Standardizing on a single error type

[kevinburkeshyp]

Hi Brian - we just got bit by #938 and I am hoping to fix the issue.

Per your comment here:

This is probably something that should be normalized within node-postgres itself rather than here in the native lib. If you wanna take a crack at it over there I'll happily review the PR!

It sounds like your proposed solution is to keep the node-postgres error API consistent, and transform error objects coming from pg-native, instead of:

• changing the error objects here to match the ones in the pg-native library
• changing the error objects in pg-native to match the error here.

does that sound right? i'll start working on a patch.

[vitaly-t]

Aligning error reporting between pg and pg.native would be a substantial breaking change.

In fact, I think it would be so serious, the library would need to change over to version 5.0.0 to accommodate the change, or else it will break the hell of people's code.

If we are to try this within 4.x branch, this universal error reporting should be made optional, off by default.

[kevinburkeshyp]

I also think this would require a major version bump. I think it's worth it, though, especially when the library advertises that 'pg-native' is a drop-in replacement.

[vitaly-t]

especially when the library advertises that 'pg-native' is a drop-in replacement.

Unfortunately, this isn't true, and not just because of the error handling, but some most important modules, like pg-query-stream do not even support the native bindings. In fact, last I tried - the connection pool didn't work for me, so I gave up on the native version long ago.

The differences are too substantial.

[kevinburkeshyp]

In fact, last I tried - the connection pool didn't work for me

can you elaborate on this?

[vitaly-t]

It just doesn't work at all, what's more elaborate than that? :)

[vitaly-t]

@kevinburkeshyp I'd like to retract my previous statement here.

See: brianc/node-pg-native#46

[vitaly-t]

After having added pg-native support into pg-promise, I could see how many differences do exist, especially within the area of errors reporting. In fact, errors are completely different.

And if someone uses errors and moves from one version of the library to another - that will be a pain.

[vitaly-t]

@kevinburke any progress here? I think this will make one useful addition to the library.

I've done a lot of implementation for custom errors. See here some nice code: https://github.com/vitaly-t/pg-promise/blob/master/lib/errors.js

In the same way, you would have:

• custom error codes and messages separate;
• custom toString to provide nice formatting for string concatenation
• override for inspect to return .toString(), to provide the same nice output into the console.

just to give you a good example of how to write custom errors for Node.js ;)

[vitaly-t]

@kevinburkeshyp if someone decides to implement this feature, the following projects should be consulted: pg-rethrow + pg-error-codes.

@brianc There seems to be a lot of helpers around to unite all this under one roof and make consistent across the entire platform. Would you consider a PR for this? If so, please label it as a feature request.

[rkaw92]

Got bitten by this peculiarity today after replacing require('pg') with require('pg').native. I had had some error handlers that checked the error.code property for known values, because the code expected some tables to already exist and treated a failed CREATE TABLE like a nominal case, but only for specified error codes ("SQL states").

Now I place error.code === '42P07' || error.sqlState === '42P07' in my handlers...

[brianc]

Since this is a breaking change I needed to hold off until doing a new major version bump. I've added this to the pg@7.0 milestone so we can fix it & normalize these in the new release.

[brianc]

I have this fixed on the 7.0 branch. Fix will be released in 7.0.