Expose error creation functions

[nirvdrum]

I don't know if you really want to expose this functionality or not, but it helps me address the problem I raised in #88. With this change, I can add custom middleware to do the following:

next => async req => {
  Sentry.setTag('graphql_query', req.getID())

  try {
    const res = await next(req)

    if (res.errors) {
      const error = createRequestError(req, res)
      Sentry.captureException(error)
    }

    return res
  } catch (e) {
    Sentry.captureException(e)
    throw e
  }
}

Or:

next => async req => {
  Sentry.setTag('graphql_query', req.getID())

  try {
    const res = await next(req)

    if (res.errors) {
      const errorMessage =
        req instanceof RelayNetworkLayerRequest
          ? formatGraphQLErrors(req, res.errors)
          : JSON.stringify(res.errors)
      Sentry.captureMessage(errorMessage)
    }

    return res
  } catch (e) {
    Sentry.captureException(e)
    throw e
  }
}

That conditional creation of the error message comes from the implementation of createRequestError. I don't really like that it needs to do an instanceof check, but it's unavoidable without changing the type definition for formatGraphQLErrors.

I also am not 100% certain that I've provided the correct type for RelayResponse#errors. As far as I can tell, I did. The type definition appears to match what I saw in the debugger. But, I don't know if the field was typed as any for any particular reason.

[nirvdrum]

Please feel free to suggest improvements or modify the commits directly as you see fit. If you don't like this PR at all, please comment on #88 with a better option, if you have the time, and I'll attempt another PR to solve the problem.

[nodkz]

Looks good! 👍
Sorry for the late response.

[nodkz]

🎉 This PR is included in version 4.5.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