A table as it was when you read it: the rows, and where each of them sat.

A snapshot is what a write is planned against, so it is the value to hold on to between reading and writing, and the value that goes stale. Sheetshow.Table.refresh/2 re-reads the id column alone and gives back the same snapshot with the positions it finds.

header is the header row as it actually reads on the tab, in the tab's own order, not the schema's. A table is written by row and column index, so it can find its columns by name and leave alone whatever a person keeps to the right of them.

rows is every row on the tab, tombstones included. Sheetshow.Table.live/1 is the live ones, which is what most code wants.