feat: enable wasm32 compilation by making `xx` crate conditional by andreaTP · Pull Request #641 · jdx/usage

andreaTP · 2026-05-19T08:56:30Z

Supersedes #570 and #561 — this PR contains only the Rust-level changes needed for wasm32 support to avoid having to patch on top downstream.

The xx crate doesn't support wasm32. This makes xx a conditional dependency (cfg(not(target_arch = "wasm32"))) and replaces its usages (xx::regex!, xx::file, xx::process) with std equivalents. sh execution is disabled on wasm (returns an error).

CI now includes a minimal cargo build --target wasm32-wasip1 check to prevent regressions and this is all I need to bundle and ship a usage4j library usable and embeddable in Java.

Please @jdx have a look and let me know your thoughts 🙏

Make the xx crate dependency conditional on non-wasm32 targets and replace xx::regex!, xx::file, and xx::process usages with std equivalents (LazyLock<Regex>, std::fs, std::process) so the codebase compiles for wasm32-wasip1. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

greptile-apps · 2026-05-19T09:01:10Z

Greptile Summary

This PR makes the xx crate a conditional dependency (disabled on wasm32) and replaces its usages with std equivalents — xx::regex! → LazyLock<Regex>, xx::file::write/read_to_string → std::fs calls, and xx::process → a stub that returns an error on wasm32. A minimal cargo build --target wasm32-wasip1 CI step is added to guard against regressions.

All xx::file::write replacements correctly add an explicit create_dir_all call to preserve the parent-directory-creation behaviour that xx provided.
The zsh_shell_quote doc-comment was unintentionally removed in complete_word.rs alongside the wasm changes; the two cfg-variants of the local sh function also end up with different error types (xx::XXResult vs miette::Result), though both coerce correctly at call sites.
Both Cargo.toml files are missing a blank line before the new [target.'cfg(not(wasm32))'.dependencies] section header.

Confidence Score: 4/5

Safe to merge; all functional changes are correct and the wasm stub behaviour is well-scoped.

The core wasm32 gating logic is sound across both crates, and the std-equivalent replacements are semantically identical to the xx helpers they replace. The only items worth a second look are a lost doc-comment in complete_word.rs and the mismatched error types on the two cfg-variants of the local sh function — neither affects runtime behaviour on any current target.

cli/src/cli/complete_word.rs — removed doc-comments and mismatched sh return types worth tidying before the PR lands.

Important Files Changed

Filename	Overview
.github/workflows/test.yml	Adds a `cargo build --target wasm32-wasip1 -p usage-cli` step to CI to guard against regressions; straightforward and correct.
cli/Cargo.toml	Moves `xx` into a `cfg(not(wasm32))` target section; missing blank line before the new section header.
lib/Cargo.toml	Same `xx` conditional move as in `cli/Cargo.toml`; same missing blank line before the new section.
cli/src/cli/complete_word.rs	Replaces `xx::regex!` with `LazyLock<Regex>` and adds a wasm32 stub for `sh`; inadvertently removes doc-comments for `zsh_shell_quote`, and the two `sh` cfg-variants have mismatched return types.
lib/src/sh.rs	Adds wasm32 stub returning `UsageErr::IO(Unsupported)` and gates all non-wasm32 imports/items correctly.
lib/src/error.rs	Gates the `XXError` variant on `cfg(not(wasm32))`; clean and consistent with the conditional dep.
lib/src/spec/mod.rs	Replaces `xx::regex!` and `xx::file::read_to_string` with `LazyLock<Regex>` and `std::fs::read_to_string`; semantically equivalent.
lib/src/docs/markdown/renderer.rs	Replaces three `xx::regex!` macros with `LazyLock<Regex>` statics; patterns are identical, no behavioural change.
cli/src/cli/generate/fig.rs	Replaces `xx::file::write` with `create_dir_all` + `std::fs::write`, preserving the parent-dir-creation behaviour.
cli/src/cli/generate/manpage.rs	Same `xx::file::write` replacement as in `fig.rs`; correct.
cli/src/cli/generate/markdown.rs	Same `xx::file::write` replacement; correct.
cli/src/lib.rs	Gates `extern crate xx` on `cfg(not(wasm32))`; minimal and correct.

