Generate CLI reference for documentation (#5685)

Loosely based on [Cargo's
format](https://github.com/rust-lang/cargo/blob/master/src/doc/src/commands/cargo-build.md)

<img width="896" alt="Screenshot 2024-08-01 at 9 44 03 AM"
src="https://github.com/user-attachments/assets/7c016bb3-2b54-46af-8ea8-ce82e07a0e30">

Future work includes:

- Grouping options
- Enforcing some sort of specific command ordering
- Showing possible values for enums
- Adding "long_about" to commands for more context

crates/uv-cli/Cargo.toml

@@ -29,7 +29,7 @@ uv-warnings = { workspace = true }
 
 anstream = { workspace = true }
 anyhow = { workspace = true }
-clap = { workspace = true, features = ["derive"] }
+clap = { workspace = true, features = ["derive", "string"] }
 clap_complete_command = { workspace = true }
 serde = { workspace = true }
 url = { workspace = true }

crates/uv-dev/Cargo.toml

@@ -23,6 +23,7 @@ pep508_rs = { workspace = true }
 pypi-types = { workspace = true }
 uv-build = { workspace = true }
 uv-cache = { workspace = true, features = ["clap"] }
+uv-cli = { workspace = true }
 uv-client = { workspace = true }
 uv-configuration = { workspace = true }
 uv-dispatch = { workspace = true }
@@ -43,6 +44,7 @@ anyhow = { workspace = true }
 clap = { workspace = true, features = ["derive", "wrap_help"] }
 fs-err = { workspace = true, features = ["tokio"] }
 itertools = { workspace = true }
+markdown = "0.3.0"
 owo-colors = { workspace = true }
 poloto = { version = "19.1.2", optional = true }
 pretty_assertions = { version = "1.4.0" }

