get method returns `error` object on document not found

Rather than reserving error for internal, API and network errors, the get API method triggers the error handling logic if the document with the specified ID is not found (due to a 404 status code being sent by Elasticsearch), despite the response actually containing some useful details.

This means you have to compare the error.message with the string "Not Found", which makes error handling logic brittle and difficult to make generic.

(this is compounded by the fact that all errors share a default type, rather than the “not found” error being comparable as error instanceof NotFoundError)

For now, I’m working around with issue by just using the Search endpoint with a match query, which I know is less than ideal for getting a specific document by id.

[spalger]

There is actually an error property exported by the elasticsearch module which includes all of the errors that the client sends back. In the case of 404 responses you can use either the 404 or NotFound class:

var elasticsearch = require('elasticsearch');

var NotFoundError = elasticsearch.errors[404];
// or
var NotFoundError = elasticsearch.errors.NotFound;

The error objects all expose their status code as well, so you can also check for 404’s with that property:

client.get(...)
.catch(function (err) {
  if (err.status === 404) {
    // do something
  } else {
    throw err;
  }
});

Finally, if you don’t want 404’s to create errors you can pass ignore: [404] and 404 repsonses will be treated as successful!

client.get({
  index: ...
  type: ...
  id: ...
  ignore: 404
})
.then(function (resp) {
  if (!resp.found) {
    // ...
  } else {
    // ...
  }
})

[spalger]

Well, in an exceptional case like that here is what would happen:

1. the host/es node that the request was directed at would be removed from the connection pool
2. other nodes would be tried until:
   • a request succeeds (aka, a response was received, even 404 is considered a success at this point)
   • the connection pool is exhausted
   • the maxRetry count is reached
3. if a request was successful
   1. the response is converted to either a user response or error based on the status code and response body
4. if a request could not be completed
   1. the final error received is wrapped in a ConnectionFault error and passed back to the caller

All of this logic is implemented by the Transport class and more details can be found there.