Re: [pbs-devel] [PATCH proxmox v3] router: cli: added `ask_for_confirmation` helper

[Wolfgang Bumiller <w.bumiller@proxmox.com>]

On Fri, Sep 08, 2023 at 01:44:09PM +0200, Gabriel Goller wrote:
> Added `ask_for_confirmation` helper that outputs a prompt and
> lets the user confirm or deny it. Also added localization using
> `libc::nl_langinfo()` to match/show a `yes` or `no` in the correct
> language.
> Implemented to close #4763.
> 
> Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
> ---
> 
> update v3:
>  - added localization using `libc::nl_langinfo`
> 
> note: skipped to version 3, so that the whole patch series has 
>       the same version number.
> 
>  proxmox-router/Cargo.toml     |  1 +
>  proxmox-router/src/cli/mod.rs | 34 +++++++++++++++++++++++++++++++++-
>  2 files changed, 34 insertions(+), 1 deletion(-)
> 
> diff --git a/proxmox-router/Cargo.toml b/proxmox-router/Cargo.toml
> index 1c08ce2..e0d67d0 100644
> --- a/proxmox-router/Cargo.toml
> +++ b/proxmox-router/Cargo.toml
> @@ -19,6 +19,7 @@ percent-encoding.workspace = true
>  serde_json.workspace = true
>  serde.workspace = true
>  unicode-width ="0.1.8"
> +regex.workspace = true
>  
>  # cli:
>  tokio = { workspace = true, features = [], optional = true }
> diff --git a/proxmox-router/src/cli/mod.rs b/proxmox-router/src/cli/mod.rs
> index 208df4a..1dcfaf0 100644
> --- a/proxmox-router/src/cli/mod.rs
> +++ b/proxmox-router/src/cli/mod.rs
> @@ -12,7 +12,10 @@
>  //! - Ability to create interactive commands (using ``rustyline``)
>  //! - Supports complex/nested commands
>  
> -use std::collections::HashMap;
> +use std::{
> +    collections::HashMap,
> +    io::{self, Write},
> +};
>  
>  use crate::ApiMethod;
>  
> @@ -62,6 +65,35 @@ pub fn init_cli_logger(env_var_name: &str, default_log_level: &str) {
>      .init();
>  }
>  
> +pub fn ask_for_confirmation(query: String) -> Result<(), io::Error> {

IMO a generic helper should take the default as parameter (to decide on
the upper/lower-casing, see below), and also return a boolean for the
result. The Err case should be exclusively for actual errors (from
`.read_line()`).

> +    use std::ffi::CStr;
> +    let yes_expr = unsafe { CStr::from_ptr(libc::nl_langinfo(libc::YESEXPR)) }
> +        .to_str()
> +        .unwrap_or("^[yY]$");
> +    let no_expr = unsafe { CStr::from_ptr(libc::nl_langinfo(libc::NOEXPR)) }
> +        .to_str()
> +        .unwrap_or("^[nN]$");
> +
> +    let n = no_expr.chars().find(|c| c.is_uppercase()).unwrap_or('N');
> +    let y = yes_expr.chars().find(|c| c.is_uppercase()).unwrap_or('Y');
> +    print!("{query} [{y}/{n}]: ");

Usually, the uppercase letter is used for the default for when typing
nothing, while the other one is lowercase.

Now I was also wondering what we should do if no upper/lowercase letter
exists, but then again, it seems yY and nN are pretty much always
included. I checked at least de_DE, ja_JP and zh_CN next to en_US, and
I'm fairly certain we can rely on that, since I'm assuming this is
exactly what `rpmatch(3)` does, which explicitly states that `y` and `n`
work regardless of locale.

> +
> +    use regex::Regex;
> +    // .unwrap() is okay, because we should always get a correct regex (see `.unwrap_or()` calls above)
> +    let yes_regex: Regex = Regex::new(yes_expr).unwrap();

It might be worth caching this in a static `OnceCell`.

> +
> +    io::stdout().flush()?;
> +    let stdin = io::stdin();
> +    let mut line = String::new();
> +    stdin.read_line(&mut line)?;
> +
> +    if yes_regex.is_match(line.trim()) {
> +        Ok(())
> +    } else {
> +        Err(io::Error::new(io::ErrorKind::Other, "Aborted"))
> +    }
> +}
> +
>  /// Define a simple CLI command.
>  pub struct CliCommand {
>      /// The Schema definition.
> -- 
> 2.39.2