Optional parameter to force column Data Type

[johha86]

To the extension class "ExcelDataReaderExtensions" was added a new optional parameter. This new parameter allow specify the value type associated to a column.

The structure of this parameter is a list of ExcelDataType.

The definition of an ExcelDataType is:

public class ExcelDataType
{
public string ColumnName { get; set; }
public Type ColumnType { get; set; }
}

[andersnm]

Hi and thank you! Unfortunately I believe this is a problematic change. In particular because of Mono, which apparently has slight differences in its DataSet implementation from the "official" ones. See #409 / #73 / #30.

Currently by default ExcelDataReader tries to autodetect the type for DataColumn.DataType and the linked issues are reporting problems when there are nulls in the data. I think we should fix the autodetector so that it always works (i.e: assign nullable types in case of nulls, also configure pr sheet), rather than providing a type override mechanism which can easily break (on mono)

If you can tell a bit more about your scenario, maybe we can come up with a better solution?

[johha86]

Hi, this is my scenary.

When I read and Excel document that contains an value of type Decimal, it always is save in the DataTable returned by the extensor method as a Double. But in this case I want it as a Decimal cause I'll save the DataTable into the database as Decimal.

This is just a specific scenario. But a very common situation is that someone want to read the value from the Excel with an specific data type. Otherwise he/she must retrieve the information from the original DataTable, perform a casting or type conversion and then create a new DataTable to be stored into the Database. This is memory consuming solution obviously.

[andersnm]

Some kind of new ExcelDataTableConfiguration.TransformValue(reader, columnIndex) callback could be a flexible way to handle this and other scenarios. F.ex I had a use case where I wanted to apply formatting on every value in the DataSet, which is similar to yours, but converting to DB types instead. Happy to accept a PR for this.

Alternatively, if you need a quick resolve you might find it easier to build the dataset manually yourself, f.ex by copying the AsDataSet method into your project and adapt+simplify as necessary. AsDataSet() is theoretically quite simple, but comes with lots of bells and whistles.