Struct ErasedError1 

pub struct ErasedError1(/* private fields */);

Type-erased error which is known not to be rusqlite::Error

For use transaction retry loops. Works well as DbError<EE1>.

Not interconvertible in either direction with anyhow::Error: That prevents:

• Smuggling a SQLITE_BUSY out of a retry loop.
• Failing to wrap something that might SQLITE_BUSY in a retry loop.

So, also, cannot impl std::error::Error, and doesn’t impl Display to avoid lossy uses via format!.

The 1 in the name stands for “within one retry”.

Implementations

impl ErasedError1

pub fn context1<C>(self, c: C) -> Self

where C: Display + Send + Sync + 'static,

pub fn new_not_sql(ae: Error) -> Self

Convert from an anyhow::Error, maybe risking races (SQLITE_BUSY)

Should not be needed in routine error handling paths. Instead, implement NotSqlError for the error type in question.

Correctness

You must be sure the error is not a rusqlite::Error internally, and definitely not SQLITE_BUSY.

pub fn unwrap_after_retries(self) -> Error

Convert to an anyhow::Error, when you’ve done needed retries

Should not be needed in routine error handling paths.

Will be needed to transform error types, next to the retry loop, from the possibly-retry-inducing errors experienced inside the loop, to the final errors seen outside it.

Correctness

It is correct to call this function on an error which remains even after needed db (transaction) retries have been done.

Using this function elsewhere can allow a SQLITE_BUSY, which ought to trigger a retry, to be returned to naive code, causing failures. Such a race bug would be hard to spot in tests.

Trait Implementations

impl Debug for ErasedError1

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<E> From<E> for ErasedError1

where E: NotSqlError, Error: From<E>,

fn from(e: E) -> Self

Converts to this type from the input type.