[pbs-devel] [PATCH proxmox-backup] fix #4301: correctly pass rate limit parameters to API

[Wolfgang Bumiller]

> On 10/20/2022 4:11 PM Stefan Hanreich <s.hanreich at proxmox.com> wrote:
> 
>  
> On 10/20/22 16:06, Wolfgang Bumiller wrote:
> > On Thu, Oct 20, 2022 at 03:50:59PM +0200, Stefan Hanreich wrote:
> >>
> >>
> >> On 10/20/22 15:44, Wolfgang Bumiller wrote:
> >>> On Thu, Oct 20, 2022 at 02:59:46PM +0200, Stefan Hanreich wrote:
> >>>>
> >>>>
> >>>> On 10/20/22 14:48, Wolfgang Bumiller wrote:
> >>>>> On Thu, Oct 20, 2022 at 01:37:31PM +0200, Stefan Hanreich wrote:
> >>>>>> With the old code the rate limit parameters got passed in their own
> >>>>>> dictionary under the limit key, but the API expects the rate-limit
> >>>>>> settings as top-level keys. This commit correctly sets the rate-limit
> >>>>>> parameters so the API actually uses them.
> >>>>>>
> >>>>>> Signed-off-by: Stefan Hanreich <s.hanreich at proxmox.com>
> >>>>>> ---
> >>>>>>     src/bin/proxmox-backup-manager.rs | 14 ++++++++++++--
> >>>>>>     1 file changed, 12 insertions(+), 2 deletions(-)
> >>>>>>
> >>>>>> diff --git a/src/bin/proxmox-backup-manager.rs b/src/bin/proxmox-backup-manager.rs
> >>>>>> index 58e7e33a..cdd1037d 100644
> >>>>>> --- a/src/bin/proxmox-backup-manager.rs
> >>>>>> +++ b/src/bin/proxmox-backup-manager.rs
> >>>>>> @@ -2,7 +2,7 @@ use std::collections::HashMap;
> >>>>>>     use std::io::{self, Write};
> >>>>>>     use std::str::FromStr;
> >>>>>> -use anyhow::Error;
> >>>>>> +use anyhow::{Error, format_err};
> >>>>>>     use serde_json::{json, Value};
> >>>>>>     use proxmox_router::{cli::*, RpcEnvironment};
> >>>>>> @@ -297,7 +297,6 @@ async fn pull_datastore(
> >>>>>>             "store": store,
> >>>>>>             "remote": remote,
> >>>>>>             "remote-store": remote_store,
> >>>>>> -        "limit": limit,
> >>>>>>         });
> >>>>>>         if remote_ns.is_some() {
> >>>>>> @@ -320,6 +319,17 @@ async fn pull_datastore(
> >>>>>>             args["remove-vanished"] = Value::from(remove_vanished);
> >>>>>>         }
> >>>>>> +    let args_map = args
> >>>>>> +        .as_object_mut()
> >>>>>> +        .ok_or_else(|| format_err!("args is not an Object"))?;
> >>>>>
> >>>>> ^ We create the `args` map only a few lines further up, so it would be
> >>>>> fine to just `.unwrap()` here. And it would be nicer to keep the access
> >>>>> short (iow move the `.as_object_mut()` down to where it's used for `.append()`)
> >>>>>
> >>>>
> >>>> Can replace it with unwrap(), shouldn't be a problem.
> >>>>
> >>>> The reason why I stored it in a variable was that with a subsequent patch I
> >>>> will also append another map to the args, but I could then just call
> >>>> .as_object_mut() twice, what do you think?
> >>>
> >>> I suppose that's fine. If you already know such things while sending a
> >>> patch without the followups being part of the same series, you can note
> >>> such things after the `---` in the patch mail, that way it won't
> >>> needlessly end up in the commit message, but still be visible to the
> >>> reviewer.
> >>>
> >>> Come to think of it, there are a lot of `args[key] = value;` assignments
> >>> which could probably benefit from already having the `&mut Map` instead
> >>> of the `Value` enum.
> >>>
> >>> Maybe should change it like this
> >>>
> >>>       -    let mut args = json!(...)
> >>>       +    let mut args_value = json(!...);
> >>>       +    let args = args_value.as_object_mut().unwrap();
> >>>
> >>> right where it is created, this way the unwrap will stay even closer to
> >>> the creation, the remaining assignments already see the object type, and
> >>> you can call append, too.
> >>> Finally, the `Some(args)` would need to be changed to `Some(args_value)`
> >>> in the `post()` call.
> >>>
> >>
> >> I will keep that in mind for future patches, thanks for the info!
> >>
> >> The problem is when assigning via args[key] = value on a Map, it actually
> >> panics if the key doesn't exist in the map [1], so I'm afraid this isn't a
> >> possibility unless I'm missing something.
> > 
> > Can be changed to use `.insert()` ;-)
> 
> yes, but I thought that would just be needlessly verbose since all 
> `args[key] = value` would be replaced with some kind of
> 
> ```
> args.insert(
>    key.to_string(),
>    value.into() / json!(value)
> );
> ```
> 
> I can definitely change it to use `insert()` in my subsequent patch if 
> you want me to - just say the word ;)

It's fine.
If the parameter list gets too large I'd probably start thinking about
putting them in a struct type shared between client & server code and
just serialize that once. (That would also get rid of the temporary
`limit` Object.)