crates/uv-dev/src/generate_cli_reference.rs

@@ -0,0 +1,262 @@
+//! Generate a Markdown-compatible reference for the uv command-line interface.
+use std::cmp::max;
+use std::path::PathBuf;
+
+use anstream::println;
+use anyhow::{bail, Result};
+use clap::{Command, CommandFactory};
+use itertools::Itertools;
+use pretty_assertions::StrComparison;
+
+use crate::generate_all::Mode;
+use crate::ROOT_DIR;
+
+use uv_cli::Cli;
+
+#[derive(clap::Args)]
+pub(crate) struct Args {
+    /// Write the generated output to stdout (rather than to `settings.md`).
+    #[arg(long, default_value_t, value_enum)]
+    pub(crate) mode: Mode,
+}
+
+pub(crate) fn main(args: &Args) -> Result<()> {
+    let reference_string = generate();
+    let filename = "cli.md";
+    let reference_path = PathBuf::from(ROOT_DIR)
+        .join("docs")
+        .join("reference")
+        .join(filename);
+
+    match args.mode {
+        Mode::DryRun => {
+            println!("{reference_string}");
+        }
+        Mode::Check => match fs_err::read_to_string(reference_path) {
+            Ok(current) => {
+                if current == reference_string {
+                    println!("Up-to-date: {filename}");
+                } else {
+                    let comparison = StrComparison::new(&current, &reference_string);
+                    bail!("{filename} changed, please run `cargo dev generate-cli-reference`:\n{comparison}");
+                }
+            }
+            Err(err) if err.kind() == std::io::ErrorKind::NotFound => {
+                bail!("{filename} not found, please run `cargo dev generate-cli-reference`");
+            }
+            Err(err) => {
+                bail!("{filename} changed, please run `cargo dev generate-cli-reference`:\n{err}");
+            }
+        },
+        Mode::Write => match fs_err::read_to_string(&reference_path) {
+            Ok(current) => {
+                if current == reference_string {
+                    println!("Up-to-date: {filename}");
+                } else {
+                    println!("Updating: {filename}");
+                    fs_err::write(reference_path, reference_string.as_bytes())?;
+                }
+            }
+            Err(err) if err.kind() == std::io::ErrorKind::NotFound => {
+                println!("Updating: {filename}");
+                fs_err::write(reference_path, reference_string.as_bytes())?;
+            }
+            Err(err) => {
+                bail!("{filename} changed, please run `cargo dev generate-cli-reference`:\n{err}");
+            }
+        },
+    }
+
+    Ok(())
+}
+
+fn generate() -> String {
+    let mut output = String::new();
+
+    let mut uv = Cli::command();
+
+    // It is very important to build the command before beginning inspection or subcommands
+    // will be missing all of the propagated options.
+    uv.build();
+
+    let mut parents = Vec::new();
+
+    output.push_str("# CLI Reference\n\n");
+    generate_command(&mut output, &uv, &mut parents);
+
+    output
+}
+
+fn generate_command<'a>(output: &mut String, command: &'a Command, parents: &mut Vec<&'a Command>) {
+    if command.is_hide_set() {
+        return;
+    }
+
+    // Generate the command header.
+    let name = if parents.is_empty() {
+        command.get_name().to_string()
+    } else {
+        format!(
+            "{} {}",
+            parents.iter().map(|cmd| cmd.get_name()).join(" "),
+            command.get_name()
+        )
+    };
+
+    // Display the top-level `uv` command at the same level as its children
+    let level = max(2, parents.len() + 1);
+    output.push_str(&format!("{} {name}\n\n", "#".repeat(level)));
+
+    // Display the command description.
+    if let Some(about) = command.get_long_about().or_else(|| command.get_about()) {
+        output.push_str(&about.to_string());
+        output.push_str("\n\n");
+    };
+
+    // Display the usage
+    {
+        // This appears to be the simplest way to get rendered usage from Clap,
+        // it is complicated to render it manually. It's annoying that it
+        // requires a mutable reference but it doesn't really matter.
+        let mut command = command.clone();
+        output.push_str("<h3 class=\"cli-reference\">Usage</h3>\n\n");
+        output.push_str(&format!(
+            "```\n{}\n```",
+            command
+                .render_usage()
+                .to_string()
+                .trim_start_matches("Usage: "),
+        ));
+        output.push_str("\n\n");
+    }
+
+    // Display a list of child commands
+    let mut subcommands = command.get_subcommands().peekable();
+    let has_subcommands = subcommands.peek().is_some();
+    if has_subcommands {
+        output.push_str("<h3 class=\"cli-reference\">Commands</h3>\n\n");
+        output.push_str("<dl class=\"cli-reference\">");
+
+        for subcommand in subcommands {
+            if subcommand.is_hide_set() {
+                continue;
+            }
+            let subcommand_name = format!("{name} {}", subcommand.get_name());
+            output.push_str(&format!(
+                "<dt><a href=\"{}\"><code>{subcommand_name}</code></a></dt>",
+                subcommand_name.replace(' ', "-")
+            ));
+            if let Some(about) = subcommand.get_about() {
+                output.push_str(&format!(
+                    "<dd>{}</dd>\n",
+                    markdown::to_html(&about.to_string())
+                ));
+            }
+        }
+
+        output.push_str("</dl>\n\n");
+    }
+
+    // Do not display options for commands with children
+    if !has_subcommands {
+        // Display positional arguments
+        let mut arguments = command
+            .get_positionals()
+            .filter(|arg| !arg.is_hide_set())
+            .peekable();
+
+        if arguments.peek().is_some() {
+            output.push_str("<h3 class=\"cli-reference\">Arguments</h3>\n\n");
+            output.push_str("<dl class=\"cli-reference\">");
+
+            for arg in arguments {
+                output.push_str("<dt>");
+                output.push_str(&format!(
+                    "<code>{}</code>",
+                    arg.get_id().to_string().to_uppercase()
+                ));
+                output.push_str("</dt>");
+                if let Some(help) = arg.get_long_help().or_else(|| arg.get_help()) {
+                    output.push_str("<dd>");
+                    output.push_str(&format!("{}\n", markdown::to_html(&help.to_string())));
+                    output.push_str("</dd>");
+                }
+            }
+
+            output.push_str("</dl>\n\n");
+        }
+
+        // Display options
+        let mut options = command
+            .get_opts()
+            .filter(|arg| !arg.is_hide_set())
+            .sorted_by_key(|opt| opt.get_id())
+            .peekable();
+
+        if options.peek().is_some() {
+            output.push_str("<h3 class=\"cli-reference\">Options</h3>\n\n");
+            output.push_str("<dl class=\"cli-reference\">");
+            for opt in command.get_opts() {
+                if opt.is_hide_set() {
+                    continue;
+                }
+
+                let Some(long) = opt.get_long() else { continue };
+
+                output.push_str("<dt>");
+                output.push_str(&format!("<code>--{long}</code>"));
+                if let Some(short) = opt.get_short() {
+                    output.push_str(&format!(", <code>-{short}</code>"));
+                }
+                if let Some(values) = opt.get_value_names() {
+                    for value in values {
+                        output.push_str(&format!(
+                            " <i>{}</i>",
+                            value.to_lowercase().replace('_', "-")
+                        ));
+                    }
+                }
+                output.push_str("</dt>");
+                if let Some(help) = opt.get_long_help().or_else(|| opt.get_help()) {
+                    output.push_str("<dd>");
+                    output.push_str(&format!("{}\n", markdown::to_html(&help.to_string())));
+                    output.push_str("</dd>");
+                }
+            }
+
+            output.push_str("</dl>");
+        }
+
+        output.push_str("\n\n");
+    }
+
+    parents.push(command);
+
+    // Recurse to all of the subcommands.
+    for subcommand in command.get_subcommands() {
+        generate_command(output, subcommand, parents);
+    }
+
+    parents.pop();
+}
+
+#[cfg(test)]
+mod tests {
+    use std::env;
+
+    use anyhow::Result;
+
+    use crate::generate_all::Mode;
+
+    use super::{main, Args};
+
+    #[test]
+    fn test_generate_cli_reference() -> Result<()> {
+        let mode = if env::var("UV_UPDATE_SCHEMA").as_deref() == Ok("1") {
+            Mode::Write
+        } else {
+            Mode::Check
+        };
+        main(&Args { mode })
+    }
+}

