix: &str) -> Ident { |
There was a problem hiding this comment.
maybe more descriptive name? field_name_with_affixes or wrap_field_name?
| fn field_name2(&self, prefix: &str, suffix: &str) -> Ident { | ||
| format!( | ||
| "{prefix}{}{suffix}", | ||
| sanitize_name(self.get_name(), ToSnakeCase::to_snake_case) |
There was a problem hiding this comment.
is it possible to use field_name to reduce duplication?
There was a problem hiding this comment.
possibly - i would need to check
| let tmp: String; | ||
| sanitize_name( | ||
| if suffix.is_empty() { | ||
| self.get_name() | ||
| } else { | ||
| tmp = format!("{}{suffix}", self.get_name()); | ||
| &tmp | ||
| }, |
There was a problem hiding this comment.
what about this to get rid of tmp String?
| let tmp: String; | |
| sanitize_name( | |
| if suffix.is_empty() { | |
| self.get_name() | |
| } else { | |
| tmp = format!("{}{suffix}", self.get_name()); | |
| &tmp | |
| }, | |
| sanitize_name( | |
| &if suffix.is_empty() { | |
| self.get_name().to_string() | |
| } else { | |
| format!("{}{suffix}", self.get_name()) | |
| }, | |
| ToShoutySnakeCase::to_shouty_snake_case, | |
| ) |
There was a problem hiding this comment.
tbh, i am not a big fan of optimizing away a single line of code at a cost of memory allocation. It does make sense for complex logic savings, but here... meh
nyurik
left a comment
nyurik
left a comment
There was a problem hiding this comment.
Thanks @trnila for looking at it. I was mostly experimenting at this stage -- this whole PR is a draft just to keep everyone posted of the direction. I keep looking at the https://github.com/yasuo-ozu/template_quote to see if something like that would offer significantly cleaner codegen... It won't be perfect, but may offer some clarity...
| let tmp: String; | ||
| sanitize_name( | ||
| if suffix.is_empty() { | ||
| self.get_name() | ||
| } else { | ||
| tmp = format!("{}{suffix}", self.get_name()); | ||
| &tmp | ||
| }, |
There was a problem hiding this comment.
tbh, i am not a big fan of optimizing away a single line of code at a cost of memory allocation. It does make sense for complex logic savings, but here... meh
| fn field_name2(&self, prefix: &str, suffix: &str) -> Ident { | ||
| format!( | ||
| "{prefix}{}{suffix}", | ||
| sanitize_name(self.get_name(), ToSnakeCase::to_snake_case) |
There was a problem hiding this comment.
possibly - i would need to check
|
Any updates on whether we will use |
|
@trnila |
|
P.s. the problem with quote is that it is too verbose - you have to construct each snippet of code ahead of things, and later add it to the result - whereas in template_quote you can do it inline - making the code shorter and more readable ... or less, depending on how crazy it gets |
2 participants
This pull request refactors the codebase to modernize code generation and streamline identifier handling, while removing legacy code that is no longer needed. The most significant changes are the migration to token-based code generation using
proc-macro2andquote, the introduction of new utilities for identifier management, and the cleanup of obsolete formatting and error-handling code.Migration to token-based code generation:
Write-based code generation inFeatureConfigwith methods that generateTokenStreams usingproc-macro2andquote, enabling more robust and idiomatic macro code generation. (src/feature_config.rs) [1] [2]Cargo.tomlto use explicit versions ofproc-macro2and a newer version ofquote.Identifier and naming utilities:
src/utils.rsfor generating and sanitizing Rust identifiers (Ident) from DBC names, including support for enum names, multiplexed variant names, and conversion traits (ToIdent,Tokens).ValTypeinsrc/signal_type.rsto implementIdentFragment, supporting its use in macro-generated identifiers. [1] [2]Removal of legacy and unused code:
PadAdapterstruct and associated formatting logic, as well as allWrite-based formatting utilities, in favor of token-based code generation. (src/pad.rs)CanErrorenum and itsDisplayimplementation from the includes module, as error handling is now generated directly in the output. (src/includes/errors.rs,src/includes/mod.rs) [1] [2]Testing and snapshot updates:
tests/snapshots.rsto use in-memory code generation and snapshot testing withinsta, reflecting the new code generation approach and ensuring test coverage for generated error types.These changes modernize the code generation pipeline, improve maintainability, and set the stage for further enhancements to the macro and codegen infrastructure.