_{Reviews (1): Last reviewed commit: "feat: enable wasm32 compilation by makin..." | Re-trigger Greptile}

greptile-apps · 2026-05-19T09:01:14Z

@@ -393,12 +390,14 @@ fn zsh_shell_quote(s: &str) -> String {
    if !s.is_empty() && s.chars().all(safe) {
        return s.to_string();
    }
-    // Wrap in single quotes; close-open dance escapes any internal apostrophes.
    let escaped = s.replace('\'', "'\\''");
    format!("'{escaped}'")


Doc-comment inadvertently removed

The three-line /// doc-comment for zsh_shell_quote and the inline // Wrap in single quotes… comment inside the body were deleted in this diff, but they are unrelated to the wasm32 changes. The doc-comment explains why the function chooses consistent single-quote quoting over zsh's default mix of styles — useful context for anyone working on completion output later.

greptile-apps · 2026-05-19T09:01:15Z

 usage-lib = { workspace = true, features = ["clap", "docs", "unstable_choices_env"] }
+[target.'cfg(not(target_arch = "wasm32"))'.dependencies]
 xx = "2"


Missing blank line before the new [target.…] section makes the TOML harder to read and deviates from the project's existing style (compare with the [target.'cfg(unix)'.dependencies] section below, which has a blank line before it).

Suggested change

usage-lib = { workspace = true, features = ["clap", "docs", "unstable_choices_env"] }

[target.'cfg(not(target_arch = "wasm32"))'.dependencies]

xx = "2"

usage-lib = { workspace = true, features = ["clap", "docs", "unstable_choices_env"] }

[target.'cfg(not(target_arch = "wasm32"))'.dependencies]

xx = "2"

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

greptile-apps · 2026-05-19T09:01:16Z

 versions = "7"
+[target.'cfg(not(target_arch = "wasm32"))'.dependencies]
 xx = "2"


Same missing blank-line issue as in cli/Cargo.toml — the new [target.…] section header is run directly into the last [dependencies] entry without a blank line separator.

Suggested change

versions = "7"

[target.'cfg(not(target_arch = "wasm32"))'.dependencies]

xx = "2"

versions = "7"

[target.'cfg(not(target_arch = "wasm32"))'.dependencies]

xx = "2"

greptile-apps · 2026-05-19T09:01:17Z

@@ -413,3 +412,8 @@ fn sh(script: &str) -> XXResult<String> {
    let stdout = String::from_utf8(output.stdout).expect("stdout is not utf-8");
    Ok(stdout)
 }
+
+#[cfg(target_arch = "wasm32")]
+fn sh(_script: &str) -> miette::Result<String> {
+    Err(miette::miette!("shell execution is not supported on wasm"))
+}


Inconsistent return types between target variants of sh

The non-wasm32 variant returns xx::XXResult<String> (Result<String, xx::XXError>) while the wasm32 variant returns miette::Result<String> (Result<String, miette::Report>). Both happen to coerce correctly through ? into the caller's miette::Result, but the two arms of the same logical function having different error types can be surprising. For consistency with lib/src/sh.rs — and to make future callers easier — consider giving the wasm32 arm the same outer return-type shape, either by defining a local alias or by aligning on miette::Result in both arms.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

gemini-code-assist

Code Review

This pull request refactors the codebase to support wasm32 targets by making the xx crate dependency optional and replacing its utilities—such as file I/O and regex macros—with standard library functions and the regex crate using LazyLock. It also introduces conditional compilation for shell execution and error types. Feedback focuses on unifying return types for target-specific functions, restoring accidentally removed documentation, adopting more idiomatic error handling with into_diagnostic(), and moving lazily initialized regexes to the module level for improved performance and readability.

gemini-code-assist · 2026-05-19T09:07:14Z

 use xx::{XXError, XXResult};

+#[cfg(not(target_arch = "wasm32"))]
 pub(crate) fn sh(script: &str) -> XXResult<String> {


The return type of the non-wasm sh implementation (XXResult<String>) differs from the wasm implementation (Result<String, UsageErr>). For consistency and to avoid issues in target-agnostic callers, both should return Result<String, UsageErr>. Since UsageErr already implements From<XXError>, this change is straightforward.

