Purpose of drop_column down is unclear in the documentation

[simonw]

This section here:

https://github.com/xataio/pgroll/blob/9d2515963865ba4b5b71fe0ff9ea85e8ce9e2ddf/docs/README.md#drop-column

A drop column operation drops a column from an existing table.

drop column operations have this structure:

{
   "drop_column": {
    "table": "name of table",
    "column": "name of column to drop",
    "down": "SQL expression"
  }
}

It's not obvious to me what the "down" key here is used for.

[exekias]

Thank you for opening this issue!

Down migrations are applied to adapt data from the latest schema version into the previous one. During the active migration period, you will have a new schema version that doesn't have the column, while the old schema (that has it) is still available.

If the deployed version of your application expects a value in that column (or there is a constraint requiring it, ie NOT NULL), down function can be used to ensure that all inserts/updates happening in the new version of the schema get a valid value backfilled into this column for the old version of the schema (and its clients) to consume.

Hope this clarifies things? I realize docs may need some more explanation, we can use this issue to ensure we update them!