Stream: git-wasmtime

Topic: wasmtime / PR #1790 wiggle: allow user-configurable error...

Wasmtime GitHub notifications bot (May 29 2020 at 20:09):

pchickey opened PR #1790 from pch/wiggle_error_transforms to master:

One issue we've had with wiggle ergonomics is that we often want to use a rich error type (a Rust enum that uses thiserror::Error, or maybe an anyhow::Error) to write the library code that ends up in an impl of the wiggle-defined module traits. However, those module traits require an error type that corresponds to the witx document.

Because Rust doesn't have try, there's not a lot of good mechanical ways to automatically replace all uses of a witx document error type like wasi's Errno with a rich type and a conversion function.
So, this PR adds a totally opt-in facility to wiggle to perform such a conversion on the user's behalf.

The new test src/wiggle/tests/errors.rs shows how the types work out on a modified version of the array test. A new optional wiggle proc-macro argument errors gives a mapping of witx identifiers (less the leading $, because proc macro parsing is enough trouble without that) to the user's desired error type, e.g.:

wiggle::from_witx!({
    witx: ["tests/arrays.witx"],
    ctx: WasiCtx,
    errors: { errno => RichError },
});

When the errors config is provided, a UserErrorConversion trait is defined in types that the user must implement for their ctx type. The method names in this enum are determined mechanically.

impl<'a> types::UserErrorConversion for WasiCtx<'a> {
    fn errno_from_rich_error(&self, _e: RichError) -> types::Errno {
        unimplemented!();
    }
}

The RichError type is then used in place of types::Errno in the witx module traits:

impl<'a> arrays::Arrays for WasiCtx<'a> {
    fn populate_excuses(&self, _excuses: &types::ExcuseArray) -> Result<(), RichError> {
        unimplemented!()
    }
   (snip)
}

Open questions:

• The syntax of the arguments is pretty arbitrary. I wanted to permit full paths to the user's rich error type, but then I ran into a problem figuring out a sensible and unique method name from that, so I punted. Ideally we'd add optional syntax for specifying the method name.
• I wasn't always consistent about what to call this feature in the code - is this an ErrorTransform, is this a UserError? The naming could stand some bikeshedding.

<!--

Please ensure that the following steps are all taken care of before submitting
the PR.

• [ ] This has been discussed in issue #..., or if not, please tell us why
  here.

• [ ] A short description of what this does, why it is needed; if the
  description becomes long, the matter should probably be discussed in an issue
  first.

• [ ] This PR contains test cases, if meaningful.

• [ ] A reviewer from the core maintainer team has been assigned for this PR.
  If you don't know who could review this, please indicate so. The list of
  suggested reviewers on the right can help you.

Please ensure all communication adheres to the code of conduct.
-->

Wasmtime GitHub notifications bot (May 29 2020 at 20:09):

pchickey requested alexcrichton and kubkon for a review on PR #1790.

Wasmtime GitHub notifications bot (May 29 2020 at 20:09):

pchickey requested alexcrichton and kubkon for a review on PR #1790.

Wasmtime GitHub notifications bot (May 29 2020 at 20:10):

pchickey edited PR #1790 from pch/wiggle_error_transforms to master:

One issue we've had with wiggle ergonomics is that we often want to use a rich error type (a Rust enum that uses thiserror::Error, or maybe an anyhow::Error) to write the library code that ends up in an impl of the wiggle-defined module traits. However, those module traits require an error type that corresponds to the witx document.

Because Rust doesn't have try, there's not a lot of good mechanical ways to automatically replace all uses of a witx document error type like wasi's Errno with a rich type and a conversion function.
So, this PR adds a totally opt-in facility to wiggle to perform such a conversion on the user's behalf.

