Database extractor
Wandora has a feature to import any relational database. Information in the database will be converted into topic map format and then imported. The conversion works for (nearly) any kind of relational database and thus it is very general. For this reason converted databases usually need post prosessing, either manually or with other tools, before using in a production environment. Rest of this page describes how the conversion is done from relational database format into topic map format.
Basic idea is that each row of each table is converted into a topic, there are some exceptions to this rule. Since primary key(s) identify the row inside the table, it is natural to use the pimary keys with the table name as the subject identifier of the topic.
Values of each normal column, that is column that is not a reference to some other (or same) table, is converted into either an occurrence or a new topic which is then associated with the topic representing the row. Creating topics for column values is natural when the value is shared by many rows, for example when the value represents a year. Creating occurrences is better suited for situations where most or all rows have unique values and creating separate topics would only make the topic map more complex without offering any additional linkage between topics. Note: In current version of Wandora all columns are converted into occurrences. The code is in place to make associations instead of occurrences but there is no user interface or any other mechanism to select which columns should be converted to associations.
Columns that reference some other column in another or same table are converted into associations. This is the natural thing to do as the referenced row is also a topic and the reference associates these two topics (rows). Note: Current version of Wandora does not support database schemas where several columns together reference another row.
Tables that only contain references are treated differently. Each row in such a table is converted into an association instead of a topic. This is done because each row links several topics without adding any other information. Thus creating an association for each row is more meaningful than creating a topic.
Tables with only one primary key that is also a foreign key are also treated slightly differently. This kind of table can usually be thought to be extending another row in the table that is referenced. For this reason subject identifier is created with the table name of the referenced table. This will cause topics of both this and the referenced tables to merge. This means that information from both tables will be in the same topic.
