Implementation Guideline

Database table is designed by following rules.

• PRIMARY KEY is minimum set of columns to specify one record uniquely.
• SERIAL (postgresql) or AUTO INCREMENT (mysql) are not used because it is hard to know the inserted value after insertion.
• Incremental integer identifier is counted from 1. Because many database systems do so and 0 can be regarded as a NULL.
• Index of some elements is counted from 0. Because most array data uses 0-origin indexing.
• INTEGER is used instead of BOOLEAN because some database system does not provide the BOOLEAN.
• UNIQUE is scarcely used. UNIQUE should be used when the system collapses if it is not used. And UNIQUE column can be PRIMARY KEY in many cases.
• NOT NULL should be used when the NULL value has a special meaning. If the column is an INTEGER, using 0 as a NULL and attaching NOT NULL restriction is mostly better.

Mostly, each database table has its accessor class.

Basically, database accessor should have minimal set of functions to get, update, insert and delete the data. More complex functions should be defined by Interface layer. But some complex functions which need to join several tables could be defined as a special case.

Usually, some of following functions are defined.

• Instantiation: db_gen
• Take: db_select_one, db_select_all, db_count
• Alteration: db_update, db_insert
• Check: db_exist_*

If some columns of a table requires to be multilingualized, two functions is implemented to take a value.

1. One is for taking the value of one language. Function name is same to column name.
2. Other is for taking the value of one language when the value is defined, otherwise one of other values of arbitrary language is taken. Function name has some_ as prefix.