new feature: provide timeout for each operation

[dentiny]

Feature Description

Hi team, I notice we have timeout for all operations via timeout layer: https://opendal.apache.org/docs/rust/opendal/layers/struct.TimeoutLayer.html

I would like to know whether you experts think it makes sense to provide per-operation timeout feature?
For example,

• list operation usually does range query, and should be expectedly taking more time than point query stat call
• read request could potentially be served with cache and don't go through any data persistent layer, while it's non-avoidable for write operations

For now, I do have workarounds to use multiple operators for different type of operations, I'm just wondering if you think the per-operation timeout makes sense, or if it's already supported somehow?

Problem and Solution

One potential way is to provide a config which contains different timeouts for each operations to override global timeout and io_timeout.

Additional Context

No response

Are you willing to contribute to the development of this feature?

• Yes, I am willing to contribute to the development of this feature.

[dheeraj12347]

Hi @dentiny and OpenDAL maintainers! I'm Dheeraj, and I'm very interested in implementing this as part of my GSoC 2026 preparation. I have experience with C++ and backend systems. I'd love to take this on. Could you assign this to me, or should I start by drafting a small RFC/proposal here?

[dheeraj12347]

"Hi @Xuanwo, I have fixed the typos and the CI is now passing. This PR is ready for your final review. Thank you!"

[dheeraj12347]

Hi @Xuanwo, I have closed PR #7176 to clear the noise. I realized my approach of injecting Layer behavior into the ops level was architecturally incorrect for OpenDAL.

I am now looking for the 'OpenDAL way' to pass per-operation configurations. Should I explore implementing a with_timeout on the Operator that passes the duration through a Context, or is there a plan to add a generic args metadata bag to OpRead/OpWrite? I'd appreciate a pointer to any existing layer that handles per-request overrides.