Geek Specs

Ignore the Trekkie reference.  I’ve started to work on the data abstraction layer for Midori, after narrowing down the supported out-of-the-box relational databases.  In no particular order, they are SqlServer, SqlServerCe, MySql, Postgress, and Sqlite. Its being written in .Net 4.0 and hopefully it should port to Mono.

The main abstraction class Midori.Data.SqlAdapter makes use of the System.Data interfaces, like System.Data.IDataReader. Obviously this class will grow as other crud operations are added.  Each RDMS will have a class that will inherit Midori.Data.SqlAdapter, like Midori.Data.Sqlite.SqliteAdapter, which will hold code specific to each database.

There is a Connection class that wraps other connection objects and supplies OnConnect and OnClose events, which should be helpful for closing the connection objects for methods that pass back datareader objects.

Interesting Notes

System.Data.Common.DbConnectionStringBuilder is a handy class for building and inspecting values of a connection string in a hash like way.  Some Ado.Net layers, like System.Data.SqlServerCe, does not implement their own connection string builder. While it is nice to have, its not a necessity, especially when the common dbconnectionstringbuilder will suffice.

Postgres or Npgsql, has some interesting behavior. The following select query has Pascal casing for the names of the columns.

SELECT 'column1' AS Column1, 10 AS Age

However when the DataReader for Postgres returns the names of the columns, they are all lowercase, all the other databases return the column aliases in the same case as specified in SQL.

So in the tests, I had to rewrite the query to use all lowercase column aliases, like so:

SELECT 'column1' AS column1, 10 AS age