- moved es install/start/stop logic into a seperate script
- `grunt test` now runs the integration tests once for each version of ES we support
- grunt can now install and run elasticearch (using grunt-run, pure js solution coming later)
- included seperate es.sh script specifically for starting or stopping elasticsearch
- url aliases, api, yaml_suite/index.js, and yaml_tests.json, are all now duplicated for 0_90 support
- the client now accepts an apiVersion argument (undocumented) which defaults to 'master' but can be '0.90'
- The yaml test runner will now check the name of the ES instance it is connecting to, preventing accidental wiping of ES
- reusing es downloads to make local testing faster
- periodic checks will kill the script on failure
- coveralls report is only sent once per build
- failures sending the coveralls report won't cause the build to fail
- removed several unneeded devDeps
- removed old get_spec.js script
- the client's ping method will now send back true as the body when the ping
succceeds, and false when it does not. When the ping fails, the error will
still be sent back and the connection's status will still be set to "dead".
- All of the client's methods now have a spec property, which will provide the
JSON spec used to run that method.
- The yaml test runner will only camelCase param names that are documented, uses
the client's method's new spec property
- Trace log events will now have their proper original query string parameters
- The "tracer" logger will now write to elasticsearch-tracer.log by default, and
will truncate the file if it already exists.
- When running the integration tests, the client will now use a tracer logger which
writes to stderr. The default level is "warning", but with the VERBOSE environment
var it becomes "trace" and the logger will write to it's default file
- Added .idea to the .gitignore, it was being published to NPM
- Cleanup of the grunt tasks. Consilidated several tiny files into seperate moderately sized ones.
- connection's ping method now accepts requestTimeout, path, and method params like
all the grown-up API calls
- ConnectionPool now managed connection timeouts. When a connection dies a
timeout object is created to track when the timeout is scheduled and the function
to call when it does. It also tracks how many times it has run to allow the timeout
to grow
- Timeouts now grow with use of `config.calcDeadTimeout` which is set to 'exponential'
by default, but can also be set to flat in order to always use the standard
deadTimeout. Exponential growth of the deadTimeout is stopped at config.maxDeadTimeout
which is set to 30 minutes by default.
- Connections no longer have a resuscitate method (too hard to spell). Now the
method is created dynamically as a part of the timeout object as it just calls
the connection's ping method and needed to access variables like revive attempts.
- Timeouts were moved to the transport layer, meaning that you need to capture the
abort method and abort the request yourself if you are handling connections
directly, ConnectionsAbstract's ping method does this.
- updated copyright
- several tempalate changes for the docs
- added a config for grunt-contrib-watch
- updated nock commit number
- fixed the coverage script
- removed the export_docs script
- added error message for legacy "es" users who don't have a version locked and have not upgraded
- host will now add auth to urls created with `#makeUrl()`
- Log class no longer looks for `config.loggers`
- The log class now properly escapes single quotes in trace logs
- Removed compiled yaml_tests.js from the repo
- Yaml suite will only log error and warning messages unless the VERBOSE env var is set
- createDefer is now a global setting, changed by modifying Transport.createDefer fubction
- wrote tests for Content-Type checking
- callbacks will now return the body and status of the request (if the request has completed) when an error occurs
- Stdio logger now adds "Elasticsearch " to the front of log messages to distinguish it from other output to stdout.
