Remove "history mode" and replace with single buffer

[hudson-newey]

At the moment we have two buffers for incoming subjects (in the data-source), and subjects with a decision (in the verification-grid).

This no longer makes sense, and we should instead use a single buffer with a verificationHead and renderHead.

The verification head should only ever increase. Representing where the user has applied decisions up to.

The renderHead can be used to indicate what items are being rendered (moving this head back in the buffer will start viewing history).

---

Notes to implementer

~Make sure that changing grid size is handled correctly~

• ~If the renderHead >= verificationHead, then the renderHead should not change~
• ~If the renderHead < verificationHead, then the renderHead should move back by the change delta. This is so that if you are on the first page of history, and change the grid size you get more items from history (instead of seeing unverified items)~

[atruskie]

I think now that we have a more concrete idea of "actions that need to be completed to advance", we don't need to do anything special about page sizes.

I.e. if page size changes and we end up rendering a partial page of verified and unverified items, it doesn't matter. All the logic for when to advance (has each item has all necessary actions applied?) remains the same..

I'm not sure about your rules, clarify for me?

If the renderHead >= verificationHead, then the renderHead should not change

Why? Users can page forward? That's one of the feature requests

If the renderHead < verificationHead, then the renderHead should move back by the change delta.

If I understand your intent, I think this is better expressed as:

• If renderHead < verificationHead then fetch a page starting at index renderHead - pageSize, else fetch a page be doing renderHead + pageSize. That way we're always fetching pages either side of the zero point.

[hudson-newey]

The two statements

If the renderHead >= verificationHead, then the renderHead should not change

If the renderHead < verificationHead, then the renderHead should move back by the change delta. This is so that if you are on the first page of history, and change the grid size you get more items from history (instead of seeing unverified items)

Are trying to address what would happen if the user changes their screen size while viewing history.

• If the user starts viewing history, then expands their window, I think that we should resolve the delta by rendering additional items that have already been verified by moving the render head back by the number of added tiles (no need to re-fetch data).
• If the user is not viewing history, then we should resolve the delta by fetching more items from the data-source.

---

I think you summarized it succinctly as

If renderHead < verificationHead then fetch a page starting at index renderHead - pageSize, else fetch a page be doing renderHead + pageSize. That way we're always fetching pages either side of the zero point.

---

I.e. if page size changes and we end up rendering a partial page of verified and unverified items, it doesn't matter. All the logic for when to advance (has each item has all necessary actions applied?) remains the same..

This will definitely be easier to implement, so (as a developer) I definitely prefer it.

(as a user) it's not how I initially expected it to work. In my mind, if I press the "previous page" button, I should only see previous items until I press a "Continue Verifying", "Jump to Unverified", or press "Next Page" until I find the page I was up to.

Raven supports changing the grid size and viewing history, so I wonder what precedence they have set?