* Do not retry a request if the body is a stream
Refactored the trnasport.request method to not use stream for gzipping
the body, but use the callback API instead. The maxRetries will be 0 in
case of a stream body and cached the Accept-Encoding header.
* Updated dependencies
* Updated test
* Updated abort behavior
- Support for aborting a request with the promise api
- Aborting a request will cause a RequestAbortedError
- Normalized Connection class errors, now every error returned is
wrapped by the client errors constructors
* Updated test
* Updated docs
* Updated code generation script
* Renamed test
* Code coverage
* Avoid calling twice transport.request
* Updated Connections handling
- The ConnectionPool.getConnection method now always returns a connection
- The ConnectionPool.update mehtod now cleans the dead list
- Deprecated the NoLivingConnectionsError
* Updated test
* Updated docs
* The NoLivingConnectionsError can still happen if the filter/selector returns no nodes
* Updated test
* Updated docs
* Catch undefined connections as well
* Updated test
* Updated ApiError type def
In very few cases, some API uses the same key for both url and query params, such as the bulk method.
The client is not designed to handle such cases since accepts both url and query keys in the same object, and the url parameter will always take precedence.
This pr fixes this edge case by adding a `querystring` key in the options object.
Fixes: https://github.com/elastic/elasticsearch-js/pull/778
```js
client.bulk({
index: 'index',
type: '_doc',
body: [...]
}, {
querystring: {
type: '_doc'
}
}, console.log)
```
- The result object contains also the metadata about the request
- The events emits the same object of the API response
- The errors, where possible, exposes the APi response object under the
meta key
With this pr we introduce the `client.child` API, which returns a new client instance that shares the connection pool with the parent client.
This feature can be handy if you need to have multiple client instances with different configurations, but with a shared connection pool.
Example:
```js
const { Client } = require('@elastic/elasticsearch')
const client = new Client({ node: 'http://localhost:9200' })
const child = client.child({
headers: { 'x-foo': 'bar' },
requestTimeout: 1000
})
client.info(console.log)
child.info(console.log)
```
**Open questions:**
* Currently, the event emitter is shared between the parent and the child(ren), is it ok?
* Currently, if you extend the parent client, the child client will have the same extensions, while if the child client adds an extension, the parent client will not be extended. Is it ok?
**Caveats:**
* You can override _any_ option except for the connection pool specific options (`ssl`, `agent`, `pingTimeout`, `Connection`, and `resurrectStrategy`).
* You can't specify a new `Connection` class.
* If you call `close` in any of the parent/child clients, every client will be closed.
_Note: the `nodeFilter` and `nodeSelector` options are now `Transport` options and no longer `ConnectionPool` options._
- Added support for different format of requestTimemout
- Changed api method result
- Now we are always returning the result in case of error
- Improved body deserialization
- Added cast to boolen for HEAD requests
- Added support for already serialized strings in the ndserializer
- Fixed qserializer in case of null object
- Updated Errors
