fix(couchdb): use url.PathEscape in DBExists for consistency with other DB* methods

[dominicbarnes]

Currently, url.PathEncode is applied for CreateDB and DestroyDB, but not for DBExists, which behave inconsistently.

I discovered this when I created a database that included special characters, but DBExists would not correctly detect a database even after being created.

For a reproducible example, I could not find a good test case to update, as the request path is never part of the test assertion. You can reproduce the issue with the following setup:

docker-compose.yml

services:
  couchdb:
    image: couchdb:3
    ports:
      - 5984:5984
    environment:
      COUCHDB_USER: admin
      COUCHDB_PASSWORD: password

go.mod

module kivik-testing

go 1.23.0

require github.com/go-kivik/kivik/v4 v4.3.0

require (
	github.com/google/uuid v1.6.0 // indirect
	golang.org/x/net v0.17.0 // indirect
	golang.org/x/sync v0.4.0 // indirect
)

main.go

package main

import (
	"context"
	"log"
	"net/http"
	"net/url"

	"github.com/go-kivik/kivik/v4"
	_ "github.com/go-kivik/kivik/v4/couchdb"
)

func main() {
	ctx := context.Background()

	couchdb, err := kivik.New("couch", "http://admin:password@localhost:5984")
	if err != nil {
		log.Fatal(err)
	}

	dbName := "foo/bar"

	// ensure database is created
	if err := couchdb.CreateDB(ctx, dbName); err != nil {
		switch kivik.HTTPStatus(err) {
		case http.StatusPreconditionFailed:
			// do nothing, database already exists
		default:
			log.Fatalf("failed to create database %s: %s", dbName, err)
		}
	}

	// check existence with existing kivik
	exists, err := couchdb.DBExists(ctx, dbName)
	if err != nil {
		log.Fatalf("failed to check database %s existence", dbName)
	}
	log.Printf("exists (got %t, expected true)", exists)

	// check existince with hack
	exists, err = couchdb.DBExists(ctx, url.PathEscape(dbName))
	if err != nil {
		log.Fatalf("failed to check database %s existence", dbName)
	}
	log.Printf("exists with hack (got %t, expected true)", exists)
}

Upon running my "test" script, I get the following output:

$ go run .
exists (got false, expected true)
exists with hack (got true, expected true)

This PR includes a fix, which is to use url.PathEscape within DBExists which mirrors CreateDB and DestroyDB and will produce consistent/correct behavior.

It would be nice to validate this fix with a test case, but I don't think the current test suite really has this kind of assertion baked in (it kinda assumes the right URL path is used, since it just returns the configured *http.Response no matter what the request).

[flimzy]

Thank you for catching this and submitting a patch! It would be good to add a test, as well, which would fail prior to the patch, but pass after. Is that something you'd like to add?

[dominicbarnes]

@flimzy yeah, let me throw something together

[dominicbarnes]

@flimzy I just pushed bb18c2a which adds a test that fails before the change and passes after.

To avoid altering any other tests, this creates a new test helper which creates a client which inspects req.URL.RawPath as part of the HTTP mock, returning an error if there is a mismatch.

This probably isn't the most elegant approach, but it certainly gets the job done short of pulling in a full HTTP mocking library. Let me know what you think, I'm happy to iterate on it further if you have any suggestions.

[dominicbarnes]

@flimzy after looking over some other tests, I found some other examples of inspecting req.URL via newCustomClient, so I've ditched the one-time helper in 8c9b10a

I think this way is more consistent with the rest of the existing tests, so let me know what you think.

[flimzy]

Looks great. Thanks for the contribution! I'll roll a patch release shortly.