The new test src/wiggle/tests/errors.rs shows how the types work out on a modified version of the array test. A new optional wiggle proc-macro argument errors gives a mapping of witx identifiers (less the leading $, because proc macro parsing is enough trouble without that) to the user's desired error type, e.g.:

wiggle::from_witx!({
    witx: ["tests/arrays.witx"],
    ctx: WasiCtx,
    errors: { errno => RichError },
});

When the errors config is provided, a UserErrorConversion trait is defined in types that the user must implement for their ctx type. The method names in this enum are determined mechanically.

impl<'a> types::UserErrorConversion for WasiCtx<'a> {
    fn errno_from_rich_error(&self, _e: RichError) -> types::Errno {
        unimplemented!();
    }
}

The RichError type is then used in place of types::Errno in the witx module traits:

impl<'a> arrays::Arrays for WasiCtx<'a> {
    fn populate_excuses(&self, _excuses: &types::ExcuseArray) -> Result<(), RichError> {
        unimplemented!()
    }
   (snip)
}

Open questions:

• The syntax of the arguments is pretty arbitrary. I wanted to permit full paths to the user's rich error type, but then I ran into a problem figuring out a sensible and unique method name from that, so I punted. Ideally we'd add optional syntax for specifying the method name, and require the syntax if a sensible name cant be determined by some simple check (currently https://docs.rs/syn/1.0.28/syn/struct.Path.html#method.get_ident)
• I wasn't always consistent about what to call this feature in the code - is this an ErrorTransform, is this a UserError? The naming could stand some bikeshedding.

<!--

Please ensure that the following steps are all taken care of before submitting
the PR.

• [ ] This has been discussed in issue #..., or if not, please tell us why
  here.

• [ ] A short description of what this does, why it is needed; if the
  description becomes long, the matter should probably be discussed in an issue
  first.

• [ ] This PR contains test cases, if meaningful.

• [ ] A reviewer from the core maintainer team has been assigned for this PR.
  If you don't know who could review this, please indicate so. The list of
  suggested reviewers on the right can help you.

Please ensure all communication adheres to the code of conduct.
-->

Wasmtime GitHub notifications bot (May 29 2020 at 20:12):

pchickey edited PR #1790 from pch/wiggle_error_transforms to master:

One issue we've had with wiggle ergonomics is that we often want to use a rich error type (a Rust enum that uses thiserror::Error, or maybe an anyhow::Error) to write the library code that ends up in an impl of the wiggle-defined module traits. However, those module traits require an error type that corresponds to the witx document.

Because Rust doesn't have try, there's not a lot of good mechanical ways to automatically replace all uses of a witx document error type like wasi's Errno with a rich type and a conversion function.
So, this PR adds a totally opt-in facility to wiggle to perform such a conversion on the user's behalf.

The new test src/wiggle/tests/errors.rs shows how the types work out on a modified version of the array test. A new optional wiggle proc-macro argument errors gives a mapping of witx identifiers (less the leading $, because proc macro parsing is enough trouble without that) to the user's desired error type, e.g.:

pub enum RichError { (snip) }
wiggle::from_witx!({
    witx: ["tests/arrays.witx"],
    ctx: WasiCtx,
    errors: { errno => RichError },
});

When the errors config is provided, a UserErrorConversion trait is defined in types that the user must implement for their ctx type. The method names in this enum are determined mechanically.

impl<'a> types::UserErrorConversion for WasiCtx<'a> {
    fn errno_from_rich_error(&self, _e: RichError) -> types::Errno {
        unimplemented!();
    }
}

The RichError type is then used in place of types::Errno in the witx module traits:

impl<'a> arrays::Arrays for WasiCtx<'a> {
    fn populate_excuses(&self, _excuses: &types::ExcuseArray) -> Result<(), RichError> {
        unimplemented!()
    }
   (snip)
}

Open questions:

• The syntax of the arguments is pretty arbitrary. I wanted to permit full paths to the user's rich error type, but then I ran into a problem figuring out a sensible and unique method name from that, so I punted. Ideally we'd add optional syntax for specifying the method name, and require the syntax if a sensible name cant be determined by some simple check (currently https://docs.rs/syn/1.0.28/syn/struct.Path.html#method.get_ident)
• I wasn't always consistent about what to call this feature in the code - is this an ErrorTransform, is this a UserError? The naming could stand some bikeshedding.

