multi-textarea data model keyed on (url, timestamp)

まず、既存の Spot datatype と、issue に記載されている現在の textarea-enhancement flow を追跡します。ドラフトがどのように識別・復元されているかを、提案されている URL/timestamp スキーマと比較します。完了の条件は、複数の textarea のドラフトを保持し、一致させるための汎用モデルについてプロジェクト内で合意できていることです。

As the code stands, there is a Spot datatype. It represents the abstract concept of where a single textarea is. e.g.

• appending the 2nd comment on issue 4
• appending the 3rd comment on issue 4
• pros: we have enough data to recreate the draft completely in the browser, and also enough data to tell if it was submitted or not
• cons:
  • a lot of manual work for each place that we handle
  • ambiguity around things like the issues/new url. You might have multiple unsubmitted drafts in progress, they don't have server-side IDs yet, what should their spot be? Inevitably it gets tied to the time that the tab happened to open, in which case why bother with the complexity of defining the "abstract place where a comment is" if you end up just doing timestamp, URL pairs?

This approach really falls down for complex multi-form things, such as this:

It would be really painful to lose the drafts in those textboxes, but it's also impossible to make a spot for each textfield in every issue/PR template that might be out there.

Another approach is this:

• anytime a textarea gets added or removed, trigger a refractory state that waits until there's been no changes for 500ms
• when that refactory period times out (the textareas have settled), the Spot is defined as:
  • primary key: (url, timestamp)
  • schema: list of textareas and their labels, as best we can tell in a generic way
  • title and icon: pulled from the tab by default
• cons: most of the time, this is not enough data to reopen the tab for the user
• pros:
  • if the user has opened a tab manually, we can find which saved drafts fit that schema
  • textareas can be enhanced by default, versus right now where they are enhanced only iff we have built an enhancer for that specific box