Validation errors are not consistent for local and remote databases.

[tohagan]

From my project unit tests I've identified that in order to get identical errors returned from a local or remote database you need the following code change

if (typeof e.unauthorized !== "undefined") {
  throw {
    name: "unauthorized",
    message: 'Name or password is incorrect.',
    reason: e.unauthorized,
    status: 401
  };
} else if (typeof e.forbidden !== "undefined") {
  throw {
    name: "forbidden",
    message: 'Forbidden by design doc validate_doc_update function',
    reason: e.forbidden,
    status: 403
  };

• renamed message field to reason
• added new message field text

[marten-de-vries]

Nice find. Hmm, is reason new in PouchDB? I thought it converted reason -> name and error -> message always. Unless that changed this might actually be a PouchDB bug?

[nolanlawson]

@marten-de-vries that is a bug. reason should not be in there.

[tohagan]

Might be a rather nasty breaking change to PouchDB to alter the error fields as there is sure to be tons of code that depends on it ... what do you think? I needed consistency so that I can write an app that switches data source to be: remote, local (live sync on/off) and db API is consistent. I'm also fetching ddocs from server.

[marten-de-vries]

@tohagan Any reason why you can't just use err.name? The value should be the same as err.reason, and it's what PouchDB recommends as well.

[marten-de-vries]

I just ported over the old pouchdb-validation test suite, seems like this is indeed a PouchDB bug. And that it's not as easy to circumvent as I assumed in my last post. Going to open a PouchDB bug.

[marten-de-vries]

As soon as pouchdb/pouchdb#4670 is fixed, the test I mentioned in that bug can be re-enabled and then we can close this.

[tohagan]

Thanks for all your test suite porting and chasing this issue Marten. Really appreciated.

[SCdF]

@marten-de-vries looks like the linked ticket (which was a dupe of pouchdb/pouchdb#5214) has now been fixed.

[marten-de-vries]

That should solve this issue without further work being required. Before closing though, I think this test (well, the equivalent in the latest master branch) should be re-enabled, so we can be sure:

https://github.com/pouchdb/pouchdb-validation/pull/5/files#diff-fb0d18ca230e188e02029b32dbb851d8R28