pgx-utils goes away!
#911
pgx-utils goes away!
#911
Conversation
- Move the `sql_entity_graph` module out of `pgx-utils` and into its own crate. - Move `pgx-utils/src/rewriter.rs` into `pgx-macros/src` and delete the `pgx-utils` crate entirely - Some dead code removal From there, the changes here are largely mechanical to account for the crate changes. A few 3rd-party dependencies have been eliminated here and there as well. `pgx-pg-sys/build.rs` now applies `#[pg_guard]` to bindgen `extern "C"` FFI blocks instead of embedding it directly. In testing this doesn't seem to hurt compilation performance and it removes a dependency in our build chain.
|
As you said yourself, we probably don't want to just rename |
There was a problem hiding this comment.
This is basically good -- admittedly I just skimmed stuff that looked like it's part of the rename/move.
A lot of this is stuff I wanted to do anyway after the JSON patch, which this will likely simplify (although it does mean that patch will need rewriting ⚰️).
There's some other follow-up stuff that I think we'll want -- (this will regress macro error messages slightly, and possibly compile-times for pgx users), but none of that is urgent or needs to block this (I'll try to get to those issues as part of a follow-up soon).
| @@ -150,9 +140,9 @@ impl PgGuardRewriter { | |||
| arg_list.extend(quote! { #ident, }); | |||
| } | |||
| } else { | |||
| eyre::bail!( | |||
| panic!( | |||
There was a problem hiding this comment.
Ideally we'd use syn::Error here, since panic will give a pretty bad error message inside a macro/build script.
It is a rename, but at least it's clear what the code is used for now. It sounds like @thomcc has further ideas that might let us eliminate the crate entirely, which would be great, but I think this is still important just for clarity. The big thing here was being able to move |
|
Fair enough, then! |
- Move the `sql_entity_graph` module out of `pgx-utils` and into its own crate. - Move `pgx-utils/src/rewriter.rs` into `pgx-macros/src` and delete the `pgx-utils` crate entirely - Some dead code removal From there, the changes here are largely mechanical to account for the crate changes. A few 3rd-party dependencies have been eliminated here and there as well. `pgx-pg-sys/build.rs` now applies `#[pg_guard]` to bindgen `extern "C"` FFI blocks instead of automatically rewriting to its output directly. In testing this doesn't seem to hurt compilation performance and it removes a dependency in our build chain. Relates to Issue pgcentralfoundation#655.
#911 introduced using the wrong signature and this UB was missed in review amidst the thousands of lines of changes. Signed-off-by: usamoi <[email protected]>
#911 introduced using the wrong signature and this UB was missed in review amidst the thousands of lines of changes. Signed-off-by: usamoi <[email protected]>
sql_entity_graphmodule out ofpgx-utilsand into its own crate.pgx-utils/src/rewriter.rsintopgx-macros/srcand delete thepgx-utilscrate entirelyFrom there, the changes here are largely mechanical to account for the crate changes. A few 3rd-party dependencies have been eliminated here and there as well.
pgx-pg-sys/build.rsnow applies#[pg_guard]to bindgenextern "C"FFI blocks instead of automatically rewriting to its output directly. In testing this doesn't seem to hurt compilation performance and it removes a dependency in our build chain.Relates to Issue #655.