Add support for describing prepared statements

[akheron]

Describing a prepared statement is useful determining inferred input parameter types and result column types for prepared statements.

Usage:

client.query({
  text: 'your statement here',
  describe: true
}).then(result => {
  // result.fields contains the output fields as usual
  // result.params (added in this PR) contains the OIDs of input parameters
})

When describe is given, the query is not bound or executed at all. The only commands sent to Posgres are Parse, Describe and Flush. Sync will be sent after the ParameterDescription message is obtained .

[akheron]

Tests fail because it seems I still need to add support to native.

Before doing that, it would be nice to know whether this is a desirable feature or not in the first place :)

[brianc]

Sorry for the delay - I like this approach. I'm cool releasing functionality in the non-native bindings w/o requiring it to be released in the native bindings as well...though it's always nice to have as much parity as possible. I'm down with merging this & doing a new minor version bump w/ the new functionality later this week. Mind rebasing on master? The tests should pass now.

[akheron]

Rebased on master, but the native tests still seem to fail 🤔 Did you do something that should prevent them from failing?

[charmander]

I think a more explicit and restricted API would be better (Client#describe(…)), since there’s not enough overlap with querying with regard to what it expects and how it behaves to justify making it an option. Plus it doesn’t get ignored by pg-native/older versions of pg and do something completely different. If it has to be implemented as a secret _describe: true option on query, that’s okay.

[charmander]

(It doing something completely different on pg-native is exactly why the tests are failing, too.)