#bls #threshold #dkg #log-messages #arpa #randcast

arpa-log

An attribute macro for logging function context with MDC

2 releases

0.2.2 Jun 17, 2024
0.2.0 Jun 3, 2024

#2 in #arpa


Used in arpa-node

MIT/Apache

6KB

As an attribute macro, log_function will

  1. Automatically log the name, input and return value of current function at debug level before it returns by trying to recognize return stmt and inserting a debug! stmt.
  2. Set the name of current function as a key in mdc at the beginning of the function and remove it before the function returns.

Note:

  1. Input and return type need to implement Debug.
  2. When dealing with async function, using #![feature(async_fn_in_trait)] is recommended. Also this is compatible with #[async_trait].
  3. To protect secrets, input and return values are ignored by default. You can specify whether to print all values, or a subset of them with semantic literal options. For example:
    use arpa_log::*;
    
    #[log_function("show-input", "except foo bar", "show-return")]
    fn show_subset_of_input_and_return_value(foo: usize, bar: usize, baz: usize) -> usize {
       foo + bar + baz
    }
    
    Then the log should be: {"message":"LogModel { fn_name: "show_subset_of_input_and_return_value", fn_args: ["foo: ignored", "bar: ignored", "baz: 3"], fn_return: "6" }","level":"DEBUG", "target":"show_subset_of_input_and_return_value","mdc":{"fn_name":"show_subset_of_input_and_return_value"}} with test logger.

Note: Logging result can be different with different logger implementation.

Dependencies

~1.5MB
~38K SLoC