# witcher [![license-badge](https://img.shields.io/crates/l/fungus.svg)](https://opensource.org/licenses/MIT) [![build](https://github.com/phR0ze/witcher/workflows/build/badge.svg?branch=main)](https://github.com/phR0ze/witcher/actions) [![codecov](https://codecov.io/gh/phR0ze/witcher/branch/main/graph/badge.svg?token=2GYE6JWAPI)](https://codecov.io/gh/phR0ze/witcher) [![crates.io](https://img.shields.io/crates/v/witcher.svg)](https://crates.io/crates/witcher) [![Minimum rustc](https://img.shields.io/badge/rustc-1.30+-lightgray.svg)](https://github.com/phR0ze/witcher#rustc-requirements) ***Track and put down bugs using simple concise error handling*** Avoid terminating execution randomly in your code with `panics` via `unwrap` and `expect` varients, or laboriously writing custom enum wrapers, or having to work with `Box` types which is messy; instead use `Result` from witcher as the return type and `wrap` errors easily for additional contextual messaging automatically propogated up the stack. witcher implements `std::error::Error` retaining downcasting and chaining. Best of all witcher provides the holy grail: `automatic simplified backtraces`. ![Display debug](docs/images/display-debug.png) ## What you get 1. ***Error handling simplified*** > by providing type matching on errors > by automatically handling conversions > by providing concise and terse user interaction > by providing conditional colored output 2. ***Error handling that tells the full story*** > by implementing `std::error::Error` > by chaining errors together > by providing contextual messaging > by providing tracing from point of origin 3. ***Safety*** > 100% safe code without any use of `unsafe` > Zero low level TraitObject manipulation > Well tested with over 90% code coverage --- ### Quick links * [Usage](#usage) * [Rustc requirments](#rustc-requirements) * [Color](#color) * [Downcasting](#downcasting) * [downcast\_ref](#downcast_ref) * [match\_err](#match_err) * [pass](#pass) * [Chaining](#chaining) * [source](#source) * [Retries](#retries) * [err\_is](#err_is) * [retry\_on](#retry_on) * [Display](#display) * [Normal](#display-normal) * [Alternate](#display-alternate) * [Debug](#display-debug) * [Alternate Debug](#display-alternate-debug) * [Contribute](#contribute) * [Git-Hook](#git-hook) * [License](#license) * [Contribution](#contribution) * [Backlog](#backlog) * [Changelog](#changelog) ## Manifesto Coming from a Golang background, most recently, I fully expected to just import the defacto standard error package in Rust similar to something like Golang's [pkg/errors](https://github.com/pkg/errors) and I'd be off to the races. Instead, as I dug, I found a rich anthropological layered history of a myriad of projects and authors all professing nobal ideals and principles all trying to solve the same issue. Rust's error handling story isn't full featured enough by itself yet. It feels a lot like Golang's before the existance of `pkg/errors`. I found a few projects clearly more used than others and saw the tide turn on once popular packages. Literally weeks of research and testing of numerous different patterns and packages later though I have still yet to find anything as simple and usable as the venerable [pkg/errors](https://github.com/pkg/errors). Thus `witcher` was born. As a side note I moved all my research on low level TraitObject manipulation, when I was going down the rabbit hole, over to [phR0ze/rust-examples](https://github.com/phR0ze/rust-examples) and am happy to say `witcher` is 100% safe code. # Usage Use the `wrap` extension method on `Result` types to wrap the error with additional contextual messaging and automatically chain errors together. `wrap` returns a `Result` so there are fewer symbols and less typing needed. #### Requires rustc >= 1.30 This minimum rustc requirement is driven by the enhancements made to [Rust's `std::error::Error` handling improvements](https://doc.rust-lang.org/std/error/trait.Error.html#method.source) 1. Import witcher in your `Cargo.toml` and keep debug symbols ```toml [dependencies] witcher = "0.1" [profile.release] debug = true ``` 2. Use the witcher prelude ```rust use witcher::prelude::*; ``` 3. Use the `Result` alias as your return type ```rust fn do_something() -> Result<()>; ``` 4. Use the `wrap` extension method on Result to provide context ```rust fn do_something() -> Result<()> { do_external_thing().wrap("Failed to slay beast") } fn do_external_thing() -> std::io::Result<()> { Err(std::io::Error::new(std::io::ErrorKind::Other, "Oh no, we missed!"))