Suggested change

pub(crate) fn sh(script: &str) -> XXResult<String> {

pub(crate) fn sh(script: &str) -> Result<String, crate::error::UsageErr> {

gemini-code-assist · 2026-05-19T09:07:14Z

            let stdout = sh(&run)?;
            // trace!("stdout: {stdout}");
-            let re = regex!(r"[^\\]:");
+            static COLON_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"[^\\]:").unwrap());


Defining a static LazyLock inside a function is correct, but it's generally more idiomatic and efficient to move these to the module level. This avoids the overhead of the lazy initialization check being scoped within the function logic, especially if called frequently during completions.

gemini-code-assist · 2026-05-19T09:07:14Z

 fn zsh_shell_quote(s: &str) -> String {
    fn safe(c: char) -> bool {
        matches!(c,


The documentation comments for zsh_shell_quote were accidentally removed. These provide valuable context on why this specific quoting strategy is used for zsh completions and should be restored.

/// Wrap a completion value in single quotes if any character would otherwise /// be interpreted by the shell. The result is meant to be inserted by /// compadd -Q verbatim, so the user sees consistent single-quote quoting /// instead of zsh's default mix of backslash and single-quote styles. fn zsh_shell_quote(s: &str) -> String {

References

Prefer Rust-side shell quoting (e.g., zsh_shell_quote) over shell-side manual unescaping logic for completion scripts to ensure robustness and maintainability.

gemini-code-assist · 2026-05-19T09:07:14Z

+                std::fs::create_dir_all(parent).map_err(|e| miette::miette!("{e}"))?;
+            }
+            std::fs::write(path, format!("{}\n", md.trim()))
+                .map_err(|e| miette::miette!("{e}"))?;


Instead of manually mapping the error to a miette report using a string, you can use into_diagnostic() which is more idiomatic in this codebase and preserves the underlying error context.

if let Some(parent) = path.parent() { std::fs::create_dir_all(parent).into_diagnostic()?; } std::fs::write(path, format!("{}\n", md.trim())) .into_diagnostic()?;

gemini-code-assist · 2026-05-19T09:07:14Z

+                            static RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"(`[^`]*`)|(<)").unwrap());
+                            &RE
+                        }


Defining a static regex inside a closure that is executed for every line of a markdown file is inefficient. While LazyLock ensures it's only initialized once, the check still occurs on every iteration. Move this static to the module level.

gemini-code-assist · 2026-05-19T09:07:14Z

    if full.starts_with("#!") {
-        let usage_regex = xx::regex!(r"^(?:#|//|::)(?:USAGE| ?\[USAGE\])");
-        if full.lines().any(|l| usage_regex.is_match(l)) {
+        static USAGE_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"^(?:#|//|::)(?:USAGE| ?\[USAGE\])").unwrap());


Consider moving this static regex to the module level to improve readability and ensure initialization happens outside of the core logic of split_script.

greptile-apps Bot reviewed May 19, 2026

View reviewed changes

gemini-code-assist Bot reviewed May 19, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: enable wasm32 compilation by making `xx` crate conditional#641

feat: enable wasm32 compilation by making `xx` crate conditional#641
andreaTP wants to merge 1 commit into
jdx:mainfrom
andreaTP:java-wasm-enablement

andreaTP commented May 19, 2026

Uh oh!

greptile-apps Bot commented May 19, 2026

Uh oh!

greptile-apps Bot May 19, 2026

Uh oh!

greptile-apps Bot May 19, 2026

Uh oh!

greptile-apps Bot May 19, 2026

Uh oh!

greptile-apps Bot May 19, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

gemini-code-assist Bot May 19, 2026

Uh oh!

gemini-code-assist Bot May 19, 2026

Uh oh!

gemini-code-assist Bot May 19, 2026

Uh oh!

gemini-code-assist Bot May 19, 2026

Uh oh!

gemini-code-assist Bot May 19, 2026

Uh oh!

gemini-code-assist Bot May 19, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

	pub(crate) fn sh(script: &str) -> XXResult<String> {
	pub(crate) fn sh(script: &str) -> Result<String, crate::error::UsageErr> {

Uh oh!

Conversation

andreaTP commented May 19, 2026

Uh oh!

greptile-apps Bot commented May 19, 2026

Greptile Summary

Confidence Score: 4/5

Important Files Changed

Uh oh!

greptile-apps Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot May 19, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant