The `timeout` parameter in docstring is misleading for following methods in elasticsearch.client.indices.py, renamed this param in query_params decorator and docstring to `request_timeout`:
1. create
2. open
3. close
4. delete
5. put_mapping
6. put_alias
7. update_aliases
8. delete_alias
9. put_template
10. delete_template
11. shrink
12. rollover
queue_size must be greater than or equal to thread_count, or else the main thread can hang trying to acquire a lock to write sentinel values to the queue during teardown.
* Adds scan test for exception & data yielded
* Adds scan test for fast route & initial search error
* Refactores scan implementation; better scroll test
* Adds tests clear_scroll & logger
To be consistent with the other part of library and to allow vendoring
without code change, use relative imports also in the
``elasticsearch.client.xpack`` module.
After Git repo is loaded to ES through bulk(), Line 180 performs an index refresh to make documents available for searching. The same operation is then repeated in Line 188; remove this duplicate refresh.
* Fixing flaky test wrt mock and threading.
Mock is apparetnly not threadsafe, and as a result will sometimes
cause tests to fail when we want to count the number of times a
mock function is called.
I've replaced the mock function with a side_effect, which allows for
locking and incrementing manually.
* put threading.Lock in globals
* testing
* more fixes
* downgrade docker-compose file version cause jenkins
* fixing matrix
* back to elasticsearch_version
* skip py3.7 for now there's a bug
* add build badge and remove 3.3
Sphinx doesn't play nice with 3.3
* sphinx 1.7 drops support for python 3.3
* updating test matrix
* remove travis
* updating badges
* In preperation for moving to Elastic's CI
Add dockerfile to build image for testing python client in a docker container
add docker-compose to create environment for running tests against specific versions of ES
Add Makefile to make it easy to execute tests and cleanup
add wait-for-elasticsearch.sh script to wait for elasticsearch before executing the tests.
* add pull and comment about pushing requring auth
* spelling errors
* adding CI directory for jenkins to pick up
fixing typo too
* fix test-matrix and create working run-tests
* add comment about why skipping python 2.6
* adding dockerfiles to create base images for ci
By passing a `Retry` object to `urlopen` instead of just `False`, it
prevents `urllib3` from having to create a Retry object with `from_int`.
`from_int` emits an undesirable log message when called with
`retries=False`, and this can be avoided by the method described above.
Nick Lang