Moving A Cell

If a user cut-pastes a cell, they are intending to move it. So the Cell ID (block ID) needs to be included in the HTML content in the clipboard.

If the cell is moved to another row, but the column is the same, this will result in a MoveBlock operation. If the content has changed, it will also have a ReplaceBlock op.

Empty Cells do not need blocks (or IDs)

If a user deletes a cell, or if they cut without pasting anywhere, this results in a deleted block.

Workflow Psuedocode

Load the document state, with the HMBlockNode hierarchy

Convert HMBlockNode to BlockNoteTree(EditorState), with the full table structure - New Code

User edits the table - draft saved as BlockNoteTree(EditorState)

On publish, convert BlockNoteTree(EditorState) back to HMBlockNode - New Code

Evaluate for operations - Code is unchanged

Publish operations into a change blob

Edge Case 1 - Concurrent Cell Insertion

Two users concurrently attempt to insert the same cell location. They can both succeed, resulting in two conflicting cells in the HMBlockNode hierarchy

We need a tie breaker, ideally last write wins.

The block hierarchy can have a revisionTime field for each block. The highest revision for the two conflicting cells will win.

Edge Case 2 - Orphan Cells

Concurrently add a row and delete a column. Final state has orphan cell which can be ignored

Edge Case 3 - Excessive Move Operations of Cells

We will have unnecessary move operations for cells. Because the cells will be ordered in column-order when converting from EditorState to HMBlockNode

Its not that big of a deal

Dealing with Row and Column Headers

Should we allow html-messy content with multi column headers and multi row headers?

No text fields on column and row blocks.

IsHeader attribute on rows and columns?

hm://z6MkqULds7oeqGPgtSpqpisK77FxhwP57ikVNFVfYfsMb2KZ/:profile will decide