Test for ExaminePage, fix two bugs it found.

[danvk]

This constructs /contigs, /spec and /genotypes responses from a VCF file.

The two bugs were:

1. The initial request for /genotypes was being throttled, which artificially slowed page load by 500ms.
2. The exploded rows from multiple samples were being dropped.

I also deleted some code that wasn't being used to boost our code coverage. I suspect that if/when we revive this code, it will be substantially different. So deleting is not a big loss. But feel free to disagree.

This PR bumps our test coverage from 25%→53%.

[tavinathanson]

Well, that solves the code quality issue w.r.t. bar charts! 💇

[tavinathanson]

The difference between these two methods isn't too clear from the names! I remember being unclear between the two when I was doing a RecordStore read-through.

I realize we don't want a super long name for a method that's called a bunch of places, but I'm still putting in a vote for updateGenotypesThrottled.

[tavinathanson]

Also, it looks like the delay is due to debounce rather than throttle? (Not that I'm suggesting you should use throttle, here.)

[ihodes]

True about delay being due to debounce rather than throttle.

[danvk]

Changed the comment to "debounce".

[danvk]

Only sorta solves the code quality problem -- that was just one of two copies of d3.bar-chart.js in this repo :(

[tavinathanson]

So what is this giving us?

[ihodes]

The key needs to be unique on all records; previously, if there were multiple samples in the VCF, the keys wouldn't be unique (since we "explode" each VCF row into N genotype rows (N = # of samples)). This fixes that, so we don't drop rows.

[ihodes]

May be better to _.pick(record, 'contig', 'position', …).join()

[danvk]

_.pick returns an object, so I'd also have to do _.values. It winds up being longer than this is. I can change if you prefer, though.

[danvk]

Isaac -- can you clarify why REF/ALT are in the key? I don't understand why they're necessary.

[ihodes]

Shoot, thought it returned a list. Do with it what you will! (Wouldn't suggest making it longer, though!)

[ihodes]

REF/ALT are necessary to make sure the variant is unique--you can have different variants at the same chromosome/position. It's still not necessarily totally theoretically unique, but it gets us closer, within reason.

[danvk]

Addressed all comments & rebased.