* Start on lua scripting
* Implement evalsha, script load, script exists, and script flush
* Type conversions from lua to resp/json.
Refactor to make luastate and luascripts persistent in the controller.
* Change controller.command and all underlying commands to return resp.Value.
Serialize only during the ouput.
* First stab at tile38 call from lua
* Change tile38 into tile38.call in Lua
* Property return errors from scripts
* Minor refactoring. No locking on script run
* Cleanup/refactoring
* Create a pool of 5 lua states, allow for more as needed. Refactor.
* Use safe map for scripts. Add a limit for max number of lua states. Refactor.
* Refactor
* Refactor script commands into atomic, read-only, and non-atomic classes.
Proper locking for all three classes.
Add tests for scripts
* More tests for scripts
* Properly escape newlines in lua-produced errors
* Better test for readonly failure
* Correctly convert ok/err messages between lua and resp.
Add pcall, sha1hex, error_reply, status_reply functions to tile38 namespace in lua.
* Add pcall test. Change writeErr to work with string argument
* Make sure eval/evalsha never attempt to write AOF
* Add eval-set and eval-get to benchmarks
* Fix eval benchmark tests, add more
* Improve benchmarks
* Optimizations and refactoring.
* Add lua memtest
* Typo
* Add dependency
* golint fixes
* gofmt fixes
* Add scripting commands to the core/commands.json
* Use ARGV for args inside lua
The tile38-benchmark tool is a new addition to the Tile38 suite alongside
tile38-server and tile38-cli. It's modeled after redis-benchmark with
matching input flags and output formatting.
It tests the common commands PING, GET, SET, and NEARBY. More commands
may be added later.
(mention #54)
Josh Baker