? } ``` ## Color Color is automatically controlled by the [gory](https://crates.io/crates/gory) based on `tty` detection. You can disable color manually by setting the `TERM_COLOR` environment variable to something falsy see [gory docs on controlling use]( https://github.com/phR0ze/gory#control-use). ```bash $ TERM_COLOR=0 cargo run -q --example simple ``` ## Downcasting We can match on error types using downcasting or with the `match_err!` macro. #### `downcast_ref` - *access std::error::Error's downcast_ref* ```rust use witcher::prelude::*; // Wrap our internal error with additional context as we move up the stack fn do_something() -> Result<()> { do_external_thing().wrap("Failed to slay beast") } // Function that returns an external error type outside our codebase fn do_external_thing() -> std::io::Result<()> { Err(std::io::Error::new(std::io::ErrorKind::Other, "Oh no, we missed!"))? } fn main() { let err = do_something().unwrap_err(); // Get the last error in the error chain which will be the root cause let root_cause = err.last(); // Match single concrete error type if let Some(e) = root_cause.downcast_ref::() { println!("Root cause is a std::io::Error: {}", e) } else { println!("{}", err) } } ``` Results: ``` $ cargo run -q --example downcast_ref Root cause is std::io::Error: Oh no, we missed! ``` #### `match_err!` - *matches on concrete error typese* ```rust use witcher::prelude::*; fn do_something() -> Result<()> { do_external_thing().wrap("Failed to slay beast") } fn do_external_thing() -> std::io::Result<()> { Err(std::io::Error::new(std::io::ErrorKind::Other, "Oh no, we missed!"))? } fn main() { let err = do_something().unwrap_err(); // Match multiple downcasted cases to handle errors differently match_err!(err.last(), { x: Error => println!("Root cause is witcher::Error: {}", x), x: std::io::Error => println!("Root cause is std::io::Error: {}", x), _ => println!("{}", err) }); } ``` Results: ``` $ cargo run -q --example downcast_match Root cause is std::io::Error: Oh no, we missed! ``` #### `pass` - *pass error through transparently* In some cases it might be nice to be able to use the Witcher common error pattern but treat particular top level errors as a pass through. WARNING: this will not work with the `match_err!` macro which uses the type directly for matching rather than indirectly when working with an `Error` type. ```rust use witcher::prelude::*; fn do_something() -> Result<()> { do_external_thing().pass() } fn do_external_thing() -> std::io::Result<()> { Err(std::io::Error::new(std::io::ErrorKind::Other, "Oh no, we missed!"))? } fn main() { let err = do_something().unwrap_err(); // Since we used `pass` we can match on the error directly match err.downcast_ref::() { Some(err) => println!("Root cause is std::io::Error: {}", err), None => println!("Root cause is witcher::Error: {}", err), } } ``` Results: ``` $ cargo run -q --example pass Root cause is std::io::Error: Oh no, we missed! ``` ## Chaining We can continue to leverage std::error::Error's `source` method for chaining of errors. The first error wrapped will retain its concrete type but errors there after in the chain have lost that information. #### `source` - *std::error::Error's source method is exposed* ```rust use witcher::prelude::*; #[derive(Debug)] struct SuperError { side: SuperErrorSideKick, } impl std::fmt::Display for SuperError { fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { write!(f, "SuperError is here!") } } impl std::error::Error for SuperError { fn source(&self) -> Option<&(dyn std::error::Error + 'static)> { Some(&self.side) } } #[derive(Debug)] struct SuperErrorSideKick; impl std::fmt::Display for SuperErrorSideKick { fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { write!(f, "SuperErrorSideKick is here!") } } impl std::error::Error for SuperErrorSideKick {} fn do_something() -> Result<()> { do_external_thing().wrap("Failed doing super hero work") } fn do_external_thing() -> std::result::Result<(), SuperError> { Err(SuperError {side: SuperErrorSideKick}) } fn main() { if let Err(err) = do_something() { // Traverse the error chain let mut source = Some(err.std()); while let Some(err) = source { match_err!(err, { // Using alternate form of display for `Error` to get just the message x: Error => println!("Found witcher::Error: {:#}", x), x: SuperError => println!("Found SuperError: {}", x), x: SuperErrorSideKick => println!("Found SuperErrorSideKick: {}", x), _ => println!