nushell/nushell · error

output-type flag

Error message

output-type flag

What it means

Same internal-invariant guard as its siblings, but for the --output-type switch: NodeOutputOptions::from_call expects has_flag("output-type") to always succeed because `query xml`'s signature() declares that switch (query_xml.rs:47-51). has_flag fails only if the engine delivers the flag with a non-boolean value over the plugin protocol. A panic here indicates the plugin binary and the engine's view of the command signature are out of sync, not a bad XPath query or bad input.

Source

Thrown at crates/nu_plugin_query/src/query_xml.rs:265

        Ok(xpath) => xpath.ok_or_else(|| {
            LabeledError::new("invalid xpath query").with_label("the query must not be empty", span)
        }),
        Err(err) => Err(LabeledError::new("invalid xpath query").with_label(err.to_string(), span)),
    }
}

struct NodeOutputOptions {
    string_value: bool,
    type_: bool,
    names: bool,
}

impl NodeOutputOptions {
    fn from_call(call: &EvaluatedCall) -> Self {
        match (
            call.has_flag("output-string-value")
                .expect("output-string-value flag"),
            call.has_flag("output-type").expect("output-type flag"),
            call.has_flag("output-names").expect("output-names flag"),
        ) {
            // no flags - old behavior - single column
            (false, false, false) => NodeOutputOptions {
                string_value: true,
                type_: false,
                names: false,
            },
            (string_value, type_, names) => NodeOutputOptions {
                string_value,
                type_,
                names,
            },
        }
    }
}

#[cfg(test)]

View on GitHub (pinned to 8e03210652)

Solutions

1. Re-register the plugin: `plugin rm query` then `plugin add /path/to/nu_plugin_query` so the engine re-reads the current signature
2. Reinstall the plugin built against the running shell: `cargo install nu_plugin_query --locked` with nushell and plugin from the same release
3. Check that flag name strings in from_call ('output-type') match the switches in signature() character-for-character
4. Propagate the error instead of expecting: map the has_flag Result into LabeledError so callers see a readable error rather than a crashed plugin

Example fix

// before
call.has_flag("output-type").expect("output-type flag"),

// after
call.has_flag("output-type")
    .map_err(|err| LabeledError::new("could not read --output-type").with_inner(err))?,

# verify the running query plugin was built for this shell
plugin list | where name == query | get 0.version
version | get version   # the two should correspond to the same release

When it happens

Trigger: `query xml '...' --output-type` executed with a stale cached plugin registration after a nushell upgrade; nu_plugin_query binary compiled against a different nu-plugin-protocol than the running shell; a fork that renamed/removed the 'output-type' switch in signature() without updating from_call.

Common situations: Mixed-version installs (system nushell + locally built plugins); plugin cache files carried over between nushell versions via a shared config dir; CI images where plugins are cached but nushell is bumped.

Related errors

• output-string-value flag
• output-names flag
• we guarantee that 1 entry is always in a list
• always ok
• we know that we have a table cause of a flag

AI-assisted analysis of nushell/nushell@8e03210652 (2026-08-17). Data as JSON: /api/errors/5ffd8f9373bde456. Report an issue: GitHub.