It's not possible.
Documentation. It is stated in Cucumber specification that examples section should be a table with single header row and all other rows with data:
The table must have a header row corresponding to the variables in the
Scenario Outline steps.
Each of the rows below will create a new Scenario, filling in the
variable values.
Technical part. Specflow uses Gherkin language parser which parses feature files into AST (Abstract Syntax Tree). This parser reads input file lines one by one and produces token for each line. From TokenScanner source code:
The scanner reads a gherkin doc (typically read from a .feature file)
and creates a token for each line. The tokens are passed to the
parser, which outputs an AST (Abstract Syntax Tree).
Parser then reads tokens one by one and determines what kind of line this token is related to. When it finds examples line, then it looks for single header line (Examples_Definition) followed by examples. If you will do something like this
|I |am | on | the |
|data |data |data |data |
|Office |Portal |link |
|data |data |data |
then the third line will be treated as a examples row with an incorrect number of cells.
Actually how the parser would be able to understand whether you have simple data row or headers row? All examples data are represented as strings. Aand Office is a completely valid value for I column. OK in this particular sample you have 3 columns in bottom table definition part. But if there would be four columns too?
|I |am | on | the |
|data |data |data |data |
|Office |Portal |link | foo |
|data |data |data |data |
There is no way to define second header row for the parser.
Readabiltiy. And frankly speaking, such multiline tables would be unreadable for people. It would be very hard to correlate third scenario outline parts from bottom and top part of examples table. But readability is the main feature of gherkin and Specflow. Consider redesigning your tests if you have to provide so many parameters for the scenario.