("unknown") }); source = err.source(); } } } ``` Results: ``` $ cargo run -q --example chain Found witcher::Error: Failed doing super hero work Found SuperError: SuperError is here! Found SuperErrorSideKick: SuperErrorSideKick is here! ``` ## Retries We can retry failing code with a few different `Result` extension functions. #### `err_is` - *will return true if an error exists and is the given type* ```rust use witcher::prelude::*; fn retry_on_concreate_error_type_using_err_is() -> Result<()> { let mut retries = 0; let mut result = do_external_thing(); while retries < 3 && result.err_is::() { retries += 1; println!("retrying using err_is #{}", retries); result = do_external_thing(); } result.wrap("Failed while attacking beast") } fn do_external_thing() -> std::io::Result<()> { Err(std::io::Error::new(std::io::ErrorKind::Other, "Oh no, we missed!")) } fn main() { println!("{}", retry_on_concreate_error_type_using_err_is().unwrap_err()); } ``` Results: ``` cargo run -q --example retry_err_is retrying using err_is #1 retrying using err_is #2 retrying using err_is #3 error: witcher::Error: Failed while attacking beast cause: std::io::error::Error: Oh no, we missed! symbol: retry_err_is::retry_on_concreate_error_type_using_err_is at: examples/retry_err_is.rs:11:5 symbol: retry_err_is::main at: examples/retry_err_is.rs:18:22 ``` #### `retry_on` - *is a cleaner simplified way to do a similar thing as our err_is example* ```rust use witcher::prelude::*; fn retry_on_concreate_error_type() -> Result<()> { do_external_thing().retry_on(3, TypeId::of::(), |i| { println!("std::io::Error: retrying! #{}", i); do_external_thing() }).wrap("Failed while attacking beast") } fn do_external_thing() -> std::io::Result<()> { Err(std::io::Error::new(std::io::ErrorKind::Other, "Oh no, we missed!")) } fn main() { println!("{}", retry_on_concreate_error_type().unwrap_err()); } ``` Results: ``` cargo run -q --example retry_on std::io::Error: retrying! #1 std::io::Error: retrying! #2 std::io::Error: retrying! #3 error: witcher::Error: Failed while attacking beast cause: std::io::error::Error: Oh no, we missed! symbol: retry_on::retry_on_concreate_error_type at: examples/retry_on.rs:4:5 symbol: retry_on::main at: examples/retry_on.rs:16:22 ``` #### `retry` - *is similar to `retry_on` but doesn't take the type of error into account* ```rust use witcher::prelude::*; fn retry() -> Result<()> { do_external_thing().retry(3, |i| { println!("std::io::Error: retrying! #{}", i); do_external_thing() }).wrap("Failed while attacking beast") } fn do_external_thing() -> std::io::Result<()> { Err(std::io::Error::new(std::io::ErrorKind::Other, "Oh no, we missed!")) } fn main() { println!("{}", retry().unwrap_err()); } ``` Results: ``` cargo run -q --example retry std::io::Error: retrying! #1 std::io::Error: retrying! #2 std::io::Error: retrying! #3 error: witcher::Error: Failed while attacking beast cause: std::io::error::Error: Oh no, we missed! symbol: retry::retry at: examples/retry.rs:4:5 symbol: retry::main at: examples/retry.rs:16:22 ``` ## Display Witcher's `Error` type implements different functionality for each of the `Display` format options. They follow a level of verbosity in witcher from least information to most i.e. ***`{} {:#} {:?} {:#?}`*** #### `Normal: {}` - *will write out the first error message only* ![Display normal](docs/images/display-normal.png) #### `Alternate: {:#}` - *will write out all error messages in the chain* ![Display alternate](docs/images/display-alternate.png) #### `Debug: {:?}` - *will write out all error messaging with simplified backtracing* ![Display debug](docs/images/display-debug.png) #### `Alternate Debug: {:#?}` - *will write out all error messaging with simplified backtracing* ![Display alternate debug](docs/images/display-alternate-debug.png) ## Contribute Pull requests are always welcome. However understand that they will be evaluated purely on whether or not the change fits with my goals/ideals for the project. ### Git-Hook Enable the git hooks to have automatic version increments ```bash cd ~/Projects/witcher git config core.hooksPath .githooks ``` ## License This project is licensed under either of: * MIT license [LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT * Apache License, Version 2.0 [LICENSE-APACHE](LICENSE-APACHE) or http://www.apache.org/licenses/LICENSE-2.0 ### Contribution Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in this project by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions. --- ## Backlog * Add rust doc comments * Mechanism for converting to JSON maybe use `{:#?}` ## Changelog * 12/30/2020 * Corrected the minimum required `rustc` badge