Include indices of duplicate rows in the Non_Unique_Primary_Key error

[radeusgd]

As discussed, it could be useful to include the indices of duplicate rows in the error indicating that the selected primary key was not unique.

To not stall that PR, since this has a lower priority, it's separated into this task.

• Add duplicate_rows : Vector Integer field to Non_Unique_Primary_Key.Error.
• Adapt its to_display_text to display what rows were duplicated.
• Fill in this field when reporting the error.
  • We should fill in row indices of one group that has duplicate entries. See below.

If we have a table

#	X
0	a
1	b
2	a
3	a
4	b
5	c

and set the primary key to be X, we have values a and b duplicated - so duplicated_rows should contain rows of one of these groups - i.e. it should either be [0, 2, 3] or [1, 4] (and it is unspecified which one). It shouldn't be [0, 1, 2, 3, 4] - if we include duplicates from all groups, in some cases we could just list all rows of a table (if every row has some duplicate) and that would not be helpful.

[radeusgd]

After further thought and discsussion with @jdunkerley we decided to display the values associated with the primary key of the first row that has a clashing primary key. That is because it can be easily used to find related rows, is easy to display and is portable (row ids were not viable in Database as they are ill-defined there).

So for the table above, the result would be: clashing_primary_key=["a"] and the following message: The primary key [X] is not unique. For example, the key [a] corresponds to more than one row.

[enso-bot]

Radosław Waśko reports a new STANDUP for yesterday (2023-05-08):

Progress: Improved Non_Unique_Primary_Key error. Improved SQL File Format detection and in general format detection. It should be finished by 2023-05-09.

Next Day: Next day I will be working on the #6543 task. Work on Date_Range