Inconsistency of type conversions with the table in README

[lwshang]

Some entries in the type conversion table in README are wrong:

kdb/q	r (wrong)	r (correct)
timespan	character	difftime
time	difftime	integer

Correct means what actually happen in rkdb.
Please check whether the README or the code should be fixed.

[sv]

Are you up for submitting a patch for readme?

…

On Mon, 9 Apr 2018, 11:55 pm Shang Linwei, ***@***.***> wrote: Some entries in the type conversion table in README are wrong: kdb/q r (wrong) r (correct) timespan character difftime time difftime integer Correct means what actually happen in rkdb. Please check whether the README or the code should be fixed. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#27>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAA3BvYHp0YgV0wmZe9HADYJLULqHLktks5tm4RjgaJpZM4TMybq> .

[lwshang]

I'm not sure whether we should just correct readme.
Because it's reasonable to convert time to difftime, so maybe we should modify the from_time_kobject function to return a difftime object.

[lwshang]

Hi @sv ,

Are you considering this issue? Do you think it is better to change the way to deal with time instead of modifying readme?

I would like to fix it and make a pull request. But I want to have a consensus on it before start to working on it.

And could please to give some instructions on how to set up the develop tools for this project? I notice that you put the source codes into several separate files like common.c, base.c and sexp2k.c then include the header and these source files inside rkdb.c which will be compiled in the end. This is very different with my previous coding practice. What IDE or editor with plugin are you using to make the develop environment efficient?

[sv]

thanks for looking into submitting pull request.
I think initially, we can just fix up documentation and discuss how we should handle time types consistently.
Issue is that according to documentation where is no support for milliseconds in difftime(i might be wrong here) and we should not lose millis when converting. There also seem to be some consesus around bringing nanotime package into the mix.

for dev environment, i just use RStudio with devtools package and rarely need to move functions between files. Agree that common.c/base.c/sexp2k.c is a bit misleading and should be just header files that get imported in rkdb.c. We can cleanup this as we go

Let me know if you need any help with raising PR