2015-01-19-2001-EST

Mon Jan 19 20:01:26 EST 2015

# journal - Identified my main Pain point
Anything concerning accessing or modifying the motionInfo.
It's ALL WRONG!
Accessing through the interface interacts with the input vacuum go routine.
Then the motionInfo is served from the interface function.
But I made a change inside the (*Player).collideWith() function to use the method and...
It blew up and deadlocks during the tests.

`ff "\.mi\." | grep "_test" | wc -l` -> `108`
Theres 108 lines of test code that dig into this internal variable instead of using the interface.
This is because using the interface subscribes you to interacting with the input vacuum loop.
While this behavior is correct in a way, it is confusing.
The way I understand it now, I don't understand why I'm deadlocking.
This is part of the problem.

I think this needs to be completely rethought and re-executed.
Using what I know about designing packages would be a good place to start.
If I can start laying down some solid walls with the package barrier these issues should start to disappear.
If they don't disappear, having a package barrier will aid me in rewriting with identifier collision.

If I learned anything from the journal rewrite
It was that porting tests to a new package was WAY more productive than
Trying to rewrite them in place.
This is absolutely because of the way I wrote the tests initially.
I don't know how much better the tests are for journal now in this regard.
The same rigidity may still exist.
Maybe that's the point of tests?

You can't write tests like those in rob pikes [APL interrupter](https://github.com/robpike/ivy) for every project can you?
If I can write tests like this, I need to.
How do I do this for my Math library?
Maybe I should use that as a case study to discover how I can do this better.
Using the package barrier Helps because it forces all external tests to use a public interface.
Even if the implementation internal to the package changes only the tests in that package should be affected.
At least, that's the way it felt when I was working on journal as I was pkgizing it.
Maybe this really is the key though, it enables quicker red/green feedback as well.
Because!!!
Even if you change the public API during a rewrite.. You only have to worrry about going green in the current package.
If you break the package's users, so what, you fix that after you've re stabilized.
Really a bingo moment going on right here.
Predicting lots of progress tomorrow.

Noticed that I believe journal removed whitespace around a `###` header in a idea that I made in the last entry.
Should take a look at that.

Mon Jan 19 20:25:22 EST 2015