crates/uv-dev/src/main.rs

@@ -20,6 +20,7 @@ use crate::build::{build, BuildArgs};
 use crate::clear_compile::ClearCompileArgs;
 use crate::compile::CompileArgs;
 use crate::generate_all::Args as GenerateAllArgs;
+use crate::generate_cli_reference::Args as GenerateCliReferenceArgs;
 use crate::generate_json_schema::Args as GenerateJsonSchemaArgs;
 use crate::generate_options_reference::Args as GenerateOptionsReferenceArgs;
 #[cfg(feature = "render")]
@@ -46,6 +47,7 @@ mod build;
 mod clear_compile;
 mod compile;
 mod generate_all;
+mod generate_cli_reference;
 mod generate_json_schema;
 mod generate_options_reference;
 mod render_benchmarks;
@@ -69,6 +71,8 @@ enum Cli {
     GenerateJSONSchema(GenerateJsonSchemaArgs),
     /// Generate the options reference for the documentation.
     GenerateOptionsReference(GenerateOptionsReferenceArgs),
+    /// Generate the CLI reference for the documentation.
+    GenerateCliReference(GenerateCliReferenceArgs),
     #[cfg(feature = "render")]
     /// Render the benchmarks.
     RenderBenchmarks(RenderBenchmarksArgs),
@@ -88,6 +92,7 @@ async fn run() -> Result<()> {
         Cli::GenerateAll(args) => generate_all::main(&args)?,
         Cli::GenerateJSONSchema(args) => generate_json_schema::main(&args)?,
         Cli::GenerateOptionsReference(args) => generate_options_reference::main(&args)?,
+        Cli::GenerateCliReference(args) => generate_cli_reference::main(&args)?,
         #[cfg(feature = "render")]
         Cli::RenderBenchmarks(args) => render_benchmarks::render_benchmarks(&args)?,
     }