Copy: Show error message on failure

[bgamari]

This ensures that the exception thrown includes the cause of the error. Sadly it seems that the only way to determine that a COPY IN has failed is to fail to parse the row count (as the result status is typically an empty string. It seems that all of the getResults calls after the putCopyEnd command inexplicably return success, despite the fact that something failed.

[lpsmith]

It would be nice to throw an exception other than the default ErrorCall exception, but it seems this is a step in the right direction anyway.

[lpsmith]

Also, it would be nice to have a demo of several failure modes of COPY IN, possibly as part of the test suite, though I think that's slightly less important than the comment above.

[bgamari]

Leon P Smith [email protected] writes:

Also, it would be nice to have a demo of several failure modes of COPY
IN, possibly as part of the test suite, though I think that's slightly
less important than the comment above.

I actually have such a thing locally. I'll try to clean it up and
integrate it into the testsuite.

[lpsmith]

Hmm, what's the testing strategy where you commit some expected output, then you can simply inspect the diffs as you work on the code? That seems like a good strategy here, though I'm certainly open to other possibilities.

[lpsmith]

This issue made me think of #181, which seems at least somewhat related. I had a small bit of difficulty finding that issue; I didn't remember that it was a PR, that it was recent, and that it was you!

[bgamari]

Leon P Smith [email protected] writes:

Hmm, what's the testing strategy where you commit some expected
output, then you can simply inspect the diffs as you work on the code?
That seems like a good strategy here, though I'm certainly open to
other possibilities.

I assume you mean the approach taken by tasty-golden? If so then sure,
that sounds sensible. In that case would you accept a dependency on
tasty-golden?

[bgamari]

Leon P Smith [email protected] writes:

This issue made me think of #181, which seems at least somewhat
related. I had a small bit of difficulty finding that issue; I didn't
remember that it was a PR, that it was recent, and that it was you!

Indeed, my trouble with #181 was in part that the underlying issue was
being masked by the lack of a sensible error message.

[lpsmith]

Sure, having the test suite depend on tasty-golden sounds fine to me. (Also, I'm less picky about test suite dependencies.)

[lpsmith]

Also, I'm not that surprised that this is the fix that was needed in #181, it was something vaguely along these lines I was sort of expecting in the first place.