If the requested query is a proplist, then send it as json. This allows
for compound queries.

Add support for complex search queries

Fix bulk indexing

new param ctimeout, updated to hackney 0.4.4

closes #9, adding http_client_options

Some servers behind firewall will only support http/https but not git
protocol due to company security policy

Change dependencies' protocol from git to https

hackney:request can also return {error, ...}

so the caller would know what went wrong

…_errors

returning elasticsearch errors

In its delete_by_query api, elasticsearch allows the query to
be passed in the query string, but also as a json payload. This
should allow that, and also does some cleanup, by treating the
absence of a payload as a payload of 0 size (content-length)

allowing payload for methods like DELETE.

adding support for percolator feature

Add upsert_doc/4 and /5

Upgrade hackney to 0.11.1

Same as in normal case is same function.

New hackney return type

Added support for creating mappings

Conflicts:
	src/erlastic_search.erl

Add _stats call

Update Hackney to 0.13.1

The server is called Elasticsearch, in one word.
The mother company has been rebranded Elastic, so links are be updated.

Name and links

From the changelog for version 1.0 :

`Elasticsearch now runs in the foreground by default. There is no more -f flag on the command line.`

https://www.elastic.co/guide/en/elasticsearch/reference/master/_system_and_settings.html#_system_and_settings

Elasticsearch is always in the foreground

Fixed missing passing options to index_doc_with_id_opts

Mapping directly to
https://www.elastic.co/guide/en/elasticsearch/reference/current/cluster-nodes-info.html

Also fixing (?) `commas/1` which had invalid syntax, at the very least for erlang v18.x

Adding `nodes_info/*` funs

The proplist representing the header JSON in the first branch of 'bulk_index_docs' contains a redundant set of braces, causing jsx to throw an exception when it attempts to encode the structure. Modify the proplist to match the format expected by jsx.

Remove redundant braces for JSON input

`put_mapping` doesn't support taking a binary `Doc` despite its type specification. Create a guarded version of the function which does.

Support binary docs as input to 'put_mapping'

We already use `jiffy` in our project to handle JSONs, which makes
integrating `erlastic_search` a pain, since it uses `jsx`, which
has a different way of representing JSONs in Erlang, for example:

```
1> SimpleJson = <<"{\"key\":\"value\"}">>.
<<"{\"key\":\"value\"}">>
2> jiffy:decode(SimpleJson).
{[{<<"key">>,<<"value">>}]}
3> jsx:decode(SimpleJson).
[{<<"key">>,<<"value">>}]
```

This patch makes it possible to use any JSON library, as long as it defines
the two following callbacks:

```erlang
-callback encode(json()) -> binary().
-callback decode(binary()) -> json().
```

This setting is done at compile time, by defining an `ERLASTIC_SEARCH_JSON_MODULE`
environment variable containing the name of the module containing the two
callbacks above.

Updated the specs to account for this change.

Thus getting rid of the unsightly `?ERLASTIC_SEARCH_JSON_MODULE` occurrences
all over the code.

Getting rid of all the internally-built JSONs, to be able to not call `jsx`
directly any more, which could be confusing together with `erls_json`.

Making it possible to use another JSON library than `jsx`

Using hackney's option `async`

Wrapping hackney's reference in an `{async, Ref}' tuple to make it
easy to match.

Had to update the specs to account for that change

Allowing asynchronous ES queries

… of a bulk item

Extend bulk operation header

Add option to configure host and port

* making it possible to index without ID but with options
* making it possible to bulk index without IDs
* fixing bulk indexes which were broken if not using `jsx`
* also accounting for the fact that `erlastic_json()` can also be a map
* the only possibly breaking change is if someone out there used that
  lib to index a doc with the atom `undefined` as ID, which seems like
  a bad idea anyway (and is against the spec that it should be a binary)

Some index changes:

Before calling search_scroll/1 function, the search_scroll/4 function should be called with index, type name and the search query. Based on the search response, "scroll_id" can be passed to search_scroll/1 to get next batch results.

Added scroll support

the app should not depend on jsx app

Fix bulk index docs

Introduce get_doc_opts/5 as an equivalent to get_doc/4 including options

'multi_search/2' allows to execute multiple search requests using
the Elasticsearch endpoint '_msearch'. The function interface is
inspired by the 'bulk_index_docs' function with the difference that
the index and type parameters are missing. If needed these parameters
must be defined through the headers, as these are optionally
according to the Elasticsearch documentation.

Introduce new search function 'multi_search/2'

…ndex with custom settings

and/or mappings, as documented in
https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-create-index.html

Also modified `erlastic_search:create_index/2` to accept either connection params
and an index name (previous behaviour) or else an index name and a body.

This patch is entirely backwards-compatible.