<!--

Please ensure that the following steps are all taken care of before submitting
the PR.

• [ ] This has been discussed in issue #..., or if not, please tell us why
  here.

• [ ] A short description of what this does, why it is needed; if the
  description becomes long, the matter should probably be discussed in an issue
  first.

• [ ] This PR contains test cases, if meaningful.

• [ ] A reviewer from the core maintainer team has been assigned for this PR.
  If you don't know who could review this, please indicate so. The list of
  suggested reviewers on the right can help you.

Please ensure all communication adheres to the code of conduct.
-->

Wasmtime GitHub notifications bot (May 29 2020 at 20:28):

alexcrichton submitted PR Review.

Wasmtime GitHub notifications bot (May 29 2020 at 20:52):

kubkon submitted PR Review.

Wasmtime GitHub notifications bot (May 30 2020 at 21:07):

pchickey updated PR #1790 from pch/wiggle_error_transforms to master:

One issue we've had with wiggle ergonomics is that we often want to use a rich error type (a Rust enum that uses thiserror::Error, or maybe an anyhow::Error) to write the library code that ends up in an impl of the wiggle-defined module traits. However, those module traits require an error type that corresponds to the witx document.

Because Rust doesn't have try, there's not a lot of good mechanical ways to automatically replace all uses of a witx document error type like wasi's Errno with a rich type and a conversion function.
So, this PR adds a totally opt-in facility to wiggle to perform such a conversion on the user's behalf.

The new test src/wiggle/tests/errors.rs shows how the types work out on a modified version of the array test. A new optional wiggle proc-macro argument errors gives a mapping of witx identifiers (less the leading $, because proc macro parsing is enough trouble without that) to the user's desired error type, e.g.:

pub enum RichError { (snip) }
wiggle::from_witx!({
    witx: ["tests/arrays.witx"],
    ctx: WasiCtx,
    errors: { errno => RichError },
});

When the errors config is provided, a UserErrorConversion trait is defined in types that the user must implement for their ctx type. The method names in this enum are determined mechanically.

impl<'a> types::UserErrorConversion for WasiCtx<'a> {
    fn errno_from_rich_error(&self, _e: RichError) -> types::Errno {
        unimplemented!();
    }
}

The RichError type is then used in place of types::Errno in the witx module traits:

impl<'a> arrays::Arrays for WasiCtx<'a> {
    fn populate_excuses(&self, _excuses: &types::ExcuseArray) -> Result<(), RichError> {
        unimplemented!()
    }
   (snip)
}

Open questions:

• The syntax of the arguments is pretty arbitrary. I wanted to permit full paths to the user's rich error type, but then I ran into a problem figuring out a sensible and unique method name from that, so I punted. Ideally we'd add optional syntax for specifying the method name, and require the syntax if a sensible name cant be determined by some simple check (currently https://docs.rs/syn/1.0.28/syn/struct.Path.html#method.get_ident)
• I wasn't always consistent about what to call this feature in the code - is this an ErrorTransform, is this a UserError? The naming could stand some bikeshedding.

<!--

Please ensure that the following steps are all taken care of before submitting
the PR.

• [ ] This has been discussed in issue #..., or if not, please tell us why
  here.

• [ ] A short description of what this does, why it is needed; if the
  description becomes long, the matter should probably be discussed in an issue
  first.

• [ ] This PR contains test cases, if meaningful.

• [ ] A reviewer from the core maintainer team has been assigned for this PR.
  If you don't know who could review this, please indicate so. The list of
  suggested reviewers on the right can help you.

Please ensure all communication adheres to the code of conduct.
-->

Wasmtime GitHub notifications bot (Jun 01 2020 at 22:12):

pchickey requested alexcrichton for a review on PR #1790.

Wasmtime GitHub notifications bot (Jun 02 2020 at 16:37):

alexcrichton submitted PR Review.

Wasmtime GitHub notifications bot (Jun 02 2020 at 16:38):

pchickey merged PR #1790.

Last updated: Dec 23 2024 at 12:05 UTC