Single level of abstraction / Better separation of concerns in query methods

[rudolfbyker]

Previously, we had .get and .post, which did something like this:

get:

• Build a query.
• If the URL is too large, throw everything away and call post.
• Perform HTTP GET.

post:

• Build a query.
• Perform HTTP POST.

The problem with this is that neither GET nor POST what one would expect:

• Neither are general HTTP GET or POST utilities.
• They work specifically with Solr Query objects.
• They expect JSON responses only.
• They send their data as query strings only (even .post, which uses application/x-www-form-urlencoded).

Therefore this can be much better encapsulated as:

doQuery:

• Take a Solr Query object, and prepare data for an HTTP query from that.
• Decide whether GET or POST should be used (note that this behaviour comes from Search Solr using POST instead of GET #129 – I did not invent it)
• Call doRequest.

doRequest:

• Build the query.
• Send the query.

This maintains a single level of abstraction per method, and therefore gives better separation of concerns and results in more understandable code.

[kibertoad]

path.resolve wouldn't do same thing in a simpler way?

[rudolfbyker]

That's possible. We can optimize in a next PR.

[rudolfbyker]

This code was moved / consolidated from e.g. here https://github.com/lbdremy/solr-node-client/pull/280/files/5531fb5405c18ccb0f15284a633474279d710544#diff-3b89c2e1c7e4e714474d6a2f17b55814a790482f4ef375d4f84d4895554c20d9L584 So it's not new. Just refactored.