BTW, I faced this problem while working in the Unitils example for the upcoming JUnit in Action 2nd edition book (http://manning.com/tahchiev/). It would be very nice if this patch is applied (or the problem fixed in some other way) soon, otherwise I will have to hack the examples to work around this issue.

Hi Felipe,

I found our original implementation of the expected data set not satisfactory, so I built a new one.
Following improvements are added:

* as you requested: if you add an empty table to the expected data set, the assertion will fail if the actual table is not empty

* better dealing with different types of values: comparing 1.00 with 1 or comparing dates used to give problems this is fixed now

* the comparison now also takes primary key columns into account: if you specify a pk value in the expected data set, this row will now be compared with the corresponding actual row with those primary key values.

* better output. The best matching differences are search and the actual content of the tables is added.
Example output:

java.lang.AssertionError: Assertion failed. Differences found between the expected data set and actual database content.
Found differences for table PUBLIC.TEST:

  Different row:
  pk, column1, column2
  "1", "value1", "xxxx"

  Best matching differences:
  column2: "xxxx" <-> "value2"


Actual database content:

PUBLIC.TEST
  COLUMN1, COLUMN2
  "value1", "value2"
  "value2", "value3"


Other example output:

java.lang.AssertionError: Assertion failed. Differences found between the expected data set and actual database content.
Expected table to be empty but found rows for table PUBLIC.TEST


Actual database content:

PUBLIC.TEST
  COLUMN1, COLUMN2
  "value1", "value2"
  "value2", "value3"


Still todo:
* make the date format configurable
* add special values like current_date
* create @NotExpectedDataSet


Could you please take a look? Remarks and suggestions are very welcome.

best regards,
Tim