Completion replace not fully replacing when in macro string literal #11020
Answered
by
pascalkuthe
RoloEdits
asked this question in
Troubleshooting
-
SummaryWhen trying to add the Not sure if its related but #10689 also seems to be an off-by-one, with that pointing to #10279 that was a fix for an off-by-one. Reproduction Steps
completion-replace = true[dependencies]
serde = { version = "1", features = ["derive"]}
chrono ={ version = "0.4", features = ["serde"]}use chrono::serde::ts_milliseconds;
use chrono::{DateTime, Utc};
use serde::Deserialize;
#[derive(Deserialize)]
struct A {
#[serde(with = "tts_milliseconds")]
time: DateTime<Utc>,
}
fn main() {
println!("Hello, world!");
}Helix log~/.cache/helix/helix.logPlatformWIndows Terminal EmulatorAlacritty Installation Methodsource Helix Version24.03-218-g9b7dffbd |
Beta Was this translation helpful? Give feedback.
Answered by
pascalkuthe
Jun 22, 2024
Replies: 1 comment 1 reply
-
|
this loooks like a rust-analyzer bug. Lsp has two modes of returning completion edits: One simple mode where the lsp has to figure out what text to replace and an explicit mode where the lsp specifies the range. RA always uses the explicit mode and all the setting does is switch between the two edits. In this case the edit RA provides is simply off by one so it's an off by one bug in rust-analyzer not in helix. Particularly these completions inside derive macros are new so I am not too surprised. It may also be a RA limitation since sered uses string literals. A better case is you start with a file where the word is partially already written out and enter insertmode (so entirely new completion). use chrono::serde::ts_milliseconds;
use chrono::{DateTime, Utc};
use serde::Deserialize;
#[derive(Deserialize)]
struct A {
#[serde(with = "ts_mill")]
time: DateTime<Utc>,
}
fn main() {
println!("Hello, world!");
}it completes |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Thanks for the explanation, still just getting into LSP, this makes a lot of sense. Originally I had thought that it would return the suggestion text minus the suggested-on text, as casing could be different than the original, for example, and so would take the general approach of forcing the expected outcome by overriding. I'll look at the current issues for RA and make a new one if I can't find one for this. |
Beta Was this translation helpful? Give feedback.
Answer selected by
RoloEdits
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
this loooks like a rust-analyzer bug. Lsp has two modes of returning completion edits: One simple mode where the lsp has to figure out what text to replace and an explicit mode where the lsp specifies the range.
RA always uses the explicit mode and all the setting does is switch between the two edits. In this case the edit RA provides is simply off by one so it's an off by one bug in rust-analyzer not in helix. Particularly these completions inside derive macros are new so I am not too surprised. It may also be a RA limitation since sered uses string literals.
A better case is you start with a file where the word is partially already written out and enter insertmode (so entirely new compl…