Submitted PR for hopefully now stable GoLang OpenWhisk runtime

[Michele Sciabarra <mi...@sciabarra.com>]

Hello, I just submitted a PR for a pretty large amount of work I did to stabilize the OpenWhisk runtime for Go.

The user-visible changes are:

- updated the runtime to golang 1.11
- changed the signature of actions to "func Main(args map[string]interface{}) map[string]interface{}"
- implemented all the mandatory tests for runtimes, both to GoLang and "pure" ActionLoop (with shell scripts)
- the runtime can still be used as a compiler to precompile and redure the startup time, but now in a simpler way as a filter (no more volume mounts)
- In the tests, there are also examples showing the base runtime (the action loop) is also a good runtime for shell script

Under the hood, I did a lot of work:

- internally now the compiler is a python script, no more a shell script, easier to understand an maintain
- I use now the same code to process the action at init and when I use the runtime as a compiler
- the compiler is now compiling also packages and vendor folder
- I optimized code in many places to avoid converting byte in string and copying files
- I fixed a lot of bugs that emerged trying to implement the test, everything from Unicode handling to environment  variables to unexpected limits in buffers and so on
- I also refactored tests to be easier to maintain (I had since I run them a lot of times)

The runtime is really very general, not just for Go. Just a python script,  a docker file, and an "epilogue" file should allow supporting a lot of languages, and I hope to release some interesting developments I have in mind in the next few months.  At least I will complete soon the Swift runtime I promised Carlos and the work I did is also making easier to support more languages. And also be used alone as a basis as an alternative to the dockerskeleton.  Examples in the unit tests shows this usage and the runtime passes those tests even without golang (just using bash).


-- 
  Michele Sciabarra
  michele@sciabarra.com

Re: Submitted PR for hopefully now stable GoLang OpenWhisk runtime

[Carlos Santana <cs...@gmail.com>]

Thanks Michele the changes look very good I did a first pass and left some
comments

On Sat, Sep 15, 2018 at 7:20 PM Michele Sciabarra <mi...@sciabarra.com>
wrote:

> Hello, I just submitted a PR for a pretty large amount of work I did to
> stabilize the OpenWhisk runtime for Go.
>
> The user-visible changes are:
>
> - updated the runtime to golang 1.11
> - changed the signature of actions to "func Main(args
> map[string]interface{}) map[string]interface{}"
> - implemented all the mandatory tests for runtimes, both to GoLang and
> "pure" ActionLoop (with shell scripts)
> - the runtime can still be used as a compiler to precompile and redure the
> startup time, but now in a simpler way as a filter (no more volume mounts)
> - In the tests, there are also examples showing the base runtime (the
> action loop) is also a good runtime for shell script
>
> Under the hood, I did a lot of work:
>
> - internally now the compiler is a python script, no more a shell script,
> easier to understand an maintain
> - I use now the same code to process the action at init and when I use the
> runtime as a compiler
> - the compiler is now compiling also packages and vendor folder
> - I optimized code in many places to avoid converting byte in string and
> copying files
> - I fixed a lot of bugs that emerged trying to implement the test,
> everything from Unicode handling to environment  variables to unexpected
> limits in buffers and so on
> - I also refactored tests to be easier to maintain (I had since I run them
> a lot of times)
>
> The runtime is really very general, not just for Go. Just a python
> script,  a docker file, and an "epilogue" file should allow supporting a
> lot of languages, and I hope to release some interesting developments I
> have in mind in the next few months.  At least I will complete soon the
> Swift runtime I promised Carlos and the work I did is also making easier to
> support more languages. And also be used alone as a basis as an alternative
> to the dockerskeleton.  Examples in the unit tests shows this usage and the
> runtime passes those tests even without golang (just using bash).
>
>
> --
>   Michele Sciabarra
>   michele@sciabarra.com
>